This is an archive of the discontinued LLVM Phabricator instance.

Differential D55655

[DAGCombiner] after simplifying demanded elements of vector operand of extract, revisit the extract
ClosedPublic

Authored by spatel on Dec 13 2018, 7:26 AM.

Details

Reviewers
andreadb
craig.topper
RKSimon
Commits
rG791ae69afe1f: [DAGCombiner] after simplifying demanded elements of vector operand of extract…
rGa7b115b392bc: [DAGCombiner] after simplifying demanded elements of vector operand of extract…
rL349058: [DAGCombiner] after simplifying demanded elements of vector operand of extract…
rL349051: [DAGCombiner] after simplifying demanded elements of vector operand of extract…
Summary

I noticed this as part of a patch for PR39973:
https://bugs.llvm.org/show_bug.cgi?id=39973

And there might be some interplay with D55600, but I haven't checked yet. I think this bug is unique to extract_element, but I'll do an audit to be sure.

These are all wins or neutral from what I can tell.

Diff Detail

Repository
rL LLVM

Event Timeline

spatel created this revision.Dec 13 2018, 7:26 AM
Herald added a subscriber: mcrosier. · View Herald TranscriptDec 13 2018, 7:26 AM
andreadb accepted this revision.Dec 13 2018, 7:33 AM

LGTM

This revision is now accepted and ready to land.Dec 13 2018, 7:33 AM
Closed by commit rL349051: [DAGCombiner] after simplifying demanded elements of vector operand of extract… (authored by spatel). · Explain WhyDec 13 2018, 7:47 AM
This revision was automatically updated to reflect the committed changes.
spatel reopened this revision.EditedDec 13 2018, 8:38 AM

Reverted at rL349056 because:

==42943==ERROR: AddressSanitizer: use-after-poison on address 0x6210000f2a10 at pc 0x0000056013aa bp 0x7ffce91a7550 sp 0x7ffce91a7548
READ of size 8 at 0x6210000f2a10 thread T0
    #0 0x56013a9 in use_empty /b/sanitizer-x86_64-linux-fast/build/llvm/include/llvm/CodeGen/SelectionDAGNodes.h:703:35
    #1 0x56013a9 in visitEXTRACT_VECTOR_ELT /b/sanitizer-x86_64-linux-fast/build/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp:15704

What is the proper way to check for node deadness here?

This revision is now accepted and ready to land.Dec 13 2018, 8:38 AM
Closed by commit rL349058: [DAGCombiner] after simplifying demanded elements of vector operand of extract… (authored by spatel). · Explain WhyDec 13 2018, 9:08 AM
This revision was automatically updated to reflect the committed changes.

Revision Contents

PathSize
llvm/
trunk/
lib/
CodeGen/
SelectionDAG/
7 lines
test/
CodeGen/
X86/
3 lines
5 lines
2 lines
2 lines
6 lines
8 lines
2 lines
10 lines

Diff 178080

llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 15,692 Lines • ▼ Show 20 Lines if (llvm::all_of(InVec->uses(), [&](SDNode *Use) {
isa<ConstantSDNode>(Use->getOperand(1)); isa<ConstantSDNode>(Use->getOperand(1));
})) { })) {
APInt DemandedElts = APInt::getNullValue(VT.getVectorNumElements()); APInt DemandedElts = APInt::getNullValue(VT.getVectorNumElements());
for (SDNode *Use : InVec->uses()) { for (SDNode *Use : InVec->uses()) {
auto *CstElt = cast<ConstantSDNode>(Use->getOperand(1)); auto *CstElt = cast<ConstantSDNode>(Use->getOperand(1));
if (CstElt->getAPIntValue().ult(VT.getVectorNumElements())) if (CstElt->getAPIntValue().ult(VT.getVectorNumElements()))
DemandedElts.setBit(CstElt->getZExtValue()); DemandedElts.setBit(CstElt->getZExtValue());
} }
if (SimplifyDemandedVectorElts(InVec, DemandedElts, true)) if (SimplifyDemandedVectorElts(InVec, DemandedElts, true)) {
// We simplified the vector operand of this extract element. If this
// extract is not dead, visit it again so it is folded properly.
if (N->getOpcode() != ISD::DELETED_NODE)
AddToWorklist(N);
return SDValue(N, 0); return SDValue(N, 0);
} }
}
bool BCNumEltsChanged = false; bool BCNumEltsChanged = false;
EVT ExtVT = VT.getVectorElementType(); EVT ExtVT = VT.getVectorElementType();
EVT LVT = ExtVT; EVT LVT = ExtVT;
// If the result of load has to be truncated, then it's not necessarily // If the result of load has to be truncated, then it's not necessarily
// profitable. // profitable.
if (NVT.bitsLT(LVT) && !TLI.isTruncateFree(LVT, NVT)) if (NVT.bitsLT(LVT) && !TLI.isTruncateFree(LVT, NVT))
▲ Show 20 Lines • Show All 3,452 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/X86/extractelement-load.ll

