diff options
| -rw-r--r-- | llvm/lib/Target/X86/X86InstrInfo.cpp | 44 | ||||
| -rw-r--r-- | llvm/test/CodeGen/X86/stack-folding-int-avx512.ll | 182 | ||||
| -rw-r--r-- | llvm/test/CodeGen/X86/stack-folding-int-avx512vl.ll | 272 |
3 files changed, 498 insertions, 0 deletions
diff --git a/llvm/lib/Target/X86/X86InstrInfo.cpp b/llvm/lib/Target/X86/X86InstrInfo.cpp index ef2dda319d7..eeab745c7d6 100644 --- a/llvm/lib/Target/X86/X86InstrInfo.cpp +++ b/llvm/lib/Target/X86/X86InstrInfo.cpp @@ -1742,8 +1742,14 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) { X86::VMINSSZrr, X86::VMINSSZrm, 0 }, { X86::VMINCSSZrr, X86::VMINCSSZrm, 0 }, { X86::VMINSSZrr_Int, X86::VMINSSZrm_Int, 0 }, + { X86::VPADDBZrr, X86::VPADDBZrm, 0 }, { X86::VPADDDZrr, X86::VPADDDZrm, 0 }, { X86::VPADDQZrr, X86::VPADDQZrm, 0 }, + { X86::VPADDSBZrr, X86::VPADDSBZrm, 0 }, + { X86::VPADDSWZrr, X86::VPADDSWZrm, 0 }, + { X86::VPADDUSBZrr, X86::VPADDUSBZrm, 0 }, + { X86::VPADDUSWZrr, X86::VPADDUSWZrm, 0 }, + { X86::VPADDWZrr, X86::VPADDWZrm, 0 }, { X86::VPERMPDZri, X86::VPERMPDZmi, 0 }, { X86::VPERMPSZrr, X86::VPERMPSZrm, 0 }, { X86::VPMAXSDZrr, X86::VPMAXSDZrm, 0 }, @@ -1760,8 +1766,14 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) { X86::VPSRAVDZrr, X86::VPSRAVDZrm, 0 }, { X86::VPSRLVDZrr, X86::VPSRLVDZrm, 0 }, { X86::VPSRLVQZrr, X86::VPSRLVQZrm, 0 }, + { X86::VPSUBBZrr, X86::VPSUBBZrm, 0 }, { X86::VPSUBDZrr, X86::VPSUBDZrm, 0 }, { X86::VPSUBQZrr, X86::VPSUBQZrm, 0 }, + { X86::VPSUBSBZrr, X86::VPSUBSBZrm, 0 }, + { X86::VPSUBSWZrr, X86::VPSUBSWZrm, 0 }, + { X86::VPSUBUSBZrr, X86::VPSUBUSBZrm, 0 }, + { X86::VPSUBUSWZrr, X86::VPSUBUSWZrm, 0 }, + { X86::VPSUBWZrr, X86::VPSUBWZrm, 0 }, { X86::VSHUFPDZrri, X86::VSHUFPDZrmi, 0 }, { X86::VSHUFPSZrri, X86::VSHUFPSZrmi, 0 }, { X86::VALIGNQZrri, X86::VALIGNQZrmi, 0 }, @@ -1860,6 +1872,38 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI) { X86::VPCMPGTQZ256rr, X86::VPCMPGTQZ256rm, 0 }, { X86::VPCMPGTWZ128rr, X86::VPCMPGTWZ128rm, 0 }, { X86::VPCMPGTWZ256rr, X86::VPCMPGTWZ256rm, 0 }, + { X86::VPADDBZ128rr, X86::VPADDBZ128rm, 0 }, + { X86::VPADDBZ256rr, X86::VPADDBZ256rm, 0 }, + { X86::VPADDDZ128rr, X86::VPADDDZ128rm, 0 }, + { X86::VPADDDZ256rr, X86::VPADDDZ256rm, 0 }, + { X86::VPADDQZ128rr, X86::VPADDQZ128rm, 0 }, + { X86::VPADDQZ256rr, X86::VPADDQZ256rm, 0 }, + { X86::VPADDSBZ128rr, X86::VPADDSBZ128rm, 0 }, + { X86::VPADDSBZ256rr, X86::VPADDSBZ256rm, 0 }, + { X86::VPADDSWZ128rr, X86::VPADDSWZ128rm, 0 }, + { X86::VPADDSWZ256rr, X86::VPADDSWZ256rm, 0 }, + { X86::VPADDUSBZ128rr, X86::VPADDUSBZ128rm, 0 }, + { X86::VPADDUSBZ256rr, X86::VPADDUSBZ256rm, 0 }, + { X86::VPADDUSWZ128rr, X86::VPADDUSWZ128rm, 0 }, + { X86::VPADDUSWZ256rr, X86::VPADDUSWZ256rm, 0 }, + { X86::VPADDWZ128rr, X86::VPADDWZ128rm, 0 }, + { X86::VPADDWZ256rr, X86::VPADDWZ256rm, 0 }, + { X86::VPSUBBZ128rr, X86::VPSUBBZ128rm, 0 }, + { X86::VPSUBBZ256rr, X86::VPSUBBZ256rm, 0 }, + { X86::VPSUBDZ128rr, X86::VPSUBDZ128rm, 0 }, + { X86::VPSUBDZ256rr, X86::VPSUBDZ256rm, 0 }, + { X86::VPSUBQZ128rr, X86::VPSUBQZ128rm, 0 }, + { X86::VPSUBQZ256rr, X86::VPSUBQZ256rm, 0 }, + { X86::VPSUBSBZ128rr, X86::VPSUBSBZ128rm, 0 }, + { X86::VPSUBSBZ256rr, X86::VPSUBSBZ256rm, 0 }, + { X86::VPSUBSWZ128rr, X86::VPSUBSWZ128rm, 0 }, + { X86::VPSUBSWZ256rr, X86::VPSUBSWZ256rm, 0 }, + { X86::VPSUBUSBZ128rr, X86::VPSUBUSBZ128rm, 0 }, + { X86::VPSUBUSBZ256rr, X86::VPSUBUSBZ256rm, 0 }, + { X86::VPSUBUSWZ128rr, X86::VPSUBUSWZ128rm, 0 }, + { X86::VPSUBUSWZ256rr, X86::VPSUBUSWZ256rm, 0 }, + { X86::VPSUBWZ128rr, X86::VPSUBWZ128rm, 0 }, + { X86::VPSUBWZ256rr, X86::VPSUBWZ256rm, 0 }, // AES foldable instructions { X86::AESDECLASTrr, X86::AESDECLASTrm, TB_ALIGN_16 }, diff --git a/llvm/test/CodeGen/X86/stack-folding-int-avx512.ll b/llvm/test/CodeGen/X86/stack-folding-int-avx512.ll new file mode 100644 index 00000000000..9ab42b40a2b --- /dev/null +++ b/llvm/test/CodeGen/X86/stack-folding-int-avx512.ll @@ -0,0 +1,182 @@ +; RUN: llc -O3 -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw < %s | FileCheck %s + +target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" +target triple = "x86_64-unknown-unknown" + +; Stack reload folding tests. +; +; 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_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 + %1 = tail call <8 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 = add <64 x i8> %a0, %a1 + ret <64 x i8> %2 +} + +define <16 x i32> @stack_fold_paddd(<16 x i32> %a0, <16 x i32> %a1) { + ;CHECK-LABEL: stack_fold_paddd + ;CHECK: vpaddd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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 = add <16 x i32> %a0, %a1 + ret <16 x i32> %2 +} + +define <8 x i64> @stack_fold_paddq(<8 x i64> %a0, <8 x i64> %a1) { + ;CHECK-LABEL: stack_fold_paddq + ;CHECK: vpaddq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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 = add <8 x i64> %a0, %a1 + ret <8 x i64> %2 +} + +define <64 x i8> @stack_fold_paddsb(<64 x i8> %a0, <64 x i8> %a1) { + ;CHECK-LABEL: stack_fold_paddsb + ;CHECK: vpaddsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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.padds.b.512(<64 x i8> %a0, <64 x i8> %a1, <64 x i8> undef, i64 -1) + ret <64 x i8> %2 +} +declare <64 x i8> @llvm.x86.avx512.mask.padds.b.512(<64 x i8>, <64 x i8>, <64 x i8>, i64) nounwind readnone + +define <32 x i16> @stack_fold_paddsw(<32 x i16> %a0, <32 x i16> %a1) { + ;CHECK-LABEL: stack_fold_paddsw + ;CHECK: vpaddsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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.padds.w.512(<32 x i16> %a0, <32 x i16> %a1, <32 x i16> undef, i32 -1) + ret <32 x i16> %2 +} +declare <32 x i16> @llvm.x86.avx512.mask.padds.