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/avx512er-intrinsics.ll

307 lines
16 KiB
LLVM
Raw Normal View History

[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512er --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X86
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512er --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X64
AVX-512: Intrinsics for ERI 3 instructions: vrcp28, vrsqrt28, vexp2, only vector forms. Intrinsics include SAE (Suppres All Exceptions) parameter. http://reviews.llvm.org/D6214 llvm-svn: 221774
2014-11-12 15:31:03 +08:00
define <16 x float> @test_rsqrt28_ps(<16 x float> %a0) {
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-LABEL: test_rsqrt28_ps:
[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
; CHECK: # %bb.0:
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-NEXT: vrsqrt28ps {sae}, %zmm0, %zmm0 # encoding: [0x62,0xf2,0x7d,0x18,0xcc,0xc0]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
AVX-512: Intrinsics for ERI 3 instructions: vrcp28, vrsqrt28, vexp2, only vector forms. Intrinsics include SAE (Suppres All Exceptions) parameter. http://reviews.llvm.org/D6214 llvm-svn: 221774
2014-11-12 15:31:03 +08:00
%res = call <16 x float> @llvm.x86.avx512.rsqrt28.ps(<16 x float> %a0, <16 x float> zeroinitializer, i16 -1, i32 8)
ret <16 x float> %res
}
define <16 x float> @test1_rsqrt28_ps(<16 x float> %a0, <16 x float> %a1) {
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-LABEL: test1_rsqrt28_ps:
[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
; CHECK: # %bb.0:
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-NEXT: movw $6, %ax # encoding: [0x66,0xb8,0x06,0x00]
; CHECK-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; CHECK-NEXT: vrsqrt28ps {sae}, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x19,0xcc,0xc8]
; CHECK-NEXT: vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
AVX-512: Intrinsics for ERI 3 instructions: vrcp28, vrsqrt28, vexp2, only vector forms. Intrinsics include SAE (Suppres All Exceptions) parameter. http://reviews.llvm.org/D6214 llvm-svn: 221774
2014-11-12 15:31:03 +08:00
%res = call <16 x float> @llvm.x86.avx512.rsqrt28.ps(<16 x float> %a0, <16 x float> %a1, i16 6, i32 8)
ret <16 x float> %res
}
define <16 x float> @test2_rsqrt28_ps(<16 x float> %a0) {
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-LABEL: test2_rsqrt28_ps:
[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
; CHECK: # %bb.0:
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-NEXT: movw $6, %ax # encoding: [0x66,0xb8,0x06,0x00]
; CHECK-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; CHECK-NEXT: vrsqrt28ps %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0xcc,0xc0]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
AVX-512: Intrinsics for ERI 3 instructions: vrcp28, vrsqrt28, vexp2, only vector forms. Intrinsics include SAE (Suppres All Exceptions) parameter. http://reviews.llvm.org/D6214 llvm-svn: 221774
2014-11-12 15:31:03 +08:00
%res = call <16 x float> @llvm.x86.avx512.rsqrt28.ps(<16 x float> %a0, <16 x float> undef, i16 6, i32 4)
ret <16 x float> %res
}
define <16 x float> @test3_rsqrt28_ps(<16 x float> %a0) {
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-LABEL: test3_rsqrt28_ps:
[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
; CHECK: # %bb.0:
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-NEXT: movw $6, %ax # encoding: [0x66,0xb8,0x06,0x00]
; CHECK-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; CHECK-NEXT: vrsqrt28ps %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0xcc,0xc0]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
AVX-512: Intrinsics for ERI 3 instructions: vrcp28, vrsqrt28, vexp2, only vector forms. Intrinsics include SAE (Suppres All Exceptions) parameter. http://reviews.llvm.org/D6214 llvm-svn: 221774
2014-11-12 15:31:03 +08:00
%res = call <16 x float> @llvm.x86.avx512.rsqrt28.ps(<16 x float> %a0, <16 x float> zeroinitializer, i16 6, i32 4)
