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/vector-idiv-udiv-128.ll

610 lines
26 KiB
LLVM
Raw Normal View History

[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=SSE --check-prefix=SSE2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=SSE --check-prefix=SSE41
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX --check-prefix=AVX1
[AVX512] Add avx512bw command lines to 128-bit idiv tests. The multiply lowering on some of the tests can take advantage of the vpmovwb to simplify the truncate. llvm-svn: 314448
2017-09-29 02:45:29 +08:00
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX --check-prefix=AVX2 --check-prefix=AVX2NOBW
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw | FileCheck %s --check-prefix=AVX --check-prefix=AVX2 --check-prefix=AVX512BW
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
;
; udiv by 7
;
define <2 x i64> @test_div7_2i64(<2 x i64> %a) nounwind {
; SSE2-LABEL: test_div7_2i64:
[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
; SSE2: # %bb.0:
[X86][SSE2] Fix asm string for movq (Move Quadword) instruction. Replace "mov{d|q}" with "movq". Differential Revision: https://reviews.llvm.org/D32220 llvm-svn: 301386
2017-04-26 15:08:44 +08:00
; SSE2-NEXT: movq %xmm0, %rcx
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE2-NEXT: movabsq $2635249153387078803, %rsi # imm = 0x2492492492492493
; SSE2-NEXT: movq %rcx, %rax
; SSE2-NEXT: mulq %rsi
; SSE2-NEXT: subq %rdx, %rcx
; SSE2-NEXT: shrq %rcx
; SSE2-NEXT: addq %rdx, %rcx
; SSE2-NEXT: shrq $2, %rcx
[X86][SSE2] Fix asm string for movq (Move Quadword) instruction. Replace "mov{d|q}" with "movq". Differential Revision: https://reviews.llvm.org/D32220 llvm-svn: 301386
2017-04-26 15:08:44 +08:00
; SSE2-NEXT: movq %rcx, %xmm1
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
[X86][SSE2] Fix asm string for movq (Move Quadword) instruction. Replace "mov{d|q}" with "movq". Differential Revision: https://reviews.llvm.org/D32220 llvm-svn: 301386
2017-04-26 15:08:44 +08:00
; SSE2-NEXT: movq %xmm0, %rcx
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE2-NEXT: movq %rcx, %rax
; SSE2-NEXT: mulq %rsi
; SSE2-NEXT: subq %rdx, %rcx
; SSE2-NEXT: shrq %rcx
; SSE2-NEXT: addq %rdx, %rcx
; SSE2-NEXT: shrq $2, %rcx
[X86][SSE2] Fix asm string for movq (Move Quadword) instruction. Replace "mov{d|q}" with "movq". Differential Revision: https://reviews.llvm.org/D32220 llvm-svn: 301386
2017-04-26 15:08:44 +08:00
; SSE2-NEXT: movq %rcx, %xmm0
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE2-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE2-NEXT: movdqa %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: test_div7_2i64:
[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
; SSE41: # %bb.0:
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE41-NEXT: pextrq $1, %xmm0, %rcx
; SSE41-NEXT: movabsq $2635249153387078803, %rsi # imm = 0x2492492492492493
; SSE41-NEXT: movq %rcx, %rax
; SSE41-NEXT: mulq %rsi
; SSE41-NEXT: subq %rdx, %rcx
; SSE41-NEXT: shrq %rcx
; SSE41-NEXT: addq %rdx, %rcx
; SSE41-NEXT: shrq $2, %rcx
[X86][SSE2] Fix asm string for movq (Move Quadword) instruction. Replace "mov{d|q}" with "movq". Differential Revision: https://reviews.llvm.org/D32220 llvm-svn: 301386
2017-04-26 15:08:44 +08:00
; SSE41-NEXT: movq %rcx, %xmm1
; SSE41-NEXT: movq %xmm0, %rcx
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE41-NEXT: movq %rcx, %rax
; SSE41-NEXT: mulq %rsi
; SSE41-NEXT: subq %rdx, %rcx
; SSE41-NEXT: shrq %rcx
; SSE41-NEXT: addq %rdx, %rcx
; SSE41-NEXT: shrq $2, %rcx
[X86][SSE2] Fix asm string for movq (Move Quadword) instruction. Replace "mov{d|q}" with "movq". Differential Revision: https://reviews.llvm.org/D32220 llvm-svn: 301386
2017-04-26 15:08:44 +08:00
; SSE41-NEXT: movq %rcx, %xmm0
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE41-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; SSE41-NEXT: retq
;
; AVX-LABEL: test_div7_2i64:
[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
; AVX: # %bb.0:
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; AVX-NEXT: vpextrq $1, %xmm0, %rcx
; AVX-NEXT: movabsq $2635249153387078803, %rsi # imm = 0x2492492492492493
; AVX-NEXT: movq %rcx, %rax
; AVX-NEXT: mulq %rsi
; AVX-NEXT: subq %rdx, %rcx
; AVX-NEXT: shrq %rcx
; AVX-NEXT: addq %rdx, %rcx
; AVX-NEXT: shrq $2, %rcx
; AVX-NEXT: vmovq %rcx, %xmm1
; AVX-NEXT: vmovq %xmm0, %rcx
; AVX-NEXT: movq %rcx, %rax
; AVX-NEXT: mulq %rsi
; AVX-NEXT: subq %rdx, %rcx
; AVX-NEXT: shrq %rcx
; AVX-NEXT: addq %rdx, %rcx
; AVX-NEXT: shrq $2, %rcx
; AVX-NEXT: vmovq %rcx, %xmm0
; AVX-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; AVX-NEXT: retq
%res = udiv <2 x i64> %a, <i64 7, i64 7>
ret <2 x i64> %res
}
define <4 x i32> @test_div7_4i32(<4 x i32> %a) nounwind {
; SSE2-LABEL: test_div7_4i32:
[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
; SSE2: # %bb.0:
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [613566757,613566757,613566757,613566757]
; SSE2-NEXT: movdqa %xmm0, %xmm2
; SSE2-NEXT: pmuludq %xmm1, %xmm2
; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
; SSE2-NEXT: pmuludq %xmm1, %xmm3
; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm3[1,3,2,3]
; SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE2-NEXT: psubd %xmm2, %xmm0