Show First 20 Lines • Show All 87 Lines • ▼ Show 20 Lines
; X32-SSE2-NEXT: movdqa (%eax), %xmm0 ; X32-SSE2-NEXT: movdqa (%eax), %xmm0
; X32-SSE2-NEXT: movd %xmm0, %eax ; X32-SSE2-NEXT: movd %xmm0, %eax
; X32-SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,0,1] ; X32-SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,0,1]
; X32-SSE2-NEXT: movd %xmm0, %edx ; X32-SSE2-NEXT: movd %xmm0, %edx
; X32-SSE2-NEXT: retl ; X32-SSE2-NEXT: retl
; ;
; X64-SSSE3-LABEL: t4: ; X64-SSSE3-LABEL: t4:
; X64-SSSE3: # %bb.0: ; X64-SSSE3: # %bb.0:
; X64-SSSE3-NEXT: movq {{.*#+}} xmm0 = mem[0],zero ; X64-SSSE3-NEXT: movq (%rdi), %rax
; X64-SSSE3-NEXT: movq %xmm0, %rax
; X64-SSSE3-NEXT: retq ; X64-SSSE3-NEXT: retq
; ;
; X64-AVX-LABEL: t4: ; X64-AVX-LABEL: t4:
; X64-AVX: # %bb.0: ; X64-AVX: # %bb.0:
; X64-AVX-NEXT: movq (%rdi), %rax ; X64-AVX-NEXT: movq (%rdi), %rax
; X64-AVX-NEXT: retq ; X64-AVX-NEXT: retq
%b = load <2 x double>, <2 x double>* %a, align 16 %b = load <2 x double>, <2 x double>* %a, align 16
%c = shufflevector <2 x double> %b, <2 x double> %b, <2 x i32> <i32 1, i32 0> %c = shufflevector <2 x double> %b, <2 x double> %b, <2 x i32> <i32 1, i32 0>
%d = bitcast <2 x double> %c to <2 x i64> %d = bitcast <2 x double> %c to <2 x i64>
%e = extractelement <2 x i64> %d, i32 1 %e = extractelement <2 x i64> %d, i32 1
ret i64 %e ret i64 %e
} }

llvm/trunk/test/CodeGen/X86/insertps-combine.ll

Show First 20 Lines • Show All 278 Lines • ▼ Show 20 Lines; AVX-NEXT: retq
%res = call <4 x float> @llvm.x86.sse41.insertps(<4 x float> %a0, <4 x float> %a1, i8 64) %res = call <4 x float> @llvm.x86.sse41.insertps(<4 x float> %a0, <4 x float> %a1, i8 64)
%ext = extractelement <4 x float> %res, i32 0 %ext = extractelement <4 x float> %res, i32 0
ret float %ext ret float %ext
} }
define float @extract_lane_insertps_6123(<4 x float> %a0, <4 x float> *%p1) { define float @extract_lane_insertps_6123(<4 x float> %a0, <4 x float> *%p1) {
; SSE-LABEL: extract_lane_insertps_6123: ; SSE-LABEL: extract_lane_insertps_6123:
; SSE: # %bb.0: ; SSE: # %bb.0:
; SSE-NEXT: movaps (%rdi), %xmm0 ; SSE-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; SSE-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1]
; SSE-NEXT: retq ; SSE-NEXT: retq
; ;
; AVX-LABEL: extract_lane_insertps_6123: ; AVX-LABEL: extract_lane_insertps_6123:
; AVX: # %bb.0: ; AVX: # %bb.0:
; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = mem[1,0] ; AVX-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; AVX-NEXT: retq ; AVX-NEXT: retq
%a1 = load <4 x float>, <4 x float> *%p1 %a1 = load <4 x float>, <4 x float> *%p1
%res = call <4 x float> @llvm.x86.sse41.insertps(<4 x float> %a0, <4 x float> %a1, i8 128) %res = call <4 x float> @llvm.x86.sse41.insertps(<4 x float> %a0, <4 x float> %a1, i8 128)
%ext = extractelement <4 x float> %res, i32 0 %ext = extractelement <4 x float> %res, i32 0
ret float %ext ret float %ext
} }
declare <4 x float> @llvm.x86.sse41.insertps(<4 x float>, <4 x float>, i8) nounwind readnone declare <4 x float> @llvm.x86.sse41.insertps(<4 x float>, <4 x float>, i8) nounwind readnone