ret <16 x float> %res
}
define <16 x float> @test4_rsqrt28_ps(<16 x float> %a0) {
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-LABEL: test4_rsqrt28_ps:
[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
; CHECK: # %bb.0:
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-NEXT: movw $6, %ax # encoding: [0x66,0xb8,0x06,0x00]
; CHECK-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; CHECK-NEXT: vrsqrt28ps {sae}, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x99,0xcc,0xc0]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
%res = call <16 x float> @llvm.x86.avx512.rsqrt28.ps(<16 x float> %a0, <16 x float> undef, i16 6, i32 8)
AVX-512: Intrinsics for ERI 3 instructions: vrcp28, vrsqrt28, vexp2, only vector forms. Intrinsics include SAE (Suppres All Exceptions) parameter. http://reviews.llvm.org/D6214 llvm-svn: 221774
2014-11-12 15:31:03 +08:00
ret <16 x float> %res
}
declare <16 x float> @llvm.x86.avx512.rsqrt28.ps(<16 x float>, <16 x float>, i16, i32) nounwind readnone
define <16 x float> @test_rcp28_ps_512(<16 x float> %a0) {
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-LABEL: test_rcp28_ps_512:
[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
; CHECK: # %bb.0:
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-NEXT: vrcp28ps {sae}, %zmm0, %zmm0 # encoding: [0x62,0xf2,0x7d,0x18,0xca,0xc0]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
AVX-512: Intrinsics for ERI 3 instructions: vrcp28, vrsqrt28, vexp2, only vector forms. Intrinsics include SAE (Suppres All Exceptions) parameter. http://reviews.llvm.org/D6214 llvm-svn: 221774
2014-11-12 15:31:03 +08:00
%res = call <16 x float> @llvm.x86.avx512.rcp28.ps(<16 x float> %a0, <16 x float> zeroinitializer, i16 -1, i32 8)
ret <16 x float> %res
}
declare <16 x float> @llvm.x86.avx512.rcp28.ps(<16 x float>, <16 x float>, i16, i32) nounwind readnone
define <8 x double> @test_rcp28_pd_512(<8 x double> %a0) {
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-LABEL: test_rcp28_pd_512:
[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
; CHECK: # %bb.0:
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-NEXT: vrcp28pd {sae}, %zmm0, %zmm0 # encoding: [0x62,0xf2,0xfd,0x18,0xca,0xc0]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
%res = call <8 x double> @llvm.x86.avx512.rcp28.pd(<8 x double> %a0, <8 x double> zeroinitializer, i8 -1, i32 8)
AVX-512: Intrinsics for ERI 3 instructions: vrcp28, vrsqrt28, vexp2, only vector forms. Intrinsics include SAE (Suppres All Exceptions) parameter. http://reviews.llvm.org/D6214 llvm-svn: 221774
2014-11-12 15:31:03 +08:00
ret <8 x double> %res
}
declare <8 x double> @llvm.x86.avx512.rcp28.pd(<8 x double>, <8 x double>, i8, i32) nounwind readnone
define <16 x float> @test_exp2_ps_512(<16 x float> %a0) {
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-LABEL: test_exp2_ps_512:
[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
; CHECK: # %bb.0:
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-NEXT: vexp2ps {sae}, %zmm0, %zmm0 # encoding: [0x62,0xf2,0x7d,0x18,0xc8,0xc0]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
AVX-512: Intrinsics for ERI 3 instructions: vrcp28, vrsqrt28, vexp2, only vector forms. Intrinsics include SAE (Suppres All Exceptions) parameter. http://reviews.llvm.org/D6214 llvm-svn: 221774
2014-11-12 15:31:03 +08:00
%res = call <16 x float> @llvm.x86.avx512.exp2.ps(<16 x float> %a0, <16 x float> zeroinitializer, i16 -1, i32 8)
ret <16 x float> %res
}
declare <16 x float> @llvm.x86.avx512.exp2.ps(<16 x float>, <16 x float>, i16, i32) nounwind readnone
define <8 x double> @test_exp2_pd_512(<8 x double> %a0) {
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-LABEL: test_exp2_pd_512:
[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
; CHECK: # %bb.0:
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-NEXT: vexp2pd {sae}, %zmm0, %zmm0 # encoding: [0x62,0xf2,0xfd,0x18,0xc8,0xc0]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
