From a455c948188c2c60e577a0ae73a98e9347d9f8e5 Mon Sep 17 00:00:00 2001 From: Craig Topper Date: Mon, 6 Feb 2017 03:18:01 +0000 Subject: [PATCH] [AVX-512] Add VPABSB/D/Q/W to load folding tables. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@294169 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/Target/X86/X86InstrInfo.cpp | 34 ++++++ test/CodeGen/X86/stack-folding-int-avx512.ll | 100 ++++++++++++++++++ .../CodeGen/X86/stack-folding-int-avx512vl.ll | 72 +++++++++++++ 3 files changed, 206 insertions(+) diff --git a/lib/Target/X86/X86InstrInfo.cpp b/lib/Target/X86/X86InstrInfo.cpp index ca29885f521..abd6160a8b2 100644 --- a/lib/Target/X86/X86InstrInfo.cpp +++ b/lib/Target/X86/X86InstrInfo.cpp @@ -881,8 +881,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) { X86::VMOVDQU64Zrr, X86::VMOVDQU64Zrm, 0 }, { X86::VMOVUPDZrr, X86::VMOVUPDZrm, 0 }, { X86::VMOVUPSZrr, X86::VMOVUPSZrm, 0 }, + { X86::VPABSBZrr, X86::VPABSBZrm, 0 }, { X86::VPABSDZrr, X86::VPABSDZrm, 0 }, { X86::VPABSQZrr, X86::VPABSQZrm, 0 }, + { X86::VPABSWZrr, X86::VPABSWZrm, 0 }, { X86::VPERMILPDZri, X86::VPERMILPDZmi, 0 }, { X86::VPERMILPSZri, X86::VPERMILPSZmi, 0 }, { X86::VPERMPDZri, X86::VPERMPDZmi, 0 }, @@ -916,6 +918,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) { X86::VMOVDQU64Z256rr, X86::VMOVDQU64Z256rm, 0 }, { X86::VMOVUPDZ256rr, X86::VMOVUPDZ256rm, 0 }, { X86::VMOVUPSZ256rr, X86::VMOVUPSZ256rm, 0 }, + { X86::VPABSBZ256rr, X86::VPABSBZ256rm, 0 }, + { X86::VPABSDZ256rr, X86::VPABSDZ256rm, 0 }, + { X86::VPABSQZ256rr, X86::VPABSQZ256rm, 0 }, + { X86::VPABSWZ256rr, X86::VPABSWZ256rm, 0 }, { X86::VPERMILPDZ256ri, X86::VPERMILPDZ256mi, 0 }, { X86::VPERMILPSZ256ri, X86::VPERMILPSZ256mi, 0 }, { X86::VPERMPDZ256ri, X86::VPERMPDZ256mi, 0 }, @@ -948,6 +954,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) { X86::VMOVDQU64Z128rr, X86::VMOVDQU64Z128rm, 0 }, { X86::VMOVUPDZ128rr, X86::VMOVUPDZ128rm, 0 }, { X86::VMOVUPSZ128rr, X86::VMOVUPSZ128rm, 0 }, + { X86::VPABSBZ128rr, X86::VPABSBZ128rm, 0 }, + { X86::VPABSDZ128rr, X86::VPABSDZ128rm, 0 }, + { X86::VPABSQZ128rr, X86::VPABSQZ128rm, 0 }, + { X86::VPABSWZ128rr, X86::VPABSWZ128rm, 0 }, { X86::VPERMILPDZ128ri, X86::VPERMILPDZ128mi, 0 }, { X86::VPERMILPSZ128ri, X86::VPERMILPSZ128mi, 0 }, { X86::VPMOVSXBDZ128rr, X86::VPMOVSXBDZ128rm, TB_NO_REVERSE }, @@ -2139,6 +2149,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) // AVX-512 masked foldable instructions { X86::VBROADCASTSSZrkz, X86::VBROADCASTSSZmkz, TB_NO_REVERSE }, { X86::VBROADCASTSDZrkz, X86::VBROADCASTSDZmkz, TB_NO_REVERSE }, + { X86::VPABSBZrrkz, X86::VPABSBZrmkz, 0 }, + { X86::VPABSDZrrkz, X86::VPABSDZrmkz, 0 }, + { X86::VPABSQZrrkz, X86::VPABSQZrmkz, 0 }, + { X86::VPABSWZrrkz, X86::VPABSWZrmkz, 0 }, { X86::VPERMILPDZrikz, X86::VPERMILPDZmikz, 0 }, { X86::VPERMILPSZrikz, X86::VPERMILPSZmikz, 0 }, { X86::VPERMPDZrikz, X86::VPERMPDZmikz, 0 }, @@ -2162,6 +2176,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) // AVX-512VL 256-bit masked foldable instructions { X86::VBROADCASTSDZ256rkz, X86::VBROADCASTSDZ256mkz, TB_NO_REVERSE }, { X86::VBROADCASTSSZ256rkz, X86::VBROADCASTSSZ256mkz, TB_NO_REVERSE }, + { X86::VPABSBZ256rrkz, X86::VPABSBZ256rmkz, 0 }, + { X86::VPABSDZ256rrkz, X86::VPABSDZ256rmkz, 0 }, + { X86::VPABSQZ256rrkz, X86::VPABSQZ256rmkz, 0 }, + { X86::VPABSWZ256rrkz, X86::VPABSWZ256rmkz, 0 }, { X86::VPERMILPDZ256rikz, X86::VPERMILPDZ256mikz, 0 }, { X86::VPERMILPSZ256rikz, X86::VPERMILPSZ256mikz, 0 }, { X86::VPERMPDZ256rikz, X86::VPERMPDZ256mikz, 0 }, @@ -2184,6 +2202,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) // AVX-512VL 128-bit masked foldable instructions { X86::VBROADCASTSSZ128rkz, X86::VBROADCASTSSZ128mkz, TB_NO_REVERSE }, + { X86::VPABSBZ128rrkz, X86::VPABSBZ128rmkz, 0 }, + { X86::VPABSDZ128rrkz, X86::VPABSDZ128rmkz, 0 }, + { X86::VPABSQZ128rrkz, X86::VPABSQZ128rmkz, 0 }, + { X86::VPABSWZ128rrkz, X86::VPABSWZ128rmkz, 0 }, { X86::VPERMILPDZ128rikz, X86::VPERMILPDZ128mikz, 0 }, { X86::VPERMILPSZ128rikz, X86::VPERMILPSZ128mikz, 0 }, { X86::VPMOVSXBDZ128rrkz, X86::VPMOVSXBDZ128rmkz, TB_NO_REVERSE }, @@ -2601,6 +2623,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) // AVX-512 masked foldable instructions { X86::VBROADCASTSSZrk, X86::VBROADCASTSSZmk, TB_NO_REVERSE }, { X86::VBROADCASTSDZrk, X86::VBROADCASTSDZmk, TB_NO_REVERSE }, + { X86::VPABSBZrrk, X86::VPABSBZrmk, 0 }, + { X86::VPABSDZrrk, X86::VPABSDZrmk, 0 }, + { X86::VPABSQZrrk, X86::VPABSQZrmk, 0 }, + { X86::VPABSWZrrk, X86::VPABSWZrmk, 0 }, { X86::VPERMILPDZrik, X86::VPERMILPDZmik, 0 }, { X86::VPERMILPSZrik, X86::VPERMILPSZmik, 0 }, { X86::VPERMPDZrik, X86::VPERMPDZmik, 0 }, @@ -2624,6 +2650,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) // AVX-512VL 256-bit masked foldable instructions { X86::VBROADCASTSSZ256rk, X86::VBROADCASTSSZ256mk, TB_NO_REVERSE }, { X86::VBROADCASTSDZ256rk, X86::VBROADCASTSDZ256mk, TB_NO_REVERSE }, + { X86::VPABSBZ256rrk, X86::VPABSBZ256rmk, 0 }, + { X86::VPABSDZ256rrk, X86::VPABSDZ256rmk, 0 }, + { X86::VPABSQZ256rrk, X86::VPABSQZ256rmk, 0 }, + { X86::VPABSWZ256rrk, X86::VPABSWZ256rmk, 0 }, { X86::VPERMILPDZ256rik, X86::VPERMILPDZ256mik, 0 }, { X86::VPERMILPSZ256rik, X86::VPERMILPSZ256mik, 0 }, { X86::VPERMPDZ256rik, X86::VPERMPDZ256mik, 0 }, @@ -2646,6 +2676,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) // AVX-512VL 128-bit masked foldable instructions { X86::VBROADCASTSSZ128rk, X86::VBROADCASTSSZ128mk, TB_NO_REVERSE }, + { X86::VPABSBZ128rrk, X86::VPABSBZ128rmk, 0 }, + { X86::VPABSDZ128rrk, X86::VPABSDZ128rmk, 