llvm/trunk/test/CodeGen/X86/masked_load.ll

Show First 20 Lines • Show All 550 Lines • ▼ Show 20 Lines
; SSE2-NEXT: packssdw %xmm0, %xmm5 ; SSE2-NEXT: packssdw %xmm0, %xmm5
; SSE2-NEXT: movd %xmm5, %eax ; SSE2-NEXT: movd %xmm5, %eax
; SSE2-NEXT: testb $1, %al ; SSE2-NEXT: testb $1, %al
; SSE2-NEXT: je LBB6_2 ; SSE2-NEXT: je LBB6_2
; SSE2-NEXT: ## %bb.1: ## %cond.load ; SSE2-NEXT: ## %bb.1: ## %cond.load
; SSE2-NEXT: movss {{.*#+}} xmm5 = mem[0],zero,zero,zero ; SSE2-NEXT: movss {{.*#+}} xmm5 = mem[0],zero,zero,zero
; SSE2-NEXT: movss {{.*#+}} xmm2 = xmm5[0],xmm2[1,2,3] ; SSE2-NEXT: movss {{.*#+}} xmm2 = xmm5[0],xmm2[1,2,3]
; SSE2-NEXT: LBB6_2: ## %else ; SSE2-NEXT: LBB6_2: ## %else
; SSE2-NEXT: packssdw %xmm0, %xmm4 ; SSE2-NEXT: psrlq $16, %xmm4
; SSE2-NEXT: movd %xmm4, %eax ; SSE2-NEXT: movd %xmm4, %eax
; SSE2-NEXT: shrl $16, %eax ; SSE2-NEXT: shrl $16, %eax
; SSE2-NEXT: testb $1, %al ; SSE2-NEXT: testb $1, %al
; SSE2-NEXT: je LBB6_4 ; SSE2-NEXT: je LBB6_4
; SSE2-NEXT: ## %bb.3: ## %cond.load1 ; SSE2-NEXT: ## %bb.3: ## %cond.load1
; SSE2-NEXT: movss {{.*#+}} xmm4 = mem[0],zero,zero,zero ; SSE2-NEXT: movss {{.*#+}} xmm4 = mem[0],zero,zero,zero
; SSE2-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,0],xmm2[0,0] ; SSE2-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,0],xmm2[0,0]
; SSE2-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm2[2,3] ; SSE2-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm2[2,3]
▲ Show 20 Lines • Show All 1,565 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/X86/masked_store.ll

Show First 20 Lines • Show All 139 Lines • ▼ Show 20 Lines
; SSE2-NEXT: movdqa %xmm4, %xmm5 ; SSE2-NEXT: movdqa %xmm4, %xmm5
; SSE2-NEXT: packssdw %xmm0, %xmm5 ; SSE2-NEXT: packssdw %xmm0, %xmm5
; SSE2-NEXT: movd %xmm5, %eax ; SSE2-NEXT: movd %xmm5, %eax
; SSE2-NEXT: testb $1, %al ; SSE2-NEXT: testb $1, %al
; SSE2-NEXT: je LBB2_2 ; SSE2-NEXT: je LBB2_2
; SSE2-NEXT: ## %bb.1: ## %cond.store ; SSE2-NEXT: ## %bb.1: ## %cond.store
; SSE2-NEXT: movd %xmm2, (%rdi) ; SSE2-NEXT: movd %xmm2, (%rdi)
; SSE2-NEXT: LBB2_2: ## %else ; SSE2-NEXT: LBB2_2: ## %else
; SSE2-NEXT: packssdw %xmm0, %xmm4 ; SSE2-NEXT: psrlq $16, %xmm4
; SSE2-NEXT: movd %xmm4, %eax ; SSE2-NEXT: movd %xmm4, %eax
; SSE2-NEXT: shrl $16, %eax ; SSE2-NEXT: shrl $16, %eax
; SSE2-NEXT: testb $1, %al ; SSE2-NEXT: testb $1, %al
; SSE2-NEXT: je LBB2_4 ; SSE2-NEXT: je LBB2_4
; SSE2-NEXT: ## %bb.3: ## %cond.store1 ; SSE2-NEXT: ## %bb.3: ## %cond.store1
; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm2[1,1,2,3] ; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm2[1,1,2,3]
; SSE2-NEXT: movd %xmm4, 4(%rdi) ; SSE2-NEXT: movd %xmm4, 4(%rdi)
; SSE2-NEXT: LBB2_4: ## %else2 ; SSE2-NEXT: LBB2_4: ## %else2
▲ Show 20 Lines • Show All 784 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/X86/shrink_vmul-widen.ll