; SSE2-NEXT: psrld $1, %xmm0
; SSE2-NEXT: paddd %xmm2, %xmm0
; SSE2-NEXT: psrld $2, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: test_div7_4i32:
[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
; SSE41: # %bb.0:
[X86] Change X86::PMULDQ/PMULUDQ opcodes to take vXi64 type as input instead of vXi32. This instruction can be thought of as reading either the even elements of a vXi32 input or the lower half of each element of a vXi64 input. We currently use the vXi32 interpretation, but vXi64 matches better with its broadcast behavior in EVEX. I'm looking at moving MULDQ/MULUDQ creation to a DAG combine so we can do it when AVX512DQ is enabled without having to go through Custom lowering. But in some of the test cases we failed to use a broadcast load due to the size difference. This should help with that. I'm also wondering if we can model these instructions in native IR and remove the intrinsics and I think using a vXi64 type will work better with that. llvm-svn: 326991
2018-03-08 16:02:52 +08:00
; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
; SSE41-NEXT: movdqa {{.*#+}} xmm2 = [613566757,613566757,613566757,613566757]
; SSE41-NEXT: pmuludq %xmm2, %xmm1
; SSE41-NEXT: pmuludq %xmm0, %xmm2
; SSE41-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; SSE41-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
; SSE41-NEXT: psubd %xmm2, %xmm0
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE41-NEXT: psrld $1, %xmm0
[X86] Change X86::PMULDQ/PMULUDQ opcodes to take vXi64 type as input instead of vXi32. This instruction can be thought of as reading either the even elements of a vXi32 input or the lower half of each element of a vXi64 input. We currently use the vXi32 interpretation, but vXi64 matches better with its broadcast behavior in EVEX. I'm looking at moving MULDQ/MULUDQ creation to a DAG combine so we can do it when AVX512DQ is enabled without having to go through Custom lowering. But in some of the test cases we failed to use a broadcast load due to the size difference. This should help with that. I'm also wondering if we can model these instructions in native IR and remove the intrinsics and I think using a vXi64 type will work better with that. llvm-svn: 326991
2018-03-08 16:02:52 +08:00
; SSE41-NEXT: paddd %xmm2, %xmm0
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE41-NEXT: psrld $2, %xmm0
; SSE41-NEXT: retq
;
; AVX1-LABEL: test_div7_4i32:
[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
; AVX1: # %bb.0:
[X86] Change X86::PMULDQ/PMULUDQ opcodes to take vXi64 type as input instead of vXi32. This instruction can be thought of as reading either the even elements of a vXi32 input or the lower half of each element of a vXi64 input. We currently use the vXi32 interpretation, but vXi64 matches better with its broadcast behavior in EVEX. I'm looking at moving MULDQ/MULUDQ creation to a DAG combine so we can do it when AVX512DQ is enabled without having to go through Custom lowering. But in some of the test cases we failed to use a broadcast load due to the size difference. This should help with that. I'm also wondering if we can model these instructions in native IR and remove the intrinsics and I think using a vXi64 type will work better with that. llvm-svn: 326991
2018-03-08 16:02:52 +08:00
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX1-NEXT: vmovdqa {{.*#+}} xmm2 = [613566757,613566757,613566757,613566757]
; AVX1-NEXT: vpmuludq %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpmuludq %xmm2, %xmm0, %xmm2
; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; AVX1-NEXT: vpsubd %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpsrld $1, %xmm0, %xmm0
; AVX1-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpsrld $2, %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: test_div7_4i32:
[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
; AVX2: # %bb.0:
[X86][AVX] Regenerate vector idiv tests with constant broadcast comments llvm-svn: 308129
2017-07-16 19:38:14 +08:00
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [613566757,613566757,613566757,613566757]
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
; AVX2-NEXT: vpmuludq %xmm2, %xmm3, %xmm2
; AVX2-NEXT: vpmuludq %xmm1, %xmm0, %xmm1
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
; AVX2-NEXT: vpsubd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrld $1, %xmm0, %xmm0
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrld $2, %xmm0, %xmm0
; AVX2-NEXT: retq
%res = udiv <4 x i32> %a, <i32 7, i32 7, i32 7, i32 7>
ret <4 x i32> %res
}
define <8 x i16> @test_div7_8i16(<8 x i16> %a) nounwind {
; SSE-LABEL: test_div7_8i16:
[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
; SSE: # %bb.0:
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE-NEXT: movdqa {{.*#+}} xmm1 = [9363,9363,9363,9363,9363,9363,9363,9363]
; SSE-NEXT: pmulhuw %xmm0, %xmm1
; SSE-NEXT: psubw %xmm1, %xmm0
; SSE-NEXT: psrlw $1, %xmm0
; SSE-NEXT: paddw %xmm1, %xmm0
; SSE-NEXT: psrlw $2, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: test_div7_8i16:
[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
; AVX: # %bb.0:
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; AVX-NEXT: vpmulhuw {{.*}}(%rip), %xmm0, %xmm1
