maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/X86/fma.ll

2012 lines
138 KiB
LLVM
Raw Normal View History

[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; RUN: llc < %s -mtriple=i386-apple-darwin10 -mattr=+avx,+fma,-fma4 -show-mc-encoding | FileCheck %s --check-prefix=FMA32
; RUN: llc < %s -mtriple=i386-apple-darwin10 -mattr=+avx,-fma,-fma4 -show-mc-encoding | FileCheck %s --check-prefix=FMACALL32
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; RUN: llc < %s -mtriple=x86_64-apple-darwin10 -mattr=+fma,-fma4 -show-mc-encoding | FileCheck %s --check-prefix=FMA64
; RUN: llc < %s -mtriple=x86_64-apple-darwin10 -mattr=-fma,-fma4 -show-mc-encoding | FileCheck %s --check-prefix=FMACALL64
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; RUN: llc < %s -mtriple=x86_64-apple-darwin10 -mattr=+avx512f,-fma4 -show-mc-encoding | FileCheck %s --check-prefix=AVX512 --check-prefix=AVX51264
; RUN: llc < %s -mtriple=x86_64-apple-darwin10 -mattr=+avx512vl,-fma4 -show-mc-encoding | FileCheck %s --check-prefix=AVX512VL
[NFC][CodeGen][X86] fma.ll, lwp-intrinsics.ll: actually spell --check-prefixes correctly :/ llvm-svn: 343588
2018-10-02 21:34:50 +08:00
; RUN: llc < %s -mtriple=i386-apple-darwin10 -mcpu=bdver2 -mattr=-fma4 -show-mc-encoding | FileCheck %s --check-prefixes=FMA32,FMA32_BDVER2
; RUN: llc < %s -mtriple=i386-apple-darwin10 -mcpu=bdver2 -mattr=-fma,-fma4 -show-mc-encoding | FileCheck %s --check-prefixes=FMACALL32,FMACALL32_BDVER2
[X86] Cleanup fma tests a little bit. NFC. Reformat, isolate 213->231 xform, actually --check-prefix CHECK, and deduplicate the FMA intrinsic tests (FMA3 in AMD-land). llvm-svn: 240615
2015-06-25 08:40:25 +08:00
define float @test_f32(float %a, float %b, float %c) #0 {
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMA32-LABEL: test_f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA32: ## %bb.0: ## %entry
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMA32-NEXT: pushl %eax ## encoding: [0x50]
; FMA32-NEXT: vmovss {{[0-9]+}}(%esp), %xmm0 ## encoding: [0xc5,0xfa,0x10,0x44,0x24,0x08]
; FMA32-NEXT: ## xmm0 = mem[0],zero,zero,zero
; FMA32-NEXT: vmovss {{[0-9]+}}(%esp), %xmm1 ## encoding: [0xc5,0xfa,0x10,0x4c,0x24,0x0c]
; FMA32-NEXT: ## xmm1 = mem[0],zero,zero,zero
; FMA32-NEXT: vfmadd213ss {{[0-9]+}}(%esp), %xmm0, %xmm1 ## encoding: [0xc4,0xe2,0x79,0xa9,0x4c,0x24,0x10]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA32-NEXT: ## xmm1 = (xmm0 * xmm1) + mem
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMA32-NEXT: vmovss %xmm1, (%esp) ## encoding: [0xc5,0xfa,0x11,0x0c,0x24]
; FMA32-NEXT: flds (%esp) ## encoding: [0xd9,0x04,0x24]
; FMA32-NEXT: popl %eax ## encoding: [0x58]
; FMA32-NEXT: retl ## encoding: [0xc3]
;
; FMACALL32-LABEL: test_f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMACALL32: ## %bb.0: ## %entry
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMACALL32-NEXT: jmp _fmaf ## TAILCALL
; FMACALL32-NEXT: ## encoding: [0xeb,A]
; FMACALL32-NEXT: ## fixup A - offset: 1, value: _fmaf-1, kind: FK_PCRel_1
;
; FMA64-LABEL: test_f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA64: ## %bb.0: ## %entry
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMA64-NEXT: vfmadd213ss %xmm2, %xmm1, %xmm0 ## encoding: [0xc4,0xe2,0x71,0xa9,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA64-NEXT: ## xmm0 = (xmm1 * xmm0) + xmm2
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMA64-NEXT: retq ## encoding: [0xc3]
;
; FMACALL64-LABEL: test_f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMACALL64: ## %bb.0: ## %entry
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMACALL64-NEXT: jmp _fmaf ## TAILCALL
; FMACALL64-NEXT: ## encoding: [0xeb,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-1, kind: FK_PCRel_1
;
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-LABEL: test_f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: vfmadd213ss %xmm2, %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xa9,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; AVX512-NEXT: ## xmm0 = (xmm1 * xmm0) + xmm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: retq ## encoding: [0xc3]
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
;
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-LABEL: test_f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512VL: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: vfmadd213ss %xmm2, %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xa9,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; AVX512VL-NEXT: ## xmm0 = (xmm1 * xmm0) + xmm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: retq ## encoding: [0xc3]
Add an intrinsic and codegen support for fused multiply-accumulate. The intent is to use this for architectures that have a native FMA instruction. llvm-svn: 134742
2011-07-09 05:39:21 +08:00
entry:
[X86] Cleanup fma tests a little bit. NFC. Reformat, isolate 213->231 xform, actually --check-prefix CHECK, and deduplicate the FMA intrinsic tests (FMA3 in AMD-land). llvm-svn: 240615
2015-06-25 08:40:25 +08:00
%call = call float @llvm.fma.f32(float %a, float %b, float %c)
Add an intrinsic and codegen support for fused multiply-accumulate. The intent is to use this for architectures that have a native FMA instruction. llvm-svn: 134742
2011-07-09 05:39:21 +08:00
ret float %call
}
[X86] Cleanup fma tests a little bit. NFC. Reformat, isolate 213->231 xform, actually --check-prefix CHECK, and deduplicate the FMA intrinsic tests (FMA3 in AMD-land). llvm-svn: 240615
2015-06-25 08:40:25 +08:00
define double @test_f64(double %a, double %b, double %c) #0 {
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMA32-LABEL: test_f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA32: ## %bb.0: ## %entry
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMA32-NEXT: subl $12, %esp ## encoding: [0x83,0xec,0x0c]
; FMA32-NEXT: vmovsd {{[0-9]+}}(%esp), %xmm0 ## encoding: [0xc5,0xfb,0x10,0x44,0x24,0x10]
; FMA32-NEXT: ## xmm0 = mem[0],zero
; FMA32-NEXT: vmovsd {{[0-9]+}}(%esp), %xmm1 ## encoding: [0xc5,0xfb,0x10,0x4c,0x24,0x18]
; FMA32-NEXT: ## xmm1 = mem[0],zero
; FMA32-NEXT: vfmadd213sd {{[0-9]+}}(%esp), %xmm0, %xmm1 ## encoding: [0xc4,0xe2,0xf9,0xa9,0x4c,0x24,0x20]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA32-NEXT: ## xmm1 = (xmm0 * xmm1) + mem
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMA32-NEXT: vmovsd %xmm1, (%esp) ## encoding: [0xc5,0xfb,0x11,0x0c,0x24]
; FMA32-NEXT: fldl (%esp) ## encoding: [0xdd,0x04,0x24]
; FMA32-NEXT: addl $12, %esp ## encoding: [0x83,0xc4,0x0c]
; FMA32-NEXT: retl ## encoding: [0xc3]
;
; FMACALL32-LABEL: test_f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMACALL32: ## %bb.0: ## %entry
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMACALL32-NEXT: jmp _fma ## TAILCALL
; FMACALL32-NEXT: ## encoding: [0xeb,A]
; FMACALL32-NEXT: ## fixup A - offset: 1, value: _fma-1, kind: FK_PCRel_1
;
; FMA64-LABEL: test_f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA64: ## %bb.0: ## %entry
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMA64-NEXT: vfmadd213sd %xmm2, %xmm1, %xmm0 ## encoding: [0xc4,0xe2,0xf1,0xa9,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA64-NEXT: ## xmm0 = (xmm1 * xmm0) + xmm2
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMA64-NEXT: retq ## encoding: [0xc3]
;
; FMACALL64-LABEL: test_f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMACALL64: ## %bb.0: ## %entry
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMACALL64-NEXT: jmp _fma ## TAILCALL
; FMACALL64-NEXT: ## encoding: [0xeb,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fma-1, kind: FK_PCRel_1
;
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-LABEL: test_f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: vfmadd213sd %xmm2, %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xa9,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; AVX512-NEXT: ## xmm0 = (xmm1 * xmm0) + xmm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: retq ## encoding: [0xc3]
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
;
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-LABEL: test_f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512VL: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: vfmadd213sd %xmm2, %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xa9,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; AVX512VL-NEXT: ## xmm0 = (xmm1 * xmm0) + xmm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: retq ## encoding: [0xc3]
Add an intrinsic and codegen support for fused multiply-accumulate. The intent is to use this for architectures that have a native FMA instruction. llvm-svn: 134742
2011-07-09 05:39:21 +08:00
entry:
[X86] Cleanup fma tests a little bit. NFC. Reformat, isolate 213->231 xform, actually --check-prefix CHECK, and deduplicate the FMA intrinsic tests (FMA3 in AMD-land). llvm-svn: 240615
2015-06-25 08:40:25 +08:00
%call = call double @llvm.fma.f64(double %a, double %b, double %c)
Add an intrinsic and codegen support for fused multiply-accumulate. The intent is to use this for architectures that have a native FMA instruction. llvm-svn: 134742
2011-07-09 05:39:21 +08:00
ret double %call
}
[X86] Cleanup fma tests a little bit. NFC. Reformat, isolate 213->231 xform, actually --check-prefix CHECK, and deduplicate the FMA intrinsic tests (FMA3 in AMD-land). llvm-svn: 240615
2015-06-25 08:40:25 +08:00
define x86_fp80 @test_f80(x86_fp80 %a, x86_fp80 %b, x86_fp80 %c) #0 {
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMA32-LABEL: test_f80:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA32: ## %bb.0: ## %entry
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMA32-NEXT: subl $60, %esp ## encoding: [0x83,0xec,0x3c]
; FMA32-NEXT: fldt {{[0-9]+}}(%esp) ## encoding: [0xdb,0x6c,0x24,0x40]
; FMA32-NEXT: fldt {{[0-9]+}}(%esp) ## encoding: [0xdb,0x6c,0x24,0x50]
; FMA32-NEXT: fldt {{[0-9]+}}(%esp) ## encoding: [0xdb,0x6c,0x24,0x60]
; FMA32-NEXT: fstpt {{[0-9]+}}(%esp) ## encoding: [0xdb,0x7c,0x24,0x20]
; FMA32-NEXT: fstpt {{[0-9]+}}(%esp) ## encoding: [0xdb,0x7c,0x24,0x10]
; FMA32-NEXT: fstpt (%esp) ## encoding: [0xdb,0x3c,0x24]
; FMA32-NEXT: calll _fmal ## encoding: [0xe8,A,A,A,A]
; FMA32-NEXT: ## fixup A - offset: 1, value: _fmal-4, kind: FK_PCRel_4
; FMA32-NEXT: addl $60, %esp ## encoding: [0x83,0xc4,0x3c]
; FMA32-NEXT: retl ## encoding: [0xc3]
;
; FMACALL32-LABEL: test_f80:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMACALL32: ## %bb.0: ## %entry
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMACALL32-NEXT: subl $60, %esp ## encoding: [0x83,0xec,0x3c]
; FMACALL32-NEXT: fldt {{[0-9]+}}(%esp) ## encoding: [0xdb,0x6c,0x24,0x40]
; FMACALL32-NEXT: fldt {{[0-9]+}}(%esp) ## encoding: [0xdb,0x6c,0x24,0x50]
; FMACALL32-NEXT: fldt {{[0-9]+}}(%esp) ## encoding: [0xdb,0x6c,0x24,0x60]
; FMACALL32-NEXT: fstpt {{[0-9]+}}(%esp) ## encoding: [0xdb,0x7c,0x24,0x20]
; FMACALL32-NEXT: fstpt {{[0-9]+}}(%esp) ## encoding: [0xdb,0x7c,0x24,0x10]
; FMACALL32-NEXT: fstpt (%esp) ## encoding: [0xdb,0x3c,0x24]
; FMACALL32-NEXT: calll _fmal ## encoding: [0xe8,A,A,A,A]
; FMACALL32-NEXT: ## fixup A - offset: 1, value: _fmal-4, kind: FK_PCRel_4
; FMACALL32-NEXT: addl $60, %esp ## encoding: [0x83,0xc4,0x3c]
; FMACALL32-NEXT: retl ## encoding: [0xc3]
;
; FMA64-LABEL: test_f80:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA64: ## %bb.0: ## %entry
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMA64-NEXT: subq $56, %rsp ## encoding: [0x48,0x83,0xec,0x38]
; FMA64-NEXT: fldt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x6c,0x24,0x40]
; FMA64-NEXT: fldt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x6c,0x24,0x50]
; FMA64-NEXT: fldt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x6c,0x24,0x60]
; FMA64-NEXT: fstpt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x7c,0x24,0x20]
; FMA64-NEXT: fstpt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x7c,0x24,0x10]
; FMA64-NEXT: fstpt (%rsp) ## encoding: [0xdb,0x3c,0x24]
; FMA64-NEXT: callq _fmal ## encoding: [0xe8,A,A,A,A]
[llvm-mc] - Produce R_X86_64_PLT32 for "call/jmp foo". For instructions like call foo and jmp foo patch changes relocation produced from R_X86_64_PC32 to R_X86_64_PLT32. Relocation can be used as a marker for 32-bit PC-relative branches. Linker will reduce PLT32 relocation to PC32 if function is defined locally. Differential revision: https://reviews.llvm.org/D43383 llvm-svn: 325569
2018-02-20 18:17:57 +08:00
; FMA64-NEXT: ## fixup A - offset: 1, value: _fmal-4, kind: reloc_branch_4byte_pcrel
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMA64-NEXT: addq $56, %rsp ## encoding: [0x48,0x83,0xc4,0x38]
; FMA64-NEXT: retq ## encoding: [0xc3]
;
; FMACALL64-LABEL: test_f80:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMACALL64: ## %bb.0: ## %entry
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMACALL64-NEXT: subq $56, %rsp ## encoding: [0x48,0x83,0xec,0x38]
; FMACALL64-NEXT: fldt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x6c,0x24,0x40]
; FMACALL64-NEXT: fldt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x6c,0x24,0x50]
; FMACALL64-NEXT: fldt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x6c,0x24,0x60]
; FMACALL64-NEXT: fstpt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x7c,0x24,0x20]
; FMACALL64-NEXT: fstpt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x7c,0x24,0x10]
; FMACALL64-NEXT: fstpt (%rsp) ## encoding: [0xdb,0x3c,0x24]
; FMACALL64-NEXT: callq _fmal ## encoding: [0xe8,A,A,A,A]
[llvm-mc] - Produce R_X86_64_PLT32 for "call/jmp foo". For instructions like call foo and jmp foo patch changes relocation produced from R_X86_64_PC32 to R_X86_64_PLT32. Relocation can be used as a marker for 32-bit PC-relative branches. Linker will reduce PLT32 relocation to PC32 if function is defined locally. Differential revision: https://reviews.llvm.org/D43383 llvm-svn: 325569
2018-02-20 18:17:57 +08:00
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmal-4, kind: reloc_branch_4byte_pcrel
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMACALL64-NEXT: addq $56, %rsp ## encoding: [0x48,0x83,0xc4,0x38]
; FMACALL64-NEXT: retq ## encoding: [0xc3]
;
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-LABEL: test_f80:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: subq $56, %rsp ## encoding: [0x48,0x83,0xec,0x38]
; AVX512-NEXT: fldt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x6c,0x24,0x40]
; AVX512-NEXT: fldt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x6c,0x24,0x50]
; AVX512-NEXT: fldt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x6c,0x24,0x60]
; AVX512-NEXT: fstpt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x7c,0x24,0x20]
; AVX512-NEXT: fstpt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x7c,0x24,0x10]
; AVX512-NEXT: fstpt (%rsp) ## encoding: [0xdb,0x3c,0x24]
; AVX512-NEXT: callq _fmal ## encoding: [0xe8,A,A,A,A]
[llvm-mc] - Produce R_X86_64_PLT32 for "call/jmp foo". For instructions like call foo and jmp foo patch changes relocation produced from R_X86_64_PC32 to R_X86_64_PLT32. Relocation can be used as a marker for 32-bit PC-relative branches. Linker will reduce PLT32 relocation to PC32 if function is defined locally. Differential revision: https://reviews.llvm.org/D43383 llvm-svn: 325569
2018-02-20 18:17:57 +08:00
; AVX512-NEXT: ## fixup A - offset: 1, value: _fmal-4, kind: reloc_branch_4byte_pcrel
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: addq $56, %rsp ## encoding: [0x48,0x83,0xc4,0x38]
; AVX512-NEXT: retq ## encoding: [0xc3]
;
; AVX512VL-LABEL: test_f80:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512VL: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: subq $56, %rsp ## encoding: [0x48,0x83,0xec,0x38]
; AVX512VL-NEXT: fldt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x6c,0x24,0x40]