AVX-512: Intrinsics for ERI 3 instructions: vrcp28, vrsqrt28, vexp2, only vector forms. Intrinsics include SAE (Suppres All Exceptions) parameter. http://reviews.llvm.org/D6214 llvm-svn: 221774
2014-11-12 15:31:03 +08:00
%res = call <8 x double> @llvm.x86.avx512.exp2.pd(<8 x double> %a0, <8 x double> zeroinitializer, i8 -1, i32 8)
ret <8 x double> %res
}
declare <8 x double> @llvm.x86.avx512.exp2.pd(<8 x double>, <8 x double>, i8, i32) nounwind readnone
define <4 x float> @test_rsqrt28_ss(<4 x float> %a0) {
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-LABEL: test_rsqrt28_ss:
[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
; CHECK: # %bb.0:
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-NEXT: vrsqrt28ss {sae}, %xmm0, %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x18,0xcd,0xc0]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
AVX-512: Intrinsics for ERI 3 instructions: vrcp28, vrsqrt28, vexp2, only vector forms. Intrinsics include SAE (Suppres All Exceptions) parameter. http://reviews.llvm.org/D6214 llvm-svn: 221774
2014-11-12 15:31:03 +08:00
%res = call <4 x float> @llvm.x86.avx512.rsqrt28.ss(<4 x float> %a0, <4 x float> %a0, <4 x float> zeroinitializer, i8 -1, i32 8) ; <<4 x float>> [#uses=1]
ret <4 x float> %res
}
declare <4 x float> @llvm.x86.avx512.rsqrt28.ss(<4 x float>, <4 x float>, <4 x float>, i8, i32) nounwind readnone
define <4 x float> @test_rcp28_ss(<4 x float> %a0) {
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-LABEL: test_rcp28_ss:
[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
; CHECK: # %bb.0:
[AVX-512] Use update_llc_test_checks.py to regenerate the avx512er intrinsic test. llvm-svn: 296268
2017-02-26 03:18:04 +08:00
; CHECK-NEXT: vrcp28ss {sae}, %xmm0, %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x18,0xcb,0xc0]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
AVX-512: Intrinsics for ERI 3 instructions: vrcp28, vrsqrt28, vexp2, only vector forms. Intrinsics include SAE (Suppres All Exceptions) parameter. http://reviews.llvm.org/D6214 llvm-svn: 221774
2014-11-12 15:31:03 +08:00
%res = call <4 x float> @llvm.x86.avx512.rcp28.ss(<4 x float> %a0, <4 x float> %a0, <4 x float> zeroinitializer, i8 -1, i32 8) ; <<4 x float>> [#uses=1]
ret <4 x float> %res
}
declare <4 x float> @llvm.x86.avx512.rcp28.ss(<4 x float>, <4 x float>, <4 x float>, i8, i32) nounwind readnone
[X86] Improve load folding of scalar rcp28 and rsqrt28 instructions using sse_load_f32/f64. llvm-svn: 318623
2017-11-19 13:42:54 +08:00
define <4 x float> @test_rcp28_ss_load(<4 x float> %a0, <4 x float>* %a1ptr) {
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-LABEL: test_rcp28_ss_load:
; X86: # %bb.0:
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT: vrcp28ss (%eax), %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x08,0xcb,0x00]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_rcp28_ss_load:
; X64: # %bb.0:
; X64-NEXT: vrcp28ss (%rdi), %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x08,0xcb,0x07]
; X64-NEXT: retq # encoding: [0xc3]
[X86] Improve load folding of scalar rcp28 and rsqrt28 instructions using sse_load_f32/f64. llvm-svn: 318623
2017-11-19 13:42:54 +08:00
%a1 = load <4 x float>, <4 x float>* %a1ptr
%res = call <4 x float> @llvm.x86.avx512.rcp28.ss(<4 x float> %a0, <4 x float> %a1, <4 x float> undef, i8 -1, i32 4) ; <<4 x float>> [#uses=1]
ret <4 x float> %res
}
define <4 x float> @test_rsqrt28_ss_load(<4 x float> %a0, <4 x float>* %a1ptr) {
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-LABEL: test_rsqrt28_ss_load:
; X86: # %bb.0:
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT: vrsqrt28ss (%eax), %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x08,0xcd,0x00]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_rsqrt28_ss_load:
; X64: # %bb.0:
; X64-NEXT: vrsqrt28ss (%rdi), %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x08,0xcd,0x07]
