2 files changed, 57 insertions, 10 deletions
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
index 54a9fbdbe82..5cde31a9162 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
@@ -384,23 +384,20 @@ static void replaceExtractElements(InsertElementInst *InsElt,
                                         ConstantVector::get(ExtendMask));
 
   // Insert the new shuffle after the vector operand of the extract is defined
-  // or at the start of the basic block, so any subsequent extracts can use it.
-  bool ReplaceAllExtUsers;
-  if (auto *ExtVecOpInst = dyn_cast<Instruction>(ExtVecOp)) {
+  // (as long as it's not a PHI) or at the start of the basic block of the
+  // extract, so any subsequent extracts in the same basic block can use it.
+  // TODO: Insert before the earliest ExtractElementInst that is replaced.
+  auto *ExtVecOpInst = dyn_cast<Instruction>(ExtVecOp);
+  if (ExtVecOpInst && !isa<PHINode>(ExtVecOpInst))
     WideVec->insertAfter(ExtVecOpInst);
-    ReplaceAllExtUsers = true;
-  } else {
-    // TODO: Insert at start of function, so it's always safe to replace all?
+  else
     IC.InsertNewInstWith(WideVec, *ExtElt->getParent()->getFirstInsertionPt());
-    ReplaceAllExtUsers = false;
-  }
 
   // Replace extracts from the original narrow vector with extracts from the new
   // wide vector.
   for (User *U : ExtVecOp->users()) {
     ExtractElementInst *OldExt = dyn_cast<ExtractElementInst>(U);
-    if (!OldExt ||
-        (!ReplaceAllExtUsers && OldExt->getParent() != WideVec->getParent()))
+    if (!OldExt || OldExt->getParent() != WideVec->getParent())
       continue;
     auto *NewExt = ExtractElementInst::Create(WideVec, OldExt->getOperand(1));
     NewExt->insertAfter(WideVec);
diff --git a/llvm/test/Transforms/InstCombine/insert-extract-shuffle.ll b/llvm/test/Transforms/InstCombine/insert-extract-shuffle.ll
index 4223660db3d..47c2a139a47 100644
--- a/llvm/test/Transforms/InstCombine/insert-extract-shuffle.ll
+++ b/llvm/test/Transforms/InstCombine/insert-extract-shuffle.ll
@@ -125,3 +125,53 @@ end:
   ret <8 x i16> %t6
 }
 
+; The widening shuffle must be inserted at a valid point (after the PHIs). 
+
+define <4 x double> @pr25999_phis1(i1 %c, <2 x double> %a, <4 x double> %b) {
+; CHECK-LABEL: @pr25999_phis1(
+; CHECK:       %tmp1 = phi <2 x double> [ %a, %bb1 ], [ %r, %bb2 ]
+; CHECK-NEXT:  %tmp2 = phi <4 x double> [ %b, %bb1 ], [ zeroinitializer, %bb2 ]
+; CHECK-NEXT:  %[[WIDEVEC:.*]] = shufflevector <2 x double> %tmp1, <2 x double> undef, <4 x i32> <i32 0, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:  %tmp4 = shufflevector <4 x double> %tmp2, <4 x double> %[[WIDEVEC]], <4 x i32> <i32 0, i32 1, i32 4, i32 3>
+; CHECK-NEXT:  ret <4 x double> %tmp4
+bb1:
+  br i1 %c, label %bb2, label %bb3
+
+bb2:
+  %r = call <2 x double> @dummy(<2 x double> %a)
+  br label %bb3
+
+bb3:
+  %tmp1 = phi <2 x double> [ %a, %bb1 ], [ %r, %bb2 ]
+  %tmp2 = phi <4 x double> [ %b, %bb1 ], [ zeroinitializer, %bb2 ]
+  %tmp3 = extractelement <2 x double> %tmp1, i32 0
+  %tmp4 = insertelement <4 x double> %tmp2, double %tmp3, i32 2
+  ret <4 x double> %tmp4
+}
+
+declare <2 x double> @dummy(<2 x double>)
+
+define <4 x double> @pr25999_phis2(i1 %c, <2 x double> %a, <4 x double> %b) {
+; CHECK-LABEL: @pr25999_phis2(
+; CHECK:       %tmp1 = phi <2 x double> [ %a, %bb1 ], [ %r, %bb2 ]
+; CHECK-NEXT:  %tmp2 = phi <4 x double> [ %b, %bb1 ], [ zeroinitializer, %bb2 ]
+; CHECK-NEXT:  %d = fadd <2 x double> %tmp1, %tmp1
+; CHECK-NEXT:  %[[WIDEVEC:.*]] = shufflevector <2 x double> %d, <2 x double> undef, <4 x i32> <i32 0, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:  %tmp4 = shufflevector <4 x double> %tmp2, <4 x double> %[[WIDEVEC]], <4 x i32> <i32 0, i32 1, i32 4, i32 3>
+; CHECK-NEXT:  ret <4 x double> %tmp4
+bb1:
+  br i1 %c, label %bb2, label %bb3
+
+bb2:
+  %r = call <2 x double> @dummy(<2 x double> %a)
+  br label %bb3
+
+bb3:
+  %tmp1 = phi <2 x double> [ %a, %bb1 ], [ %r, %bb2 ]
+  %tmp2 = phi <4 x double> [ %b, %bb1 ], [ zeroinitializer, %bb2 ]
+  %d = fadd <2 x double> %tmp1, %tmp1
+  %tmp3 = extractelement <2 x double> %d, i32 0
+  %tmp4 = insertelement <4 x double> %tmp2, double %tmp3, i32 2
+  ret <4 x double> %tmp4
+}
+