; AVX512VL-NEXT: fldt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x6c,0x24,0x50]
; AVX512VL-NEXT: fldt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x6c,0x24,0x60]
; AVX512VL-NEXT: fstpt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x7c,0x24,0x20]
; AVX512VL-NEXT: fstpt {{[0-9]+}}(%rsp) ## encoding: [0xdb,0x7c,0x24,0x10]
; AVX512VL-NEXT: fstpt (%rsp) ## encoding: [0xdb,0x3c,0x24]
; AVX512VL-NEXT: callq _fmal ## encoding: [0xe8,A,A,A,A]
[llvm-mc] - Produce R_X86_64_PLT32 for "call/jmp foo". For instructions like call foo and jmp foo patch changes relocation produced from R_X86_64_PC32 to R_X86_64_PLT32. Relocation can be used as a marker for 32-bit PC-relative branches. Linker will reduce PLT32 relocation to PC32 if function is defined locally. Differential revision: https://reviews.llvm.org/D43383 llvm-svn: 325569
2018-02-20 18:17:57 +08:00
; AVX512VL-NEXT: ## fixup A - offset: 1, value: _fmal-4, kind: reloc_branch_4byte_pcrel
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: addq $56, %rsp ## encoding: [0x48,0x83,0xc4,0x38]
; AVX512VL-NEXT: retq ## encoding: [0xc3]
Add an intrinsic and codegen support for fused multiply-accumulate. The intent is to use this for architectures that have a native FMA instruction. llvm-svn: 134742
2011-07-09 05:39:21 +08:00
entry:
[X86] Cleanup fma tests a little bit. NFC. Reformat, isolate 213->231 xform, actually --check-prefix CHECK, and deduplicate the FMA intrinsic tests (FMA3 in AMD-land). llvm-svn: 240615
2015-06-25 08:40:25 +08:00
%call = call x86_fp80 @llvm.fma.f80(x86_fp80 %a, x86_fp80 %b, x86_fp80 %c)
Add an intrinsic and codegen support for fused multiply-accumulate. The intent is to use this for architectures that have a native FMA instruction. llvm-svn: 134742
2011-07-09 05:39:21 +08:00
ret x86_fp80 %call
}
[X86] Cleanup fma tests a little bit. NFC. Reformat, isolate 213->231 xform, actually --check-prefix CHECK, and deduplicate the FMA intrinsic tests (FMA3 in AMD-land). llvm-svn: 240615
2015-06-25 08:40:25 +08:00
define float @test_f32_cst() #0 {
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMA32-LABEL: test_f32_cst:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA32: ## %bb.0: ## %entry
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMA32-NEXT: flds LCPI3_0 ## encoding: [0xd9,0x05,A,A,A,A]
; FMA32-NEXT: ## fixup A - offset: 2, value: LCPI3_0, kind: FK_Data_4
; FMA32-NEXT: retl ## encoding: [0xc3]
;
; FMACALL32-LABEL: test_f32_cst:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMACALL32: ## %bb.0: ## %entry
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMACALL32-NEXT: flds LCPI3_0 ## encoding: [0xd9,0x05,A,A,A,A]
; FMACALL32-NEXT: ## fixup A - offset: 2, value: LCPI3_0, kind: FK_Data_4
; FMACALL32-NEXT: retl ## encoding: [0xc3]
;
; FMA64-LABEL: test_f32_cst:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA64: ## %bb.0: ## %entry
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMA64-NEXT: vmovss {{.*}}(%rip), %xmm0 ## encoding: [0xc5,0xfa,0x10,0x05,A,A,A,A]
; FMA64-NEXT: ## fixup A - offset: 4, value: LCPI3_0-4, kind: reloc_riprel_4byte
; FMA64-NEXT: ## xmm0 = mem[0],zero,zero,zero
; FMA64-NEXT: retq ## encoding: [0xc3]
;
; FMACALL64-LABEL: test_f32_cst:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMACALL64: ## %bb.0: ## %entry
[X86] Use update_llc_test_checks.py to regenerate a test and add command lines to demonstrate that we don't pick EVEX encoded instruction when AVX512 and FMA3 are both enabled. This bug only exists on the scalar llvm.fma instrinsics. Looks like we don't test the llvm.fma intrinsics very thoroughly. In fact I don't see any tests for the vector versions. llvm-svn: 298045
2017-03-17 14:00:01 +08:00
; FMACALL64-NEXT: movss {{.*}}(%rip), %xmm0 ## encoding: [0xf3,0x0f,0x10,0x05,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 4, value: LCPI3_0-4, kind: reloc_riprel_4byte
; FMACALL64-NEXT: ## xmm0 = mem[0],zero,zero,zero
; FMACALL64-NEXT: retq ## encoding: [0xc3]
;
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-LABEL: test_f32_cst:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: vmovss {{.*}}(%rip), %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfa,0x10,0x05,A,A,A,A]
; AVX512-NEXT: ## fixup A - offset: 4, value: LCPI3_0-4, kind: reloc_riprel_4byte
; AVX512-NEXT: ## xmm0 = mem[0],zero,zero,zero
; AVX512-NEXT: retq ## encoding: [0xc3]
;
; AVX512VL-LABEL: test_f32_cst:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512VL: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: vmovss {{.*}}(%rip), %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfa,0x10,0x05,A,A,A,A]
; AVX512VL-NEXT: ## fixup A - offset: 4, value: LCPI3_0-4, kind: reloc_riprel_4byte
; AVX512VL-NEXT: ## xmm0 = mem[0],zero,zero,zero
; AVX512VL-NEXT: retq ## encoding: [0xc3]
Teach SelectionDAG to constant fold all-constant FMA nodes the same way that it constant folds FADD, FMUL, etc. llvm-svn: 181555
2013-05-10 06:27:13 +08:00
entry:
[X86] Cleanup fma tests a little bit. NFC. Reformat, isolate 213->231 xform, actually --check-prefix CHECK, and deduplicate the FMA intrinsic tests (FMA3 in AMD-land). llvm-svn: 240615
2015-06-25 08:40:25 +08:00
%call = call float @llvm.fma.f32(float 3.0, float 3.0, float 3.0)
Teach SelectionDAG to constant fold all-constant FMA nodes the same way that it constant folds FADD, FMUL, etc. llvm-svn: 181555
2013-05-10 06:27:13 +08:00
ret float %call
}
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
define <4 x float> @test_v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c) #0 {
; FMA32-LABEL: test_v4f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA32: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA32-NEXT: vfmadd213ps %xmm2, %xmm1, %xmm0 ## encoding: [0xc4,0xe2,0x71,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA32-NEXT: ## xmm0 = (xmm1 * xmm0) + xmm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA32-NEXT: retl ## encoding: [0xc3]
;
; FMA64-LABEL: test_v4f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA64: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA64-NEXT: vfmadd213ps %xmm2, %xmm1, %xmm0 ## encoding: [0xc4,0xe2,0x71,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA64-NEXT: ## xmm0 = (xmm1 * xmm0) + xmm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA64-NEXT: retq ## encoding: [0xc3]
;
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-LABEL: test_v4f32:
; FMACALL64: ## %bb.0: ## %entry
; FMACALL64-NEXT: subq $88, %rsp ## encoding: [0x48,0x83,0xec,0x58]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x54,0x24,0x30]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x4c,0x24,0x10]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x20]
; FMACALL64-NEXT: shufps $231, %xmm0, %xmm0 ## encoding: [0x0f,0xc6,0xc0,0xe7]
; FMACALL64-NEXT: ## xmm0 = xmm0[3,1,2,3]
; FMACALL64-NEXT: shufps $231, %xmm1, %xmm1 ## encoding: [0x0f,0xc6,0xc9,0xe7]
; FMACALL64-NEXT: ## xmm1 = xmm1[3,1,2,3]
; FMACALL64-NEXT: shufps $231, %xmm2, %xmm2 ## encoding: [0x0f,0xc6,0xd2,0xe7]
; FMACALL64-NEXT: ## xmm2 = xmm2[3,1,2,3]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
; FMACALL64-NEXT: movaps %xmm0, (%rsp) ## 16-byte Spill
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x04,0x24]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x20]
; FMACALL64-NEXT: movhlps %xmm0, %xmm0 ## encoding: [0x0f,0x12,0xc0]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x10]
; FMACALL64-NEXT: movhlps %xmm1, %xmm1 ## encoding: [0x0f,0x12,0xc9]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x54,0x24,0x30]
; FMACALL64-NEXT: movhlps %xmm2, %xmm2 ## encoding: [0x0f,0x12,0xd2]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
; FMACALL64-NEXT: unpcklps (%rsp), %xmm0 ## 16-byte Folded Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x14,0x04,0x24]
; FMACALL64-NEXT: ## xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; FMACALL64-NEXT: movaps %xmm0, (%rsp) ## 16-byte Spill
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x04,0x24]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x20]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x10]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x54,0x24,0x30]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x40]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x20]
; FMACALL64-NEXT: shufps $229, %xmm0, %xmm0 ## encoding: [0x0f,0xc6,0xc0,0xe5]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1,2,3]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x10]
; FMACALL64-NEXT: shufps $229, %xmm1, %xmm1 ## encoding: [0x0f,0xc6,0xc9,0xe5]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1,2,3]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x54,0x24,0x30]
; FMACALL64-NEXT: shufps $229, %xmm2, %xmm2 ## encoding: [0x0f,0xc6,0xd2,0xe5]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1,2,3]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x40]
; FMACALL64-NEXT: unpcklps %xmm0, %xmm1 ## encoding: [0x0f,0x14,0xc8]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; FMACALL64-NEXT: unpcklpd (%rsp), %xmm1 ## 16-byte Folded Reload
; FMACALL64-NEXT: ## encoding: [0x66,0x0f,0x14,0x0c,0x24]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],mem[0]
; FMACALL64-NEXT: movaps %xmm1, %xmm0 ## encoding: [0x0f,0x28,0xc1]
; FMACALL64-NEXT: addq $88, %rsp ## encoding: [0x48,0x83,0xc4,0x58]
; FMACALL64-NEXT: retq ## encoding: [0xc3]
;
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-LABEL: test_v4f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: vfmadd213ps %xmm2, %xmm1, %xmm0 ## encoding: [0xc4,0xe2,0x71,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; AVX512-NEXT: ## xmm0 = (xmm1 * xmm0) + xmm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: retq ## encoding: [0xc3]
;
; AVX512VL-LABEL: test_v4f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512VL: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: vfmadd213ps %xmm2, %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; AVX512VL-NEXT: ## xmm0 = (xmm1 * xmm0) + xmm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: retq ## encoding: [0xc3]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
;
; FMACALL32_BDVER2-LABEL: test_v4f32:
; FMACALL32_BDVER2: ## %bb.0: ## %entry
; FMACALL32_BDVER2-NEXT: subl $108, %esp ## encoding: [0x83,0xec,0x6c]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x54,0x24,0x40]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x4c,0x24,0x30]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x44,0x24,0x20]
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm2, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x54,0x24,0x08,0x02]
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm1, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x4c,0x24,0x04,0x02]
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm0, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x04,0x24,0x02]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0x7c,0x24,0x60]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x44,0x24,0x40]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x54,0x24,0x30]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x4c,0x24,0x20]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x01]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm2, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x54,0x24,0x04,0x01]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm1, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x0c,0x24,0x01]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0x7c,0x24,0x54]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x44,0x24,0x40]
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xfa,0x11,0x44,0x24,0x08]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x44,0x24,0x30]
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xfa,0x11,0x44,0x24,0x04]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x44,0x24,0x20]
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, (%esp) ## encoding: [0xc5,0xfa,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x44,0x24,0x40]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x54,0x24,0x30]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x4c,0x24,0x20]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x03]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm2, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x54,0x24,0x04,0x03]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm1, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x0c,0x24,0x03]
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x1c]
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0x6c,0x24,0x54]
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x18]
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0x6c,0x24,0x60]
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x14]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x10]
; FMACALL32_BDVER2-NEXT: vmovss {{[0-9]+}}(%esp), %xmm0 ## encoding: [0xc5,0xfa,0x10,0x44,0x24,0x1c]
; FMACALL32_BDVER2-NEXT: ## xmm0 = mem[0],zero,zero,zero
; FMACALL32_BDVER2-NEXT: vinsertps $16, {{[0-9]+}}(%esp), %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x21,0x44,0x24,0x18,0x10]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0],mem[0],xmm0[2,3]
; FMACALL32_BDVER2-NEXT: vinsertps $32, {{[0-9]+}}(%esp), %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x21,0x44,0x24,0x14,0x20]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0,1],mem[0],xmm0[3]
; FMACALL32_BDVER2-NEXT: vinsertps $48, {{[0-9]+}}(%esp), %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x21,0x44,0x24,0x10,0x30]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0,1,2],mem[0]
; FMACALL32_BDVER2-NEXT: addl $108, %esp ## encoding: [0x83,0xc4,0x6c]
; FMACALL32_BDVER2-NEXT: retl ## encoding: [0xc3]
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
entry:
%call = call <4 x float> @llvm.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c)
ret <4 x float> %call
}
define <8 x float> @test_v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c) #0 {
; FMA32-LABEL: test_v8f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA32: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA32-NEXT: vfmadd213ps %ymm2, %ymm1, %ymm0 ## encoding: [0xc4,0xe2,0x75,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA32-NEXT: ## ymm0 = (ymm1 * ymm0) + ymm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA32-NEXT: retl ## encoding: [0xc3]
;
; FMA64-LABEL: test_v8f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA64: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA64-NEXT: vfmadd213ps %ymm2, %ymm1, %ymm0 ## encoding: [0xc4,0xe2,0x75,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA64-NEXT: ## ymm0 = (ymm1 * ymm0) + ymm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA64-NEXT: retq ## encoding: [0xc3]
;
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-LABEL: test_v8f32:
; FMACALL64: ## %bb.0: ## %entry
; FMACALL64-NEXT: subq $136, %rsp ## encoding: [0x48,0x81,0xec,0x88,0x00,0x00,0x00]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x6c,0x24,0x50]
; FMACALL64-NEXT: movaps %xmm4, (%rsp) ## 16-byte Spill
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x24,0x24]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x5c,0x24,0x40]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x54,0x24,0x60]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x4c,0x24,0x30]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x10]
; FMACALL64-NEXT: shufps $231, %xmm0, %xmm0 ## encoding: [0x0f,0xc6,0xc0,0xe7]
; FMACALL64-NEXT: ## xmm0 = xmm0[3,1,2,3]
; FMACALL64-NEXT: movaps %xmm2, %xmm1 ## encoding: [0x0f,0x28,0xca]
; FMACALL64-NEXT: shufps $231, %xmm2, %xmm1 ## encoding: [0x0f,0xc6,0xca,0xe7]
; FMACALL64-NEXT: ## xmm1 = xmm1[3,1],xmm2[2,3]
; FMACALL64-NEXT: movaps %xmm4, %xmm2 ## encoding: [0x0f,0x28,0xd4]
; FMACALL64-NEXT: shufps $231, %xmm4, %xmm2 ## encoding: [0x0f,0xc6,0xd4,0xe7]
