From 2259a26a5dc348b3c2465ec0154c62e32c4372e0 Mon Sep 17 00:00:00 2001 From: Rafael Espindola Date: Tue, 29 Apr 2014 22:20:40 +0000 Subject: Also handle ConstantAggregateZero when optimizing vpermilvar*. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207582 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/Transforms/InstCombine/InstCombineCalls.cpp | 42 +++++++++++++------------ 1 file changed, 22 insertions(+), 20 deletions(-) (limited to 'lib') diff --git a/lib/Transforms/InstCombine/InstCombineCalls.cpp b/lib/Transforms/InstCombine/InstCombineCalls.cpp index cc43675360..eafdf6e7b0 100644 --- a/lib/Transforms/InstCombine/InstCombineCalls.cpp +++ b/lib/Transforms/InstCombine/InstCombineCalls.cpp @@ -724,11 +724,10 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) { case Intrinsic::x86_avx_vpermilvar_pd_256: { // Convert vpermil* to shufflevector if the mask is constant. Value *V = II->getArgOperand(1); + unsigned Size = cast(V->getType())->getNumElements(); + assert(Size == 8 || Size == 4 || Size == 2); + uint32_t Indexes[8]; if (auto C = dyn_cast(V)) { - unsigned Size = C->getNumElements(); - assert(Size == 8 || Size == 4 || Size == 2); - uint32_t Indexes[8]; - // The intrinsics only read one or two bits, clear the rest. for (unsigned I = 0; I < Size; ++I) { uint32_t Index = C->getElementAsInteger(I) & 0x3; @@ -737,23 +736,26 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) { Index >>= 1; Indexes[I] = Index; } - - // The _256 variants are a bit trickier since the mask bits always index - // into the corresponding 128 half. In order to convert to a generic - // shuffle, we have to make that explicit. - if (II->getIntrinsicID() == Intrinsic::x86_avx_vpermilvar_ps_256 || - II->getIntrinsicID() == Intrinsic::x86_avx_vpermilvar_pd_256) { - for (unsigned I = Size / 2; I < Size; ++I) - Indexes[I] += Size / 2; - } - auto NewC = - ConstantDataVector::get(C->getContext(), makeArrayRef(Indexes, Size)); - auto V1 = II->getArgOperand(0); - auto V2 = UndefValue::get(V1->getType()); - auto Shuffle = Builder->CreateShuffleVector(V1, V2, NewC); - return ReplaceInstUsesWith(CI, Shuffle); + } else if (isa(V)) { + for (unsigned I = 0; I < Size; ++I) + Indexes[I] = 0; + } else { + break; } - break; + // The _256 variants are a bit trickier since the mask bits always index + // into the corresponding 128 half. In order to convert to a generic + // shuffle, we have to make that explicit. + if (II->getIntrinsicID() == Intrinsic::x86_avx_vpermilvar_ps_256 || + II->getIntrinsicID() == Intrinsic::x86_avx_vpermilvar_pd_256) { + for (unsigned I = Size / 2; I < Size; ++I) + Indexes[I] += Size / 2; + } + auto NewC = + ConstantDataVector::get(V->getContext(), makeArrayRef(Indexes, Size)); + auto V1 = II->getArgOperand(0); + auto V2 = UndefValue::get(V1->getType()); + auto Shuffle = Builder->CreateShuffleVector(V1, V2, NewC); + return ReplaceInstUsesWith(CI, Shuffle); } case Intrinsic::ppc_altivec_vperm: -- cgit v1.2.3