replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[InstCombine][X86] Add DemandedElts support for 512-bit PMULDQ/PMULUDQ instructions 

PMULDQ/PMULUDQ vXi64 instructions only use the even numbered v2Xi32 input elements which SimplifyDemandedVectorElts should try and use.

This builds on r290554 which added supported for 128 and 256-bit.

llvm-svn: 290582
This commit is contained in:
Craig Topper 2016-12-27 05:30:09 +00:00
parent 162504578b
commit 72f2d4e8d6
3 changed files with 35 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
View File

@ -2048,7 +2048,9 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) {
case Intrinsic::x86_sse2_pmulu_dq:
case Intrinsic::x86_sse41_pmuldq:
case Intrinsic::x86_avx2_pmul_dq:
case Intrinsic::x86_avx2_pmulu_dq: {
case Intrinsic::x86_avx2_pmulu_dq:
case Intrinsic::x86_avx512_pmul_dq_512:
case Intrinsic::x86_avx512_pmulu_dq_512: {
unsigned VWidth = II->getType()->getVectorNumElements();
APInt UndefElts(VWidth, 0);
APInt DemandedElts = APInt::getAllOnesValue(VWidth);

4
llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
View File

@ -1434,7 +1434,9 @@ Value *InstCombiner::SimplifyDemandedVectorElts(Value *V, APInt DemandedElts,
case Intrinsic::x86_sse2_pmulu_dq:
case Intrinsic::x86_sse41_pmuldq:
case Intrinsic::x86_avx2_pmul_dq:
case Intrinsic::x86_avx2_pmulu_dq: {
case Intrinsic::x86_avx2_pmulu_dq:
case Intrinsic::x86_avx512_pmul_dq_512:
case Intrinsic::x86_avx512_pmulu_dq_512: {
Value *Op0 = II->getArgOperand(0);
Value *Op1 = II->getArgOperand(1);
unsigned InnerVWidth = Op0->getType()->getVectorNumElements();

29
llvm/test/Transforms/InstCombine/x86-muldq.ll
View File

@ -31,6 +31,18 @@ define <4 x i64> @test_demanded_elts_pmuludq_256(<8 x i32> %a0, <8 x i32> %a1) {
ret <4 x i64> %3
}
define <8 x i64> @test_demanded_elts_pmuludq_512(<16 x i32> %a0, <16 x i32> %a1) {
; CHECK-LABEL: @test_demanded_elts_pmuludq_512(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <16 x i32> %a1, <16 x i32> undef, <16 x i32> <i32 1, i32 undef, i32 3, i32 undef, i32 5, i32 undef, i32 7, i32 undef, i32 9, i32 undef, i32 11, i32 undef, i32 13, i32 undef, i32 15, i32 undef>
; CHECK-NEXT: [[TMP2:%.*]] = call <8 x i64> @llvm.x86.avx512.pmulu.dq.512(<16 x i32> %a0, <16 x i32> [[TMP1]])
; CHECK-NEXT: ret <8 x i64> [[TMP2]]
;
%1 = shufflevector <16 x i32> %a0, <16 x i32> undef, <16 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6, i32 8, i32 8, i32 10, i32 10, i32 12, i32 12, i32 14, i32 14>
%2 = shufflevector <16 x i32> %a1, <16 x i32> undef, <16 x i32> <i32 1, i32 1, i32 3, i32 3, i32 5, i32 5, i32 7, i32 7, i32 9, i32 9, i32 11, i32 11, i32 13, i32 13, i32 15, i32 15>
%3 = call <8 x i64> @llvm.x86.avx512.pmulu.dq.512(<16 x i32> %1, <16 x i32> %2)
ret <8 x i64> %3
}
define <2 x i64> @test_demanded_elts_pmuldq_128(<4 x i32> %a0, <4 x i32> %a1) {
; CHECK-LABEL: @test_demanded_elts_pmuldq_128(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <4 x i32> %a1, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 3, i32 undef>
@ -57,8 +69,25 @@ define <4 x i64> @test_demanded_elts_pmuluq_256(<8 x i32> %a0, <8 x i32> %a1) {
ret <4 x i64> %4
}
define <8 x i64> @test_demanded_elts_pmuluq_512(<16 x i32> %a0, <16 x i32> %a1) {
; CHECK-LABEL: @test_demanded_elts_pmuluq_512(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <16 x i32> %a1, <16 x i32> undef, <16 x i32> <i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 7, i32 undef, i32 9, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 15, i32 undef>
; CHECK-NEXT: [[TMP2:%.*]] = call <8 x i64> @llvm.x86.avx512.pmul.dq.512(<16 x i32> %a0, <16 x i32> [[TMP1]])
; CHECK-NEXT: [[TMP3:%.*]] = shufflevector <8 x i64> [[TMP2]], <8 x i64> undef, <8 x i32> <i32 0, i32 0, i32 3, i32 3, i32 4, i32 4, i32 7, i32 7>
; CHECK-NEXT: ret <8 x i64> [[TMP3]]
;
%1 = shufflevector <16 x i32> %a0, <16 x i32> undef, <16 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6, i32 8, i32 8, i32 10, i32 10, i32 12, i32 12, i32 14, i32 14>
%2 = shufflevector <16 x i32> %a1, <16 x i32> undef, <16 x i32> <i32 1, i32 1, i32 3, i32 3, i32 5, i32 5, i32 7, i32 7, i32 9, i32 9, i32 11, i32 11, i32 13, i32 13, i32 15, i32 15>
%3 = call <8 x i64> @llvm.x86.avx512.pmul.dq.512(<16 x i32> %1, <16 x i32> %2)
%4 = shufflevector <8 x i64> %3, <8 x i64> undef, <8 x i32> <i32 0, i32 0, i32 3, i32 3, i32 4, i32 4, i32 7, i32 7>
ret <8 x i64> %4
}
declare <2 x i64> @llvm.x86.sse2.pmulu.dq(<4 x i32>, <4 x i32>) nounwind readnone
declare <2 x i64> @llvm.x86.sse41.pmuldq(<4 x i32>, <4 x i32>) nounwind readnone
declare <4 x i64> @llvm.x86.avx2.pmulu.dq(<8 x i32>, <8 x i32>) nounwind readnone
declare <4 x i64> @llvm.x86.avx2.pmul.dq(<8 x i32>, <8 x i32>) nounwind readnone
declare <8 x i64> @llvm.x86.avx512.pmulu.dq.512(<16 x i32>, <16 x i32>) nounwind readnone
declare <8 x i64> @llvm.x86.avx512.pmul.dq.512(<16 x i32>, <16 x i32>) nounwind readnone