0 }, + { X86::VPABSQZ128rrk, X86::VPABSQZ128rmk, 0 }, + { X86::VPABSWZ128rrk, X86::VPABSWZ128rmk, 0 }, { X86::VPERMILPDZ128rik, X86::VPERMILPDZ128mik, 0 }, { X86::VPERMILPSZ128rik, X86::VPERMILPSZ128mik, 0 }, { X86::VPMOVSXBDZ128rrk, X86::VPMOVSXBDZ128rmk, TB_NO_REVERSE }, diff --git a/test/CodeGen/X86/stack-folding-int-avx512.ll b/test/CodeGen/X86/stack-folding-int-avx512.ll index 20572b37308..0bbbe7b54b3 100644 --- a/test/CodeGen/X86/stack-folding-int-avx512.ll +++ b/test/CodeGen/X86/stack-folding-int-avx512.ll @@ -8,6 +8,106 @@ target triple = "x86_64-unknown-unknown" ; By including a nop call with sideeffects we can force a partial register spill of the ; relevant registers and check that the reload is correctly folded into the instruction. +define <64 x i8> @stack_fold_pabsb(<64 x i8> %a0) { + ;CHECK-LABEL: stack_fold_pabsb + ;CHECK: vpabsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <64 x i8> @llvm.x86.avx512.mask.pabs.b.512(<64 x i8> %a0, <64 x i8> undef, i64 -1) + ret <64 x i8> %2 +} +declare <64 x i8> @llvm.x86.avx512.mask.pabs.b.512(<64 x i8>, <64 x i8>, i64) nounwind readnone + +define <64 x i8> @stack_fold_pabsb_mask(<64 x i8> %passthru, <64 x i8> %a0, i64 %mask) { + ;CHECK-LABEL: stack_fold_pabsb_mask + ;CHECK: vpabsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <64 x i8> @llvm.x86.avx512.mask.pabs.b.512(<64 x i8> %a0, <64 x i8> %passthru, i64 %mask) + ret <64 x i8> %2 +} + +define <64 x i8> @stack_fold_pabsb_maskz(<64 x i8> %a0, i64 %mask) { + ;CHECK-LABEL: stack_fold_pabsb_maskz + ;CHECK: vpabsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <64 x i8> @llvm.x86.avx512.mask.pabs.b.512(<64 x i8> %a0, <64 x i8> zeroinitializer, i64 %mask) + ret <64 x i8> %2 +} + +define <16 x i32> @stack_fold_pabsd(<16 x i32> %a0) { + ;check-label: stack_fold_pabsd + ;check: vpabsd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte folded reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <16 x i32> @llvm.x86.avx512.mask.pabs.d.512(<16 x i32> %a0, <16 x i32> undef, i16 -1) + ret <16 x i32> %2 +} +declare <16 x i32> @llvm.x86.avx512.mask.pabs.d.512(<16 x i32>, <16 x i32>, i16) nounwind readnone + +define <16 x i32> @stack_fold_pabsd_mask(<16 x i32> %passthru, <16 x i32> %a0, i16 %mask) { + ;check-label: stack_fold_pabsd + ;check: vpabsd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte folded reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <16 x i32> @llvm.x86.avx512.mask.pabs.d.512(<16 x i32> %a0, <16 x i32> %passthru, i16 %mask) + ret <16 x i32> %2 +} + +define <16 x i32> @stack_fold_pabsd_maskz(<16 x i32> %a0, i16 %mask) { + ;check-label: stack_fold_pabsd + ;check: vpabsd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte folded reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <16 