1 files changed, 24 insertions, 1 deletions
diff --git a/lib/Transforms/InstCombine/InstCombineCalls.cpp b/lib/Transforms/InstCombine/InstCombineCalls.cpp
index df217f19ac..785d19e45b 100644
--- a/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -725,9 +725,32 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) {
     // Convert vpermil* to shufflevector if the mask is constant.
     Value *V = II->getArgOperand(1);
     if (auto C = dyn_cast<ConstantDataVector>(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;
+	if (II->getIntrinsicID() == Intrinsic::x86_avx_vpermilvar_pd ||
+	    II->getIntrinsicID() == Intrinsic::x86_avx_vpermilvar_pd_256)
+	  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, C);
+      auto Shuffle = Builder->CreateShuffleVector(V1, V2, NewC);
       return ReplaceInstUsesWith(CI, Shuffle);
     }
     break;