; FMACALL64-NEXT: ## xmm2 = xmm2[3,1],xmm4[2,3]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x20]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x10]
; FMACALL64-NEXT: movhlps %xmm0, %xmm0 ## encoding: [0x0f,0x12,0xc0]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x60]
; FMACALL64-NEXT: movhlps %xmm1, %xmm1 ## encoding: [0x0f,0x12,0xc9]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1]
; FMACALL64-NEXT: movaps (%rsp), %xmm2 ## 16-byte Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x14,0x24]
; FMACALL64-NEXT: movhlps %xmm2, %xmm2 ## encoding: [0x0f,0x12,0xd2]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Folded Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x14,0x44,0x24,0x20]
; FMACALL64-NEXT: ## xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x70]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x10]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x60]
; FMACALL64-NEXT: movaps (%rsp), %xmm2 ## 16-byte Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x14,0x24]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x20]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x10]
; FMACALL64-NEXT: shufps $229, %xmm0, %xmm0 ## encoding: [0x0f,0xc6,0xc0,0xe5]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1,2,3]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x60]
; FMACALL64-NEXT: shufps $229, %xmm1, %xmm1 ## encoding: [0x0f,0xc6,0xc9,0xe5]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1,2,3]
; FMACALL64-NEXT: movaps (%rsp), %xmm2 ## 16-byte Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x14,0x24]
; FMACALL64-NEXT: shufps $229, %xmm2, %xmm2 ## encoding: [0x0f,0xc6,0xd2,0xe5]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1,2,3]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x20]
; FMACALL64-NEXT: unpcklps %xmm0, %xmm1 ## encoding: [0x0f,0x14,0xc8]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Folded Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x66,0x0f,0x14,0x4c,0x24,0x70]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],mem[0]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x4c,0x24,0x20]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x30]
; FMACALL64-NEXT: shufps $231, %xmm0, %xmm0 ## encoding: [0x0f,0xc6,0xc0,0xe7]
; FMACALL64-NEXT: ## xmm0 = xmm0[3,1,2,3]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x40]
; FMACALL64-NEXT: shufps $231, %xmm1, %xmm1 ## encoding: [0x0f,0xc6,0xc9,0xe7]
; FMACALL64-NEXT: ## xmm1 = xmm1[3,1,2,3]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x54,0x24,0x50]
; FMACALL64-NEXT: shufps $231, %xmm2, %xmm2 ## encoding: [0x0f,0xc6,0xd2,0xe7]
; FMACALL64-NEXT: ## xmm2 = xmm2[3,1,2,3]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
; FMACALL64-NEXT: movaps %xmm0, (%rsp) ## 16-byte Spill
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x04,0x24]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x30]
; FMACALL64-NEXT: movhlps %xmm0, %xmm0 ## encoding: [0x0f,0x12,0xc0]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x40]
; FMACALL64-NEXT: movhlps %xmm1, %xmm1 ## encoding: [0x0f,0x12,0xc9]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x54,0x24,0x50]
; FMACALL64-NEXT: movhlps %xmm2, %xmm2 ## encoding: [0x0f,0x12,0xd2]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
; FMACALL64-NEXT: unpcklps (%rsp), %xmm0 ## 16-byte Folded Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x14,0x04,0x24]
; FMACALL64-NEXT: ## xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; FMACALL64-NEXT: movaps %xmm0, (%rsp) ## 16-byte Spill
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x04,0x24]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x30]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x40]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x54,0x24,0x50]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x10]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x30]
; FMACALL64-NEXT: shufps $229, %xmm0, %xmm0 ## encoding: [0x0f,0xc6,0xc0,0xe5]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1,2,3]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x40]
; FMACALL64-NEXT: shufps $229, %xmm1, %xmm1 ## encoding: [0x0f,0xc6,0xc9,0xe5]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1,2,3]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x54,0x24,0x50]
; FMACALL64-NEXT: shufps $229, %xmm2, %xmm2 ## encoding: [0x0f,0xc6,0xd2,0xe5]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1,2,3]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x10]
; FMACALL64-NEXT: unpcklps %xmm0, %xmm1 ## encoding: [0x0f,0x14,0xc8]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; FMACALL64-NEXT: unpcklpd (%rsp), %xmm1 ## 16-byte Folded Reload
; FMACALL64-NEXT: ## encoding: [0x66,0x0f,0x14,0x0c,0x24]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],mem[0]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x20]
; FMACALL64-NEXT: addq $136, %rsp ## encoding: [0x48,0x81,0xc4,0x88,0x00,0x00,0x00]
; FMACALL64-NEXT: retq ## encoding: [0xc3]
;
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-LABEL: test_v8f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: vfmadd213ps %ymm2, %ymm1, %ymm0 ## encoding: [0xc4,0xe2,0x75,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; AVX512-NEXT: ## ymm0 = (ymm1 * ymm0) + ymm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: retq ## encoding: [0xc3]
;
; AVX512VL-LABEL: test_v8f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512VL: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: vfmadd213ps %ymm2, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0x75,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; AVX512VL-NEXT: ## ymm0 = (ymm1 * ymm0) + ymm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: retq ## encoding: [0xc3]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
;
; FMACALL32_BDVER2-LABEL: test_v8f32:
; FMACALL32_BDVER2: ## %bb.0: ## %entry
; FMACALL32_BDVER2-NEXT: subl $316, %esp ## encoding: [0x81,0xec,0x3c,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: ## imm = 0x13C
; FMACALL32_BDVER2-NEXT: vextractf128 $1, %ymm2, %xmm3 ## encoding: [0xc4,0xe3,0x7d,0x19,0xd3,0x01]
; FMACALL32_BDVER2-NEXT: vmovups %ymm2, {{[-0-9]+}}(%e{{[sb]}}p) ## 32-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x11,0x94,0x24,0x00,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractf128 $1, %ymm1, %xmm2 ## encoding: [0xc4,0xe3,0x7d,0x19,0xca,0x01]
; FMACALL32_BDVER2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%e{{[sb]}}p) ## 32-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x11,0x8c,0x24,0xe0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractf128 $1, %ymm0, %xmm1 ## encoding: [0xc4,0xe3,0x7d,0x19,0xc1,0x01]
; FMACALL32_BDVER2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 32-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x11,0x84,0x24,0xc0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x5c,0x24,0x60]
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm3, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x5c,0x24,0x08,0x02]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x54,0x24,0x50]
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm2, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x54,0x24,0x04,0x02]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x4c,0x24,0x40]
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm1, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x0c,0x24,0x02]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0xb4,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x44,0x24,0x60]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x54,0x24,0x50]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x4c,0x24,0x40]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x01]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm2, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x54,0x24,0x04,0x01]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm1, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x0c,0x24,0x01]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0xa8,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x44,0x24,0x60]
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xfa,0x11,0x44,0x24,0x08]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x44,0x24,0x50]
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xfa,0x11,0x44,0x24,0x04]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x44,0x24,0x40]
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, (%esp) ## encoding: [0xc5,0xfa,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0x9c,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x10,0x84,0x24,0x00,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x03]
; FMACALL32_BDVER2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x10,0x84,0x24,0xe0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x04,0x03]
; FMACALL32_BDVER2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x10,0x84,0x24,0xc0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm0, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x04,0x24,0x03]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0x90,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x10,0x84,0x24,0x00,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x02]
; FMACALL32_BDVER2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x10,0x84,0x24,0xe0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x04,0x02]
; FMACALL32_BDVER2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x10,0x84,0x24,0xc0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm0, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x04,0x24,0x02]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0x84,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x10,0x84,0x24,0x00,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x01]
; FMACALL32_BDVER2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x10,0x84,0x24,0xe0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x04,0x01]
; FMACALL32_BDVER2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x10,0x84,0x24,0xc0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm0, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x04,0x24,0x01]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0x7c,0x24,0x78]
; FMACALL32_BDVER2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x10,0x84,0x24,0x00,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xfa,0x11,0x44,0x24,0x08]
; FMACALL32_BDVER2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x10,0x84,0x24,0xe0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xfa,0x11,0x44,0x24,0x04]
; FMACALL32_BDVER2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x10,0x84,0x24,0xc0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, (%esp) ## encoding: [0xc5,0xfa,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x44,0x24,0x60]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x54,0x24,0x50]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x4c,0x24,0x40]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x03]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm2, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x54,0x24,0x04,0x03]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm1, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x0c,0x24,0x03]
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x3c]
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0x6c,0x24,0x78]
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x38]
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0x84,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x34]
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0x90,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x30]
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0x9c,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x2c]
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0xa8,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x28]
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0xb4,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x24]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x20]
; FMACALL32_BDVER2-NEXT: vmovss {{[0-9]+}}(%esp), %xmm0 ## encoding: [0xc5,0xfa,0x10,0x44,0x24,0x3c]
; FMACALL32_BDVER2-NEXT: ## xmm0 = mem[0],zero,zero,zero
; FMACALL32_BDVER2-NEXT: vinsertps $16, {{[0-9]+}}(%esp), %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x21,0x44,0x24,0x38,0x10]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0],mem[0],xmm0[2,3]
; FMACALL32_BDVER2-NEXT: vmovss {{[0-9]+}}(%esp), %xmm1 ## encoding: [0xc5,0xfa,0x10,0x4c,0x24,0x2c]
; FMACALL32_BDVER2-NEXT: ## xmm1 = mem[0],zero,zero,zero
; FMACALL32_BDVER2-NEXT: vinsertps $16, {{[0-9]+}}(%esp), %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x21,0x4c,0x24,0x28,0x10]
; FMACALL32_BDVER2-NEXT: ## xmm1 = xmm1[0],mem[0],xmm1[2,3]
; FMACALL32_BDVER2-NEXT: vinsertps $32, {{[0-9]+}}(%esp), %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x21,0x44,0x24,0x34,0x20]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0,1],mem[0],xmm0[3]
; FMACALL32_BDVER2-NEXT: vinsertps $32, {{[0-9]+}}(%esp), %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x21,0x4c,0x24,0x24,0x20]
; FMACALL32_BDVER2-NEXT: ## xmm1 = xmm1[0,1],mem[0],xmm1[3]
; FMACALL32_BDVER2-NEXT: vinsertps $48, {{[0-9]+}}(%esp), %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x21,0x44,0x24,0x30,0x30]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0,1,2],mem[0]
; FMACALL32_BDVER2-NEXT: vinsertps $48, {{[0-9]+}}(%esp), %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x21,0x4c,0x24,0x20,0x30]
; FMACALL32_BDVER2-NEXT: ## xmm1 = xmm1[0,1,2],mem[0]
; FMACALL32_BDVER2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 ## encoding: [0xc4,0xe3,0x7d,0x18,0xc1,0x01]
; FMACALL32_BDVER2-NEXT: addl $316, %esp ## encoding: [0x81,0xc4,0x3c,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: ## imm = 0x13C
; FMACALL32_BDVER2-NEXT: retl ## encoding: [0xc3]
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
entry:
%call = call <8 x float> @llvm.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c)
ret <8 x float> %call
}
define <16 x float> @test_v16f32(<16 x float> %a, <16 x float> %b, <16 x float> %c) #0 {
; FMA32-LABEL: test_v16f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA32: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA32-NEXT: pushl %ebp ## encoding: [0x55]
; FMA32-NEXT: movl %esp, %ebp ## encoding: [0x89,0xe5]
; FMA32-NEXT: andl $-32, %esp ## encoding: [0x83,0xe4,0xe0]
; FMA32-NEXT: subl $32, %esp ## encoding: [0x83,0xec,0x20]
; FMA32-NEXT: vfmadd213ps 8(%ebp), %ymm2, %ymm0 ## encoding: [0xc4,0xe2,0x6d,0xa8,0x45,0x08]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA32-NEXT: ## ymm0 = (ymm2 * ymm0) + mem
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA32-NEXT: vfmadd213ps 40(%ebp), %ymm3, %ymm1 ## encoding: [0xc4,0xe2,0x65,0xa8,0x4d,0x28]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA32-NEXT: ## ymm1 = (ymm3 * ymm1) + mem
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA32-NEXT: movl %ebp, %esp ## encoding: [0x89,0xec]
; FMA32-NEXT: popl %ebp ## encoding: [0x5d]
; FMA32-NEXT: retl ## encoding: [0xc3]
;
; FMA64-LABEL: test_v16f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA64: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA64-NEXT: vfmadd213ps %ymm4, %ymm2, %ymm0 ## encoding: [0xc4,0xe2,0x6d,0xa8,0xc4]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA64-NEXT: ## ymm0 = (ymm2 * ymm0) + ymm4
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA64-NEXT: vfmadd213ps %ymm5, %ymm3, %ymm1 ## encoding: [0xc4,0xe2,0x65,0xa8,0xcd]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA64-NEXT: ## ymm1 = (ymm3 * ymm1) + ymm5
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA64-NEXT: retq ## encoding: [0xc3]
;
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-LABEL: test_v16f32:
; FMACALL64: ## %bb.0: ## %entry