; AVX-NEXT: vpsubw %xmm1, %xmm0, %xmm0
; AVX-NEXT: vpsrlw $1, %xmm0, %xmm0
; AVX-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX-NEXT: vpsrlw $2, %xmm0, %xmm0
; AVX-NEXT: retq
%res = udiv <8 x i16> %a, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
ret <8 x i16> %res
}
define <16 x i8> @test_div7_16i8(<16 x i8> %a) nounwind {
; SSE2-LABEL: test_div7_16i8:
[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
; SSE2: # %bb.0:
[X86][SSE] Allow matchVectorShuffleWithUNPCK to recognise ZERO inputs Add support for specifying an UNPCK input as ZERO, particularly improves ZEXT cases with non-zero offsets llvm-svn: 295169
2017-02-15 19:46:15 +08:00
; SSE2-NEXT: pxor %xmm1, %xmm1
; SSE2-NEXT: movdqa %xmm0, %xmm2
; SSE2-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [37,0,37,0,37,0,37,0,37,0,37,0,37,0,37,0]
; SSE2-NEXT: pmullw %xmm3, %xmm2
; SSE2-NEXT: psrlw $8, %xmm2
; SSE2-NEXT: movdqa %xmm0, %xmm4
; SSE2-NEXT: punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3],xmm4[4],xmm1[4],xmm4[5],xmm1[5],xmm4[6],xmm1[6],xmm4[7],xmm1[7]
; SSE2-NEXT: pmullw %xmm3, %xmm4
; SSE2-NEXT: psrlw $8, %xmm4
; SSE2-NEXT: packuswb %xmm2, %xmm4
; SSE2-NEXT: psubb %xmm4, %xmm0
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; SSE2-NEXT: psrlw $1, %xmm0
; SSE2-NEXT: pand {{.*}}(%rip), %xmm0
[X86][SSE] Allow matchVectorShuffleWithUNPCK to recognise ZERO inputs Add support for specifying an UNPCK input as ZERO, particularly improves ZEXT cases with non-zero offsets llvm-svn: 295169
2017-02-15 19:46:15 +08:00
; SSE2-NEXT: paddb %xmm4, %xmm0
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; SSE2-NEXT: psrlw $2, %xmm0
; SSE2-NEXT: pand {{.*}}(%rip), %xmm0
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE2-NEXT: retq
;
; SSE41-LABEL: test_div7_16i8:
[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
; SSE41: # %bb.0:
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; SSE41-NEXT: pmovzxbw {{.*#+}} xmm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
[X86][SSE] Improve VSEXT/VZEXT constant folding. Generalize VSEXT/VZEXT constant folding to work with any target constant bits source not just BUILD_VECTOR . llvm-svn: 294873
2017-02-12 05:55:24 +08:00
; SSE41-NEXT: movdqa {{.*#+}} xmm2 = [37,37,37,37,37,37,37,37]
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; SSE41-NEXT: pmullw %xmm2, %xmm1
; SSE41-NEXT: psrlw $8, %xmm1
; SSE41-NEXT: pshufd {{.*#+}} xmm3 = xmm0[2,3,0,1]
; SSE41-NEXT: pmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
; SSE41-NEXT: pmullw %xmm2, %xmm3
; SSE41-NEXT: psrlw $8, %xmm3
; SSE41-NEXT: packuswb %xmm3, %xmm1
; SSE41-NEXT: psubb %xmm1, %xmm0
; SSE41-NEXT: psrlw $1, %xmm0
; SSE41-NEXT: pand {{.*}}(%rip), %xmm0
; SSE41-NEXT: paddb %xmm1, %xmm0
; SSE41-NEXT: psrlw $2, %xmm0
; SSE41-NEXT: pand {{.*}}(%rip), %xmm0
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE41-NEXT: retq
;
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; AVX1-LABEL: test_div7_16i8:
[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
; AVX1: # %bb.0:
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
[X86][SSE] Improve VSEXT/VZEXT constant folding. Generalize VSEXT/VZEXT constant folding to work with any target constant bits source not just BUILD_VECTOR . llvm-svn: 294873
2017-02-12 05:55:24 +08:00
; AVX1-NEXT: vmovdqa {{.*#+}} xmm2 = [37,37,37,37,37,37,37,37]
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; AVX1-NEXT: vpmullw %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpsrlw $8, %xmm1, %xmm1
; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[2,3,0,1]
; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
; AVX1-NEXT: vpmullw %xmm2, %xmm3, %xmm2
; AVX1-NEXT: vpsrlw $8, %xmm2, %xmm2
; AVX1-NEXT: vpackuswb %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpsubb %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpsrlw $1, %xmm0, %xmm0
; AVX1-NEXT: vpand {{.*}}(%rip), %xmm0, %xmm0
; AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpsrlw $2, %xmm0, %xmm0
; AVX1-NEXT: vpand {{.*}}(%rip), %xmm0, %xmm0
; AVX1-NEXT: retq
;
[X86] Make use of vpmovwb when possible in LowerMULH If we have BWI, we can truncate in a much simpler way by using vpmovwb. This even works without VLX by using the wider zmm->ymm truncate with a subvector extract. Differential Revision: https://reviews.llvm.org/D38375 llvm-svn: 314457
2017-09-29 04:10:34 +08:00
; AVX2NOBW-LABEL: test_div7_16i8:
[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
; AVX2NOBW: # %bb.0:
[X86] Make use of vpmovwb when possible in LowerMULH If we have BWI, we can truncate in a much simpler way by using vpmovwb. This even works without VLX by using the wider zmm->ymm truncate with a subvector extract. Differential Revision: https://reviews.llvm.org/D38375 llvm-svn: 314457
2017-09-29 04:10:34 +08:00
; AVX2NOBW-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
; AVX2NOBW-NEXT: vpmullw {{.*}}(%rip), %ymm1, %ymm1
; AVX2NOBW-NEXT: vpsrlw $8, %ymm1, %ymm1
; AVX2NOBW-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2NOBW-NEXT: vpackuswb %xmm2, %xmm1, %xmm1
; AVX2NOBW-NEXT: vpsubb %xmm1, %xmm0, %xmm0
; AVX2NOBW-NEXT: vpsrlw $1, %xmm0, %xmm0
; AVX2NOBW-NEXT: vpand {{.*}}(%rip), %xmm0, %xmm0
; AVX2NOBW-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX2NOBW-NEXT: vpsrlw $2, %xmm0, %xmm0