x i32> @llvm.x86.avx512.mask.pabs.d.512(<16 x i32> %a0, <16 x i32> zeroinitializer, i16 %mask) + ret <16 x i32> %2 +} + +define <8 x i64> @stack_fold_pabsq(<8 x i64> %a0) { + ;CHECK-LABEL: stack_fold_pabsq + ;CHECK: vpabsq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <8 x i64> @llvm.x86.avx512.mask.pabs.q.512(<8 x i64> %a0, <8 x i64> undef, i8 -1) + ret <8 x i64> %2 +} +declare <8 x i64> @llvm.x86.avx512.mask.pabs.q.512(<8 x i64>, <8 x i64>, i8) nounwind readnone + +define <8 x i64> @stack_fold_pabsq_mask(<8 x i64> %passthru, <8 x i64> %a0, i8 %mask) { + ;CHECK-LABEL: stack_fold_pabsq_mask + ;CHECK: vpabsq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <8 x i64> @llvm.x86.avx512.mask.pabs.q.512(<8 x i64> %a0, <8 x i64> %passthru, i8 %mask) + ret <8 x i64> %2 +} + +define <8 x i64> @stack_fold_pabsq_maskz(<8 x i64> %a0, i8 %mask) { + ;CHECK-LABEL: stack_fold_pabsq_maskz + ;CHECK: vpabsq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <8 x i64> @llvm.x86.avx512.mask.pabs.q.512(<8 x i64> %a0, <8 x i64> zeroinitializer, i8 %mask) + ret <8 x i64> %2 +} + +define <32 x i16> @stack_fold_pabsw(<32 x i16> %a0) { + ;CHECK-LABEL: stack_fold_pabsw + ;CHECK: vpabsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <32 x i16> @llvm.x86.avx512.mask.pabs.w.512(<32 x i16> %a0, <32 x i16> undef, i32 -1) + ret <32 x i16> %2 +} +declare <32 x i16> @llvm.x86.avx512.mask.pabs.w.512(<32 x i16>, <32 x i16>, i32) nounwind readnone + +define <32 x i16> @stack_fold_pabsw_mask(<32 x i16> %passthru, <32 x i16> %a0, i32 %mask) { + ;CHECK-LABEL: stack_fold_pabsw_mask + ;CHECK: vpabsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <32 x i16> @llvm.x86.avx512.mask.pabs.w.512(<32 x i16> %a0, <32 x i16> %passthru, i32 %mask) + ret <32 x i16> %2 +} + +define <32 x i16> @stack_fold_pabsw_maskz(<32 x i16> %a0, i32 %mask) { + ;CHECK-LABEL: stack_fold_pabsw_maskz + ;CHECK: vpabsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <32 x i16> @llvm.x86.avx512.mask.pabs.w.512(<32 x i16> %a0, <32 x i16> zeroinitializer, i32 %mask) + ret <32 x i16> %2 +} + define <64 x i8> @stack_fold_paddb(<64 x i8> %a0, <64 x i8> %a1) { ;CHECK-LABEL: stack_fold_paddb ;CHECK: vpaddb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload diff --git a/test/CodeGen/X86/stack-folding-int-avx512vl.ll b/test/CodeGen/X86/stack-folding-int-avx512vl.ll index 06cd19e9577..ff82be346b0 100644 --- a/test/CodeGen/X86/stack-folding-int-avx512vl.ll +++ b/test/CodeGen/X86/stack-folding-int-avx512vl.ll @@ -8,6 +8,78 @@ target triple = "x86_64-unknown-unknown" ; By including a nop call with sideeffects we can force a partial register spill of the ; relevant registers and check that the reload is correctly folded into the instruction. +define <16 