; FMACALL64-NEXT: subq $168, %rsp ## encoding: [0x48,0x81,0xec,0xa8,0x00,0x00,0x00]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0xbc,0x24,0x80,0x00,0x00,0x00]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x74,0x24,0x20]
; FMACALL64-NEXT: movaps %xmm5, (%rsp) ## 16-byte Spill
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x2c,0x24]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x64,0x24,0x10]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x5c,0x24,0x70]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x54,0x24,0x40]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x4c,0x24,0x30]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x50]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xb0,0x00,0x00,0x00]
; FMACALL64-NEXT: shufps $231, %xmm0, %xmm0 ## encoding: [0x0f,0xc6,0xc0,0xe7]
; FMACALL64-NEXT: ## xmm0 = xmm0[3,1,2,3]
; FMACALL64-NEXT: movaps %xmm4, %xmm1 ## encoding: [0x0f,0x28,0xcc]
; FMACALL64-NEXT: shufps $231, %xmm4, %xmm1 ## encoding: [0x0f,0xc6,0xcc,0xe7]
; FMACALL64-NEXT: ## xmm1 = xmm1[3,1],xmm4[2,3]
; FMACALL64-NEXT: shufps $231, %xmm2, %xmm2 ## encoding: [0x0f,0xc6,0xd2,0xe7]
; FMACALL64-NEXT: ## xmm2 = xmm2[3,1,2,3]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x60]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x50]
; FMACALL64-NEXT: movhlps %xmm0, %xmm0 ## encoding: [0x0f,0x12,0xc0]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x10]
; FMACALL64-NEXT: movhlps %xmm1, %xmm1 ## encoding: [0x0f,0x12,0xc9]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xb0,0x00,0x00,0x00]
; FMACALL64-NEXT: movhlps %xmm2, %xmm2 ## encoding: [0x0f,0x12,0xd2]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Folded Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x14,0x44,0x24,0x60]
; FMACALL64-NEXT: ## xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x84,0x24,0x90,0x00,0x00,0x00]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x50]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x10]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xb0,0x00,0x00,0x00]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x60]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x50]
; FMACALL64-NEXT: shufps $229, %xmm0, %xmm0 ## encoding: [0x0f,0xc6,0xc0,0xe5]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1,2,3]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x10]
; FMACALL64-NEXT: shufps $229, %xmm1, %xmm1 ## encoding: [0x0f,0xc6,0xc9,0xe5]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1,2,3]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xb0,0x00,0x00,0x00]
; FMACALL64-NEXT: shufps $229, %xmm2, %xmm2 ## encoding: [0x0f,0xc6,0xd2,0xe5]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1,2,3]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x60]
; FMACALL64-NEXT: unpcklps %xmm0, %xmm1 ## encoding: [0x0f,0x14,0xc8]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Folded Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x66,0x0f,0x14,0x8c,0x24,0x90,0x00,0x00,0x00]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],mem[0]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x4c,0x24,0x60]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x30]
; FMACALL64-NEXT: shufps $231, %xmm0, %xmm0 ## encoding: [0x0f,0xc6,0xc0,0xe7]
; FMACALL64-NEXT: ## xmm0 = xmm0[3,1,2,3]
; FMACALL64-NEXT: movaps (%rsp), %xmm1 ## 16-byte Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x0c,0x24]
; FMACALL64-NEXT: shufps $231, %xmm1, %xmm1 ## encoding: [0x0f,0xc6,0xc9,0xe7]
; FMACALL64-NEXT: ## xmm1 = xmm1[3,1,2,3]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xc0,0x00,0x00,0x00]
; FMACALL64-NEXT: shufps $231, %xmm2, %xmm2 ## encoding: [0x0f,0xc6,0xd2,0xe7]
; FMACALL64-NEXT: ## xmm2 = xmm2[3,1,2,3]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x10]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x30]
; FMACALL64-NEXT: movhlps %xmm0, %xmm0 ## encoding: [0x0f,0x12,0xc0]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1]
; FMACALL64-NEXT: movaps (%rsp), %xmm1 ## 16-byte Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x0c,0x24]
; FMACALL64-NEXT: movhlps %xmm1, %xmm1 ## encoding: [0x0f,0x12,0xc9]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xc0,0x00,0x00,0x00]
; FMACALL64-NEXT: movhlps %xmm2, %xmm2 ## encoding: [0x0f,0x12,0xd2]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Folded Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x14,0x44,0x24,0x10]
; FMACALL64-NEXT: ## xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x50]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x30]
; FMACALL64-NEXT: movaps (%rsp), %xmm1 ## 16-byte Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x0c,0x24]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xc0,0x00,0x00,0x00]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x10]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x30]
; FMACALL64-NEXT: shufps $229, %xmm0, %xmm0 ## encoding: [0x0f,0xc6,0xc0,0xe5]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1,2,3]
; FMACALL64-NEXT: movaps (%rsp), %xmm1 ## 16-byte Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x0c,0x24]
; FMACALL64-NEXT: shufps $229, %xmm1, %xmm1 ## encoding: [0x0f,0xc6,0xc9,0xe5]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1,2,3]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xc0,0x00,0x00,0x00]
; FMACALL64-NEXT: shufps $229, %xmm2, %xmm2 ## encoding: [0x0f,0xc6,0xd2,0xe5]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1,2,3]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x10]
; FMACALL64-NEXT: unpcklps %xmm0, %xmm1 ## encoding: [0x0f,0x14,0xc8]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Folded Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x66,0x0f,0x14,0x4c,0x24,0x50]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],mem[0]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x4c,0x24,0x10]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x40]
; FMACALL64-NEXT: shufps $231, %xmm0, %xmm0 ## encoding: [0x0f,0xc6,0xc0,0xe7]
; FMACALL64-NEXT: ## xmm0 = xmm0[3,1,2,3]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x20]
; FMACALL64-NEXT: shufps $231, %xmm1, %xmm1 ## encoding: [0x0f,0xc6,0xc9,0xe7]
; FMACALL64-NEXT: ## xmm1 = xmm1[3,1,2,3]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xd0,0x00,0x00,0x00]
; FMACALL64-NEXT: shufps $231, %xmm2, %xmm2 ## encoding: [0x0f,0xc6,0xd2,0xe7]
; FMACALL64-NEXT: ## xmm2 = xmm2[3,1,2,3]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
; FMACALL64-NEXT: movaps %xmm0, (%rsp) ## 16-byte Spill
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x04,0x24]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x40]
; FMACALL64-NEXT: movhlps %xmm0, %xmm0 ## encoding: [0x0f,0x12,0xc0]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x20]
; FMACALL64-NEXT: movhlps %xmm1, %xmm1 ## encoding: [0x0f,0x12,0xc9]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xd0,0x00,0x00,0x00]
; FMACALL64-NEXT: movhlps %xmm2, %xmm2 ## encoding: [0x0f,0x12,0xd2]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
; FMACALL64-NEXT: unpcklps (%rsp), %xmm0 ## 16-byte Folded Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x14,0x04,0x24]
; FMACALL64-NEXT: ## xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x30]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x40]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x20]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xd0,0x00,0x00,0x00]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
; FMACALL64-NEXT: movaps %xmm0, (%rsp) ## 16-byte Spill
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x04,0x24]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x40]
; FMACALL64-NEXT: shufps $229, %xmm0, %xmm0 ## encoding: [0x0f,0xc6,0xc0,0xe5]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1,2,3]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x20]
; FMACALL64-NEXT: shufps $229, %xmm1, %xmm1 ## encoding: [0x0f,0xc6,0xc9,0xe5]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1,2,3]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xd0,0x00,0x00,0x00]
; FMACALL64-NEXT: shufps $229, %xmm2, %xmm2 ## encoding: [0x0f,0xc6,0xd2,0xe5]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1,2,3]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
; FMACALL64-NEXT: movaps (%rsp), %xmm1 ## 16-byte Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x0c,0x24]
; FMACALL64-NEXT: unpcklps %xmm0, %xmm1 ## encoding: [0x0f,0x14,0xc8]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Folded Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x66,0x0f,0x14,0x4c,0x24,0x30]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],mem[0]
; FMACALL64-NEXT: movaps %xmm1, (%rsp) ## 16-byte Spill
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x0c,0x24]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x70]
; FMACALL64-NEXT: shufps $231, %xmm0, %xmm0 ## encoding: [0x0f,0xc6,0xc0,0xe7]
; FMACALL64-NEXT: ## xmm0 = xmm0[3,1,2,3]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x8c,0x24,0x80,0x00,0x00,0x00]
; FMACALL64-NEXT: shufps $231, %xmm1, %xmm1 ## encoding: [0x0f,0xc6,0xc9,0xe7]
; FMACALL64-NEXT: ## xmm1 = xmm1[3,1,2,3]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xe0,0x00,0x00,0x00]
; FMACALL64-NEXT: shufps $231, %xmm2, %xmm2 ## encoding: [0x0f,0xc6,0xd2,0xe7]
; FMACALL64-NEXT: ## xmm2 = xmm2[3,1,2,3]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x20]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x70]
; FMACALL64-NEXT: movhlps %xmm0, %xmm0 ## encoding: [0x0f,0x12,0xc0]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x8c,0x24,0x80,0x00,0x00,0x00]
; FMACALL64-NEXT: movhlps %xmm1, %xmm1 ## encoding: [0x0f,0x12,0xc9]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xe0,0x00,0x00,0x00]
; FMACALL64-NEXT: movhlps %xmm2, %xmm2 ## encoding: [0x0f,0x12,0xd2]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Folded Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x14,0x44,0x24,0x20]
; FMACALL64-NEXT: ## xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x20]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x70]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x8c,0x24,0x80,0x00,0x00,0x00]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xe0,0x00,0x00,0x00]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x40]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x70]
; FMACALL64-NEXT: shufps $229, %xmm0, %xmm0 ## encoding: [0x0f,0xc6,0xc0,0xe5]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1,2,3]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x8c,0x24,0x80,0x00,0x00,0x00]
; FMACALL64-NEXT: shufps $229, %xmm1, %xmm1 ## encoding: [0x0f,0xc6,0xc9,0xe5]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1,2,3]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xe0,0x00,0x00,0x00]
; FMACALL64-NEXT: shufps $229, %xmm2, %xmm2 ## encoding: [0x0f,0xc6,0xd2,0xe5]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1,2,3]
; FMACALL64-NEXT: callq _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: reloc_branch_4byte_pcrel
[X86] Handle COPYs of physregs better (regalloc hints) Enable enableMultipleCopyHints() on X86. Original Patch by @jonpa: While enabling the mischeduler for SystemZ, it was discovered that for some reason a test needed one extra seemingly needless COPY (test/CodeGen/SystemZ/call-03.ll). The handling for that is resulted in this patch, which improves the register coalescing by providing not just one copy hint, but a sorted list of copy hints. On SystemZ, this gives ~12500 less register moves on SPEC, as well as marginally less spilling. Instead of improving just the SystemZ backend, the improvement has been implemented in common-code (calculateSpillWeightAndHint(). This gives a lot of test failures, but since this should be a general improvement I hope that the involved targets will help and review the test updates. Differential Revision: https://reviews.llvm.org/D38128 llvm-svn: 342578
2018-09-20 02:59:08 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 ## 16-byte Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x5c,0x24,0x40]
; FMACALL64-NEXT: unpcklps %xmm0, %xmm3 ## encoding: [0x0f,0x14,0xd8]
; FMACALL64-NEXT: ## xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; FMACALL64-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 ## 16-byte Folded Reload
; FMACALL64-NEXT: ## encoding: [0x66,0x0f,0x14,0x5c,0x24,0x20]
; FMACALL64-NEXT: ## xmm3 = xmm3[0],mem[0]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x60]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x10]
; FMACALL64-NEXT: movaps (%rsp), %xmm2 ## 16-byte Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x14,0x24]
; FMACALL64-NEXT: addq $168, %rsp ## encoding: [0x48,0x81,0xc4,0xa8,0x00,0x00,0x00]
; FMACALL64-NEXT: retq ## encoding: [0xc3]
;
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-LABEL: test_v16f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: vfmadd213ps %zmm2, %zmm1, %zmm0 ## encoding: [0x62,0xf2,0x75,0x48,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; AVX512-NEXT: ## zmm0 = (zmm1 * zmm0) + zmm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: retq ## encoding: [0xc3]
;
; AVX512VL-LABEL: test_v16f32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512VL: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: vfmadd213ps %zmm2, %zmm1, %zmm0 ## encoding: [0x62,0xf2,0x75,0x48,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; AVX512VL-NEXT: ## zmm0 = (zmm1 * zmm0) + zmm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: retq ## encoding: [0xc3]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
;
; FMACALL32_BDVER2-LABEL: test_v16f32:
; FMACALL32_BDVER2: ## %bb.0: ## %entry
; FMACALL32_BDVER2-NEXT: pushl %ebp ## encoding: [0x55]
; FMACALL32_BDVER2-NEXT: movl %esp, %ebp ## encoding: [0x89,0xe5]
; FMACALL32_BDVER2-NEXT: andl $-32, %esp ## encoding: [0x83,0xe4,0xe0]
; FMACALL32_BDVER2-NEXT: subl $448, %esp ## encoding: [0x81,0xec,0xc0,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: ## imm = 0x1C0
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 56(%ebp), %xmm4 ## encoding: [0xc5,0xf8,0x28,0x65,0x38]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps %ymm2, {{[-0-9]+}}(%e{{[sb]}}p) ## 32-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x29,0x94,0x24,0x60,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractf128 $1, %ymm3, %xmm2 ## encoding: [0xc4,0xe3,0x7d,0x19,0xda,0x01]
; FMACALL32_BDVER2-NEXT: vmovaps %ymm1, {{[-0-9]+}}(%e{{[sb]}}p) ## 32-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x29,0x8c,0x24,0x80,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractf128 $1, %ymm1, %xmm1 ## encoding: [0xc4,0xe3,0x7d,0x19,0xc9,0x01]