; AVX2NOBW-NEXT: vpand {{.*}}(%rip), %xmm0, %xmm0
; AVX2NOBW-NEXT: vzeroupper
; AVX2NOBW-NEXT: retq
;
; AVX512BW-LABEL: test_div7_16i8:
[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
; AVX512BW: # %bb.0:
[X86] Make use of vpmovwb when possible in LowerMULH If we have BWI, we can truncate in a much simpler way by using vpmovwb. This even works without VLX by using the wider zmm->ymm truncate with a subvector extract. Differential Revision: https://reviews.llvm.org/D38375 llvm-svn: 314457
2017-09-29 04:10:34 +08:00
; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
; AVX512BW-NEXT: vpmullw {{.*}}(%rip), %ymm1, %ymm1
; AVX512BW-NEXT: vpsrlw $8, %ymm1, %ymm1
; AVX512BW-NEXT: vpmovwb %zmm1, %ymm1
; AVX512BW-NEXT: vpsubb %xmm1, %xmm0, %xmm0
; AVX512BW-NEXT: vpsrlw $1, %xmm0, %xmm0
; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm0, %xmm0
; AVX512BW-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX512BW-NEXT: vpsrlw $2, %xmm0, %xmm0
; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm0, %xmm0
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
%res = udiv <16 x i8> %a, <i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7>
ret <16 x i8> %res
}
;
; urem by 7
;
define <2 x i64> @test_rem7_2i64(<2 x i64> %a) nounwind {
; SSE2-LABEL: test_rem7_2i64:
[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
; SSE2: # %bb.0:
[X86][SSE2] Fix asm string for movq (Move Quadword) instruction. Replace "mov{d|q}" with "movq". Differential Revision: https://reviews.llvm.org/D32220 llvm-svn: 301386
2017-04-26 15:08:44 +08:00
; SSE2-NEXT: movq %xmm0, %rcx
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE2-NEXT: movabsq $2635249153387078803, %rsi # imm = 0x2492492492492493
; SSE2-NEXT: movq %rcx, %rax
; SSE2-NEXT: mulq %rsi
; SSE2-NEXT: movq %rcx, %rax
; SSE2-NEXT: subq %rdx, %rax
; SSE2-NEXT: shrq %rax
; SSE2-NEXT: addq %rdx, %rax
; SSE2-NEXT: shrq $2, %rax
; SSE2-NEXT: leaq (,%rax,8), %rdx
; SSE2-NEXT: subq %rax, %rdx
; SSE2-NEXT: subq %rdx, %rcx
[X86][SSE2] Fix asm string for movq (Move Quadword) instruction. Replace "mov{d|q}" with "movq". Differential Revision: https://reviews.llvm.org/D32220 llvm-svn: 301386
2017-04-26 15:08:44 +08:00
; SSE2-NEXT: movq %rcx, %xmm1
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
[X86][SSE2] Fix asm string for movq (Move Quadword) instruction. Replace "mov{d|q}" with "movq". Differential Revision: https://reviews.llvm.org/D32220 llvm-svn: 301386
2017-04-26 15:08:44 +08:00
; SSE2-NEXT: movq %xmm0, %rcx
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE2-NEXT: movq %rcx, %rax
; SSE2-NEXT: mulq %rsi
; SSE2-NEXT: movq %rcx, %rax
; SSE2-NEXT: subq %rdx, %rax
; SSE2-NEXT: shrq %rax
; SSE2-NEXT: addq %rdx, %rax
; SSE2-NEXT: shrq $2, %rax
; SSE2-NEXT: leaq (,%rax,8), %rdx
; SSE2-NEXT: subq %rax, %rdx
; SSE2-NEXT: subq %rdx, %rcx
[X86][SSE2] Fix asm string for movq (Move Quadword) instruction. Replace "mov{d|q}" with "movq". Differential Revision: https://reviews.llvm.org/D32220 llvm-svn: 301386
2017-04-26 15:08:44 +08:00
; SSE2-NEXT: movq %rcx, %xmm0
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE2-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE2-NEXT: movdqa %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: test_rem7_2i64:
[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
; SSE41: # %bb.0:
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE41-NEXT: pextrq $1, %xmm0, %rcx
; SSE41-NEXT: movabsq $2635249153387078803, %rsi # imm = 0x2492492492492493
; SSE41-NEXT: movq %rcx, %rax
; SSE41-NEXT: mulq %rsi
; SSE41-NEXT: movq %rcx, %rax
; SSE41-NEXT: subq %rdx, %rax
; SSE41-NEXT: shrq %rax
; SSE41-NEXT: addq %rdx, %rax
; SSE41-NEXT: shrq $2, %rax
; SSE41-NEXT: leaq (,%rax,8), %rdx
; SSE41-NEXT: subq %rax, %rdx
; SSE41-NEXT: subq %rdx, %rcx
[X86][SSE2] Fix asm string for movq (Move Quadword) instruction. Replace "mov{d|q}" with "movq". Differential Revision: https://reviews.llvm.org/D32220 llvm-svn: 301386
2017-04-26 15:08:44 +08:00
; SSE41-NEXT: movq %rcx, %xmm1
; SSE41-NEXT: movq %xmm0, %rcx
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE41-NEXT: movq %rcx, %rax
; SSE41-NEXT: mulq %rsi
; SSE41-NEXT: movq %rcx, %rax
; SSE41-NEXT: subq %rdx, %rax
; SSE41-NEXT: shrq %rax
; SSE41-NEXT: addq %rdx, %rax
; SSE41-NEXT: shrq $2, %rax
; SSE41-NEXT: leaq (,%rax,8), %rdx
; SSE41-NEXT: subq %rax, %rdx
; SSE41-NEXT: subq %rdx, %rcx
[X86][SSE2] Fix asm string for movq (Move Quadword) instruction. Replace "mov{d|q}" with "movq". Differential Revision: https://reviews.llvm.org/D32220 llvm-svn: 301386
2017-04-26 15:08:44 +08:00
; SSE41-NEXT: movq %rcx, %xmm0
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE41-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; SSE41-NEXT: retq
;
; AVX-LABEL: test_rem7_2i64:
[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
; AVX: # %bb.0:
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; AVX-NEXT: vpextrq $1, %xmm0, %rcx
; AVX-NEXT: movabsq $2635249153387078803, %rsi # imm = 0x2492492492492493
; AVX-NEXT: movq %rcx, %rax
; AVX-NEXT: mulq %rsi
; AVX-NEXT: movq %rcx, %rax
; AVX-NEXT: subq %rdx, %rax
; AVX-NEXT: shrq %rax
; AVX-NEXT: addq %rdx, %rax
; AVX-NEXT: shrq $2, %rax
; AVX-NEXT: leaq (,%rax,8), %rdx
