From: Craig Topper Date: Wed, 15 Feb 2017 06:51:39 +0000 (+0000) Subject: [AVX-512] Add PACKSS/PACKUS instructions to load folding tables. X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=fc3e8436201537bbd694e1990f3b0a6a88017c73;p=llvm [AVX-512] Add PACKSS/PACKUS instructions to load folding tables. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@295154 91177308-0d34-0410-b5e6-96231b3b80d8 --- diff --git a/lib/Target/X86/X86InstrInfo.cpp b/lib/Target/X86/X86InstrInfo.cpp index 8335476f0b5..d69c9d7a71f 100644 --- a/lib/Target/X86/X86InstrInfo.cpp +++ b/lib/Target/X86/X86InstrInfo.cpp @@ -1894,6 +1894,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) { X86::VMULSSZrr_Int, X86::VMULSSZrm_Int, TB_NO_REVERSE }, { X86::VORPDZrr, X86::VORPDZrm, 0 }, { X86::VORPSZrr, X86::VORPSZrm, 0 }, + { X86::VPACKSSDWZrr, X86::VPACKSSDWZrm, 0 }, + { X86::VPACKSSWBZrr, X86::VPACKSSWBZrm, 0 }, + { X86::VPACKUSDWZrr, X86::VPACKUSDWZrm, 0 }, + { X86::VPACKUSWBZrr, X86::VPACKUSWBZrm, 0 }, { X86::VPADDBZrr, X86::VPADDBZrm, 0 }, { X86::VPADDDZrr, X86::VPADDDZrm, 0 }, { X86::VPADDQZrr, X86::VPADDQZrm, 0 }, @@ -2068,6 +2072,14 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) { X86::VORPDZ256rr, X86::VORPDZ256rm, 0 }, { X86::VORPSZ128rr, X86::VORPSZ128rm, 0 }, { X86::VORPSZ256rr, X86::VORPSZ256rm, 0 }, + { X86::VPACKSSDWZ256rr, X86::VPACKSSDWZ256rm, 0 }, + { X86::VPACKSSDWZ128rr, X86::VPACKSSDWZ128rm, 0 }, + { X86::VPACKSSWBZ256rr, X86::VPACKSSWBZ256rm, 0 }, + { X86::VPACKSSWBZ128rr, X86::VPACKSSWBZ128rm, 0 }, + { X86::VPACKUSDWZ256rr, X86::VPACKUSDWZ256rm, 0 }, + { X86::VPACKUSDWZ128rr, X86::VPACKUSDWZ128rm, 0 }, + { X86::VPACKUSWBZ256rr, X86::VPACKUSWBZ256rm, 0 }, + { X86::VPACKUSWBZ128rr, X86::VPACKUSWBZ128rm, 0 }, { X86::VPADDBZ128rr, X86::VPADDBZ128rm, 0 }, { X86::VPADDBZ256rr, X86::VPADDBZ256rm, 0 }, { X86::VPADDDZ128rr, X86::VPADDDZ128rm, 0 }, @@ -2562,6 +2574,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) { X86::VMULSSZrr_Intkz, X86::VMULSSZrm_Intkz, TB_NO_REVERSE }, { X86::VORPDZrrkz, X86::VORPDZrmkz, 0 }, { X86::VORPSZrrkz, X86::VORPSZrmkz, 0 }, + { X86::VPACKSSDWZrrkz, X86::VPACKSSDWZrmkz, 0 }, + { X86::VPACKSSWBZrrkz, X86::VPACKSSWBZrmkz, 0 }, + { X86::VPACKUSDWZrrkz, X86::VPACKUSDWZrmkz, 0 }, + { X86::VPACKUSWBZrrkz, X86::VPACKUSWBZrmkz, 0 }, { X86::VPADDBZrrkz, X86::VPADDBZrmkz, 0 }, { X86::VPADDDZrrkz, X86::VPADDDZrmkz, 0 }, { X86::VPADDQZrrkz, X86::VPADDQZrmkz, 0 }, @@ -2687,6 +2703,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) { X86::VMULPSZ256rrkz, X86::VMULPSZ256rmkz, 0 }, { X86::VORPDZ256rrkz, X86::VORPDZ256rmkz, 0 }, { X86::VORPSZ256rrkz, X86::VORPSZ256rmkz, 0 }, + { X86::VPACKSSDWZ256rrkz, X86::VPACKSSDWZ256rmkz, 0 }, + { X86::VPACKSSWBZ256rrkz, X86::VPACKSSWBZ256rmkz, 0 }, + { X86::VPACKUSDWZ256rrkz, X86::VPACKUSDWZ256rmkz, 0 }, + { X86::VPACKUSWBZ256rrkz, X86::VPACKUSWBZ256rmkz, 0 }, { X86::VPADDBZ256rrkz, X86::VPADDBZ256rmkz, 0 }, { X86::VPADDDZ256rrkz, X86::VPADDDZ256rmkz, 0 }, { X86::VPADDQZ256rrkz, X86::VPADDQZ256rmkz, 0 }, @@ -2806,6 +2826,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) { X86::VMULPSZ128rrkz, X86::VMULPSZ128rmkz, 0 }, { X86::VORPDZ128rrkz, X86::VORPDZ128rmkz, 0 }, { X86::VORPSZ128rrkz, X86::VORPSZ128rmkz, 0 }, + { X86::VPACKSSDWZ128rrkz, X86::VPACKSSDWZ128rmkz, 0 }, + { X86::VPACKSSWBZ128rrkz, X86::VPACKSSWBZ128rmkz, 0 }, + { X86::VPACKUSDWZ128rrkz, X86::VPACKUSDWZ128rmkz, 0 }, + { X86::VPACKUSWBZ128rrkz, X86::VPACKUSWBZ128rmkz, 0 }, { X86::VPADDBZ128rrkz, X86::VPADDBZ128rmkz, 0 }, { X86::VPADDDZ128rrkz, X86::VPADDDZ128rmkz, 0 }, { X86::VPADDQZ128rrkz, X86::VPADDQZ128rmkz, 0 }, @@ -3069,6 +3093,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) { X86::VMULSSZrr_Intk, X86::VMULSSZrm_Intk, TB_NO_REVERSE }, { X86::VORPDZrrk, X86::VORPDZrmk, 0 }, { X86::VORPSZrrk, X86::VORPSZrmk, 0 }, + { X86::VPACKSSDWZrrk, X86::VPACKSSDWZrmk, 0 }, + { X86::VPACKSSWBZrrk, X86::VPACKSSWBZrmk, 0 }, + { X86::VPACKUSDWZrrk, X86::VPACKUSDWZrmk, 0 }, + { X86::VPACKUSWBZrrk, X86::VPACKUSWBZrmk, 0 }, { X86::VPADDBZrrk, X86::VPADDBZrmk, 0 }, { X86::VPADDDZrrk, X86::VPADDDZrmk, 0 }, { X86::VPADDQZrrk, X86::VPADDQZrmk, 0 }, @@ -3207,6 +3235,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) { X86::VMULPSZ256rrk, X86::VMULPSZ256rmk, 0 }, { X86::VORPDZ256rrk, X86::VORPDZ256rmk, 0 }, { X86::VORPSZ256rrk, X86::VORPSZ256rmk, 0 }, + { X86::VPACKSSDWZ256rrk, X86::VPACKSSDWZ256rmk, 0 }, + { X86::VPACKSSWBZ256rrk, X86::VPACKSSWBZ256rmk, 0 }, + { X86::VPACKUSDWZ256rrk, X86::VPACKUSDWZ256rmk, 0 }, + { X86::VPACKUSWBZ256rrk, X86::VPACKUSWBZ256rmk, 0 }, { X86::VPADDBZ256rrk, X86::VPADDBZ256rmk, 0 }, { X86::VPADDDZ256rrk, X86::VPADDDZ256rmk, 0 }, { X86::VPADDQZ256rrk, X86::VPADDQZ256rmk, 0 }, @@ -3340,6 +3372,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) { X86::VMULPSZ128rrk, X86::VMULPSZ128rmk, 0 }, { X86::VORPDZ128rrk, X86::VORPDZ128rmk, 0 }, { X86::VORPSZ128rrk, X86::VORPSZ128rmk, 0 }, + { X86::VPACKSSDWZ128rrk, X86::VPACKSSDWZ128rmk, 0 }, + { X86::VPACKSSWBZ128rrk, X86::VPACKSSWBZ128rmk, 0 }, + { X86::VPACKUSDWZ128rrk, X86::VPACKUSDWZ128rmk, 0 }, + { X86::VPACKUSWBZ128rrk, X86::VPACKUSWBZ128rmk, 0 }, { X86::VPADDBZ128rrk, X86::VPADDBZ128rmk, 0 }, { X86::VPADDDZ128rrk, X86::VPADDDZ128rmk, 0 }, { X86::VPADDQZ128rrk, X86::VPADDQZ128rmk, 0 }, diff --git a/test/CodeGen/X86/stack-folding-int-avx512.ll b/test/CodeGen/X86/stack-folding-int-avx512.ll index df8b54e4875..04a7d115901 100644 --- a/test/CodeGen/X86/stack-folding-int-avx512.ll +++ b/test/CodeGen/X86/stack-folding-int-avx512.ll @@ -278,6 +278,59 @@ define <32 x i16> @stack_fold_pabsw_maskz(<32 x i16> %a0, i32 %mask) { ret <32 x i16> %2 } +define <32 x i16> @stack_fold_packssdw(<16 x i32> %a0, <16 x i32> %a1) { + ;CHECK-LABEL: stack_fold_packssdw + ;CHECK: vpackssdw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{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.packssdw.512(<16 x i32> %a0, <16 x i32> %a1, <32 x i16> undef, i32 -1) + ret <32 x i16> %2 +} +declare <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32>, <16 x i32>, <32 x i16>, i32) nounwind readnone + +define <64 x i8> @stack_fold_packsswb(<32 x i16> %a0, <32 x i16> %a1) { + ;CHECK-LABEL: stack_fold_packsswb + ;CHECK: vpacksswb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{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.packsswb.512(<32 x i16> %a0, <32 x i16> %a1, <64 x i8> undef, i64 -1) + ret <64 x i8> %2 +} +declare <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16>, <32 x i16>, <64 x i8>, i64) nounwind readnone + +define <32 x i16> @stack_fold_packusdw(<16 x i32> %a0, <16 x i32> %a1) { + ;CHECK-LABEL: stack_fold_packusdw + ;CHECK: vpackusdw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{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.packusdw.512(<16 x i32> %a0, <16 x i32> %a1, <32 x i16> undef, i32 -1) + ret <32 x i16> %2 +} +declare <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32>, <16 x i32>, <32 x i16>, i32) nounwind readnone + +define <32 x i16> @stack_fold_packusdw_mask(<32 x i16>* %passthru, <16 x i32> %a0, <16 x i32> %a1, i32 %mask) { + ;CHECK-LABEL: stack_fold_packusdw_mask + ;CHECK: vpackusdw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{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 = load <32 x i16>, <32 x i16>* %passthru + %3 = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a0, <16 x i32> %a1, <32 x i16> %2, i32 %mask) + ret <32 x i16> %3 +} + +define <32 x i16> @stack_fold_packusdw_maskz(<16 x i32> %a0, <16 x i32> %a1, i32 %mask) { + ;CHECK-LABEL: stack_fold_packusdw_maskz + ;CHECK: vpackusdw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{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.packusdw.512(<16 x i32> %a0, <16 x i32> %a1, <32 x i16> zeroinitializer, i32 %mask) + ret <32 x i16> %2 +} + +define <64 x i8> @stack_fold_packuswb(<32 x i16> %a0, <32 x i16> %a1) { + ;CHECK-LABEL: stack_fold_packuswb + ;CHECK: vpackuswb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{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.packuswb.512(<32 x i16> %a0, <32 x i16> %a1, <64 x i8> undef, i64 -1) + ret <64 x i8> %2 +} +declare <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16>, <32 x i16>, <64 x i8>, i64) nounwind readnone + 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 4c7c0daf4d1..7ce798f778a 100644 --- a/test/CodeGen/X86/stack-folding-int-avx512vl.ll +++ b/test/CodeGen/X86/stack-folding-int-avx512vl.ll @@ -193,6 +193,78 @@ define <16 x i16> @stack_fold_pabsw_ymm(<16 x i16> %a0) { } declare <16 x i16> @llvm.x86.avx2.pabs.w(<16 x i16>) nounwind readnone +define <8 x i16> @stack_fold_packssdw(<4 x i32> %a0, <4 x i32> %a1) { + ;CHECK-LABEL: stack_fold_packssdw + ;CHECK: vpackssdw {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{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.sse2.packssdw.128(<4 x i32> %a0, <4 x i32> %a1) + ret <8 x i16> %2 +} +declare <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32>, <4 x i32>) nounwind readnone + +define <16 x i16> @stack_fold_packssdw_ymm(<8 x i32> %a0, <8 x i32> %a1) { + ;CHECK-LABEL: stack_fold_packssdw_ymm + ;CHECK: vpackssdw {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}}, {{%ymm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{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.packssdw(<8 x i32> %a0, <8 x i32> %a1) + ret <16 x i16> %2 +} +declare <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32>, <8 x i32>) nounwind readnone + +define <16 x i8> @stack_fold_packsswb(<8 x i16> %a0, <8 x i16> %a1) { + ;CHECK-LABEL: stack_fold_packsswb + ;CHECK: vpacksswb {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{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.sse2.packsswb.128(<8 x i16> %a0, <8 x i16> %a1) + ret <16 x i8> %2 +} +declare <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16>, <8 x i16>) nounwind readnone + +define <32 x i8> @stack_fold_packsswb_ymm(<16 x i16> %a0, <16 x i16> %a1) { + ;CHECK-LABEL: stack_fold_packsswb_ymm + ;CHECK: vpacksswb {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}}, {{%ymm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{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.packsswb(<16 x i16> %a0, <16 x i16> %a1) + ret <32 x i8> %2 +} +declare <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16>, <16 x i16>) nounwind readnone + +define <8 x i16> @stack_fold_packusdw(<4 x i32> %a0, <4 x i32> %a1) { + ;CHECK-LABEL: stack_fold_packusdw + ;CHECK: vpackusdw {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{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.sse41.packusdw(<4 x i32> %a0, <4 x i32> %a1) + ret <8 x i16> %2 +} +declare <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32>, <4 x i32>) nounwind readnone + +define <16 x i16> @stack_fold_packusdw_ymm(<8 x i32> %a0, <8 x i32> %a1) { + ;CHECK-LABEL: stack_fold_packusdw_ymm + ;CHECK: vpackusdw {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}}, {{%ymm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{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.packusdw(<8 x i32> %a0, <8 x i32> %a1) + ret <16 x i16> %2 +} +declare <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32>, <8 x i32>) nounwind readnone + +define <16 x i8> @stack_fold_packuswb(<8 x i16> %a0, <8 x i16> %a1) { + ;CHECK-LABEL: stack_fold_packuswb + ;CHECK: vpackuswb {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{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.sse2.packuswb.128(<8 x i16> %a0, <8 x i16> %a1) + ret <16 x i8> %2 +} +declare <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16>, <8 x i16>) nounwind readnone + +define <32 x i8> @stack_fold_packuswb_ymm(<16 x i16> %a0, <16 x i16> %a1) { + ;CHECK-LABEL: stack_fold_packuswb_ymm + ;CHECK: vpackuswb {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}}, {{%ymm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload + %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{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.packuswb(<16 x i16> %a0, <16 x i16> %a1) + ret <32 x i8> %2 +} +declare <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16>, <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