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32) + +define <64 x i8> @stack_fold_paddusb(<64 x i8> %a0, <64 x i8> %a1) { + ;CHECK-LABEL: stack_fold_paddusb + ;CHECK: vpaddusb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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.paddus.b.512(<64 x i8> %a0, <64 x i8> %a1, <64 x i8> undef, i64 -1) + ret <64 x i8> %2 +} +declare <64 x i8> @llvm.x86.avx512.mask.paddus.b.512(<64 x i8>, <64 x i8>, <64 x i8>, i64) nounwind readnone + +define <32 x i16> @stack_fold_paddusw(<32 x i16> %a0, <32 x i16> %a1) { + ;CHECK-LABEL: stack_fold_paddusw + ;CHECK: vpaddusw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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.paddus.w.512(<32 x i16> %a0, <32 x i16> %a1, <32 x i16> undef, i32 -1) + ret <32 x i16> %2 +} +declare <32 x i16> @llvm.x86.avx512.mask.paddus.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32) + +define <32 x i16> @stack_fold_paddw(<32 x i16> %a0, <32 x i16> %a1) { + ;CHECK-LABEL: stack_fold_paddw + ;CHECK: vpaddw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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 = add <32 x i16> %a0, %a1 + ret <32 x i16> %2 +} + +define i64 @stack_fold_pcmpeqb(<64 x i8> %a0, <64 x i8> %a1) { + ;CHECK-LABEL: stack_fold_pcmpeqb + ;CHECK: vpcmpeqb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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 = icmp eq <64 x i8> %a0, %a1 + %3 = bitcast <64 x i1> %2 to i64 + ret i64 %3 +} + +define i16 @stack_fold_pcmpeqd(<16 x i32> %a0, <16 x i32> %a1) { + ;CHECK-LABEL: stack_fold_pcmpeqd + ;CHECK: vpcmpeqd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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 = icmp eq <16 x i32> %a0, %a1 + %3 = bitcast <16 x i1> %2 to i16 + ret i16 %3 +} + +define i8 @stack_fold_pcmpeqq(<8 x i64> %a0, <8 x i64> %a1) { + ;CHECK-LABEL: stack_fold_pcmpeqq + ;CHECK: vpcmpeqq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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 = icmp eq <8 x i64> %a0, %a1 + %3 = bitcast <8 x i1> %2 to i8 + ret i8 %3 +} + +define i32 @stack_fold_pcmpeqw(<32 x i16> %a0, <32 x i16> %a1) { + ;CHECK-LABEL: stack_fold_pcmpeqw + ;CHECK: vpcmpeqw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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 = icmp eq <32 x i16> %a0, %a1 + %3 = bitcast <32 x i1> %2 to i32 + ret i32 %3 +} + +define <64 x i8> @stack_fold_psubb(<64 x i8> %a0, <64 x i8> %a1) { + ;CHECK-LABEL: stack_fold_psubb + ;CHECK: vpsubb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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 = sub <64 x i8> %a0, %a1 + ret <64 x i8> %2 +} + +define <16 x i32> @stack_fold_psubd(<16 x i32> %a0, <16 x i32> %a1) { + ;CHECK-LABEL: stack_fold_psubd + ;CHECK: vpsubd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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 = sub <16 x i32> %a0, %a1 + ret <16 x i32> %2 +} + +define <8 x i64> @stack_fold_psubq(<8 x i64> %a0, <8 x i64> %a1) { + ;CHECK-LABEL: stack_fold_psubq + ;CHECK: