InferAddressSpaces: Fix bug about replacing addrspacecast

InferAddressSpaces assumes the pointee type of addrspacecast
is the same as the operand, which is not always true and causes
invalid IR.

This bug cause build failure in HCC.

This patch fixes that.

Differential Revision: https://reviews.llvm.org/D39432

llvm-svn: 316957

diff --git a/llvm/lib/Transforms/Scalar/InferAddressSpaces.cpp b/llvm/lib/Transforms/Scalar/InferAddressSpaces.cpp
index fba80ab..7d66c0f 100644 (file)
--- a/llvm/lib/Transforms/Scalar/InferAddressSpaces.cpp
+++ b/llvm/lib/Transforms/Scalar/InferAddressSpaces.cpp
@@ -971,6 +971,11 @@ bool InferAddressSpaces::rewriteWithNewAddressSpaces(
         if (AddrSpaceCastInst *ASC = dyn_cast<AddrSpaceCastInst>(CurUser)) {
           unsigned NewAS = NewV->getType()->getPointerAddressSpace();
           if (ASC->getDestAddressSpace() == NewAS) {
+            if (ASC->getType()->getPointerElementType() !=
+                NewV->getType()->getPointerElementType()) {
+              NewV = CastInst::Create(Instruction::BitCast, NewV,
+                                      ASC->getType(), "", ASC);
+            }
             ASC->replaceAllUsesWith(NewV);
             DeadInstructions.push_back(ASC);
             continue;

diff --git a/llvm/test/Transforms/InferAddressSpaces/AMDGPU/infer-addrspacecast.ll b/llvm/test/Transforms/InferAddressSpaces/AMDGPU/infer-addrspacecast.ll
index 73829e8..74a2595 100644 (file)
--- a/llvm/test/Transforms/InferAddressSpaces/AMDGPU/infer-addrspacecast.ll
+++ b/llvm/test/Transforms/InferAddressSpaces/AMDGPU/infer-addrspacecast.ll
@@ -15,6 +15,19 @@ define void @addrspacecast_gep_addrspacecast(i32 addrspace(3)* %ptr) {
   ret void
 }
 
+; CHECK-LABEL: @addrspacecast_different_pointee_type(
+; CHECK: [[GEP:%.*]] = getelementptr i32, i32 addrspace(3)* %ptr, i64 9
+; CHECK: [[CAST:%.*]] = bitcast i32 addrspace(3)* [[GEP]] to i8 addrspace(3)*
+; CHECK-NEXT: store i8 8, i8 addrspace(3)* [[CAST]], align 8
+; CHECK-NEXT: ret void
+define void @addrspacecast_different_pointee_type(i32 addrspace(3)* %ptr) {
+  %asc0 = addrspacecast i32 addrspace(3)* %ptr to i32 addrspace(4)*
+  %gep0 = getelementptr i32, i32 addrspace(4)* %asc0, i64 9
+  %asc1 = addrspacecast i32 addrspace(4)* %gep0 to i8 addrspace(3)*
+  store i8 8, i8 addrspace(3)* %asc1, align 8
+  ret void
+}
+
 ; CHECK-LABEL: @addrspacecast_to_memory(
 ; CHECK: %gep0 = getelementptr i32, i32 addrspace(3)* %ptr, i64 9
 ; CHECK-NEXT: store volatile i32 addrspace(3)* %gep0, i32 addrspace(3)* addrspace(1)* undef