; FMACALL32_BDVER2-NEXT: vmovaps %ymm3, {{[-0-9]+}}(%e{{[sb]}}p) ## 32-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x29,0x5c,0x24,0x60]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 32-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x29,0x84,0x24,0x80,0x01,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x94,0x24,0xc0,0x00,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm2, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x54,0x24,0x04,0x02]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x8c,0x24,0xb0,0x00,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm1, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x0c,0x24,0x02]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm4, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x64,0x24,0x08,0x02]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0x54,0x01,0x00,0x00]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 56(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x38]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 ## 16-byte Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x94,0x24,0xc0,0x00,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 ## 16-byte Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x8c,0x24,0xb0,0x00,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x01]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm2, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x54,0x24,0x04,0x01]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm1, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x0c,0x24,0x01]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 56(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x38]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0x48,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xfa,0x11,0x44,0x24,0x08]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x84,0x24,0xc0,0x00,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xfa,0x11,0x44,0x24,0x04]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x84,0x24,0xb0,0x00,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, (%esp) ## encoding: [0xc5,0xfa,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 40(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x28]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0x3c,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x03]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x44,0x24,0x60]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x04,0x03]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0x80,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm0, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x04,0x24,0x03]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 40(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x28]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0x30,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x02]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x44,0x24,0x60]
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x04,0x02]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0x80,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm0, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x04,0x24,0x02]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 40(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x28]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0x24,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x01]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x44,0x24,0x60]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x04,0x01]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0x80,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm0, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x04,0x24,0x01]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 40(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x28]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0x18,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xfa,0x11,0x44,0x24,0x08]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x44,0x24,0x60]
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xfa,0x11,0x44,0x24,0x04]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0x80,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, (%esp) ## encoding: [0xc5,0xfa,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 24(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x18]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0x0c,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x03]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0x60,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractf128 $1, %ymm0, %xmm0 ## encoding: [0xc4,0xe3,0x7d,0x19,0xc0,0x01]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x84,0x24,0x80,0x00,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x04,0x03]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0x80,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractf128 $1, %ymm0, %xmm0 ## encoding: [0xc4,0xe3,0x7d,0x19,0xc0,0x01]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x44,0x24,0x60]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm0, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x04,0x24,0x03]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0x00,0x01,0x00,0x00]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 24(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x18]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 ## 16-byte Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x94,0x24,0x80,0x00,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 ## 16-byte Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x4c,0x24,0x60]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x02]
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm2, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x54,0x24,0x04,0x02]
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm1, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x0c,0x24,0x02]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0xf4,0x00,0x00,0x00]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 24(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x18]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 ## 16-byte Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x94,0x24,0x80,0x00,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 ## 16-byte Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x4c,0x24,0x60]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x01]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm2, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x54,0x24,0x04,0x01]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm1, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x0c,0x24,0x01]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 24(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x18]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0xe8,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xfa,0x11,0x44,0x24,0x08]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x84,0x24,0x80,0x00,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xfa,0x11,0x44,0x24,0x04]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x44,0x24,0x60]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, (%esp) ## encoding: [0xc5,0xfa,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 8(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x08]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0x80,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x03]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0x60,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x04,0x03]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0x80,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm0, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x04,0x24,0x03]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 8(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x08]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0x7c,0x24,0x60]
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x02]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0x60,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x04,0x02]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0x80,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $2, %xmm0, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x04,0x24,0x02]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 8(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x08]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0xdc,0x00,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x01]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0x60,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x04,0x01]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0x80,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractps $1, %xmm0, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x04,0x24,0x01]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 8(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x08]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0xd0,0x00,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xfa,0x11,0x44,0x24,0x08]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0x60,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xfa,0x11,0x44,0x24,0x04]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0x80,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovss %xmm0, (%esp) ## encoding: [0xc5,0xfa,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 56(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x38]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 ## 16-byte Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x94,0x24,0xc0,0x00,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 ## 16-byte Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x8c,0x24,0xb0,0x00,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x44,0x24,0x08,0x03]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm2, {{[0-9]+}}(%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x54,0x24,0x04,0x03]
; FMACALL32_BDVER2-NEXT: vextractps $3, %xmm1, (%esp) ## encoding: [0xc4,0xe3,0x79,0x17,0x0c,0x24,0x03]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x3c]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0xd0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x38]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0xdc,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x34]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0x6c,0x24,0x60]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x30]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0x80,0x00,0x00,0x00]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x2c]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0xe8,0x00,0x00,0x00]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x28]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0xf4,0x00,0x00,0x00]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x24]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0x00,0x01,0x00,0x00]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x20]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0x0c,0x01,0x00,0x00]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x5c]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0x18,0x01,0x00,0x00]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x58]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0x24,0x01,0x00,0x00]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x54]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0x30,0x01,0x00,0x00]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x50]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0x3c,0x01,0x00,0x00]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x4c]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0x48,0x01,0x00,0x00]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x48]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0x54,0x01,0x00,0x00]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x44]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: calll _fmaf ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fmaf-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: fstps {{[0-9]+}}(%esp) ## encoding: [0xd9,0x5c,0x24,0x40]
; FMACALL32_BDVER2-NEXT: vmovss {{[0-9]+}}(%esp), %xmm0 ## encoding: [0xc5,0xfa,0x10,0x44,0x24,0x3c]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm0 = mem[0],zero,zero,zero
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vinsertps $16, {{[0-9]+}}(%esp), %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x21,0x44,0x24,0x38,0x10]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0],mem[0],xmm0[2,3]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovss {{[0-9]+}}(%esp), %xmm1 ## encoding: [0xc5,0xfa,0x10,0x4c,0x24,0x2c]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm1 = mem[0],zero,zero,zero
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vinsertps $16, {{[0-9]+}}(%esp), %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x21,0x4c,0x24,0x28,0x10]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm1 = xmm1[0],mem[0],xmm1[2,3]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovss {{[0-9]+}}(%esp), %xmm2 ## encoding: [0xc5,0xfa,0x10,0x54,0x24,0x4c]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm2 = mem[0],zero,zero,zero
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vinsertps $16, {{[0-9]+}}(%esp), %xmm2, %xmm2 ## encoding: [0xc4,0xe3,0x69,0x21,0x54,0x24,0x48,0x10]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm2 = xmm2[0],mem[0],xmm2[2,3]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vinsertps $32, {{[0-9]+}}(%esp), %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x21,0x44,0x24,0x34,0x20]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0,1],mem[0],xmm0[3]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vinsertps $32, {{[0-9]+}}(%esp), %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x21,0x4c,0x24,0x24,0x20]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm1 = xmm1[0,1],mem[0],xmm1[3]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vinsertps $32, {{[0-9]+}}(%esp), %xmm2, %xmm2 ## encoding: [0xc4,0xe3,0x69,0x21,0x54,0x24,0x44,0x20]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm2 = xmm2[0,1],mem[0],xmm2[3]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vinsertps $48, {{[0-9]+}}(%esp), %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x21,0x44,0x24,0x30,0x30]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0,1,2],mem[0]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vinsertps $48, {{[0-9]+}}(%esp), %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x21,0x4c,0x24,0x20,0x30]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm1 = xmm1[0,1,2],mem[0]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vinsertps $48, {{[0-9]+}}(%esp), %xmm2, %xmm2 ## encoding: [0xc4,0xe3,0x69,0x21,0x54,0x24,0x40,0x30]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm2 = xmm2[0,1,2],mem[0]
; FMACALL32_BDVER2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 ## encoding: [0xc4,0xe3,0x7d,0x18,0xc1,0x01]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovss {{[0-9]+}}(%esp), %xmm1 ## encoding: [0xc5,0xfa,0x10,0x4c,0x24,0x5c]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm1 = mem[0],zero,zero,zero
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vinsertps $16, {{[0-9]+}}(%esp), %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x21,0x4c,0x24,0x58,0x10]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm1 = xmm1[0],mem[0],xmm1[2,3]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vinsertps $32, {{[0-9]+}}(%esp), %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x21,0x4c,0x24,0x54,0x20]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm1 = xmm1[0,1],mem[0],xmm1[3]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vinsertps $48, {{[0-9]+}}(%esp), %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x21,0x4c,0x24,0x50,0x30]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm1 = xmm1[0,1,2],mem[0]
; FMACALL32_BDVER2-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1 ## encoding: [0xc4,0xe3,0x75,0x18,0xca,0x01]
; FMACALL32_BDVER2-NEXT: movl %ebp, %esp ## encoding: [0x89,0xec]
; FMACALL32_BDVER2-NEXT: popl %ebp ## encoding: [0x5d]
; FMACALL32_BDVER2-NEXT: retl ## encoding: [0xc3]
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
entry:
%call = call <16 x float> @llvm.fma.v16f32(<16 x float> %a, <16 x float> %b, <16 x float> %c)
ret <16 x float> %call
}
define <2 x double> @test_v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c) #0 {
; FMA32-LABEL: test_v2f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA32: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA32-NEXT: vfmadd213pd %xmm2, %xmm1, %xmm0 ## encoding: [0xc4,0xe2,0xf1,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA32-NEXT: ## xmm0 = (xmm1 * xmm0) + xmm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA32-NEXT: retl ## encoding: [0xc3]
;