vpsubq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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 = sub <8 x i64> %a0, %a1 + ret <8 x i64> %2 +} + +define <64 x i8> @stack_fold_psubsb(<64 x i8> %a0, <64 x i8> %a1) { + ;CHECK-LABEL: stack_fold_psubsb + ;CHECK: vpsubsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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.psubs.b.512(<64 x i8> %a0, <64 x i8> %a1, <64 x i8> undef, i64 -1) + ret <64 x i8> %2 +} +declare <64 x i8> @llvm.x86.avx512.mask.psubs.b.512(<64 x i8>, <64 x i8>, <64 x i8>, i64) nounwind readnone + +define <32 x i16> @stack_fold_psubsw(<32 x i16> %a0, <32 x i16> %a1) { + ;CHECK-LABEL: stack_fold_psubsw + ;CHECK: vpsubsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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.psubs.w.512(<32 x i16> %a0, <32 x i16> %a1, <32 x i16> undef, i32 -1) + ret <32 x i16> %2 +} +declare <32 x i16> @llvm.x86.avx512.mask.psubs.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32) + +define <64 x i8> @stack_fold_psubusb(<64 x i8> %a0, <64 x i8> %a1) { + ;CHECK-LABEL: stack_fold_psubusb + ;CHECK: vpsubusb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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.psubus.b.512(<64 x i8> %a0, <64 x i8> %a1, <64 x i8> undef, i64 -1) + ret <64 x i8> %2 +} +declare <64 x i8> @llvm.x86.avx512.mask.psubus.b.512(<64 x i8>, <64 x i8>, <64 x i8>, i64) nounwind readnone + +define <32 x i16> @stack_fold_psubusw(<32 x i16> %a0, <32 x i16> %a1) { + ;CHECK-LABEL: stack_fold_psubusw + ;CHECK: vpsubusw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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.psubus.w.512(<32 x i16> %a0, <32 x i16> %a1, <32 x i16> undef, i32 -1) + ret <32 x i16> %2 +} +declare <32 x i16> @llvm.x86.avx512.mask.psubus.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32) + +define <32 x i16> @stack_fold_psubw(<32 x i16> %a0, <32 x i16> %a1) { + ;CHECK-LABEL: stack_fold_psubw + ;CHECK: vpsubw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload + %1 = tail call <8 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 = sub <32 x i16> %a0, %a1 + ret <32 x i16> %2 +} + diff --git a/llvm/test/CodeGen/X86/stack-folding-int-avx512vl.ll b/llvm/test/CodeGen/X86/stack-folding-int-avx512vl.ll index f92f9593f13..3e145955e38 100644 --- a/llvm/test/CodeGen/X86/stack-folding-int-avx512vl.ll +++ b/llvm/test/CodeGen/X86/stack-folding-int-avx512vl.ll @@ -8,6 +8,142 @@ 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_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 + %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 = add <16 x i8> %a0, %a1 + ret <16 x i8> %2 +} + +define <32 x i8> @stack_fold_paddb_ymm(<32 x i8> %a0, <32 x i8> %a1) { + ;CHECK-LABEL: stack_fold_paddb_ymm + ;CHECK: vpaddb {{-?[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 = add <32 x i8> %a0, %a1 + ret <32 x i8> %2 +} + +define <4 x i32> @stack_fold_paddd(<4 x i32> %a0, <4 x i32> %a1) { + ;CHECK-LABEL: stack_fold_paddd + ;CHECK: vpaddd {{-?