Show First 20 Lines • Show All 1,169 Lines • ▼ Show 20 Lines
; X86-SSE-NEXT: psrad $16, %xmm0 ; X86-SSE-NEXT: psrad $16, %xmm0
; X86-SSE-NEXT: movd {{.*#+}} xmm1 = mem[0],zero,zero,zero ; X86-SSE-NEXT: movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
; X86-SSE-NEXT: pxor %xmm2, %xmm2 ; X86-SSE-NEXT: pxor %xmm2, %xmm2
; X86-SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3] ; X86-SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; X86-SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3] ; X86-SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
; X86-SSE-NEXT: pmuludq %xmm0, %xmm1 ; X86-SSE-NEXT: pmuludq %xmm0, %xmm1
; X86-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3] ; X86-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
; X86-SSE-NEXT: pmuludq %xmm2, %xmm0 ; X86-SSE-NEXT: pmuludq %xmm2, %xmm0
; X86-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; X86-SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; X86-SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; X86-SSE-NEXT: movq %xmm1, (%esi,%ecx,4) ; X86-SSE-NEXT: movq %xmm1, (%esi,%ecx,4)
; X86-SSE-NEXT: popl %esi ; X86-SSE-NEXT: popl %esi
; X86-SSE-NEXT: retl ; X86-SSE-NEXT: retl
; ;
; X86-AVX-LABEL: mul_2xi16_sext_zext: ; X86-AVX-LABEL: mul_2xi16_sext_zext:
; X86-AVX: # %bb.0: # %entry ; X86-AVX: # %bb.0: # %entry
; X86-AVX-NEXT: pushl %esi ; X86-AVX-NEXT: pushl %esi
Show All 18 Lines
; X64-SSE-NEXT: psrad $16, %xmm0 ; X64-SSE-NEXT: psrad $16, %xmm0
; X64-SSE-NEXT: movd {{.*#+}} xmm1 = mem[0],zero,zero,zero ; X64-SSE-NEXT: movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
; X64-SSE-NEXT: pxor %xmm2, %xmm2 ; X64-SSE-NEXT: pxor %xmm2, %xmm2
; X64-SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3] ; X64-SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; X64-SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3] ; X64-SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
; X64-SSE-NEXT: pmuludq %xmm0, %xmm1 ; X64-SSE-NEXT: pmuludq %xmm0, %xmm1
; X64-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3] ; X64-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
; X64-SSE-NEXT: pmuludq %xmm2, %xmm0 ; X64-SSE-NEXT: pmuludq %xmm2, %xmm0
; X64-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; X64-SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; X64-SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; X64-SSE-NEXT: movq %xmm1, (%rax,%rdx,4) ; X64-SSE-NEXT: movq %xmm1, (%rax,%rdx,4)
; X64-SSE-NEXT: retq ; X64-SSE-NEXT: retq
; ;
; X64-AVX-LABEL: mul_2xi16_sext_zext: ; X64-AVX-LABEL: mul_2xi16_sext_zext:
; X64-AVX: # %bb.0: # %entry ; X64-AVX: # %bb.0: # %entry
; X64-AVX-NEXT: movq {{.*}}(%rip), %rax ; X64-AVX-NEXT: movq {{.*}}(%rip), %rax
; X64-AVX-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero ; X64-AVX-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
▲ Show 20 Lines • Show All 738 Lines • ▼ Show 20 Lines
; X86-SSE-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero ; X86-SSE-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X86-SSE-NEXT: pxor %xmm1, %xmm1 ; X86-SSE-NEXT: pxor %xmm1, %xmm1
; X86-SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] ; X86-SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; X86-SSE-NEXT: movdqa {{.*#+}} xmm1 = <0,65536,u,u> ; X86-SSE-NEXT: movdqa {{.*#+}} xmm1 = <0,65536,u,u>
; X86-SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3] ; X86-SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
; X86-SSE-NEXT: pmuludq %xmm1, %xmm0 ; X86-SSE-NEXT: pmuludq %xmm1, %xmm0
; X86-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3] ; X86-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
; X86-SSE-NEXT: pmuludq %xmm2, %xmm1 ; X86-SSE-NEXT: pmuludq %xmm2, %xmm1
; X86-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; X86-SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; X86-SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; X86-SSE-NEXT: movq %xmm0, (%edx,%eax,4) ; X86-SSE-NEXT: movq %xmm0, (%edx,%eax,4)
; X86-SSE-NEXT: retl ; X86-SSE-NEXT: retl
; ;
; X86-AVX-LABEL: mul_2xi16_varconst3: ; X86-AVX-LABEL: mul_2xi16_varconst3:
; X86-AVX: # %bb.0: # %entry ; X86-AVX: # %bb.0: # %entry
; X86-AVX-NEXT: movl {{[0-9]+}}(%esp), %eax ; X86-AVX-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-AVX-NEXT: movl {{[0-9]+}}(%esp), %ecx ; X86-AVX-NEXT: movl {{[0-9]+}}(%esp), %ecx
Show All 10 Lines
; X64-SSE-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero ; X64-SSE-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X64-SSE-NEXT: pxor %xmm1, %xmm1 ; X64-SSE-NEXT: pxor %xmm1, %xmm1
; X64-SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] ; X64-SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; X64-SSE-NEXT: movdqa {{.*#+}} xmm1 = <0,65536,u,u> ; X64-SSE-NEXT: movdqa {{.*#+}} xmm1 = <0,65536,u,u>
; X64-SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3] ; X64-SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
; X64-SSE-NEXT: pmuludq %xmm1, %xmm0 ; X64-SSE-NEXT: pmuludq %xmm1, %xmm0
; X64-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3] ; X64-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
; X64-SSE-NEXT: pmuludq %xmm2, %xmm1 ; X64-SSE-NEXT: pmuludq %xmm2, %xmm1
; X64-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; X64-SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; X64-SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; X64-SSE-NEXT: movq %xmm0, (%rax,%rsi,4) ; X64-SSE-NEXT: movq %xmm0, (%rax,%rsi,4)
; X64-SSE-NEXT: retq ; X64-SSE-NEXT: retq
; ;
; X64-AVX-LABEL: mul_2xi16_varconst3: ; X64-AVX-LABEL: mul_2xi16_varconst3:
; X64-AVX: # %bb.0: # %entry ; X64-AVX: # %bb.0: # %entry
; X64-AVX-NEXT: movq {{.*}}(%rip), %rax ; X64-AVX-NEXT: movq {{.*}}(%rip), %rax
; X64-AVX-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero ; X64-AVX-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
Show All 28 Lines
; X86-SSE-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero ; X86-SSE-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X86-SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,1,4,5,6,7] ; X86-SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,1,4,5,6,7]
; X86-SSE-NEXT: psrad $16, %xmm0 ; X86-SSE-NEXT: psrad $16, %xmm0
; X86-SSE-NEXT: movdqa {{.*#+}} xmm1 = <0,32768,u,u> ; X86-SSE-NEXT: movdqa {{.*#+}} xmm1 = <0,32768,u,u>
; X86-SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3] ; X86-SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
; X86-SSE-NEXT: pmuludq %xmm1, %xmm0 ; X86-SSE-NEXT: pmuludq %xmm1, %xmm0
; X86-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3] ; X86-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
; X86-SSE-NEXT: pmuludq %xmm2, %xmm1 ; X86-SSE-NEXT: pmuludq %xmm2, %xmm1
; X86-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; X86-SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; X86-SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; X86-SSE-NEXT: movq %xmm0, (%edx,%eax,4) ; X86-SSE-NEXT: movq %xmm0, (%edx,%eax,4)
; X86-SSE-NEXT: retl ; X86-SSE-NEXT: retl
; ;
; X86-AVX-LABEL: mul_2xi16_varconst4: ; X86-AVX-LABEL: mul_2xi16_varconst4:
; X86-AVX: # %bb.0: # %entry ; X86-AVX: # %bb.0: # %entry
; X86-AVX-NEXT: movl {{[0-9]+}}(%esp), %eax ; X86-AVX-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-AVX-NEXT: movl {{[0-9]+}}(%esp), %ecx ; X86-AVX-NEXT: movl {{[0-9]+}}(%esp), %ecx
Show All 10 Lines
; X64-SSE-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero ; X64-SSE-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X64-SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,1,4,5,6,7] ; X64-SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,1,4,5,6,7]
; X64-SSE-NEXT: psrad $16, %xmm0 ; X64-SSE-NEXT: psrad $16, %xmm0
; X64-SSE-NEXT: movdqa {{.*#+}} xmm1 = <0,32768,u,u> ; X64-SSE-NEXT: movdqa {{.*#+}} xmm1 = <0,32768,u,u>
; X64-SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3] ; X64-SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
; X64-SSE-NEXT: pmuludq %xmm1, %xmm0 ; X64-SSE-NEXT: pmuludq %xmm1, %xmm0
; X64-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3] ; X64-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
; X64-SSE-NEXT: pmuludq %xmm2, %xmm1 ; X64-SSE-NEXT: pmuludq %xmm2, %xmm1
; X64-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; X64-SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; X64-SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; X64-SSE-NEXT: movq %xmm0, (%rax,%rsi,4) ; X64-SSE-NEXT: movq %xmm0, (%rax,%rsi,4)
; X64-SSE-NEXT: retq ; X64-SSE-NEXT: retq
; ;
; X64-AVX-LABEL: mul_2xi16_varconst4: ; X64-AVX-LABEL: mul_2xi16_varconst4:
; X64-AVX: # %bb.0: # %entry ; X64-AVX: # %bb.0: # %entry
; X64-AVX-NEXT: movq {{.*}}(%rip), %rax ; X64-AVX-NEXT: movq {{.*}}(%rip), %rax
; X64-AVX-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero ; X64-AVX-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
▲ Show 20 Lines • Show All 514 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/X86/vec_extract-sse4.ll