; AVX-NEXT: subq %rax, %rdx
; AVX-NEXT: subq %rdx, %rcx
; AVX-NEXT: vmovq %rcx, %xmm1
; AVX-NEXT: vmovq %xmm0, %rcx
; AVX-NEXT: movq %rcx, %rax
; AVX-NEXT: mulq %rsi
; AVX-NEXT: movq %rcx, %rax
; AVX-NEXT: subq %rdx, %rax
; AVX-NEXT: shrq %rax
; AVX-NEXT: addq %rdx, %rax
; AVX-NEXT: shrq $2, %rax
; AVX-NEXT: leaq (,%rax,8), %rdx
; AVX-NEXT: subq %rax, %rdx
; AVX-NEXT: subq %rdx, %rcx
; AVX-NEXT: vmovq %rcx, %xmm0
; AVX-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; AVX-NEXT: retq
%res = urem <2 x i64> %a, <i64 7, i64 7>
ret <2 x i64> %res
}
define <4 x i32> @test_rem7_4i32(<4 x i32> %a) nounwind {
; SSE2-LABEL: test_rem7_4i32:
[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
; SSE2: # %bb.0:
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [613566757,613566757,613566757,613566757]
; SSE2-NEXT: movdqa %xmm0, %xmm2
; SSE2-NEXT: pmuludq %xmm1, %xmm2
; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
; SSE2-NEXT: pmuludq %xmm1, %xmm3
; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm3[1,3,2,3]
; SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE2-NEXT: movdqa %xmm0, %xmm1
; SSE2-NEXT: psubd %xmm2, %xmm1
; SSE2-NEXT: psrld $1, %xmm1
; SSE2-NEXT: paddd %xmm2, %xmm1
; SSE2-NEXT: psrld $2, %xmm1
; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [7,7,7,7]
; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
; SSE2-NEXT: pmuludq %xmm2, %xmm1
; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; SSE2-NEXT: pmuludq %xmm2, %xmm3
; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; SSE2-NEXT: psubd %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: test_rem7_4i32:
[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
; SSE41: # %bb.0:
[X86] Change X86::PMULDQ/PMULUDQ opcodes to take vXi64 type as input instead of vXi32. This instruction can be thought of as reading either the even elements of a vXi32 input or the lower half of each element of a vXi64 input. We currently use the vXi32 interpretation, but vXi64 matches better with its broadcast behavior in EVEX. I'm looking at moving MULDQ/MULUDQ creation to a DAG combine so we can do it when AVX512DQ is enabled without having to go through Custom lowering. But in some of the test cases we failed to use a broadcast load due to the size difference. This should help with that. I'm also wondering if we can model these instructions in native IR and remove the intrinsics and I think using a vXi64 type will work better with that. llvm-svn: 326991
2018-03-08 16:02:52 +08:00
; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
; SSE41-NEXT: movdqa {{.*#+}} xmm2 = [613566757,613566757,613566757,613566757]
; SSE41-NEXT: pmuludq %xmm2, %xmm1
; SSE41-NEXT: pmuludq %xmm0, %xmm2
; SSE41-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; SSE41-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
; SSE41-NEXT: movdqa %xmm0, %xmm1
; SSE41-NEXT: psubd %xmm2, %xmm1
; SSE41-NEXT: psrld $1, %xmm1
; SSE41-NEXT: paddd %xmm2, %xmm1
; SSE41-NEXT: psrld $2, %xmm1
; SSE41-NEXT: pmulld {{.*}}(%rip), %xmm1
; SSE41-NEXT: psubd %xmm1, %xmm0
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE41-NEXT: retq
;
; AVX1-LABEL: test_rem7_4i32:
[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
; AVX1: # %bb.0:
[X86] Change X86::PMULDQ/PMULUDQ opcodes to take vXi64 type as input instead of vXi32. This instruction can be thought of as reading either the even elements of a vXi32 input or the lower half of each element of a vXi64 input. We currently use the vXi32 interpretation, but vXi64 matches better with its broadcast behavior in EVEX. I'm looking at moving MULDQ/MULUDQ creation to a DAG combine so we can do it when AVX512DQ is enabled without having to go through Custom lowering. But in some of the test cases we failed to use a broadcast load due to the size difference. This should help with that. I'm also wondering if we can model these instructions in native IR and remove the intrinsics and I think using a vXi64 type will work better with that. llvm-svn: 326991
2018-03-08 16:02:52 +08:00
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX1-NEXT: vmovdqa {{.*#+}} xmm2 = [613566757,613566757,613566757,613566757]
; AVX1-NEXT: vpmuludq %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpmuludq %xmm2, %xmm0, %xmm2
; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; AVX1-NEXT: vpsubd %xmm1, %xmm0, %xmm2
; AVX1-NEXT: vpsrld $1, %xmm2, %xmm2
; AVX1-NEXT: vpaddd %xmm1, %xmm2, %xmm1
; AVX1-NEXT: vpsrld $2, %xmm1, %xmm1
; AVX1-NEXT: vpmulld {{.*}}(%rip), %xmm1, %xmm1
; AVX1-NEXT: vpsubd %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: test_rem7_4i32:
[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
; AVX2: # %bb.0:
[X86][AVX] Regenerate vector idiv tests with constant broadcast comments llvm-svn: 308129
2017-07-16 19:38:14 +08:00
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [613566757,613566757,613566757,613566757]
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
; AVX2-NEXT: vpmuludq %xmm2, %xmm3, %xmm2
; AVX2-NEXT: vpmuludq %xmm1, %xmm0, %xmm1
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
; AVX2-NEXT: vpsubd %xmm1, %xmm0, %xmm2
; AVX2-NEXT: vpsrld $1, %xmm2, %xmm2
; AVX2-NEXT: vpaddd %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vpsrld $2, %xmm1, %xmm1