[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 = add <4 x i32> %a0, %a1 + ret <4 x i32> %2 +} + +define <8 x i32> @stack_fold_paddd_ymm(<8 x i32> %a0, <8 x i32> %a1) { + ;CHECK-LABEL: stack_fold_paddd_ymm + ;CHECK: vpaddd {{-?[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 = add <8 x i32> %a0, %a1 + ret <8 x i32> %2 +} + +define <2 x i64> @stack_fold_paddq(<2 x i64> %a0, <2 x i64> %a1) { + ;CHECK-LABEL: stack_fold_paddq + ;CHECK: vpaddq {{-?[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 = add <2 x i64> %a0, %a1 + ret <2 x i64> %2 +} + +define <4 x i64> @stack_fold_paddq_ymm(<4 x i64> %a0, <4 x i64> %a1) { + ;CHECK-LABEL: stack_fold_paddq_ymm + ;CHECK: vpaddq {{-?[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 = add <4 x i64> %a0, %a1 + ret <4 x i64> %2 +} + +define <16 x i8> @stack_fold_paddsb(<16 x i8> %a0, <16 x i8> %a1) { + ;CHECK-LABEL: stack_fold_paddsb + ;CHECK: vpaddsb {{-?[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.padds.b(<16 x i8> %a0, <16 x i8> %a1) + ret <16 x i8> %2 +} +declare <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8>, <16 x i8>) nounwind readnone + +define <32 x i8> @stack_fold_paddsb_ymm(<32 x i8> %a0, <32 x i8> %a1) { + ;CHECK-LABEL: stack_fold_paddsb_ymm + ;CHECK: vpaddsb {{-?[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.padds.b(<32 x i8> %a0, <32 x i8> %a1) + ret <32 x i8> %2 +} +declare <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8>, <32 x i8>) nounwind readnone + +define <8 x i16> @stack_fold_paddsw(<8 x i16> %a0, <8 x i16> %a1) { + ;CHECK-LABEL: stack_fold_paddsw + ;CHECK: vpaddsw {{-?[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.padds.w(<8 x i16> %a0, <8 x i16> %a1) + ret <8 x i16> %2 +} +declare <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16>, <8 x i16>) nounwind readnone + +define <16 x i16> @stack_fold_paddsw_ymm(<16 x i16> %a0, <16 x i16> %a1) { + ;CHECK-LABEL: stack_fold_paddsw_ymm + ;CHECK: vpaddsw {{-?[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.padds.w(<16 x i16> %a0, <16 x i16> %a1) + ret <16 x i16> %2 +} +declare <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16>, <16 x i16>) nounwind readnone + +define <16 x i8> @stack_fold_paddusb(<16 x i8> %a0, <16 x i8> %a1) { + ;CHECK-LABEL: stack_fold_paddusb + ;CHECK: vpaddusb {{-?[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.paddus.b(<16 x i8> %a0, <16 x i8> %a1) + ret <16 x i8> %2 +} +declare <16 x i8> @llvm.x86.sse2.paddus.b(<16 x i8>, <16 x i8>) nounwind readnone + +define <32 x i8> @stack_fold_paddusb_ymm(<32 x i8> %a0, <32 x i8> %a1) { + ;CHECK-LABEL: stack_fold_paddusb_ymm + ;CHECK: vpaddusb {{-?[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.paddus.b(<32 x i8> %a0, <32 x i8> %a1) + ret <32 x i8> %2 +} +declare <32 x i8> @llvm.x86.avx2.paddus.b(<32 x i8>, <32 x i8>) nounwind readnone + +define <8 x i16> @stack_fold_paddusw(<8 x i16> %a0, <8 x i16> %a1) { + ;CHECK-LABEL: stack_fold_paddusw + ;CHECK: vpaddusw {{-?