Show All 19 Lines; X64-NEXT: retq
%tmp = extractelement <4 x float> %X, i32 3 %tmp = extractelement <4 x float> %X, i32 3
store float %tmp, float* %R store float %tmp, float* %R
ret void ret void
} }
define float @t2(<4 x float>* %P1) nounwind { define float @t2(<4 x float>* %P1) nounwind {
; X32-LABEL: t2: ; X32-LABEL: t2:
; X32: # %bb.0: ; X32: # %bb.0:
; X32-NEXT: pushl %eax
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
; X32-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; X32-NEXT: flds 8(%eax)
; X32-NEXT: movss %xmm0, (%esp)
; X32-NEXT: flds (%esp)
; X32-NEXT: popl %eax
; X32-NEXT: retl ; X32-NEXT: retl
; ;
; X64-LABEL: t2: ; X64-LABEL: t2:
; X64: # %bb.0: ; X64: # %bb.0:
; X64-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero ; X64-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X64-NEXT: retq ; X64-NEXT: retq
%X = load <4 x float>, <4 x float>* %P1 %X = load <4 x float>, <4 x float>* %P1
%tmp = extractelement <4 x float> %X, i32 2 %tmp = extractelement <4 x float> %X, i32 2
ret float %tmp ret float %tmp
} }
define void @t3(i32* %R, <4 x i32>* %P1) nounwind { define void @t3(i32* %R, <4 x i32>* %P1) nounwind {
; X32-LABEL: t3: ; X32-LABEL: t3:
Show All 33 Lines

