summaryrefslogtreecommitdiffstats
path: root/llvm/lib/Transforms
diff options
context:
space:
mode:
authorSimon Pilgrim <llvm-dev@redking.me.uk>2017-01-17 11:35:03 +0000
committerSimon Pilgrim <llvm-dev@redking.me.uk>2017-01-17 11:35:03 +0000
commitd4eb800b03e0353ce880092d2f1775d5aa37bdd8 (patch)
tree9ffb302a366eb6e7fc147dbcd2a7455636863d23 /llvm/lib/Transforms
parent92b375311551dde9500ce73f5a8e77d389aa0cab (diff)
downloadbcm5719-llvm-d4eb800b03e0353ce880092d2f1775d5aa37bdd8.tar.gz
bcm5719-llvm-d4eb800b03e0353ce880092d2f1775d5aa37bdd8.zip
[InstCombine][X86][AVX] Add DemandedElts support for VPERMILPD/VPERMILPS instructions
Simplify a vpermilvar shuffle mask based on the elements of the mask that are actually demanded. llvm-svn: 292209
Diffstat (limited to 'llvm/lib/Transforms')
-rw-r--r--llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp12
-rw-r--r--llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp10
2 files changed, 20 insertions, 2 deletions
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index c68a5ca9134..8d896109e9f 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -2347,10 +2347,20 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) {
case Intrinsic::x86_avx512_vpermilvar_ps_512:
case Intrinsic::x86_avx_vpermilvar_pd:
case Intrinsic::x86_avx_vpermilvar_pd_256:
- case Intrinsic::x86_avx512_vpermilvar_pd_512:
+ case Intrinsic::x86_avx512_vpermilvar_pd_512: {
if (Value *V = simplifyX86vpermilvar(*II, *Builder))
return replaceInstUsesWith(*II, V);
+
+ unsigned VWidth = II->getType()->getVectorNumElements();
+ APInt UndefElts(VWidth, 0);
+ APInt DemandedElts = APInt::getAllOnesValue(VWidth);
+ if (Value *V = SimplifyDemandedVectorElts(II, DemandedElts, UndefElts)) {
+ if (V != II)
+ return replaceInstUsesWith(*II, V);
+ return II;
+ }
break;
+ }
case Intrinsic::x86_avx2_permd:
case Intrinsic::x86_avx2_permps:
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
index 95100d074b4..bf88a7517c1 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
@@ -1472,9 +1472,17 @@ Value *InstCombiner::SimplifyDemandedVectorElts(Value *V, APInt DemandedElts,
break;
}
+ // PSHUFB
case Intrinsic::x86_ssse3_pshuf_b_128:
case Intrinsic::x86_avx2_pshuf_b:
- case Intrinsic::x86_avx512_pshuf_b_512: {
+ case Intrinsic::x86_avx512_pshuf_b_512:
+ // PERMILVAR
+ case Intrinsic::x86_avx_vpermilvar_ps:
+ case Intrinsic::x86_avx_vpermilvar_ps_256:
+ case Intrinsic::x86_avx512_vpermilvar_ps_512:
+ case Intrinsic::x86_avx_vpermilvar_pd:
+ case Intrinsic::x86_avx_vpermilvar_pd_256:
+ case Intrinsic::x86_avx512_vpermilvar_pd_512: {
Value *Op1 = II->getArgOperand(1);
TmpV = SimplifyDemandedVectorElts(Op1, DemandedElts, UndefElts,
Depth + 1);
OpenPOWER on IntegriCloud