; X64-NEXT: retq # encoding: [0xc3]
[X86] Improve load folding of scalar rcp28 and rsqrt28 instructions using sse_load_f32/f64. llvm-svn: 318623
2017-11-19 13:42:54 +08:00
%a1 = load <4 x float>, <4 x float>* %a1ptr
%res = call <4 x float> @llvm.x86.avx512.rsqrt28.ss(<4 x float> %a0, <4 x float> %a1, <4 x float> undef, i8 -1, i32 4) ; <<4 x float>> [#uses=1]
ret <4 x float> %res
}
[X86][AVX512] Refine some avx512er intrinsics tests. NFC. The modified tests should test the masked intrinsics. Currently the mask is constant, which with a future patch (https://reviews.llvm.org/D32805) will cause the intrinsics to be replaced with an unmasked version. This patch changes the constant mask to be a variable one. llvm-svn: 302529
2017-05-09 22:03:51 +08:00
define <4 x float> @test_rsqrt28_ss_maskz(<4 x float> %a0, i8 %mask) {
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-LABEL: test_rsqrt28_ss_maskz:
; X86: # %bb.0:
[X86] In getScalarMaskingNode, replace scalar_to_vector with a bitcast to v8i1 and an extract_subvector to convert i8 to v1i1. The bitcast can be nicely merged with any i8 loads that exist for argument passing in 32 mode for example. llvm-svn: 347380
2018-11-21 15:01:22 +08:00
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT: vrsqrt28ss {sae}, %xmm0, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x99,0xcd,0xc0]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_rsqrt28_ss_maskz:
; X64: # %bb.0:
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT: vrsqrt28ss {sae}, %xmm0, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x99,0xcd,0xc0]
; X64-NEXT: retq # encoding: [0xc3]
[X86][AVX512] Refine some avx512er intrinsics tests. NFC. The modified tests should test the masked intrinsics. Currently the mask is constant, which with a future patch (https://reviews.llvm.org/D32805) will cause the intrinsics to be replaced with an unmasked version. This patch changes the constant mask to be a variable one. llvm-svn: 302529
2017-05-09 22:03:51 +08:00
%res = call <4 x float> @llvm.x86.avx512.rsqrt28.ss(<4 x float> %a0, <4 x float> %a0, <4 x float> zeroinitializer, i8 %mask, i32 8) ;
AVX-512: Scalar ERI intrinsics including SAE mode and memory operand. Added AVX512_maskable_scalar template, that should cover all scalar instructions in the future. The main difference between AVX512_maskable_scalar<> and AVX512_maskable<> is using X86select instead of vselect. I need it, because I can't create vselect node for MVT::i1 mask for scalar instruction. http://reviews.llvm.org/D6378 llvm-svn: 222820
2014-11-26 18:46:49 +08:00
ret <4 x float> %res
}
[X86][AVX512] Refine some avx512er intrinsics tests. NFC. The modified tests should test the masked intrinsics. Currently the mask is constant, which with a future patch (https://reviews.llvm.org/D32805) will cause the intrinsics to be replaced with an unmasked version. This patch changes the constant mask to be a variable one. llvm-svn: 302529
2017-05-09 22:03:51 +08:00
define <4 x float> @test_rsqrt28_ss_mask(<4 x float> %a0, <4 x float> %b0, <4 x float> %c0, i8 %mask) {
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-LABEL: test_rsqrt28_ss_mask:
; X86: # %bb.0:
[X86] In getScalarMaskingNode, replace scalar_to_vector with a bitcast to v8i1 and an extract_subvector to convert i8 to v1i1. The bitcast can be nicely merged with any i8 loads that exist for argument passing in 32 mode for example. llvm-svn: 347380
2018-11-21 15:01:22 +08:00
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT: vrsqrt28ss {sae}, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x19,0xcd,0xd1]
; X86-NEXT: vmovaps %xmm2, %xmm0 # encoding: [0xc5,0xf8,0x28,0xc2]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_rsqrt28_ss_mask:
; X64: # %bb.0:
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT: vrsqrt28ss {sae}, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x19,0xcd,0xd1]
; X64-NEXT: vmovaps %xmm2, %xmm0 # encoding: [0xc5,0xf8,0x28,0xc2]