llvm/trunk/test/CodeGen/X86/vector-reduce-mul-widen.ll

Show First 20 Lines • Show All 1,857 Lines • ▼ Show 20 Lines
; AVX2-NEXT: vpmullw %ymm2, %ymm0, %ymm0 ; AVX2-NEXT: vpmullw %ymm2, %ymm0, %ymm0
; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0 ; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 ; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpackuswb %xmm1, %xmm0, %xmm0 ; AVX2-NEXT: vpackuswb %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrlw $8, %xmm0, %xmm1 ; AVX2-NEXT: vpsrlw $8, %xmm0, %xmm1
; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero ; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero ; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
; AVX2-NEXT: vpmullw %xmm1, %xmm0, %xmm0 ; AVX2-NEXT: vpmullw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpand {{.*}}(%rip), %xmm0, %xmm0
; AVX2-NEXT: vpackuswb %xmm0, %xmm0, %xmm0
; AVX2-NEXT: vpextrb $0, %xmm0, %eax ; AVX2-NEXT: vpextrb $0, %xmm0, %eax
; AVX2-NEXT: # kill: def $al killed $al killed $eax ; AVX2-NEXT: # kill: def $al killed $al killed $eax
; AVX2-NEXT: vzeroupper ; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq ; AVX2-NEXT: retq
; ;
; AVX512BW-LABEL: test_v16i8: ; AVX512BW-LABEL: test_v16i8:
; AVX512BW: # %bb.0: ; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512BW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
▲ Show 20 Lines • Show All 1,504 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/X86/vector-reduce-mul.ll