[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.paddus.w(<8 x i16> %a0, <8 x i16> %a1) + ret <8 x i16> %2 +} +declare <8 x i16> @llvm.x86.sse2.paddus.w(<8 x i16>, <8 x i16>) nounwind readnone + +define <16 x i16> @stack_fold_paddusw_ymm(<16 x i16> %a0, <16 x i16> %a1) { + ;CHECK-LABEL: stack_fold_paddusw_ymm + ;CHECK: vpaddusw {{-?[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.paddus.w(<16 x i16> %a0, <16 x i16> %a1) + ret <16 x i16> %2 +} +declare <16 x i16> @llvm.x86.avx2.paddus.w(<16 x i16>, <16 x i16>) nounwind readnone + +define <8 x i16> @stack_fold_paddw(<8 x i16> %a0, <8 x i16> %a1) { + ;CHECK-LABEL: stack_fold_paddw + ;CHECK: vpaddw {{-?[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 = add <8 x i16> %a0, %a1 + ret <8 x i16> %2 +} + +define <16 x i16> @stack_fold_paddw_ymm(<16 x i16> %a0, <16 x i16> %a1) { + ;CHECK-LABEL: stack_fold_paddw_ymm + ;CHECK: vpaddw {{-?[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 = add <16 x i16> %a0, %a1 + ret <16 x i16> %2 +} + define i16 @stack_fold_pcmpeqb(<16 x i8> %a0, <16 x i8> %a1) { ;CHECK-LABEL: stack_fold_pcmpeqb ;CHECK: vpcmpeqb {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 16-byte Folded Reload @@ -46,3 +182,139 @@ define i8 @stack_fold_pcmpeqw(<8 x i16> %a0, <8 x i16> %a1) { ret i8 %3 } +define <16 x i8> @stack_fold_psubb(<16 x i8> %a0, <16 x i8> %a1) { + ;CHECK-LABEL: stack_fold_psubb + ;CHECK: vpsubb {{-?[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 = sub <16 x i8> %a0, %a1 + ret <16 x i8> %2 +} + +define <32 x i8> @stack_fold_psubb_ymm(<32 x i8> %a0, <32 x i8> %a1) { + ;CHECK-LABEL: stack_fold_psubb_ymm + ;CHECK: vpsubb {{-?[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 = sub <32 x i8> %a0, %a1 + ret <32 x i8> %2 +} + +define <4 x i32> @stack_fold_psubd(<4 x i32> %a0, <4 x i32> %a1) { + ;CHECK-LABEL: stack_fold_psubd + ;CHECK: vpsubd {{-?[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 = sub <4 x i32> %a0, %a1 + ret <4 x i32> %2 +} + +define <8 x i32> @stack_fold_psubd_ymm(<8 x i32> %a0, <8 x i32> %a1) { + ;CHECK-LABEL: stack_fold_psubd_ymm + ;CHECK: vpsubd {{-?[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 = sub <8 x i32> %a0, %a1 + ret <8 x i32> %2 +} + +define <2 x i64> @stack_fold_psubq(<2 x i64> %a0, <2 x i64> %a1) { + ;CHECK-LABEL: stack_fold_psubq + ;CHECK: vpsubq {{-?[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 = sub <2 x i64> %a0, %a1 + ret <2 x i64> %2 +} + +define <4 x i64> @stack_fold_psubq_ymm(<4 x i64> %a0, <4 x i64> %a1) { + ;CHECK-LABEL: stack_fold_psubq_ymm + ;CHECK: vpsubq {{-?[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 = sub <4 x i64> %a0, %a1 + ret <4 x i64> %2 +} + +define <16 x i8> @stack_fold_psubsb(<16 x i8> %a0, <16 x i8> %a1) { + ;CHECK-LABEL: stack_fold_psubsb + ;CHECK: vpsubsb {{-?