; X64-NEXT: retq # encoding: [0xc3]
[X86][AVX512] Refine some avx512er intrinsics tests. NFC. The modified tests should test the masked intrinsics. Currently the mask is constant, which with a future patch (https://reviews.llvm.org/D32805) will cause the intrinsics to be replaced with an unmasked version. This patch changes the constant mask to be a variable one. llvm-svn: 302529
2017-05-09 22:03:51 +08:00
%res = call <4 x float> @llvm.x86.avx512.rsqrt28.ss(<4 x float> %a0, <4 x float> %b0, <4 x float> %c0, i8 %mask, i32 8) ;
AVX-512: Scalar ERI intrinsics including SAE mode and memory operand. Added AVX512_maskable_scalar template, that should cover all scalar instructions in the future. The main difference between AVX512_maskable_scalar<> and AVX512_maskable<> is using X86select instead of vselect. I need it, because I can't create vselect node for MVT::i1 mask for scalar instruction. http://reviews.llvm.org/D6378 llvm-svn: 222820
2014-11-26 18:46:49 +08:00
ret <4 x float> %res
}
[X86] Improve load folding of scalar rcp28 and rsqrt28 instructions using sse_load_f32/f64. llvm-svn: 318623
2017-11-19 13:42:54 +08:00
define <2 x double> @test_rcp28_sd_mask_load(<2 x double> %a0, <2 x double>* %a1ptr, <2 x double> %a2, i8 %mask) {
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-LABEL: test_rcp28_sd_mask_load:
; X86: # %bb.0:
[X86] In getScalarMaskingNode, replace scalar_to_vector with a bitcast to v8i1 and an extract_subvector to convert i8 to v1i1. The bitcast can be nicely merged with any i8 loads that exist for argument passing in 32 mode for example. llvm-svn: 347380
2018-11-21 15:01:22 +08:00
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT: vrcp28sd %xmm0, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0xcb,0xc8]
; X86-NEXT: vmovapd %xmm1, %xmm0 # encoding: [0xc5,0xf9,0x28,0xc1]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_rcp28_sd_mask_load:
; X64: # %bb.0:
; X64-NEXT: kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT: vrcp28sd %xmm0, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0xcb,0xc8]
; X64-NEXT: vmovapd %xmm1, %xmm0 # encoding: [0xc5,0xf9,0x28,0xc1]
; X64-NEXT: retq # encoding: [0xc3]
[X86] Improve load folding of scalar rcp28 and rsqrt28 instructions using sse_load_f32/f64. llvm-svn: 318623
2017-11-19 13:42:54 +08:00
%a1 = load <2 x double>, <2 x double>* %a1ptr
%res = call <2 x double> @llvm.x86.avx512.rcp28.sd(<2 x double> %a0, <2 x double> %a0, <2 x double> %a2, i8 %mask, i32 4) ;
ret <2 x double> %res
}
declare <2 x double> @llvm.x86.avx512.rcp28.sd(<2 x double>, <2 x double>, <2 x double>, i8, i32) nounwind readnone
define <2 x double> @test_rsqrt28_sd_maskz_load(<2 x double> %a0, <2 x double>* %a1ptr, i8 %mask) {
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-LABEL: test_rsqrt28_sd_maskz_load:
; X86: # %bb.0:
[X86] In getScalarMaskingNode, replace scalar_to_vector with a bitcast to v8i1 and an extract_subvector to convert i8 to v1i1. The bitcast can be nicely merged with any i8 loads that exist for argument passing in 32 mode for example. llvm-svn: 347380
2018-11-21 15:01:22 +08:00
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT: vrsqrt28sd %xmm0, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0xcd,0xc0]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_rsqrt28_sd_maskz_load:
; X64: # %bb.0:
; X64-NEXT: kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT: vrsqrt28sd %xmm0, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0xcd,0xc0]
; X64-NEXT: retq # encoding: [0xc3]
[X86] Improve load folding of scalar rcp28 and rsqrt28 instructions using sse_load_f32/f64. llvm-svn: 318623
2017-11-19 13:42:54 +08:00
%a1 = load <2 x double>, <2 x double>* %a1ptr
%res = call <2 x double> @llvm.x86.avx512.rsqrt28.sd(<2 x double> %a0, <2 x double> %a0, <2 x double> zeroinitializer, i8 %mask, i32 4) ;
ret <2 x double> %res
}
