[ScalarizeMaskedMemIntrin] Handle the case where the mask is an all zero vector.

This shouldn't really happen in practice I hope, but we tried to handle other constant cases. We missed this one because we checked for ConstantVector without realizing that zero becomes ConstantAggregateZero instead.

So instead just check for Constant and use getAggregateElement which will do the dirty work for us.

llvm-svn: 343270

2 files changed, 10 insertions, 26 deletions

diff --git a/llvm/lib/CodeGen/ScalarizeMaskedMemIntrin.cpp b/llvm/lib/CodeGen/ScalarizeMaskedMemIntrin.cpp
index 549697c7139..35580df42f1 100644
--- a/llvm/lib/CodeGen/ScalarizeMaskedMemIntrin.cpp
+++ b/llvm/lib/CodeGen/ScalarizeMaskedMemIntrin.cpp
@@ -151,9 +151,9 @@ static void scalarizeMaskedLoad(CallInst *CI) {
   // The result vector
   Value *VResult = UndefVal;
 
-  if (isa<ConstantVector>(Mask)) {
+  if (isa<Constant>(Mask)) {
     for (unsigned Idx = 0; Idx < VectorWidth; ++Idx) {
-      if (cast<ConstantVector>(Mask)->getOperand(Idx)->isNullValue())
+      if (cast<Constant>(Mask)->getAggregateElement(Idx)->isNullValue())
         continue;
       Value *Gep =
           Builder.CreateInBoundsGEP(EltTy, FirstEltPtr, Builder.getInt32(Idx));
@@ -280,9 +280,9 @@ static void scalarizeMaskedStore(CallInst *CI) {
   Value *FirstEltPtr = Builder.CreateBitCast(Ptr, NewPtrType);
   unsigned VectorWidth = VecType->getNumElements();
 
-  if (isa<ConstantVector>(Mask)) {
+  if (isa<Constant>(Mask)) {
     for (unsigned Idx = 0; Idx < VectorWidth; ++Idx) {
-      if (cast<ConstantVector>(Mask)->getOperand(Idx)->isNullValue())
+      if (cast<Constant>(Mask)->getAggregateElement(Idx)->isNullValue())
         continue;
       Value *OneElt = Builder.CreateExtractElement(Src, Builder.getInt32(Idx));
       Value *Gep =
@@ -385,9 +385,9 @@ static void scalarizeMaskedGather(CallInst *CI) {
   unsigned VectorWidth = VecType->getNumElements();
 
   // Shorten the way if the mask is a vector of constants.
-  if (isa<ConstantVector>(Mask)) {
+  if (isa<Constant>(Mask)) {
     for (unsigned Idx = 0; Idx < VectorWidth; ++Idx) {
-      if (cast<ConstantVector>(Mask)->getOperand(Idx)->isNullValue())
+      if (cast<Constant>(Mask)->getAggregateElement(Idx)->isNullValue())
         continue;
       Value *Ptr = Builder.CreateExtractElement(Ptrs, Builder.getInt32(Idx),
                                                 "Ptr" + Twine(Idx));
@@ -504,9 +504,9 @@ static void scalarizeMaskedScatter(CallInst *CI) {
   unsigned VectorWidth = Src->getType()->getVectorNumElements();
 
   // Shorten the way if the mask is a vector of constants.
-  if (isa<ConstantVector>(Mask)) {
+  if (isa<Constant>(Mask)) {
     for (unsigned Idx = 0; Idx < VectorWidth; ++Idx) {
-      if (cast<ConstantVector>(Mask)->getOperand(Idx)->isNullValue())
+      if (cast<ConstantVector>(Mask)->getAggregateElement(Idx)->isNullValue())
         continue;
       Value *OneElt = Builder.CreateExtractElement(Src, Builder.getInt32(Idx),
                                                    "Elt" + Twine(Idx));
diff --git a/llvm/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-load.ll b/llvm/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-load.ll
index d50230a67d8..cd575f8bfaa 100644
--- a/llvm/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-load.ll
+++ b/llvm/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-load.ll
@@ -41,24 +41,8 @@ define <2 x i64> @scalarize_v2i64_ones_mask(<2 x i64>* %p, <2 x i64> %passthru)
 define <2 x i64> @scalarize_v2i64_zero_mask(<2 x i64>* %p, <2 x i64> %passthru) {
 ; CHECK-LABEL: @scalarize_v2i64_zero_mask(
 ; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64>* [[P:%.*]] to i64*
-; CHECK-NEXT:    br i1 false, label [[COND_LOAD:%.*]], label [[ELSE:%.*]]
-; CHECK:       cond.load:
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, i64* [[TMP1]], i32 0
-; CHECK-NEXT:    [[TMP3:%.*]] = load i64, i64* [[TMP2]], align 8
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <2 x i64> undef, i64 [[TMP3]], i32 0
-; CHECK-NEXT:    br label [[ELSE]]
-; CHECK:       else:
-; CHECK-NEXT:    [[RES_PHI_ELSE:%.*]] = phi <2 x i64> [ [[TMP4]], [[COND_LOAD]] ], [ undef, [[TMP0:%.*]] ]
-; CHECK-NEXT:    br i1 false, label [[COND_LOAD1:%.*]], label [[ELSE2:%.*]]
-; CHECK:       cond.load1:
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, i64* [[TMP1]], i32 1
-; CHECK-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
-; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <2 x i64> [[RES_PHI_ELSE]], i64 [[TMP6]], i32 1
-; CHECK-NEXT:    br label [[ELSE2]]
-; CHECK:       else2:
-; CHECK-NEXT:    [[RES_PHI_SELECT:%.*]] = phi <2 x i64> [ [[TMP7]], [[COND_LOAD1]] ], [ [[RES_PHI_ELSE]], [[ELSE]] ]
-; CHECK-NEXT:    [[TMP8:%.*]] = select <2 x i1> zeroinitializer, <2 x i64> [[RES_PHI_SELECT]], <2 x i64> [[PASSTHRU:%.*]]
-; CHECK-NEXT:    ret <2 x i64> [[TMP8]]
+; CHECK-NEXT:    [[TMP2:%.*]] = select <2 x i1> zeroinitializer, <2 x i64> undef, <2 x i64> [[PASSTHRU:%.*]]
+; CHECK-NEXT:    ret <2 x i64> [[TMP2]]
 ;
   %ret = call <2 x i64> @llvm.masked.load.v2i64.p0v2i64(<2 x i64>* %p, i32 8, <2 x i1> <i1 false, i1 false>, <2 x i64> %passthru)
   ret <2 x i64> %ret