; FMA64-LABEL: test_v2f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA64: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA64-NEXT: vfmadd213pd %xmm2, %xmm1, %xmm0 ## encoding: [0xc4,0xe2,0xf1,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA64-NEXT: ## xmm0 = (xmm1 * xmm0) + xmm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA64-NEXT: retq ## encoding: [0xc3]
;
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-LABEL: test_v2f64:
; FMACALL64: ## %bb.0: ## %entry
; FMACALL64-NEXT: subq $72, %rsp ## encoding: [0x48,0x83,0xec,0x48]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x54,0x24,0x20]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x4c,0x24,0x10]
; FMACALL64-NEXT: movaps %xmm0, (%rsp) ## 16-byte Spill
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x04,0x24]
; FMACALL64-NEXT: callq _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x30]
; FMACALL64-NEXT: movaps (%rsp), %xmm0 ## 16-byte Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x04,0x24]
; FMACALL64-NEXT: movhlps %xmm0, %xmm0 ## encoding: [0x0f,0x12,0xc0]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x10]
; FMACALL64-NEXT: movhlps %xmm1, %xmm1 ## encoding: [0x0f,0x12,0xc9]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x54,0x24,0x20]
; FMACALL64-NEXT: movhlps %xmm2, %xmm2 ## encoding: [0x0f,0x12,0xd2]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1]
; FMACALL64-NEXT: callq _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x30]
; FMACALL64-NEXT: movlhps %xmm0, %xmm1 ## encoding: [0x0f,0x16,0xc8]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],xmm0[0]
; FMACALL64-NEXT: movaps %xmm1, %xmm0 ## encoding: [0x0f,0x28,0xc1]
; FMACALL64-NEXT: addq $72, %rsp ## encoding: [0x48,0x83,0xc4,0x48]
; FMACALL64-NEXT: retq ## encoding: [0xc3]
;
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-LABEL: test_v2f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: vfmadd213pd %xmm2, %xmm1, %xmm0 ## encoding: [0xc4,0xe2,0xf1,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; AVX512-NEXT: ## xmm0 = (xmm1 * xmm0) + xmm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: retq ## encoding: [0xc3]
;
; AVX512VL-LABEL: test_v2f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512VL: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: vfmadd213pd %xmm2, %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; AVX512VL-NEXT: ## xmm0 = (xmm1 * xmm0) + xmm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: retq ## encoding: [0xc3]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
;
; FMACALL32_BDVER2-LABEL: test_v2f64:
; FMACALL32_BDVER2: ## %bb.0: ## %entry
; FMACALL32_BDVER2-NEXT: subl $108, %esp ## encoding: [0x83,0xec,0x6c]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x44,0x24,0x50]
; FMACALL32_BDVER2-NEXT: vmovlhps %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf8,0x16,0xc1]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0],xmm1[0]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x54,0x24,0x30]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x4c,0x24,0x40]
; FMACALL32_BDVER2-NEXT: vmovlps %xmm2, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xf8,0x13,0x54,0x24,0x10]
; FMACALL32_BDVER2-NEXT: vmovups %xmm0, (%esp) ## encoding: [0xc5,0xf8,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: calll _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: vmovapd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf9,0x28,0x44,0x24,0x30]
; FMACALL32_BDVER2-NEXT: vmovhpd %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xf9,0x17,0x44,0x24,0x10]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x44,0x24,0x40]
; FMACALL32_BDVER2-NEXT: vmovlps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 ## 16-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x12,0x44,0x24,0x58]
; FMACALL32_BDVER2-NEXT: ## xmm0 = mem[0,1],xmm0[2,3]
; FMACALL32_BDVER2-NEXT: vmovups %xmm0, (%esp) ## encoding: [0xc5,0xf8,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: fstpl {{[0-9]+}}(%esp) ## encoding: [0xdd,0x5c,0x24,0x28]
; FMACALL32_BDVER2-NEXT: calll _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstpl {{[0-9]+}}(%esp) ## encoding: [0xdd,0x5c,0x24,0x20]
; FMACALL32_BDVER2-NEXT: vmovsd {{[0-9]+}}(%esp), %xmm0 ## encoding: [0xc5,0xfb,0x10,0x44,0x24,0x28]
; FMACALL32_BDVER2-NEXT: ## xmm0 = mem[0],zero
; FMACALL32_BDVER2-NEXT: vmovhpd {{[0-9]+}}(%esp), %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0x16,0x44,0x24,0x20]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0],mem[0]
; FMACALL32_BDVER2-NEXT: addl $108, %esp ## encoding: [0x83,0xc4,0x6c]
; FMACALL32_BDVER2-NEXT: retl ## encoding: [0xc3]
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
entry:
%call = call <2 x double> @llvm.fma.v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c)
ret <2 x double> %call
}
define <4 x double> @test_v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c) #0 {
; FMA32-LABEL: test_v4f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA32: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA32-NEXT: vfmadd213pd %ymm2, %ymm1, %ymm0 ## encoding: [0xc4,0xe2,0xf5,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA32-NEXT: ## ymm0 = (ymm1 * ymm0) + ymm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA32-NEXT: retl ## encoding: [0xc3]
;
; FMA64-LABEL: test_v4f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA64: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA64-NEXT: vfmadd213pd %ymm2, %ymm1, %ymm0 ## encoding: [0xc4,0xe2,0xf5,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA64-NEXT: ## ymm0 = (ymm1 * ymm0) + ymm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA64-NEXT: retq ## encoding: [0xc3]
;
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-LABEL: test_v4f64:
; FMACALL64: ## %bb.0: ## %entry
; FMACALL64-NEXT: subq $120, %rsp ## encoding: [0x48,0x83,0xec,0x78]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x6c,0x24,0x40]
; FMACALL64-NEXT: movaps %xmm4, (%rsp) ## 16-byte Spill
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x24,0x24]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x5c,0x24,0x30]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x54,0x24,0x60]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x4c,0x24,0x20]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x50]
; FMACALL64-NEXT: movaps %xmm2, %xmm1 ## encoding: [0x0f,0x28,0xca]
; FMACALL64-NEXT: movaps %xmm4, %xmm2 ## encoding: [0x0f,0x28,0xd4]
; FMACALL64-NEXT: callq _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x10]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x50]
; FMACALL64-NEXT: movhlps %xmm0, %xmm0 ## encoding: [0x0f,0x12,0xc0]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x60]
; FMACALL64-NEXT: movhlps %xmm1, %xmm1 ## encoding: [0x0f,0x12,0xc9]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1]
; FMACALL64-NEXT: movaps (%rsp), %xmm2 ## 16-byte Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x14,0x24]
; FMACALL64-NEXT: movhlps %xmm2, %xmm2 ## encoding: [0x0f,0x12,0xd2]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1]
; FMACALL64-NEXT: callq _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x10]
; FMACALL64-NEXT: movlhps %xmm0, %xmm1 ## encoding: [0x0f,0x16,0xc8]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],xmm0[0]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x4c,0x24,0x10]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x20]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x30]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x54,0x24,0x40]
; FMACALL64-NEXT: callq _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: reloc_branch_4byte_pcrel
; FMACALL64-NEXT: movaps %xmm0, (%rsp) ## 16-byte Spill
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x04,0x24]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x20]
; FMACALL64-NEXT: movhlps %xmm0, %xmm0 ## encoding: [0x0f,0x12,0xc0]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x30]
; FMACALL64-NEXT: movhlps %xmm1, %xmm1 ## encoding: [0x0f,0x12,0xc9]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x54,0x24,0x40]
; FMACALL64-NEXT: movhlps %xmm2, %xmm2 ## encoding: [0x0f,0x12,0xd2]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1]
; FMACALL64-NEXT: callq _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: reloc_branch_4byte_pcrel
; FMACALL64-NEXT: movaps (%rsp), %xmm1 ## 16-byte Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x0c,0x24]
; FMACALL64-NEXT: movlhps %xmm0, %xmm1 ## encoding: [0x0f,0x16,0xc8]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],xmm0[0]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x10]
; FMACALL64-NEXT: addq $120, %rsp ## encoding: [0x48,0x83,0xc4,0x78]
; FMACALL64-NEXT: retq ## encoding: [0xc3]
;
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-LABEL: test_v4f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: vfmadd213pd %ymm2, %ymm1, %ymm0 ## encoding: [0xc4,0xe2,0xf5,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; AVX512-NEXT: ## ymm0 = (ymm1 * ymm0) + ymm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: retq ## encoding: [0xc3]
;
; AVX512VL-LABEL: test_v4f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512VL: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: vfmadd213pd %ymm2, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf5,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; AVX512VL-NEXT: ## ymm0 = (ymm1 * ymm0) + ymm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: retq ## encoding: [0xc3]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
;
; FMACALL32_BDVER2-LABEL: test_v4f64:
; FMACALL32_BDVER2: ## %bb.0: ## %entry
; FMACALL32_BDVER2-NEXT: subl $252, %esp ## encoding: [0x81,0xec,0xfc,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractf128 $1, %ymm2, %xmm3 ## encoding: [0xc4,0xe3,0x7d,0x19,0xd3,0x01]
; FMACALL32_BDVER2-NEXT: vmovups %ymm2, {{[-0-9]+}}(%e{{[sb]}}p) ## 32-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x11,0x94,0x24,0xc0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 32-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x11,0x84,0x24,0x80,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractf128 $1, %ymm1, %xmm2 ## encoding: [0xc4,0xe3,0x7d,0x19,0xca,0x01]
; FMACALL32_BDVER2-NEXT: vextractf128 $1, %ymm0, %xmm0 ## encoding: [0xc4,0xe3,0x7d,0x19,0xc0,0x01]
; FMACALL32_BDVER2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%e{{[sb]}}p) ## 32-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x11,0x8c,0x24,0xa0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x44,0x24,0x60]
; FMACALL32_BDVER2-NEXT: vmovlhps %xmm2, %xmm0, %xmm0 ## encoding: [0xc5,0xf8,0x16,0xc2]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0],xmm2[0]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x5c,0x24,0x70]
; FMACALL32_BDVER2-NEXT: vmovlps %xmm3, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xf8,0x13,0x5c,0x24,0x10]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x54,0x24,0x50]
; FMACALL32_BDVER2-NEXT: vmovups %xmm0, (%esp) ## encoding: [0xc5,0xf8,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0x7c,0x24,0x44]
; FMACALL32_BDVER2-NEXT: vmovupd {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfd,0x10,0x84,0x24,0xc0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovhpd %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xf9,0x17,0x44,0x24,0x10]
; FMACALL32_BDVER2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x10,0x84,0x24,0x80,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vunpckhpd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 ## 16-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf9,0x15,0x84,0x24,0xa0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[1],mem[1]
; FMACALL32_BDVER2-NEXT: vmovups %xmm0, (%esp) ## encoding: [0xc5,0xf8,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0x7c,0x24,0x38]
; FMACALL32_BDVER2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x10,0x84,0x24,0xc0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovlps %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xf8,0x13,0x44,0x24,0x10]
; FMACALL32_BDVER2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x10,0x84,0x24,0x80,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vunpcklpd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 ## 16-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf9,0x14,0x84,0x24,0xa0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0],mem[0]
; FMACALL32_BDVER2-NEXT: vmovups %xmm0, (%esp) ## encoding: [0xc5,0xf8,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: vmovapd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf9,0x28,0x44,0x24,0x70]
; FMACALL32_BDVER2-NEXT: vmovhpd %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xf9,0x17,0x44,0x24,0x10]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x44,0x24,0x50]
; FMACALL32_BDVER2-NEXT: vmovlps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 ## 16-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x12,0x44,0x24,0x68]
; FMACALL32_BDVER2-NEXT: ## xmm0 = mem[0,1],xmm0[2,3]
; FMACALL32_BDVER2-NEXT: vmovups %xmm0, (%esp) ## encoding: [0xc5,0xf8,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: fstpl {{[0-9]+}}(%esp) ## encoding: [0xdd,0x5c,0x24,0x30]
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0x6c,0x24,0x38]
; FMACALL32_BDVER2-NEXT: fstpl {{[0-9]+}}(%esp) ## encoding: [0xdd,0x5c,0x24,0x28]
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0x6c,0x24,0x44]
; FMACALL32_BDVER2-NEXT: fstpl {{[0-9]+}}(%esp) ## encoding: [0xdd,0x5c,0x24,0x20]
; FMACALL32_BDVER2-NEXT: calll _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstpl {{[0-9]+}}(%esp) ## encoding: [0xdd,0x5c,0x24,0x18]
; FMACALL32_BDVER2-NEXT: vmovsd {{[0-9]+}}(%esp), %xmm0 ## encoding: [0xc5,0xfb,0x10,0x44,0x24,0x30]
; FMACALL32_BDVER2-NEXT: ## xmm0 = mem[0],zero
; FMACALL32_BDVER2-NEXT: vmovsd {{[0-9]+}}(%esp), %xmm1 ## encoding: [0xc5,0xfb,0x10,0x4c,0x24,0x20]
; FMACALL32_BDVER2-NEXT: ## xmm1 = mem[0],zero
; FMACALL32_BDVER2-NEXT: vmovhpd {{[0-9]+}}(%esp), %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0x16,0x44,0x24,0x28]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0],mem[0]
; FMACALL32_BDVER2-NEXT: vmovhpd {{[0-9]+}}(%esp), %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x16,0x4c,0x24,0x18]
; FMACALL32_BDVER2-NEXT: ## xmm1 = xmm1[0],mem[0]
; FMACALL32_BDVER2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 ## encoding: [0xc4,0xe3,0x7d,0x18,0xc1,0x01]
; FMACALL32_BDVER2-NEXT: addl $252, %esp ## encoding: [0x81,0xc4,0xfc,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: retl ## encoding: [0xc3]
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
entry:
%call = call <4 x double> @llvm.fma.v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c)
ret <4 x double> %call
}
define <8 x double> @test_v8f64(<8 x double> %a, <8 x double> %b, <8 x double> %c) #0 {
; FMA32-LABEL: test_v8f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA32: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA32-NEXT: pushl %ebp ## encoding: [0x55]
; FMA32-NEXT: movl %esp, %ebp ## encoding: [0x89,0xe5]
; FMA32-NEXT: andl $-32, %esp ## encoding: [0x83,0xe4,0xe0]
; FMA32-NEXT: subl $32, %esp ## encoding: [0x83,0xec,0x20]
; FMA32-NEXT: vfmadd213pd 8(%ebp), %ymm2, %ymm0 ## encoding: [0xc4,0xe2,0xed,0xa8,0x45,0x08]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA32-NEXT: ## ymm0 = (ymm2 * ymm0) + mem
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA32-NEXT: vfmadd213pd 40(%ebp), %ymm3, %ymm1 ## encoding: [0xc4,0xe2,0xe5,0xa8,0x4d,0x28]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA32-NEXT: ## ymm1 = (ymm3 * ymm1) + mem
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA32-NEXT: movl %ebp, %esp ## encoding: [0x89,0xec]
; FMA32-NEXT: popl %ebp ## encoding: [0x5d]
; FMA32-NEXT: retl ## encoding: [0xc3]
;