[X86][AVX512] Refine some avx512er intrinsics tests. NFC. The modified tests should test the masked intrinsics. Currently the mask is constant, which with a future patch (https://reviews.llvm.org/D32805) will cause the intrinsics to be replaced with an unmasked version. This patch changes the constant mask to be a variable one. llvm-svn: 302529
2017-05-09 22:03:51 +08:00
define <2 x double> @test_rsqrt28_sd_maskz(<2 x double> %a0, i8 %mask) {
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-LABEL: test_rsqrt28_sd_maskz:
; X86: # %bb.0:
[X86] In getScalarMaskingNode, replace scalar_to_vector with a bitcast to v8i1 and an extract_subvector to convert i8 to v1i1. The bitcast can be nicely merged with any i8 loads that exist for argument passing in 32 mode for example. llvm-svn: 347380
2018-11-21 15:01:22 +08:00
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT: vrsqrt28sd {sae}, %xmm0, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x99,0xcd,0xc0]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_rsqrt28_sd_maskz:
; X64: # %bb.0:
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT: vrsqrt28sd {sae}, %xmm0, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x99,0xcd,0xc0]
; X64-NEXT: retq # encoding: [0xc3]
[X86][AVX512] Refine some avx512er intrinsics tests. NFC. The modified tests should test the masked intrinsics. Currently the mask is constant, which with a future patch (https://reviews.llvm.org/D32805) will cause the intrinsics to be replaced with an unmasked version. This patch changes the constant mask to be a variable one. llvm-svn: 302529
2017-05-09 22:03:51 +08:00
%res = call <2 x double> @llvm.x86.avx512.rsqrt28.sd(<2 x double> %a0, <2 x double> %a0, <2 x double> zeroinitializer, i8 %mask, i32 8) ;
AVX-512: Scalar ERI intrinsics including SAE mode and memory operand. Added AVX512_maskable_scalar template, that should cover all scalar instructions in the future. The main difference between AVX512_maskable_scalar<> and AVX512_maskable<> is using X86select instead of vselect. I need it, because I can't create vselect node for MVT::i1 mask for scalar instruction. http://reviews.llvm.org/D6378 llvm-svn: 222820
2014-11-26 18:46:49 +08:00
ret <2 x double> %res
}
[X86][AVX512] Refine some avx512er intrinsics tests. NFC. The modified tests should test the masked intrinsics. Currently the mask is constant, which with a future patch (https://reviews.llvm.org/D32805) will cause the intrinsics to be replaced with an unmasked version. This patch changes the constant mask to be a variable one. llvm-svn: 302529
2017-05-09 22:03:51 +08:00
define <2 x double> @test_rsqrt28_sd_mask(<2 x double> %a0, <2 x double> %b0, <2 x double> %c0, i8 %mask) {
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-LABEL: test_rsqrt28_sd_mask:
; X86: # %bb.0:
[X86] In getScalarMaskingNode, replace scalar_to_vector with a bitcast to v8i1 and an extract_subvector to convert i8 to v1i1. The bitcast can be nicely merged with any i8 loads that exist for argument passing in 32 mode for example. llvm-svn: 347380
2018-11-21 15:01:22 +08:00
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT: vrsqrt28sd {sae}, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x19,0xcd,0xd1]
; X86-NEXT: vmovapd %xmm2, %xmm0 # encoding: [0xc5,0xf9,0x28,0xc2]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_rsqrt28_sd_mask:
; X64: # %bb.0:
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT: vrsqrt28sd {sae}, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x19,0xcd,0xd1]
; X64-NEXT: vmovapd %xmm2, %xmm0 # encoding: [0xc5,0xf9,0x28,0xc2]
; X64-NEXT: retq # encoding: [0xc3]
[X86][AVX512] Refine some avx512er intrinsics tests. NFC. The modified tests should test the masked intrinsics. Currently the mask is constant, which with a future patch (https://reviews.llvm.org/D32805) will cause the intrinsics to be replaced with an unmasked version. This patch changes the constant mask to be a variable one. llvm-svn: 302529
2017-05-09 22:03:51 +08:00
