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/avx512-bugfix-26264.ll

42 lines
2.0 KiB
LLVM
Raw Normal View History

AVX512: Fix truncate v32i8 to v32i1 lowering implementation. Enable truncate 128/256bit packed byte/word with AVX512BW but without AVX512VL, use 512bit instructions. Differential Revision: http://reviews.llvm.org/D16531 llvm-svn: 259044
2016-01-28 21:19:25 +08:00
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=x86_64-apple-darwin -mattr=avx512f,avx512bw < %s | FileCheck %s --check-prefix=AVX512BW
define <32 x double> @test_load_32f64(<32 x double>* %ptrs, <32 x i1> %mask, <32 x double> %src0) {
; AVX512BW-LABEL: test_load_32f64:
[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: Fix truncate v32i8 to v32i1 lowering implementation. Enable truncate 128/256bit packed byte/word with AVX512BW but without AVX512VL, use 512bit instructions. Differential Revision: http://reviews.llvm.org/D16531 llvm-svn: 259044
2016-01-28 21:19:25 +08:00
; AVX512BW-NEXT: vpsllw $7, %ymm0, %ymm0
; AVX512BW-NEXT: vpmovb2m %zmm0, %k1
[AVX-512] Teach two address instruction pass to replace masked move instructions with blendm instructions when its beneficial. Isel now selects masked move instructions for vselect instead of blendm. But sometimes it beneficial to register allocation to remove the tied register constraint by using blendm instructions. This also picks up cases where the masked move was created due to a masked load intrinsic. Differential Revision: https://reviews.llvm.org/D28454 llvm-svn: 292005
2017-01-14 15:50:52 +08:00
; AVX512BW-NEXT: vblendmpd (%rdi), %zmm1, %zmm0 {%k1}
AVX512: Fix truncate v32i8 to v32i1 lowering implementation. Enable truncate 128/256bit packed byte/word with AVX512BW but without AVX512VL, use 512bit instructions. Differential Revision: http://reviews.llvm.org/D16531 llvm-svn: 259044
2016-01-28 21:19:25 +08:00
; AVX512BW-NEXT: kshiftrd $16, %k1, %k2
[AVX-512] Teach two address instruction pass to replace masked move instructions with blendm instructions when its beneficial. Isel now selects masked move instructions for vselect instead of blendm. But sometimes it beneficial to register allocation to remove the tied register constraint by using blendm instructions. This also picks up cases where the masked move was created due to a masked load intrinsic. Differential Revision: https://reviews.llvm.org/D28454 llvm-svn: 292005
2017-01-14 15:50:52 +08:00
; AVX512BW-NEXT: vblendmpd 128(%rdi), %zmm3, %zmm5 {%k2}
AVX512: Fix truncate v32i8 to v32i1 lowering implementation. Enable truncate 128/256bit packed byte/word with AVX512BW but without AVX512VL, use 512bit instructions. Differential Revision: http://reviews.llvm.org/D16531 llvm-svn: 259044
2016-01-28 21:19:25 +08:00
; AVX512BW-NEXT: kshiftrw $8, %k1, %k1
[AVX-512] Teach two address instruction pass to replace masked move instructions with blendm instructions when its beneficial. Isel now selects masked move instructions for vselect instead of blendm. But sometimes it beneficial to register allocation to remove the tied register constraint by using blendm instructions. This also picks up cases where the masked move was created due to a masked load intrinsic. Differential Revision: https://reviews.llvm.org/D28454 llvm-svn: 292005
2017-01-14 15:50:52 +08:00
; AVX512BW-NEXT: vblendmpd 64(%rdi), %zmm2, %zmm1 {%k1}
AVX512: Fix truncate v32i8 to v32i1 lowering implementation. Enable truncate 128/256bit packed byte/word with AVX512BW but without AVX512VL, use 512bit instructions. Differential Revision: http://reviews.llvm.org/D16531 llvm-svn: 259044
2016-01-28 21:19:25 +08:00
; AVX512BW-NEXT: kshiftrw $8, %k2, %k1
[AVX-512] Teach two address instruction pass to replace masked move instructions with blendm instructions when its beneficial. Isel now selects masked move instructions for vselect instead of blendm. But sometimes it beneficial to register allocation to remove the tied register constraint by using blendm instructions. This also picks up cases where the masked move was created due to a masked load intrinsic. Differential Revision: https://reviews.llvm.org/D28454 llvm-svn: 292005