[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.psubs.b(<16 x i8> %a0, <16 x i8> %a1) + ret <16 x i8> %2 +} +declare <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8>, <16 x i8>) nounwind readnone + +define <32 x i8> @stack_fold_psubsb_ymm(<32 x i8> %a0, <32 x i8> %a1) { + ;CHECK-LABEL: stack_fold_psubsb_ymm + ;CHECK: vpsubsb {{-?[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.psubs.b(<32 x i8> %a0, <32 x i8> %a1) + ret <32 x i8> %2 +} +declare <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8>, <32 x i8>) nounwind readnone + +define <8 x i16> @stack_fold_psubsw(<8 x i16> %a0, <8 x i16> %a1) { + ;CHECK-LABEL: stack_fold_psubsw + ;CHECK: vpsubsw {{-?[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.psubs.w(<8 x i16> %a0, <8 x i16> %a1) + ret <8 x i16> %2 +} +declare <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16>, <8 x i16>) nounwind readnone + +define <16 x i16> @stack_fold_psubsw_ymm(<16 x i16> %a0, <16 x i16> %a1) { + ;CHECK-LABEL: stack_fold_psubsw_ymm + ;CHECK: vpsubsw {{-?[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.psubs.w(<16 x i16> %a0, <16 x i16> %a1) + ret <16 x i16> %2 +} +declare <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16>, <16 x i16>) nounwind readnone + +define <16 x i8> @stack_fold_psubusb(<16 x i8> %a0, <16 x i8> %a1) { + ;CHECK-LABEL: stack_fold_psubusb + ;CHECK: vpsubusb {{-?[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.psubus.b(<16 x i8> %a0, <16 x i8> %a1) + ret <16 x i8> %2 +} +declare <16 x i8> @llvm.x86.sse2.psubus.b(<16 x i8>, <16 x i8>) nounwind readnone + +define <32 x i8> @stack_fold_psubusb_ymm(<32 x i8> %a0, <32 x i8> %a1) { + ;CHECK-LABEL: stack_fold_psubusb_ymm + ;CHECK: vpsubusb {{-?[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.psubus.b(<32 x i8> %a0, <32 x i8> %a1) + ret <32 x i8> %2 +} +declare <32 x i8> @llvm.x86.avx2.psubus.b(<32 x i8>, <32 x i8>) nounwind readnone + +define <8 x i16> @stack_fold_psubusw(<8 x i16> %a0, <8 x i16> %a1) { + ;CHECK-LABEL: stack_fold_psubusw + ;CHECK: vpsubusw {{-?[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.psubus.w(<8 x i16> %a0, <8 x i16> %a1) + ret <8 x i16> %2 +} +declare <8 x i16> @llvm.x86.sse2.psubus.w(<8 x i16>, <8 x i16>) nounwind readnone + +define <16 x i16> @stack_fold_psubusw_ymm(<16 x i16> %a0, <16 x i16> %a1) { + ;CHECK-LABEL: stack_fold_psubusw_ymm + ;CHECK: vpsubusw {{-?[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.psubus.w(<16 x i16> %a0, <16 x i16> %a1) + ret <16 x i16> %2 +} +declare <16 x i16> @llvm.x86.avx2.psubus.w(<16 x i16>, <16 x i16>) nounwind readnone + +define <8 x i16> @stack_fold_psubw(<8 x i16> %a0, <8 x i16> %a1) { + ;CHECK-LABEL: stack_fold_psubw + ;CHECK: vpsubw {{-?[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 = sub <8 x i16> %a0, %a1 + ret <8 x i16> %2 +} + +define <16 x i16> @stack_fold_psubw_ymm(<16 x i16> %a0, <16 x i16> %a1) { + ;CHECK-LABEL: stack_fold_psubw_ymm + ;CHECK: vpsubw {{-?[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 = sub <16 x i16> %a0, %a1 + ret <16 x i16> %2 +} + |