%res = call <2 x double> @llvm.x86.avx512.rsqrt28.sd(<2 x double> %a0, <2 x double> %b0, <2 x double> %c0, i8 %mask, i32 8) ;
[AVX-512] Add an additional test case to show the execution domain for vrqsrtsd is wrong. llvm-svn: 296269
2017-02-26 03:18:08 +08:00
ret <2 x double> %res
}
AVX-512: Scalar ERI intrinsics including SAE mode and memory operand. Added AVX512_maskable_scalar template, that should cover all scalar instructions in the future. The main difference between AVX512_maskable_scalar<> and AVX512_maskable<> is using X86select instead of vselect. I need it, because I can't create vselect node for MVT::i1 mask for scalar instruction. http://reviews.llvm.org/D6378 llvm-svn: 222820
2014-11-26 18:46:49 +08:00
declare <2 x double> @llvm.x86.avx512.rsqrt28.sd(<2 x double>, <2 x double>, <2 x double>, i8, i32) nounwind readnone
[X86][AVX512] Refine some avx512er intrinsics tests. NFC. The modified tests should test the masked intrinsics. Currently the mask is constant, which with a future patch (https://reviews.llvm.org/D32805) will cause the intrinsics to be replaced with an unmasked version. This patch changes the constant mask to be a variable one. llvm-svn: 302529
2017-05-09 22:03:51 +08:00
define <2 x double> @test_rsqrt28_sd_maskz_mem(<2 x double> %a0, double* %ptr, i8 %mask) {
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-LABEL: test_rsqrt28_sd_maskz_mem:
; X86: # %bb.0:
[X86] In getScalarMaskingNode, replace scalar_to_vector with a bitcast to v8i1 and an extract_subvector to convert i8 to v1i1. The bitcast can be nicely merged with any i8 loads that exist for argument passing in 32 mode for example. llvm-svn: 347380
2018-11-21 15:01:22 +08:00
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT: vrsqrt28sd (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0xcd,0x00]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_rsqrt28_sd_maskz_mem:
; X64: # %bb.0:
; X64-NEXT: kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT: vrsqrt28sd (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0xcd,0x07]
; X64-NEXT: retq # encoding: [0xc3]
[opaque pointer type] Add textual IR support for explicit type parameter to load instruction Essentially the same as the GEP change in r230786. A similar migration script can be used to update test cases, though a few more test case improvements/changes were required this time around: (r229269-r229278) import fileinput import sys import re pat = re.compile(r"((?:=|:|^)\s*load (?:atomic )?(?:volatile )?(.*?))(| addrspace\(\d+\) *)\*($| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$)") for line in sys.stdin: sys.stdout.write(re.sub(pat, r"\1, \2\3*\4", line)) Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7649 llvm-svn: 230794
2015-02-28 05:17:42 +08:00
%mem = load double , double * %ptr, align 8
AVX-512: Scalar ERI intrinsics including SAE mode and memory operand. Added AVX512_maskable_scalar template, that should cover all scalar instructions in the future. The main difference between AVX512_maskable_scalar<> and AVX512_maskable<> is using X86select instead of vselect. I need it, because I can't create vselect node for MVT::i1 mask for scalar instruction. http://reviews.llvm.org/D6378 llvm-svn: 222820
2014-11-26 18:46:49 +08:00
%mem_v = insertelement <2 x double> undef, double %mem, i32 0
[X86][AVX512] Refine some avx512er intrinsics tests. NFC. The modified tests should test the masked intrinsics. Currently the mask is constant, which with a future patch (https://reviews.llvm.org/D32805) will cause the intrinsics to be replaced with an unmasked version. This patch changes the constant mask to be a variable one. llvm-svn: 302529
2017-05-09 22:03:51 +08:00
%res = call <2 x double> @llvm.x86.avx512.rsqrt28.sd(<2 x double> %a0, <2 x double> %mem_v, <2 x double> zeroinitializer, i8 %mask, i32 4) ;