Show First 20 Lines • Show All 802 Lines • ▼ Show 20 Lines
; AVX512BWVL-NEXT: vpmuludq %xmm1, %xmm0, %xmm0 ; AVX512BWVL-NEXT: vpmuludq %xmm1, %xmm0, %xmm0
; AVX512BWVL-NEXT: vmovd %xmm0, %eax ; AVX512BWVL-NEXT: vmovd %xmm0, %eax
; AVX512BWVL-NEXT: retq ; AVX512BWVL-NEXT: retq
; ;
; AVX512DQ-LABEL: test_v2i32: ; AVX512DQ-LABEL: test_v2i32:
; AVX512DQ: # %bb.0: ; AVX512DQ: # %bb.0:
; AVX512DQ-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512DQ-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; AVX512DQ-NEXT: vpmovzxdq {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero
; AVX512DQ-NEXT: vpmullq %zmm1, %zmm0, %zmm0 ; AVX512DQ-NEXT: vpmullq %zmm1, %zmm0, %zmm0
; AVX512DQ-NEXT: vmovd %xmm0, %eax ; AVX512DQ-NEXT: vmovd %xmm0, %eax
; AVX512DQ-NEXT: vzeroupper ; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq ; AVX512DQ-NEXT: retq
; ;
; AVX512DQVL-LABEL: test_v2i32: ; AVX512DQVL-LABEL: test_v2i32:
; AVX512DQVL: # %bb.0: ; AVX512DQVL: # %bb.0:
; AVX512DQVL-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512DQVL-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; AVX512DQVL-NEXT: vpmovzxdq {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero
; AVX512DQVL-NEXT: vpmullq %xmm1, %xmm0, %xmm0 ; AVX512DQVL-NEXT: vpmullq %xmm1, %xmm0, %xmm0
; AVX512DQVL-NEXT: vmovd %xmm0, %eax ; AVX512DQVL-NEXT: vmovd %xmm0, %eax
; AVX512DQVL-NEXT: retq ; AVX512DQVL-NEXT: retq
%1 = call i32 @llvm.experimental.vector.reduce.mul.i32.v2i32(<2 x i32> %a0) %1 = call i32 @llvm.experimental.vector.reduce.mul.i32.v2i32(<2 x i32> %a0)
ret i32 %1 ret i32 %1
} }
define i32 @test_v4i32(<4 x i32> %a0) { define i32 @test_v4i32(<4 x i32> %a0) {
▲ Show 20 Lines • Show All 344 Lines • ▼ Show 20 Lines
; AVX512BWVL-NEXT: vmovd %xmm0, %eax ; AVX512BWVL-NEXT: vmovd %xmm0, %eax
; AVX512BWVL-NEXT: # kill: def $ax killed $ax killed $eax ; AVX512BWVL-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512BWVL-NEXT: retq ; AVX512BWVL-NEXT: retq
; ;
; AVX512DQ-LABEL: test_v2i16: ; AVX512DQ-LABEL: test_v2i16:
; AVX512DQ: # %bb.0: ; AVX512DQ: # %bb.0:
; AVX512DQ-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512DQ-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; AVX512DQ-NEXT: vpmovzxdq {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero
; AVX512DQ-NEXT: vpmullq %zmm1, %zmm0, %zmm0 ; AVX512DQ-NEXT: vpmullq %zmm1, %zmm0, %zmm0
; AVX512DQ-NEXT: vmovd %xmm0, %eax ; AVX512DQ-NEXT: vmovd %xmm0, %eax
; AVX512DQ-NEXT: # kill: def $ax killed $ax killed $eax ; AVX512DQ-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512DQ-NEXT: vzeroupper ; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq ; AVX512DQ-NEXT: retq
; ;
; AVX512DQVL-LABEL: test_v2i16: ; AVX512DQVL-LABEL: test_v2i16:
; AVX512DQVL: # %bb.0: ; AVX512DQVL: # %bb.0:
; AVX512DQVL-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512DQVL-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; AVX512DQVL-NEXT: vpmovzxdq {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero
; AVX512DQVL-NEXT: vpmullq %xmm1, %xmm0, %xmm0 ; AVX512DQVL-NEXT: vpmullq %xmm1, %xmm0, %xmm0
; AVX512DQVL-NEXT: vmovd %xmm0, %eax ; AVX512DQVL-NEXT: vmovd %xmm0, %eax
; AVX512DQVL-NEXT: # kill: def $ax killed $ax killed $eax ; AVX512DQVL-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512DQVL-NEXT: retq ; AVX512DQVL-NEXT: retq
%1 = call i16 @llvm.experimental.vector.reduce.mul.i16.v2i16(<2 x i16> %a0) %1 = call i16 @llvm.experimental.vector.reduce.mul.i16.v2i16(<2 x i16> %a0)
ret i16 %1 ret i16 %1
} }
define i16 @test_v4i16(<4 x i16> %a0) { define i16 @test_v4i16(<4 x i16> %a0) {
; SSE2-LABEL: test_v4i16: ; SSE2-LABEL: test_v4i16:
; SSE2: # %bb.0: ; SSE2: # %bb.0:
; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[3,3,1,1] ; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[3,3,1,1]
; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3] ; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
; SSE2-NEXT: pmuludq %xmm2, %xmm3 ; SSE2-NEXT: pmuludq %xmm2, %xmm3
; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3] ; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