[X86][AVX] Regenerate vector idiv tests with constant broadcast comments llvm-svn: 308129
2017-07-16 19:38:14 +08:00
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm2 = [7,7,7,7]
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; AVX2-NEXT: vpmulld %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpsubd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
%res = urem <4 x i32> %a, <i32 7, i32 7, i32 7, i32 7>
ret <4 x i32> %res
}
define <8 x i16> @test_rem7_8i16(<8 x i16> %a) nounwind {
; SSE-LABEL: test_rem7_8i16:
[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
; SSE: # %bb.0:
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE-NEXT: movdqa {{.*#+}} xmm1 = [9363,9363,9363,9363,9363,9363,9363,9363]
; SSE-NEXT: pmulhuw %xmm0, %xmm1
; SSE-NEXT: movdqa %xmm0, %xmm2
; SSE-NEXT: psubw %xmm1, %xmm2
; SSE-NEXT: psrlw $1, %xmm2
; SSE-NEXT: paddw %xmm1, %xmm2
; SSE-NEXT: psrlw $2, %xmm2
; SSE-NEXT: pmullw {{.*}}(%rip), %xmm2
; SSE-NEXT: psubw %xmm2, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: test_rem7_8i16:
[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
; AVX: # %bb.0:
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; AVX-NEXT: vpmulhuw {{.*}}(%rip), %xmm0, %xmm1
; AVX-NEXT: vpsubw %xmm1, %xmm0, %xmm2
; AVX-NEXT: vpsrlw $1, %xmm2, %xmm2
; AVX-NEXT: vpaddw %xmm1, %xmm2, %xmm1
; AVX-NEXT: vpsrlw $2, %xmm1, %xmm1
; AVX-NEXT: vpmullw {{.*}}(%rip), %xmm1, %xmm1
; AVX-NEXT: vpsubw %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
%res = urem <8 x i16> %a, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
ret <8 x i16> %res
}
define <16 x i8> @test_rem7_16i8(<16 x i8> %a) nounwind {
; SSE2-LABEL: test_rem7_16i8:
[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
; SSE2: # %bb.0:
[X86][SSE] Allow matchVectorShuffleWithUNPCK to recognise ZERO inputs Add support for specifying an UNPCK input as ZERO, particularly improves ZEXT cases with non-zero offsets llvm-svn: 295169
2017-02-15 19:46:15 +08:00
; SSE2-NEXT: pxor %xmm1, %xmm1
; SSE2-NEXT: movdqa %xmm0, %xmm2
; SSE2-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [37,0,37,0,37,0,37,0,37,0,37,0,37,0,37,0]
; SSE2-NEXT: pmullw %xmm3, %xmm2
; SSE2-NEXT: psrlw $8, %xmm2
; SSE2-NEXT: movdqa %xmm0, %xmm4
; SSE2-NEXT: punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3],xmm4[4],xmm1[4],xmm4[5],xmm1[5],xmm4[6],xmm1[6],xmm4[7],xmm1[7]
; SSE2-NEXT: pmullw %xmm3, %xmm4
; SSE2-NEXT: psrlw $8, %xmm4
; SSE2-NEXT: packuswb %xmm2, %xmm4
[X86][SSE] Add support for combining SSE VSHLI/VSRLI uniform constant shifts. Fixes some missed constant folding opportunities and allows us to combine shuffles that end with a logical bit shift. llvm-svn: 289429
2016-12-12 21:33:58 +08:00
; SSE2-NEXT: movdqa %xmm0, %xmm1
[X86][SSE] Allow matchVectorShuffleWithUNPCK to recognise ZERO inputs Add support for specifying an UNPCK input as ZERO, particularly improves ZEXT cases with non-zero offsets llvm-svn: 295169
2017-02-15 19:46:15 +08:00
; SSE2-NEXT: psubb %xmm4, %xmm1
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; SSE2-NEXT: psrlw $1, %xmm1
; SSE2-NEXT: pand {{.*}}(%rip), %xmm1
[X86][SSE] Allow matchVectorShuffleWithUNPCK to recognise ZERO inputs Add support for specifying an UNPCK input as ZERO, particularly improves ZEXT cases with non-zero offsets llvm-svn: 295169
2017-02-15 19:46:15 +08:00
; SSE2-NEXT: paddb %xmm4, %xmm1
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; SSE2-NEXT: psrlw $2, %xmm1
; SSE2-NEXT: pand {{.*}}(%rip), %xmm1
; SSE2-NEXT: movdqa %xmm1, %xmm2
[X86] Remove SRAs from v16i8 multiply lowering on sse2 targets Previously we unpacked the even bytes of each input into the high byte of 16-bit elements then did an v8i16 arithmetic shift right by 8 bits to fill the upper bits of each word with sign bits. Then we did the v8i16 multiply and then masked to zero the upper 8-bits of each result. The similar was done for all the odd bytes. The results are then packed together with packuswb Since we are masking each multiply result element to 8-bits, and those 8-bits are determined only by the lower 8-bits of each of the inputs, we don't need to fill the upper bits with sign bits. So we can just unpack into the low byte of each element and treat the upper bits as garbage. This is what gcc also does. Differential Revision: https://reviews.llvm.org/D44267 llvm-svn: 327093
2018-03-09 09:22:31 +08:00
; SSE2-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm0[8],xmm2[9],xmm0[9],xmm2[10],xmm0[10],xmm2[11],xmm0[11],xmm2[12],xmm0[12],xmm2[13],xmm0[13],xmm2[14],xmm0[14],xmm2[15],xmm0[15]
; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; SSE2-NEXT: pmullw %xmm3, %xmm2
; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
; SSE2-NEXT: pand %xmm4, %xmm2
[X86] Remove SRAs from v16i8 multiply lowering on sse2 targets Previously we unpacked the even bytes of each input into the high byte of 16-bit elements then did an v8i16 arithmetic shift right by 8 bits to fill the upper bits of each word with sign bits. Then we did the v8i16 multiply and then masked to zero the upper 8-bits of each result. The similar was done for all the odd bytes. The results are then packed together with packuswb Since we are masking each multiply result element to 8-bits, and those 8-bits are determined only by the lower 8-bits of each of the inputs, we don't need to fill the upper bits with sign bits. So we can just unpack into the low byte of each element and treat the upper bits as garbage. This is what gcc also does. Differential Revision: https://reviews.llvm.org/D44267 llvm-svn: 327093