; FMA64-LABEL: test_v8f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; FMA64: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA64-NEXT: vfmadd213pd %ymm4, %ymm2, %ymm0 ## encoding: [0xc4,0xe2,0xed,0xa8,0xc4]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA64-NEXT: ## ymm0 = (ymm2 * ymm0) + ymm4
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA64-NEXT: vfmadd213pd %ymm5, %ymm3, %ymm1 ## encoding: [0xc4,0xe2,0xe5,0xa8,0xcd]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMA64-NEXT: ## ymm1 = (ymm3 * ymm1) + ymm5
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; FMA64-NEXT: retq ## encoding: [0xc3]
;
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-LABEL: test_v8f64:
; FMACALL64: ## %bb.0: ## %entry
; FMACALL64-NEXT: subq $152, %rsp ## encoding: [0x48,0x81,0xec,0x98,0x00,0x00,0x00]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x7c,0x24,0x70]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x74,0x24,0x20]
; FMACALL64-NEXT: movaps %xmm5, (%rsp) ## 16-byte Spill
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x2c,0x24]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x64,0x24,0x10]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x5c,0x24,0x60]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x54,0x24,0x50]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x4c,0x24,0x40]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x84,0x24,0x80,0x00,0x00,0x00]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xa0,0x00,0x00,0x00]
; FMACALL64-NEXT: movaps %xmm4, %xmm1 ## encoding: [0x0f,0x28,0xcc]
; FMACALL64-NEXT: callq _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x30]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x84,0x24,0x80,0x00,0x00,0x00]
; FMACALL64-NEXT: movhlps %xmm0, %xmm0 ## encoding: [0x0f,0x12,0xc0]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x10]
; FMACALL64-NEXT: movhlps %xmm1, %xmm1 ## encoding: [0x0f,0x12,0xc9]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xa0,0x00,0x00,0x00]
; FMACALL64-NEXT: movhlps %xmm2, %xmm2 ## encoding: [0x0f,0x12,0xd2]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1]
; FMACALL64-NEXT: callq _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x30]
; FMACALL64-NEXT: movlhps %xmm0, %xmm1 ## encoding: [0x0f,0x16,0xc8]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],xmm0[0]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x4c,0x24,0x30]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x40]
; FMACALL64-NEXT: movaps (%rsp), %xmm1 ## 16-byte Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x0c,0x24]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xb0,0x00,0x00,0x00]
; FMACALL64-NEXT: callq _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x10]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x40]
; FMACALL64-NEXT: movhlps %xmm0, %xmm0 ## encoding: [0x0f,0x12,0xc0]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1]
; FMACALL64-NEXT: movaps (%rsp), %xmm1 ## 16-byte Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x0c,0x24]
; FMACALL64-NEXT: movhlps %xmm1, %xmm1 ## encoding: [0x0f,0x12,0xc9]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xb0,0x00,0x00,0x00]
; FMACALL64-NEXT: movhlps %xmm2, %xmm2 ## encoding: [0x0f,0x12,0xd2]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1]
; FMACALL64-NEXT: callq _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x10]
; FMACALL64-NEXT: movlhps %xmm0, %xmm1 ## encoding: [0x0f,0x16,0xc8]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],xmm0[0]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x4c,0x24,0x10]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x50]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x20]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xc0,0x00,0x00,0x00]
; FMACALL64-NEXT: callq _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: reloc_branch_4byte_pcrel
; FMACALL64-NEXT: movaps %xmm0, (%rsp) ## 16-byte Spill
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x04,0x24]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x50]
; FMACALL64-NEXT: movhlps %xmm0, %xmm0 ## encoding: [0x0f,0x12,0xc0]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x20]
; FMACALL64-NEXT: movhlps %xmm1, %xmm1 ## encoding: [0x0f,0x12,0xc9]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xc0,0x00,0x00,0x00]
; FMACALL64-NEXT: movhlps %xmm2, %xmm2 ## encoding: [0x0f,0x12,0xd2]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1]
; FMACALL64-NEXT: callq _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: reloc_branch_4byte_pcrel
; FMACALL64-NEXT: movaps (%rsp), %xmm1 ## 16-byte Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x0c,0x24]
; FMACALL64-NEXT: movlhps %xmm0, %xmm1 ## encoding: [0x0f,0x16,0xc8]
; FMACALL64-NEXT: ## xmm1 = xmm1[0],xmm0[0]
; FMACALL64-NEXT: movaps %xmm1, (%rsp) ## 16-byte Spill
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x0c,0x24]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x60]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x70]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xd0,0x00,0x00,0x00]
; FMACALL64-NEXT: callq _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x29,0x44,0x24,0x20]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x60]
; FMACALL64-NEXT: movhlps %xmm0, %xmm0 ## encoding: [0x0f,0x12,0xc0]
; FMACALL64-NEXT: ## xmm0 = xmm0[1,1]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x70]
; FMACALL64-NEXT: movhlps %xmm1, %xmm1 ## encoding: [0x0f,0x12,0xc9]
; FMACALL64-NEXT: ## xmm1 = xmm1[1,1]
; FMACALL64-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 ## encoding: [0x0f,0x28,0x94,0x24,0xd0,0x00,0x00,0x00]
; FMACALL64-NEXT: movhlps %xmm2, %xmm2 ## encoding: [0x0f,0x12,0xd2]
; FMACALL64-NEXT: ## xmm2 = xmm2[1,1]
; FMACALL64-NEXT: callq _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL64-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: reloc_branch_4byte_pcrel
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x5c,0x24,0x20]
; FMACALL64-NEXT: movlhps %xmm0, %xmm3 ## encoding: [0x0f,0x16,0xd8]
; FMACALL64-NEXT: ## xmm3 = xmm3[0],xmm0[0]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x44,0x24,0x30]
[X86] Regenerate fma.ll checks using current version of the script which produces different regular expressions on spills and reloads. NFC llvm-svn: 337354
2018-07-18 15:08:28 +08:00
; FMACALL64-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 ## 16-byte Reload
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x4c,0x24,0x10]
; FMACALL64-NEXT: movaps (%rsp), %xmm2 ## 16-byte Reload
; FMACALL64-NEXT: ## encoding: [0x0f,0x28,0x14,0x24]
; FMACALL64-NEXT: addq $152, %rsp ## encoding: [0x48,0x81,0xc4,0x98,0x00,0x00,0x00]
; FMACALL64-NEXT: retq ## encoding: [0xc3]
;
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-LABEL: test_v8f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: vfmadd213pd %zmm2, %zmm1, %zmm0 ## encoding: [0x62,0xf2,0xf5,0x48,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; AVX512-NEXT: ## zmm0 = (zmm1 * zmm0) + zmm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512-NEXT: retq ## encoding: [0xc3]
;
; AVX512VL-LABEL: test_v8f64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512VL: ## %bb.0: ## %entry
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: vfmadd213pd %zmm2, %zmm1, %zmm0 ## encoding: [0x62,0xf2,0xf5,0x48,0xa8,0xc2]
[X86] Add comments to the end of FMA3 instructions to make the operation clear Summary: There are 3 different operand orders for FMA instructions so figuring out the exact operation being performed requires a lot of thought. This patch adds a comment to the end of the assembly line to print the exact operation. I think I've got all the instructions in here except the ones with builtin rounding. I didn't update all tests, but I assume we can get them as we regenerate tests in the future. Reviewers: spatel, v_klochkov, RKSimon Reviewed By: spatel Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D44345 llvm-svn: 327225
2018-03-11 05:30:46 +08:00
; AVX512VL-NEXT: ## zmm0 = (zmm1 * zmm0) + zmm2
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
; AVX512VL-NEXT: retq ## encoding: [0xc3]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
;
; FMACALL32_BDVER2-LABEL: test_v8f64:
; FMACALL32_BDVER2: ## %bb.0: ## %entry
; FMACALL32_BDVER2-NEXT: pushl %ebp ## encoding: [0x55]
; FMACALL32_BDVER2-NEXT: movl %esp, %ebp ## encoding: [0x89,0xe5]
; FMACALL32_BDVER2-NEXT: andl $-32, %esp ## encoding: [0x83,0xe4,0xe0]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: subl $352, %esp ## encoding: [0x81,0xec,0x60,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: ## imm = 0x160
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 32-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x29,0x84,0x24,0xe0,0x00,0x00,0x00]
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 56(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x38]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps %ymm2, {{[-0-9]+}}(%e{{[sb]}}p) ## 32-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x29,0x94,0x24,0x00,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractf128 $1, %ymm3, %xmm2 ## encoding: [0xc4,0xe3,0x7d,0x19,0xda,0x01]
; FMACALL32_BDVER2-NEXT: vmovaps %ymm3, {{[-0-9]+}}(%e{{[sb]}}p) ## 32-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x29,0x9c,0x24,0xc0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovaps %ymm1, {{[-0-9]+}}(%e{{[sb]}}p) ## 32-byte Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x29,0x8c,0x24,0xa0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x94,0x24,0x30,0x01,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovlps %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xf8,0x13,0x44,0x24,0x10]
; FMACALL32_BDVER2-NEXT: vextractf128 $1, %ymm1, %xmm0 ## encoding: [0xc4,0xe3,0x7d,0x19,0xc8,0x01]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x84,0x24,0x40,0x01,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovlhps %xmm2, %xmm0, %xmm0 ## encoding: [0xc5,0xf8,0x16,0xc2]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0],xmm2[0]
; FMACALL32_BDVER2-NEXT: vmovups %xmm0, (%esp) ## encoding: [0xc5,0xf8,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovapd 40(%ebp), %xmm0 ## encoding: [0xc5,0xf9,0x28,0x45,0x28]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0x94,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovhpd %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xf9,0x17,0x44,0x24,0x10]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0xa0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vunpckhpd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 ## 16-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf9,0x15,0x84,0x24,0xc0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[1],mem[1]
; FMACALL32_BDVER2-NEXT: vmovups %xmm0, (%esp) ## encoding: [0xc5,0xf8,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 40(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x28]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0x88,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovlps %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xf8,0x13,0x44,0x24,0x10]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0xa0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vunpcklpd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 ## 16-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf9,0x14,0x84,0x24,0xc0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0],mem[0]
; FMACALL32_BDVER2-NEXT: vmovups %xmm0, (%esp) ## encoding: [0xc5,0xf8,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovapd 24(%ebp), %xmm0 ## encoding: [0xc5,0xf9,0x28,0x45,0x18]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0xc0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovhpd %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xf9,0x17,0x44,0x24,0x10]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0x00,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vextractf128 $1, %ymm0, %xmm1 ## encoding: [0xc4,0xe3,0x7d,0x19,0xc1,0x01]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0xe0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x4c,0x24,0x20]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vextractf128 $1, %ymm0, %xmm0 ## encoding: [0xc4,0xe3,0x7d,0x19,0xc0,0x01]
; FMACALL32_BDVER2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x29,0x44,0x24,0x30]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vunpckhpd %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0x15,0xc1]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[1],xmm1[1]
; FMACALL32_BDVER2-NEXT: vmovups %xmm0, (%esp) ## encoding: [0xc5,0xf8,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 24(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x18]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xbc,0x24,0xa0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vmovlps %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xf8,0x13,0x44,0x24,0x10]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x44,0x24,0x30]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vunpcklpd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 ## 16-byte Folded Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf9,0x14,0x44,0x24,0x20]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0],mem[0]
; FMACALL32_BDVER2-NEXT: vmovups %xmm0, (%esp) ## encoding: [0xc5,0xf8,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: calll _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovapd 8(%ebp), %xmm0 ## encoding: [0xc5,0xf9,0x28,0x45,0x08]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0x7c,0x24,0x30]
; FMACALL32_BDVER2-NEXT: vmovhpd %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xf9,0x17,0x44,0x24,0x10]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0xe0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vunpckhpd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 ## 16-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf9,0x15,0x84,0x24,0x00,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[1],mem[1]
; FMACALL32_BDVER2-NEXT: vmovups %xmm0, (%esp) ## encoding: [0xc5,0xf8,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovaps 8(%ebp), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x45,0x08]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: fstpt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Spill
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0x7c,0x24,0x20]
; FMACALL32_BDVER2-NEXT: vmovlps %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xf8,0x13,0x44,0x24,0x10]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 ## 32-byte Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xfc,0x28,0x84,0x24,0xe0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: vunpcklpd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 ## 16-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf9,0x14,0x84,0x24,0x00,0x01,0x00,0x00]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0],mem[0]
; FMACALL32_BDVER2-NEXT: vmovups %xmm0, (%esp) ## encoding: [0xc5,0xf8,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: vzeroupper ## encoding: [0xc5,0xf8,0x77]
; FMACALL32_BDVER2-NEXT: calll _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: FK_PCRel_4
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: vmovapd 56(%ebp), %xmm0 ## encoding: [0xc5,0xf9,0x28,0x45,0x38]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovhpd %xmm0, {{[0-9]+}}(%esp) ## encoding: [0xc5,0xf9,0x17,0x44,0x24,0x10]
; FMACALL32_BDVER2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x28,0x84,0x24,0x30,0x01,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: vmovlps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 ## 16-byte Folded Reload