; SSE2-NEXT: pmuludq %xmm0, %xmm1 ; SSE2-NEXT: pmuludq %xmm0, %xmm1
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3] ; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] ; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; SSE2-NEXT: pmuludq %xmm0, %xmm1 ; SSE2-NEXT: pmuludq %xmm0, %xmm1
; SSE2-NEXT: movd %xmm1, %eax ; SSE2-NEXT: pextrw $0, %xmm1, %eax
; SSE2-NEXT: # kill: def $ax killed $ax killed $eax ; SSE2-NEXT: # kill: def $ax killed $ax killed $eax
; SSE2-NEXT: retq ; SSE2-NEXT: retq
; ;
; SSE41-LABEL: test_v4i16: ; SSE41-LABEL: test_v4i16:
; SSE41: # %bb.0: ; SSE41: # %bb.0:
; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; SSE41-NEXT: pmulld %xmm0, %xmm1 ; SSE41-NEXT: pmulld %xmm0, %xmm1
; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3] ; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
▲ Show 20 Lines • Show All 431 Lines • ▼ Show 20 Lines
; AVX512BWVL-NEXT: vpextrb $0, %xmm0, %eax ; AVX512BWVL-NEXT: vpextrb $0, %xmm0, %eax
; AVX512BWVL-NEXT: # kill: def $al killed $al killed $eax ; AVX512BWVL-NEXT: # kill: def $al killed $al killed $eax
; AVX512BWVL-NEXT: retq ; AVX512BWVL-NEXT: retq
; ;
; AVX512DQ-LABEL: test_v2i8: ; AVX512DQ-LABEL: test_v2i8:
; AVX512DQ: # %bb.0: ; AVX512DQ: # %bb.0:
; AVX512DQ-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 ; AVX512DQ-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; AVX512DQ-NEXT: vpmovzxdq {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero
; AVX512DQ-NEXT: vpmullq %zmm1, %zmm0, %zmm0 ; AVX512DQ-NEXT: vpmullq %zmm1, %zmm0, %zmm0
; AVX512DQ-NEXT: vpextrb $0, %xmm0, %eax ; AVX512DQ-NEXT: vpextrb $0, %xmm0, %eax
; AVX512DQ-NEXT: # kill: def $al killed $al killed $eax ; AVX512DQ-NEXT: # kill: def $al killed $al killed $eax
; AVX512DQ-NEXT: vzeroupper ; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq ; AVX512DQ-NEXT: retq
; ;
; AVX512DQVL-LABEL: test_v2i8: ; AVX512DQVL-LABEL: test_v2i8:
; AVX512DQVL: # %bb.0: ; AVX512DQVL: # %bb.0:
; AVX512DQVL-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512DQVL-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; AVX512DQVL-NEXT: vpmovzxdq {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero
; AVX512DQVL-NEXT: vpmullq %xmm1, %xmm0, %xmm0 ; AVX512DQVL-NEXT: vpmullq %xmm1, %xmm0, %xmm0
; AVX512DQVL-NEXT: vpextrb $0, %xmm0, %eax ; AVX512DQVL-NEXT: vpextrb $0, %xmm0, %eax
; AVX512DQVL-NEXT: # kill: def $al killed $al killed $eax ; AVX512DQVL-NEXT: # kill: def $al killed $al killed $eax
; AVX512DQVL-NEXT: retq ; AVX512DQVL-NEXT: retq
%1 = call i8 @llvm.experimental.vector.reduce.mul.i8.v2i8(<2 x i8> %a0) %1 = call i8 @llvm.experimental.vector.reduce.mul.i8.v2i8(<2 x i8> %a0)
ret i8 %1 ret i8 %1
} }
▲ Show 20 Lines • Show All 214 Lines • ▼ Show 20 Lines
; AVX2-NEXT: vpmullw %ymm2, %ymm0, %ymm0 ; AVX2-NEXT: vpmullw %ymm2, %ymm0, %ymm0
; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0 ; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 ; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpackuswb %xmm1, %xmm0, %xmm0 ; AVX2-NEXT: vpackuswb %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrlw $8, %xmm0, %xmm1 ; AVX2-NEXT: vpsrlw $8, %xmm0, %xmm1
; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero ; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero ; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
; AVX2-NEXT: vpmullw %xmm1, %xmm0, %xmm0 ; AVX2-NEXT: vpmullw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpand {{.*}}(%rip), %xmm0, %xmm0
; AVX2-NEXT: vpackuswb %xmm0, %xmm0, %xmm0
; AVX2-NEXT: vpextrb $0, %xmm0, %eax ; AVX2-NEXT: vpextrb $0, %xmm0, %eax
; AVX2-NEXT: # kill: def $al killed $al killed $eax ; AVX2-NEXT: # kill: def $al killed $al killed $eax
; AVX2-NEXT: vzeroupper ; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq ; AVX2-NEXT: retq
; ;
; AVX512BW-LABEL: test_v16i8: ; AVX512BW-LABEL: test_v16i8:
; AVX512BW: # %bb.0: ; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512BW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
▲ Show 20 Lines • Show All 1,504 Lines • Show Last 20 Lines