2018-03-09 09:22:31 +08:00
; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; SSE2-NEXT: pmullw %xmm3, %xmm1
; SSE2-NEXT: pand %xmm4, %xmm1
; SSE2-NEXT: packuswb %xmm2, %xmm1
; SSE2-NEXT: psubb %xmm1, %xmm0
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE2-NEXT: retq
;
; SSE41-LABEL: test_rem7_16i8:
[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
; SSE41: # %bb.0:
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; SSE41-NEXT: pmovzxbw {{.*#+}} xmm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
[X86][SSE] Improve VSEXT/VZEXT constant folding. Generalize VSEXT/VZEXT constant folding to work with any target constant bits source not just BUILD_VECTOR . llvm-svn: 294873
2017-02-12 05:55:24 +08:00
; SSE41-NEXT: movdqa {{.*#+}} xmm2 = [37,37,37,37,37,37,37,37]
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; SSE41-NEXT: pmullw %xmm2, %xmm1
; SSE41-NEXT: psrlw $8, %xmm1
; SSE41-NEXT: pshufd {{.*#+}} xmm3 = xmm0[2,3,0,1]
; SSE41-NEXT: pmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
; SSE41-NEXT: pmullw %xmm2, %xmm3
; SSE41-NEXT: psrlw $8, %xmm3
; SSE41-NEXT: packuswb %xmm3, %xmm1
; SSE41-NEXT: movdqa %xmm0, %xmm2
; SSE41-NEXT: psubb %xmm1, %xmm2
; SSE41-NEXT: psrlw $1, %xmm2
; SSE41-NEXT: pand {{.*}}(%rip), %xmm2
; SSE41-NEXT: paddb %xmm1, %xmm2
; SSE41-NEXT: psrlw $2, %xmm2
; SSE41-NEXT: pand {{.*}}(%rip), %xmm2
[X86] Remove sse41 specific code from lowering v16i8 multiply With the SRAs removed from the SSE2 code in D44267, then there doesn't appear to be any advantage to the sse41 code. The punpcklbw instruction and pmovsx seem to have the same latency and throughput on most CPUs. And the SSE41 code requires moving the upper 64-bits into the lower 64-bit before the sign extend can be done. The unpckhbw in sse2 code can do better than that. llvm-svn: 327869
2018-03-20 01:31:41 +08:00
; SSE41-NEXT: pmovzxbw {{.*#+}} xmm1 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
; SSE41-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
; SSE41-NEXT: movdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; SSE41-NEXT: pmullw %xmm3, %xmm2
[X86] Remove sse41 specific code from lowering v16i8 multiply With the SRAs removed from the SSE2 code in D44267, then there doesn't appear to be any advantage to the sse41 code. The punpcklbw instruction and pmovsx seem to have the same latency and throughput on most CPUs. And the SSE41 code requires moving the upper 64-bits into the lower 64-bit before the sign extend can be done. The unpckhbw in sse2 code can do better than that. llvm-svn: 327869
2018-03-20 01:31:41 +08:00
; SSE41-NEXT: movdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; SSE41-NEXT: pand %xmm4, %xmm2
[X86] Remove sse41 specific code from lowering v16i8 multiply With the SRAs removed from the SSE2 code in D44267, then there doesn't appear to be any advantage to the sse41 code. The punpcklbw instruction and pmovsx seem to have the same latency and throughput on most CPUs. And the SSE41 code requires moving the upper 64-bits into the lower 64-bit before the sign extend can be done. The unpckhbw in sse2 code can do better than that. llvm-svn: 327869
2018-03-20 01:31:41 +08:00
; SSE41-NEXT: pmullw %xmm3, %xmm1
; SSE41-NEXT: pand %xmm4, %xmm1
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; SSE41-NEXT: packuswb %xmm2, %xmm1
; SSE41-NEXT: psubb %xmm1, %xmm0
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
; SSE41-NEXT: retq
;
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; AVX1-LABEL: test_rem7_16i8:
[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
; AVX1: # %bb.0:
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
[X86][SSE] Improve VSEXT/VZEXT constant folding. Generalize VSEXT/VZEXT constant folding to work with any target constant bits source not just BUILD_VECTOR . llvm-svn: 294873
2017-02-12 05:55:24 +08:00
; AVX1-NEXT: vmovdqa {{.*#+}} xmm2 = [37,37,37,37,37,37,37,37]
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; AVX1-NEXT: vpmullw %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpsrlw $8, %xmm1, %xmm1
; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[2,3,0,1]
; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
; AVX1-NEXT: vpmullw %xmm2, %xmm3, %xmm2
; AVX1-NEXT: vpsrlw $8, %xmm2, %xmm2
; AVX1-NEXT: vpackuswb %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpsubb %xmm1, %xmm0, %xmm2
; AVX1-NEXT: vpsrlw $1, %xmm2, %xmm2
; AVX1-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
; AVX1-NEXT: vpaddb %xmm1, %xmm2, %xmm1
; AVX1-NEXT: vpsrlw $2, %xmm1, %xmm1
; AVX1-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
[X86] Remove sse41 specific code from lowering v16i8 multiply With the SRAs removed from the SSE2 code in D44267, then there doesn't appear to be any advantage to the sse41 code. The punpcklbw instruction and pmovsx seem to have the same latency and throughput on most CPUs. And the SSE41 code requires moving the upper 64-bits into the lower 64-bit before the sign extend can be done. The unpckhbw in sse2 code can do better than that. llvm-svn: 327869
