Index: llvm/trunk/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp =================================================================== --- llvm/trunk/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp +++ llvm/trunk/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp @@ -21,6 +21,7 @@ #include "llvm/ADT/StringRef.h" #include "llvm/ADT/Triple.h" #include "llvm/ADT/Twine.h" +#include "llvm/Analysis/CaptureTracking.h" #include "llvm/Analysis/ValueTracking.h" #include "llvm/IR/Attributes.h" #include "llvm/IR/BasicBlock.h" @@ -608,6 +609,9 @@ } if (UseInst->getOpcode() == Instruction::AddrSpaceCast) { + // Give up if the pointer may be captured. + if (PointerMayBeCaptured(UseInst, true, true)) + return false; // Don't collect the users of this. WorkList.push_back(User); continue; Index: llvm/trunk/test/CodeGen/AMDGPU/addrspacecast-captured.ll =================================================================== --- llvm/trunk/test/CodeGen/AMDGPU/addrspacecast-captured.ll +++ llvm/trunk/test/CodeGen/AMDGPU/addrspacecast-captured.ll @@ -0,0 +1,47 @@ +; RUN: opt -S -mtriple=amdgcn-unknown-amdhsa -mcpu=kaveri -amdgpu-promote-alloca < %s | FileCheck %s + +; Nothing should be done if the addrspacecast is captured. + +declare void @consume_ptr2int(i32) #0 + +; CHECK-LABEL: @addrspacecast_captured( +; CHECK: %data = alloca i32, align 4 +; CHECK: %cast = addrspacecast i32* %data to i32 addrspace(4)* +; CHECK: %ptr2int = ptrtoint i32 addrspace(4)* %cast to i32 +; CHECK: store i32 %ptr2int, i32 addrspace(1)* %out +define void @addrspacecast_captured(i32 addrspace(1)* %out) #0 { +entry: + %data = alloca i32, align 4 + %cast = addrspacecast i32* %data to i32 addrspace(4)* + %ptr2int = ptrtoint i32 addrspace(4)* %cast to i32 + store i32 %ptr2int, i32 addrspace(1)* %out + ret void +} + +; CHECK-LABEL: @addrspacecast_captured_store( +; CHECK: %data = alloca i32, align 4 +; CHECK: %cast = addrspacecast i32* %data to i32 addrspace(4)* +; CHECK: store i32 addrspace(4)* %cast, i32 addrspace(4)* addrspace(1)* %out +define void @addrspacecast_captured_store(i32 addrspace(4)* addrspace(1)* %out) #0 { +entry: + %data = alloca i32, align 4 + %cast = addrspacecast i32* %data to i32 addrspace(4)* + store i32 addrspace(4)* %cast, i32 addrspace(4)* addrspace(1)* %out + ret void +} + +; CHECK-LABEL: @addrspacecast_captured_call( +; CHECK: %data = alloca i32, align 4 +; CHECK: %cast = addrspacecast i32* %data to i32 addrspace(4)* +; CHECK: %ptr2int = ptrtoint i32 addrspace(4)* %cast to i32 +; CHECK: call void @consume_ptr2int(i32 %ptr2int) +define void @addrspacecast_captured_call() #0 { +entry: + %data = alloca i32, align 4 + %cast = addrspacecast i32* %data to i32 addrspace(4)* + %ptr2int = ptrtoint i32 addrspace(4)* %cast to i32 + call void @consume_ptr2int(i32 %ptr2int) + ret void +} + +attributes #0 = { nounwind }