x i8> @stack_fold_pabsb(<16 x i8> %a0) { + ;CHECK-LABEL: stack_fold_pabsb + ;CHECK: vpabsb {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <16 x i8> @llvm.x86.ssse3.pabs.b.128(<16 x i8> %a0) + ret <16 x i8> %2 +} +declare <16 x i8> @llvm.x86.ssse3.pabs.b.128(<16 x i8>) nounwind readnone + +define <32 x i8> @stack_fold_pabsb_ymm(<32 x i8> %a0) { + ;CHECK-LABEL: stack_fold_pabsb_ymm + ;CHECK: vpabsb {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <32 x i8> @llvm.x86.avx2.pabs.b(<32 x i8> %a0) + ret <32 x i8> %2 +} +declare <32 x i8> @llvm.x86.avx2.pabs.b(<32 x i8>) nounwind readnone + +define <4 x i32> @stack_fold_pabsd(<4 x i32> %a0) { + ;CHECK-LABEL: stack_fold_pabsd + ;CHECK: vpabsd {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <4 x i32> @llvm.x86.ssse3.pabs.d.128(<4 x i32> %a0) + ret <4 x i32> %2 +} +declare <4 x i32> @llvm.x86.ssse3.pabs.d.128(<4 x i32>) nounwind readnone + +define <8 x i32> @stack_fold_pabsd_ymm(<8 x i32> %a0) { + ;CHECK-LABEL: stack_fold_pabsd_ymm + ;CHECK: vpabsd {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <8 x i32> @llvm.x86.avx2.pabs.d(<8 x i32> %a0) + ret <8 x i32> %2 +} +declare <8 x i32> @llvm.x86.avx2.pabs.d(<8 x i32>) nounwind readnone + +define <2 x i64> @stack_fold_pabsq(<2 x i64> %a0) { + ;CHECK-LABEL: stack_fold_pabsq + ;CHECK: vpabsq {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <2 x i64> @llvm.x86.avx512.mask.pabs.q.128(<2 x i64> %a0, <2 x i64> undef, i8 -1) + ret <2 x i64> %2 +} +declare <2 x i64> @llvm.x86.avx512.mask.pabs.q.128(<2 x i64>, <2 x i64>, i8) nounwind readnone + +define <4 x i64> @stack_fold_pabsq_ymm(<4 x i64> %a0) { + ;CHECK-LABEL: stack_fold_pabsq_ymm + ;CHECK: vpabsq {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <4 x i64> @llvm.x86.avx512.mask.pabs.q.256(<4 x i64> %a0, <4 x i64> undef, i8 -1) + ret <4 x i64> %2 +} +declare <4 x i64> @llvm.x86.avx512.mask.pabs.q.256(<4 x i64>, <4 x i64>, i8) nounwind readnone + +define <8 x i16> @stack_fold_pabsw(<8 x i16> %a0) { + ;CHECK-LABEL: stack_fold_pabsw + ;CHECK: vpabsw {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <8 x i16> @llvm.x86.ssse3.pabs.w.128(<8 x i16> %a0) + ret <8 x i16> %2 +} +declare <8 x i16> @llvm.x86.ssse3.pabs.w.128(<8 x i16>) nounwind readnone + +define <16 x i16> @stack_fold_pabsw_ymm(<16 x i16> %a0) { + ;CHECK-LABEL: stack_fold_pabsw_ymm + ;CHECK: vpabsw {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"() + %2 = call <16 x i16> @llvm.x86.avx2.pabs.w(<16 x i16> %a0) + ret <16 x i16> %2 +} +declare <16 x i16> @llvm.x86.avx2.pabs.w(<16 x i16>) nounwind readnone + define <16 x i8> @stack_fold_paddb(<16 x i8> %a0, <16 x i8> %a1) { ;CHECK-LABEL: stack_fold_paddb ;CHECK: vpaddb {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload -- 2.50.1