2017-01-14 15:50:52 +08:00
; AVX512BW-NEXT: vblendmpd 192(%rdi), %zmm4, %zmm3 {%k1}
; AVX512BW-NEXT: vmovapd %zmm5, %zmm2
AVX512: Fix truncate v32i8 to v32i1 lowering implementation. Enable truncate 128/256bit packed byte/word with AVX512BW but without AVX512VL, use 512bit instructions. Differential Revision: http://reviews.llvm.org/D16531 llvm-svn: 259044
2016-01-28 21:19:25 +08:00
; AVX512BW-NEXT: retq
Support arbitrary addrspace pointers in masked load/store intrinsics This is a resubmittion of 263158 change after fixing the existing problem with intrinsics mangling (see LTO and intrinsics mangling llvm-dev thread for details). This patch fixes the problem which occurs when loop-vectorize tries to use @llvm.masked.load/store intrinsic for a non-default addrspace pointer. It fails with "Calling a function with a bad signature!" assertion in CallInst constructor because it tries to pass a non-default addrspace pointer to the pointer argument which has default addrspace. The fix is to add pointer type as another overloaded type to @llvm.masked.load/store intrinsics. Reviewed By: reames Differential Revision: http://reviews.llvm.org/D17270 llvm-svn: 274043
2016-06-29 02:27:25 +08:00
%res = call <32 x double> @llvm.masked.load.v32f64.p0v32f64(<32 x double>* %ptrs, i32 4, <32 x i1> %mask, <32 x double> %src0)
AVX512: Fix truncate v32i8 to v32i1 lowering implementation. Enable truncate 128/256bit packed byte/word with AVX512BW but without AVX512VL, use 512bit instructions. Differential Revision: http://reviews.llvm.org/D16531 llvm-svn: 259044
2016-01-28 21:19:25 +08:00
ret <32 x double> %res
}
define <32 x i64> @test_load_32i64(<32 x i64>* %ptrs, <32 x i1> %mask, <32 x i64> %src0) {
; AVX512BW-LABEL: test_load_32i64:
[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: Fix truncate v32i8 to v32i1 lowering implementation. Enable truncate 128/256bit packed byte/word with AVX512BW but without AVX512VL, use 512bit instructions. Differential Revision: http://reviews.llvm.org/D16531 llvm-svn: 259044
2016-01-28 21:19:25 +08:00
; AVX512BW-NEXT: vpsllw $7, %ymm0, %ymm0
; AVX512BW-NEXT: vpmovb2m %zmm0, %k1
[AVX-512] Teach two address instruction pass to replace masked move instructions with blendm instructions when its beneficial. Isel now selects masked move instructions for vselect instead of blendm. But sometimes it beneficial to register allocation to remove the tied register constraint by using blendm instructions. This also picks up cases where the masked move was created due to a masked load intrinsic. Differential Revision: https://reviews.llvm.org/D28454 llvm-svn: 292005
2017-01-14 15:50:52 +08:00
; AVX512BW-NEXT: vpblendmq (%rdi), %zmm1, %zmm0 {%k1}
AVX512: Fix truncate v32i8 to v32i1 lowering implementation. Enable truncate 128/256bit packed byte/word with AVX512BW but without AVX512VL, use 512bit instructions. Differential Revision: http://reviews.llvm.org/D16531 llvm-svn: 259044
2016-01-28 21:19:25 +08:00
; AVX512BW-NEXT: kshiftrd $16, %k1, %k2
[AVX-512] Teach two address instruction pass to replace masked move instructions with blendm instructions when its beneficial. Isel now selects masked move instructions for vselect instead of blendm. But sometimes it beneficial to register allocation to remove the tied register constraint by using blendm instructions. This also picks up cases where the masked move was created due to a masked load intrinsic. Differential Revision: https://reviews.llvm.org/D28454 llvm-svn: 292005
2017-01-14 15:50:52 +08:00
; AVX512BW-NEXT: vpblendmq 128(%rdi), %zmm3, %zmm5 {%k2}
AVX512: Fix truncate v32i8 to v32i1 lowering implementation. Enable truncate 128/256bit packed byte/word with AVX512BW but without AVX512VL, use 512bit instructions. Differential Revision: http://reviews.llvm.org/D16531 llvm-svn: 259044
2016-01-28 21:19:25 +08:00
; AVX512BW-NEXT: kshiftrw $8, %k1, %k1