AVX-512: Scalar ERI intrinsics including SAE mode and memory operand. Added AVX512_maskable_scalar template, that should cover all scalar instructions in the future. The main difference between AVX512_maskable_scalar<> and AVX512_maskable<> is using X86select instead of vselect. I need it, because I can't create vselect node for MVT::i1 mask for scalar instruction. http://reviews.llvm.org/D6378 llvm-svn: 222820
2014-11-26 18:46:49 +08:00
ret <2 x double> %res
}
[X86][AVX512] Refine some avx512er intrinsics tests. NFC. The modified tests should test the masked intrinsics. Currently the mask is constant, which with a future patch (https://reviews.llvm.org/D32805) will cause the intrinsics to be replaced with an unmasked version. This patch changes the constant mask to be a variable one. llvm-svn: 302529
2017-05-09 22:03:51 +08:00
define <2 x double> @test_rsqrt28_sd_maskz_mem_offset(<2 x double> %a0, double* %ptr, i8 %mask) {
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-LABEL: test_rsqrt28_sd_maskz_mem_offset:
; X86: # %bb.0:
[X86] In getScalarMaskingNode, replace scalar_to_vector with a bitcast to v8i1 and an extract_subvector to convert i8 to v1i1. The bitcast can be nicely merged with any i8 loads that exist for argument passing in 32 mode for example. llvm-svn: 347380
2018-11-21 15:01:22 +08:00
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT: vrsqrt28sd 144(%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0xcd,0x40,0x12]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_rsqrt28_sd_maskz_mem_offset:
; X64: # %bb.0:
; X64-NEXT: kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT: vrsqrt28sd 144(%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0xcd,0x47,0x12]
; X64-NEXT: retq # encoding: [0xc3]
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-28 03:29:02 +08:00
%ptr1 = getelementptr double, double* %ptr, i32 18
[opaque pointer type] Add textual IR support for explicit type parameter to load instruction Essentially the same as the GEP change in r230786. A similar migration script can be used to update test cases, though a few more test case improvements/changes were required this time around: (r229269-r229278) import fileinput import sys import re pat = re.compile(r"((?:=|:|^)\s*load (?:atomic )?(?:volatile )?(.*?))(| addrspace\(\d+\) *)\*($| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$)") for line in sys.stdin: sys.stdout.write(re.sub(pat, r"\1, \2\3*\4", line)) Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7649 llvm-svn: 230794
2015-02-28 05:17:42 +08:00
%mem = load double , double * %ptr1, align 8
AVX-512: Scalar ERI intrinsics including SAE mode and memory operand. Added AVX512_maskable_scalar template, that should cover all scalar instructions in the future. The main difference between AVX512_maskable_scalar<> and AVX512_maskable<> is using X86select instead of vselect. I need it, because I can't create vselect node for MVT::i1 mask for scalar instruction. http://reviews.llvm.org/D6378 llvm-svn: 222820
2014-11-26 18:46:49 +08:00
%mem_v = insertelement <2 x double> undef, double %mem, i32 0
[X86][AVX512] Refine some avx512er intrinsics tests. NFC. The modified tests should test the masked intrinsics. Currently the mask is constant, which with a future patch (https://reviews.llvm.org/D32805) will cause the intrinsics to be replaced with an unmasked version. This patch changes the constant mask to be a variable one. llvm-svn: 302529
2017-05-09 22:03:51 +08:00
%res = call <2 x double> @llvm.x86.avx512.rsqrt28.sd(<2 x double> %a0, <2 x double> %mem_v, <2 x double> zeroinitializer, i8 %mask, i32 4) ;
AVX-512: Scalar ERI intrinsics including SAE mode and memory operand. Added AVX512_maskable_scalar template, that should cover all scalar instructions in the future. The main difference between AVX512_maskable_scalar<> and AVX512_maskable<> is using X86select instead of vselect. I need it, because I can't create vselect node for MVT::i1 mask for scalar instruction. http://reviews.llvm.org/D6378 llvm-svn: 222820
2014-11-26 18:46:49 +08:00
ret <2 x double> %res
}