Index: lib/Transforms/Scalar/SROA.cpp =================================================================== --- lib/Transforms/Scalar/SROA.cpp +++ lib/Transforms/Scalar/SROA.cpp @@ -3571,10 +3571,11 @@ int Idx = 0, Size = Offsets.Splits.size(); for (;;) { auto *PartTy = Type::getIntNTy(Ty->getContext(), PartSize * 8); - auto *PartPtrTy = PartTy->getPointerTo(LI->getPointerAddressSpace()); + auto AS = LI->getPointerAddressSpace(); + auto *PartPtrTy = PartTy->getPointerTo(AS); LoadInst *PLoad = IRB.CreateAlignedLoad( getAdjustedPtr(IRB, DL, BasePtr, - APInt(DL.getPointerSizeInBits(), PartOffset), + APInt(DL.getPointerSizeInBits(AS), PartOffset), PartPtrTy, BasePtr->getName() + "."), getAdjustedAlignment(LI, PartOffset, DL), /*IsVolatile*/ false, LI->getName()); Index: test/Transforms/SROA/non-zero-alloca-address-space.ll =================================================================== --- /dev/null +++ test/Transforms/SROA/non-zero-alloca-address-space.ll @@ -0,0 +1,31 @@ +; RUN: opt < %s -sroa -S | FileCheck %s +target datalayout = "e-p:64:64:64-p1:64:64:64-p5:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-n8:16:32:64-A5" + +; Test load from and store to non-zero address space. +define void @test_load_store_diff_addr_space([2 x float] addrspace(1)* %complex1, [2 x float] addrspace(1)* %complex2) { +; CHECK-LABEL: @test_load_store_diff_addr_space +; CHECK-NOT: alloca +; CHECK: load i32, i32 addrspace(1)* +; CHECK: load i32, i32 addrspace(1)* +; CHECK: store i32 %{{.*}}, i32 addrspace(1)* +; CHECK: store i32 %{{.*}}, i32 addrspace(1)* + %a0 = alloca [2 x i64], align 8, addrspace(5) + %a = getelementptr [2 x i64], [2 x i64] addrspace(5)* %a0, i32 0, i32 0 + %a.cast = bitcast i64 addrspace(5)* %a to [2 x float] addrspace(5)* + %a.gep1 = getelementptr [2 x float], [2 x float] addrspace(5)* %a.cast, i32 0, i32 0 + %a.gep2 = getelementptr [2 x float], [2 x float] addrspace(5)* %a.cast, i32 0, i32 1 + %complex1.gep = getelementptr [2 x float], [2 x float] addrspace(1)* %complex1, i32 0, i32 0 + %p1 = bitcast float addrspace(1)* %complex1.gep to i64 addrspace(1)* + %v1 = load i64, i64 addrspace(1)* %p1 + store i64 %v1, i64 addrspace(5)* %a + %f1 = load float, float addrspace(5)* %a.gep1 + %f2 = load float, float addrspace(5)* %a.gep2 + %sum = fadd float %f1, %f2 + store float %sum, float addrspace(5)* %a.gep1 + store float %sum, float addrspace(5)* %a.gep2 + %v2 = load i64, i64 addrspace(5)* %a + %complex2.gep = getelementptr [2 x float], [2 x float] addrspace(1)* %complex2, i32 0, i32 0 + %p2 = bitcast float addrspace(1)* %complex2.gep to i64 addrspace(1)* + store i64 %v2, i64 addrspace(1)* %p2 + ret void +}