2018-03-20 01:31:41 +08:00
; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm2 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; AVX1-NEXT: vpmullw %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
; AVX1-NEXT: vpand %xmm4, %xmm2, %xmm2
[X86] Remove sse41 specific code from lowering v16i8 multiply With the SRAs removed from the SSE2 code in D44267, then there doesn't appear to be any advantage to the sse41 code. The punpcklbw instruction and pmovsx seem to have the same latency and throughput on most CPUs. And the SSE41 code requires moving the upper 64-bits into the lower 64-bit before the sign extend can be done. The unpckhbw in sse2 code can do better than that. llvm-svn: 327869
2018-03-20 01:31:41 +08:00
; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; AVX1-NEXT: vpmullw %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vpand %xmm4, %xmm1, %xmm1
[X86] Remove sse41 specific code from lowering v16i8 multiply With the SRAs removed from the SSE2 code in D44267, then there doesn't appear to be any advantage to the sse41 code. The punpcklbw instruction and pmovsx seem to have the same latency and throughput on most CPUs. And the SSE41 code requires moving the upper 64-bits into the lower 64-bit before the sign extend can be done. The unpckhbw in sse2 code can do better than that. llvm-svn: 327869
2018-03-20 01:31:41 +08:00
; AVX1-NEXT: vpackuswb %xmm2, %xmm1, %xmm1
[X86][SSE] Add MULHS/MULHU custom lowering for i8 vectors Currently this is to mainly to prevent scalarization of integer division by constants. Differential Revision: http://reviews.llvm.org/D18307 llvm-svn: 264511
2016-03-26 23:27:20 +08:00
; AVX1-NEXT: vpsubb %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
;
[AVX512] Add avx512bw command lines to 128-bit idiv tests. The multiply lowering on some of the tests can take advantage of the vpmovwb to simplify the truncate. llvm-svn: 314448
2017-09-29 02:45:29 +08:00
; AVX2NOBW-LABEL: test_rem7_16i8:
[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
; AVX2NOBW: # %bb.0:
[AVX512] Add avx512bw command lines to 128-bit idiv tests. The multiply lowering on some of the tests can take advantage of the vpmovwb to simplify the truncate. llvm-svn: 314448
2017-09-29 02:45:29 +08:00
; AVX2NOBW-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
; AVX2NOBW-NEXT: vpmullw {{.*}}(%rip), %ymm1, %ymm1
; AVX2NOBW-NEXT: vpsrlw $8, %ymm1, %ymm1
; AVX2NOBW-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2NOBW-NEXT: vpackuswb %xmm2, %xmm1, %xmm1
; AVX2NOBW-NEXT: vpsubb %xmm1, %xmm0, %xmm2
; AVX2NOBW-NEXT: vpsrlw $1, %xmm2, %xmm2
; AVX2NOBW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
; AVX2NOBW-NEXT: vpaddb %xmm1, %xmm2, %xmm1
; AVX2NOBW-NEXT: vpsrlw $2, %xmm1, %xmm1
; AVX2NOBW-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
; AVX2NOBW-NEXT: vpmovsxbw %xmm1, %ymm1
; AVX2NOBW-NEXT: vpmullw {{.*}}(%rip), %ymm1, %ymm1
; AVX2NOBW-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2NOBW-NEXT: vmovdqa {{.*#+}} xmm3 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
; AVX2NOBW-NEXT: vpshufb %xmm3, %xmm2, %xmm2
; AVX2NOBW-NEXT: vpshufb %xmm3, %xmm1, %xmm1
; AVX2NOBW-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX2NOBW-NEXT: vpsubb %xmm1, %xmm0, %xmm0
; AVX2NOBW-NEXT: vzeroupper
; AVX2NOBW-NEXT: retq
;
; AVX512BW-LABEL: test_rem7_16i8:
[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
; AVX512BW: # %bb.0:
[AVX512] Add avx512bw command lines to 128-bit idiv tests. The multiply lowering on some of the tests can take advantage of the vpmovwb to simplify the truncate. llvm-svn: 314448
2017-09-29 02:45:29 +08:00
; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
; AVX512BW-NEXT: vpmullw {{.*}}(%rip), %ymm1, %ymm1
; AVX512BW-NEXT: vpsrlw $8, %ymm1, %ymm1
[X86] Make use of vpmovwb when possible in LowerMULH If we have BWI, we can truncate in a much simpler way by using vpmovwb. This even works without VLX by using the wider zmm->ymm truncate with a subvector extract. Differential Revision: https://reviews.llvm.org/D38375 llvm-svn: 314457
2017-09-29 04:10:34 +08:00
; AVX512BW-NEXT: vpmovwb %zmm1, %ymm1
[AVX512] Add avx512bw command lines to 128-bit idiv tests. The multiply lowering on some of the tests can take advantage of the vpmovwb to simplify the truncate. llvm-svn: 314448
2017-09-29 02:45:29 +08:00
; AVX512BW-NEXT: vpsubb %xmm1, %xmm0, %xmm2
; AVX512BW-NEXT: vpsrlw $1, %xmm2, %xmm2
; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
; AVX512BW-NEXT: vpaddb %xmm1, %xmm2, %xmm1
; AVX512BW-NEXT: vpsrlw $2, %xmm1, %xmm1
; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
; AVX512BW-NEXT: vpmovsxbw %xmm1, %ymm1
; AVX512BW-NEXT: vpmullw {{.*}}(%rip), %ymm1, %ymm1
; AVX512BW-NEXT: vpmovwb %zmm1, %ymm1
; AVX512BW-NEXT: vpsubb %xmm1, %xmm0, %xmm0
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
[X86][SSE] Add vector integer division by constant tests Expanded tests and split into sdiv/srem and udiv/urem cases for 128 and 256 bit vectors. llvm-svn: 263917
2016-03-21 05:46:58 +08:00
%res = urem <16 x i8> %a, <i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7>
ret <16 x i8> %res
}