[x86] allow vector load narrowing with multi-use values This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs. Apart from 2-3 strange cases, these are all wins. I've structured this to be no-functional-change-intended for any target except for x86 because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those targets have existing regression tests (4, 4, 10 files respectively) that would be affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show any regression test diffs. The trade-off is deciding if an extra vector load is better than a single wide load + extract_subvector. For x86, this is almost always better (on paper at least) because we often can fold loads into subsequent ops and not increase the official instruction count. There's also some unknown -- but potentially large -- benefit from using narrower vector ops if wide ops are implemented with multiple uops and/or frequency throttling is avoided. Differential Revision: https://reviews.llvm.org/D54073 llvm-svn: 346595
2018-11-11 04:05:31 +08:00
; FMACALL32_BDVER2-NEXT: ## encoding: [0xc5,0xf8,0x12,0x84,0x24,0x48,0x01,0x00,0x00]
AMD BdVer2 (Piledriver) Initial Scheduler model Summary: # Overview This is somewhat partial. * Latencies are good {F7371125} * All of these remaining inconsistencies //appear// to be noise/noisy/flaky. * NumMicroOps are somewhat good {F7371158} * Most of the remaining inconsistencies are from `Ld` / `Ld_ReadAfterLd` classes * Actual unit occupation (pipes, `ResourceCycles`) are undiscovered lands, i did not really look there. They are basically verbatum copy from `btver2` * Many `InstRW`. And there are still inconsistencies left... To be noted: I think this is the first new schedule profile produced with the new next-gen tools like llvm-exegesis! # Benchmark I realize that isn't what was suggested, but i'll start with some "internal" public real-world benchmark i understand - [[ https://github.com/darktable-org/rawspeed | RawSpeed raw image decoding library ]]. Diff (the exact clang from trunk without/with this patch): ``` Comparing /home/lebedevri/rawspeed/build-old/src/utilities/rsbench/rsbench to /home/lebedevri/rawspeed/build-new/src/utilities/rsbench/rsbench Benchmark Time CPU Time Old Time New CPU Old CPU New ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_mean -0.0607 -0.0604 234 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_median -0.0630 -0.0626 233 219 233 219 Canon/EOS 5D Mark II/09.canon.sraw1.cr2/threads:8/real_time_stddev +0.2581 +0.2587 1 2 1 2 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_mean -0.0770 -0.0767 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_median -0.0767 -0.0763 144 133 144 133 Canon/EOS 5D Mark II/10.canon.sraw2.cr2/threads:8/real_time_stddev -0.4170 -0.4156 1 0 1 0 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_mean -0.0271 -0.0270 463 450 463 450 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_median -0.0093 -0.0093 453 449 453 449 Canon/EOS 5DS/2K4A9927.CR2/threads:8/real_time_stddev -0.7280 -0.7280 13 4 13 4 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_mean -0.0065 -0.0065 569 565 569 565 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_median -0.0077 -0.0077 569 564 569 564 Canon/EOS 5DS/2K4A9928.CR2/threads:8/real_time_stddev +1.0077 +1.0068 2 5 2 5 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_pvalue 0.0220 0.0199 U Test, Repetitions: 25 vs 25 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_mean +0.0006 +0.0007 312 312 312 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_median +0.0031 +0.0032 311 312 311 312 Canon/EOS 5DS/2K4A9929.CR2/threads:8/real_time_stddev -0.7069 -0.7072 4 1 4 1 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_pvalue 0.0004 0.0004 U Test, Repetitions: 25 vs 25 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_mean -0.0015 -0.0015 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_median -0.0010 -0.0011 141 141 141 141 Canon/EOS 10D/CRW_7673.CRW/threads:8/real_time_stddev -0.1486 -0.1456 0 0 0 0 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_pvalue 0.6139 0.8766 U Test, Repetitions: 25 vs 25 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_mean -0.0008 -0.0005 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_median -0.0006 -0.0002 60 60 60 60 Canon/EOS 40D/_MG_0154.CR2/threads:8/real_time_stddev -0.1467 -0.1390 0 0 0 0 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_pvalue 0.0137 0.0137 U Test, Repetitions: 25 vs 25 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_mean +0.0002 +0.0002 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_median -0.0015 -0.0014 275 275 275 275 Canon/EOS 77D/IMG_4049.CR2/threads:8/real_time_stddev +3.3687 +3.3587 0 2 0 2 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_pvalue 0.4041 0.3933 U Test, Repetitions: 25 vs 25 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_mean +0.0004 +0.0004 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_median -0.0000 -0.0000 67 67 67 67 Canon/PowerShot G1/crw_1693.crw/threads:8/real_time_stddev +0.1947 +0.1995 0 0 0 0 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_pvalue 0.0074 0.0001 U Test, Repetitions: 25 vs 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_mean -0.0092 +0.0074 547 542 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_median -0.0054 +0.0115 544 541 25 25 Fujifilm/GFX 50S/20170525_0037TEST.RAF/threads:8/real_time_stddev -0.4086 -0.3486 8 5 0 0 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_pvalue 0.3320 0.0000 U Test, Repetitions: 25 vs 25 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_mean +0.0015 +0.0204 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_median +0.0001 +0.0203 218 218 12 12 Fujifilm/X-Pro2/_DSF3051.RAF/threads:8/real_time_stddev +0.2259 +0.2023 1 1 0 0 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_pvalue 0.0000 0.0001 U Test, Repetitions: 25 vs 25 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_mean -0.0209 -0.0179 96 94 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_median -0.0182 -0.0155 95 93 90 88 GoPro/HERO6 Black/GOPR9172.GPR/threads:8/real_time_stddev -0.6164 -0.2703 2 1 2 1 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_mean -0.0098 -0.0098 176 175 176 175 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_median -0.0126 -0.0126 176 174 176 174 Kodak/DCS Pro 14nx/D7465857.DCR/threads:8/real_time_stddev +6.9789 +6.9157 0 2 0 2 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_mean -0.0237 -0.0238 474 463 474 463 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_median -0.0267 -0.0267 473 461 473 461 Nikon/D850/Nikon-D850-14bit-lossless-compressed.NEF/threads:8/real_time_stddev +0.7179 +0.7178 3 5 3 5 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_pvalue 0.6837 0.6554 U Test, Repetitions: 25 vs 25 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_mean -0.0014 -0.0013 1375 1373 1375 1373 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_median +0.0018 +0.0019 1371 1374 1371 1374 Olympus/E-M1MarkII/Olympus_EM1mk2__HIRES_50MP.ORF/threads:8/real_time_stddev -0.7457 -0.7382 11 3 10 3 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_mean -0.0080 -0.0289 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_median -0.0070 -0.0287 22 22 10 10 Panasonic/DC-G9/P1000476.RW2/threads:8/real_time_stddev +1.0977 +0.6614 0 0 0 0 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_mean +0.0132 +0.0967 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_median +0.0132 +0.0956 35 36 10 11 Panasonic/DC-GH5/_T012014.RW2/threads:8/real_time_stddev -0.0407 -0.1695 0 0 0 0 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_mean +0.0331 +0.1307 13 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_median +0.0430 +0.1373 12 13 6 6 Panasonic/DC-GH5S/P1022085.RW2/threads:8/real_time_stddev -0.9006 -0.8847 1 0 0 0 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_pvalue 0.0016 0.0010 U Test, Repetitions: 25 vs 25 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_mean -0.0023 -0.0024 395 394 395 394 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_median -0.0029 -0.0030 395 394 395 393 Pentax/645Z/IMGP2837.PEF/threads:8/real_time_stddev -0.0275 -0.0375 1 1 1 1 Phase One/P65/CF027310.IIQ/threads:8/real_time_pvalue 0.0232 0.0000 U Test, Repetitions: 25 vs 25 Phase One/P65/CF027310.IIQ/threads:8/real_time_mean -0.0047 +0.0039 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_median -0.0050 +0.0037 114 113 28 28 Phase One/P65/CF027310.IIQ/threads:8/real_time_stddev -0.0599 -0.2683 1 1 0 0 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_mean +0.0206 +0.0207 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_median +0.0204 +0.0205 405 414 405 414 Samsung/NX1/2016-07-23-142101_sam_9364.srw/threads:8/real_time_stddev +0.2155 +0.2212 1 1 1 1 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_mean -0.0109 -0.0108 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_median -0.0104 -0.0103 147 145 147 145 Samsung/NX30/2015-03-07-163604_sam_7204.srw/threads:8/real_time_stddev -0.4919 -0.4800 0 0 0 0 Samsung/NX3000/_3184416.SRW/threads:8/real_time_pvalue 0.0000 0.0000 U Test, Repetitions: 25 vs 25 Samsung/NX3000/_3184416.SRW/threads:8/real_time_mean -0.0149 -0.0147 220 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_median -0.0173 -0.0169 221 217 220 217 Samsung/NX3000/_3184416.SRW/threads:8/real_time_stddev +1.0337 +1.0341 1 3 1 3 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_pvalue 0.0001 0.0001 U Test, Repetitions: 25 vs 25 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_mean -0.0019 -0.0019 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_median -0.0021 -0.0021 194 193 194 193 Sony/DSLR-A350/DSC05472.ARW/threads:8/real_time_stddev -0.4441 -0.4282 0 0 0 0 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_pvalue 0.0000 0.4263 U Test, Repetitions: 25 vs 25 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_mean +0.0258 -0.0006 81 83 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_median +0.0235 -0.0011 81 82 19 19 Sony/ILCE-7RM2/14-bit-compressed.ARW/threads:8/real_time_stddev +0.1634 +0.1070 1 1 0 0 ``` {F7443905} If we look at the `_mean`s, the time column, the biggest win is `-7.7%` (`Canon/EOS 5D Mark II/10.canon.sraw2.cr2`), and the biggest loose is `+3.3%` (`Panasonic/DC-GH5S/P1022085.RW2`); Overall: mean `-0.7436%`, median `-0.23%`, `cbrt(sum(time^3))` = `-8.73%` Looks good so far i'd say. llvm-exegesis details: {F7371117} {F7371125} {F7371128} {F7371144} {F7371158} Reviewers: craig.topper, RKSimon, andreadb, courbet, avt77, spatel, GGanesh Reviewed By: andreadb Subscribers: javed.absar, gbedwell, jfb, llvm-commits Differential Revision: https://reviews.llvm.org/D52779 llvm-svn: 345463
2018-10-28 04:46:30 +08:00
; FMACALL32_BDVER2-NEXT: ## xmm0 = mem[0,1],xmm0[2,3]
; FMACALL32_BDVER2-NEXT: vmovups %xmm0, (%esp) ## encoding: [0xc5,0xf8,0x11,0x04,0x24]
; FMACALL32_BDVER2-NEXT: fstpl {{[0-9]+}}(%esp) ## encoding: [0xdd,0x5c,0x24,0x60]
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0x6c,0x24,0x20]
; FMACALL32_BDVER2-NEXT: fstpl {{[0-9]+}}(%esp) ## encoding: [0xdd,0x5c,0x24,0x58]
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0x6c,0x24,0x30]
; FMACALL32_BDVER2-NEXT: fstpl {{[0-9]+}}(%esp) ## encoding: [0xdd,0x5c,0x24,0x50]
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0xa0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: fstpl {{[0-9]+}}(%esp) ## encoding: [0xdd,0x5c,0x24,0x48]
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0xc0,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: fstpl {{[0-9]+}}(%esp) ## encoding: [0xdd,0x9c,0x24,0x80,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0x88,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: fstpl {{[0-9]+}}(%esp) ## encoding: [0xdd,0x5c,0x24,0x78]
; FMACALL32_BDVER2-NEXT: fldt {{[-0-9]+}}(%e{{[sb]}}p) ## 10-byte Folded Reload
; FMACALL32_BDVER2-NEXT: ## encoding: [0xdb,0xac,0x24,0x94,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: fstpl {{[0-9]+}}(%esp) ## encoding: [0xdd,0x5c,0x24,0x70]
; FMACALL32_BDVER2-NEXT: calll _fma ## encoding: [0xe8,A,A,A,A]
; FMACALL32_BDVER2-NEXT: ## fixup A - offset: 1, value: _fma-4, kind: FK_PCRel_4
; FMACALL32_BDVER2-NEXT: fstpl {{[0-9]+}}(%esp) ## encoding: [0xdd,0x5c,0x24,0x68]
; FMACALL32_BDVER2-NEXT: vmovsd {{[0-9]+}}(%esp), %xmm0 ## encoding: [0xc5,0xfb,0x10,0x44,0x24,0x60]
; FMACALL32_BDVER2-NEXT: ## xmm0 = mem[0],zero
; FMACALL32_BDVER2-NEXT: vmovsd {{[0-9]+}}(%esp), %xmm1 ## encoding: [0xc5,0xfb,0x10,0x4c,0x24,0x50]
; FMACALL32_BDVER2-NEXT: ## xmm1 = mem[0],zero
; FMACALL32_BDVER2-NEXT: vmovhpd {{[0-9]+}}(%esp), %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0x16,0x44,0x24,0x58]
; FMACALL32_BDVER2-NEXT: ## xmm0 = xmm0[0],mem[0]
; FMACALL32_BDVER2-NEXT: vmovhpd {{[0-9]+}}(%esp), %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x16,0x4c,0x24,0x48]
; FMACALL32_BDVER2-NEXT: ## xmm1 = xmm1[0],mem[0]
; FMACALL32_BDVER2-NEXT: vmovsd {{[0-9]+}}(%esp), %xmm2 ## encoding: [0xc5,0xfb,0x10,0x54,0x24,0x70]
; FMACALL32_BDVER2-NEXT: ## xmm2 = mem[0],zero
; FMACALL32_BDVER2-NEXT: vmovhpd {{[0-9]+}}(%esp), %xmm2, %xmm2 ## encoding: [0xc5,0xe9,0x16,0x54,0x24,0x68]
; FMACALL32_BDVER2-NEXT: ## xmm2 = xmm2[0],mem[0]
; FMACALL32_BDVER2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 ## encoding: [0xc4,0xe3,0x7d,0x18,0xc1,0x01]
; FMACALL32_BDVER2-NEXT: vmovsd {{[0-9]+}}(%esp), %xmm1 ## encoding: [0xc5,0xfb,0x10,0x8c,0x24,0x80,0x00,0x00,0x00]
; FMACALL32_BDVER2-NEXT: ## xmm1 = mem[0],zero
; FMACALL32_BDVER2-NEXT: vmovhpd {{[0-9]+}}(%esp), %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x16,0x4c,0x24,0x78]
; FMACALL32_BDVER2-NEXT: ## xmm1 = xmm1[0],mem[0]
; FMACALL32_BDVER2-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1 ## encoding: [0xc4,0xe3,0x75,0x18,0xca,0x01]
; FMACALL32_BDVER2-NEXT: movl %ebp, %esp ## encoding: [0x89,0xec]
; FMACALL32_BDVER2-NEXT: popl %ebp ## encoding: [0x5d]
; FMACALL32_BDVER2-NEXT: retl ## encoding: [0xc3]
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
entry:
%call = call <8 x double> @llvm.fma.v8f64(<8 x double> %a, <8 x double> %b, <8 x double> %c)
ret <8 x double> %call
}
[X86] Cleanup fma tests a little bit. NFC. Reformat, isolate 213->231 xform, actually --check-prefix CHECK, and deduplicate the FMA intrinsic tests (FMA3 in AMD-land). llvm-svn: 240615
2015-06-25 08:40:25 +08:00
declare float @llvm.fma.f32(float, float, float)
declare double @llvm.fma.f64(double, double, double)
declare x86_fp80 @llvm.fma.f80(x86_fp80, x86_fp80, x86_fp80)
Add a testcase for the changes in r199938. <rdar://problem/15611947> llvm-svn: 200027
2014-01-25 03:00:19 +08:00
[AVX-512] Make VEX encoded FMA instructions available when AVX512 is enabled regardless of whether +fma was added on the command line. We weren't able to handle isel of the 128/256-bit FMA instructions when AVX512F was enabled but VLX and FMA weren't. I didn't mask FeatureAVX512 imply FeatureFMA as I wasn't sure I wanted disabling FMA to also disable AVX512. Instead we just can't prevent FMA instructions if AVX512 is enabled. Another option would be to promote 128/256-bit to 512-bit, do the operation and extract it. But that requires a lot of extra isel patterns. Since no CPUs exist that support AVX512, but not FMA just using the VEX instructions seems better. llvm-svn: 298051
2017-03-17 15:37:31 +08:00
declare <4 x float> @llvm.fma.v4f32(<4 x float>, <4 x float>, <4 x float>)
declare <8 x float> @llvm.fma.v8f32(<8 x float>, <8 x float>, <8 x float>)
declare <16 x float> @llvm.fma.v16f32(<16 x float>, <16 x float>, <16 x float>)
declare <2 x double> @llvm.fma.v2f64(<2 x double>, <2 x double>, <2 x double>)
declare <4 x double> @llvm.fma.v4f64(<4 x double>, <4 x double>, <4 x double>)
declare <8 x double> @llvm.fma.v8f64(<8 x double>, <8 x double>, <8 x double>)
[X86] Cleanup fma tests a little bit. NFC. Reformat, isolate 213->231 xform, actually --check-prefix CHECK, and deduplicate the FMA intrinsic tests (FMA3 in AMD-land). llvm-svn: 240615
2015-06-25 08:40:25 +08:00
attributes #0 = { nounwind }