[AVX-512] Teach two address instruction pass to replace masked move instructions with blendm instructions when its beneficial. Isel now selects masked move instructions for vselect instead of blendm. But sometimes it beneficial to register allocation to remove the tied register constraint by using blendm instructions. This also picks up cases where the masked move was created due to a masked load intrinsic. Differential Revision: https://reviews.llvm.org/D28454 llvm-svn: 292005
2017-01-14 15:50:52 +08:00
; AVX512BW-NEXT: vpblendmq 64(%rdi), %zmm2, %zmm1 {%k1}
AVX512: Fix truncate v32i8 to v32i1 lowering implementation. Enable truncate 128/256bit packed byte/word with AVX512BW but without AVX512VL, use 512bit instructions. Differential Revision: http://reviews.llvm.org/D16531 llvm-svn: 259044
2016-01-28 21:19:25 +08:00
; AVX512BW-NEXT: kshiftrw $8, %k2, %k1
[AVX-512] Teach two address instruction pass to replace masked move instructions with blendm instructions when its beneficial. Isel now selects masked move instructions for vselect instead of blendm. But sometimes it beneficial to register allocation to remove the tied register constraint by using blendm instructions. This also picks up cases where the masked move was created due to a masked load intrinsic. Differential Revision: https://reviews.llvm.org/D28454 llvm-svn: 292005
2017-01-14 15:50:52 +08:00
; AVX512BW-NEXT: vpblendmq 192(%rdi), %zmm4, %zmm3 {%k1}
; AVX512BW-NEXT: vmovdqa64 %zmm5, %zmm2
AVX512: Fix truncate v32i8 to v32i1 lowering implementation. Enable truncate 128/256bit packed byte/word with AVX512BW but without AVX512VL, use 512bit instructions. Differential Revision: http://reviews.llvm.org/D16531 llvm-svn: 259044
2016-01-28 21:19:25 +08:00
; AVX512BW-NEXT: retq
Support arbitrary addrspace pointers in masked load/store intrinsics This is a resubmittion of 263158 change after fixing the existing problem with intrinsics mangling (see LTO and intrinsics mangling llvm-dev thread for details). This patch fixes the problem which occurs when loop-vectorize tries to use @llvm.masked.load/store intrinsic for a non-default addrspace pointer. It fails with "Calling a function with a bad signature!" assertion in CallInst constructor because it tries to pass a non-default addrspace pointer to the pointer argument which has default addrspace. The fix is to add pointer type as another overloaded type to @llvm.masked.load/store intrinsics. Reviewed By: reames Differential Revision: http://reviews.llvm.org/D17270 llvm-svn: 274043
2016-06-29 02:27:25 +08:00
%res = call <32 x i64> @llvm.masked.load.v32i64.p0v32i64(<32 x i64>* %ptrs, i32 4, <32 x i1> %mask, <32 x i64> %src0)
AVX512: Fix truncate v32i8 to v32i1 lowering implementation. Enable truncate 128/256bit packed byte/word with AVX512BW but without AVX512VL, use 512bit instructions. Differential Revision: http://reviews.llvm.org/D16531 llvm-svn: 259044
2016-01-28 21:19:25 +08:00
ret <32 x i64> %res
}
Support arbitrary addrspace pointers in masked load/store intrinsics This is a resubmittion of 263158 change after fixing the existing problem with intrinsics mangling (see LTO and intrinsics mangling llvm-dev thread for details). This patch fixes the problem which occurs when loop-vectorize tries to use @llvm.masked.load/store intrinsic for a non-default addrspace pointer. It fails with "Calling a function with a bad signature!" assertion in CallInst constructor because it tries to pass a non-default addrspace pointer to the pointer argument which has default addrspace. The fix is to add pointer type as another overloaded type to @llvm.masked.load/store intrinsics. Reviewed By: reames Differential Revision: http://reviews.llvm.org/D17270 llvm-svn: 274043
2016-06-29 02:27:25 +08:00
declare <32 x i64> @llvm.masked.load.v32i64.p0v32i64(<32 x i64>* %ptrs, i32, <32 x i1> %mask, <32 x i64> %src0)
declare <32 x double> @llvm.masked.load.v32f64.p0v32f64(<32 x double>* %ptrs, i32, <32 x i1> %mask, <32 x double> %src0)