Mercurial > hg > CbC > CbC_llvm

comparison llvm/test/CodeGen/AMDGPU/promote-alloca-addrspacecast.ll @ 252:1f2b6ac9f198 llvm-original

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
LLVM16-1
author Shinji KONO <kono@ie.u-ryukyu.ac.jp>
date Fri, 18 Aug 2023 09:04:13 +0900
parents 1d019706d866
children
comparison
equal deleted inserted replaced
237:c80f45b162ad 252:1f2b6ac9f198
1 ; RUN: opt -S -mtriple=amdgcn-unknown-amdhsa -mcpu=kaveri -amdgpu-promote-alloca < %s | FileCheck %s 1 ; RUN: opt -S -mtriple=amdgcn-unknown-amdhsa -mcpu=kaveri -passes=amdgpu-promote-alloca < %s | FileCheck %s
2 2
3 ; The types of the users of the addrspacecast should not be changed. 3 ; The types of the users of the addrspacecast should not be changed.
4 4
5 ; CHECK-LABEL: @invalid_bitcast_addrspace( 5 ; CHECK-LABEL: @invalid_bitcast_addrspace(
6 ; CHECK: getelementptr inbounds [256 x [1 x i32]], [256 x [1 x i32]] addrspace(3)* @invalid_bitcast_addrspace.data, i32 0, i32 %14 6 ; CHECK: [[GEP:%[0-9]+]] = getelementptr inbounds [256 x [1 x i32]], ptr addrspace(3) @invalid_bitcast_addrspace.data, i32 0, i32 %{{[0-9]+}}
7 ; CHECK: bitcast [1 x i32] addrspace(3)* %{{[0-9]+}} to half addrspace(3)* 7 ; CHECK: [[ASC:%[a-z0-9]+]] = addrspacecast ptr addrspace(3) [[GEP]] to ptr
8 ; CHECK: addrspacecast half addrspace(3)* %tmp to half addrspace(4)* 8 ; CHECK: [[LOAD:%[a-z0-9]+]] = load <2 x i16>, ptr [[ASC]]
9 ; CHECK: bitcast half addrspace(4)* %tmp1 to <2 x i16> addrspace(4)* 9 ; CHECK: bitcast <2 x i16> [[LOAD]] to <2 x half>
10 define amdgpu_kernel void @invalid_bitcast_addrspace() #0 { 10 define amdgpu_kernel void @invalid_bitcast_addrspace() #0 {
11 entry: 11 entry:
12 %data = alloca [1 x i32], align 4 12 %data = alloca [1 x i32], addrspace(5)
13 %tmp = bitcast [1 x i32]* %data to half* 13 %tmp1 = addrspacecast ptr addrspace(5) %data to ptr
14 %tmp1 = addrspacecast half* %tmp to half addrspace(4)* 14 %tmp3 = load <2 x i16>, ptr %tmp1, align 2
15 %tmp2 = bitcast half addrspace(4)* %tmp1 to <2 x i16> addrspace(4)*
16 %tmp3 = load <2 x i16>, <2 x i16> addrspace(4)* %tmp2, align 2
17 %tmp4 = bitcast <2 x i16> %tmp3 to <2 x half> 15 %tmp4 = bitcast <2 x i16> %tmp3 to <2 x half>
18 ret void 16 ret void
19 } 17 }
20 18
21 attributes #0 = { nounwind "amdgpu-flat-work-group-size"="1,256" } 19 attributes #0 = { nounwind "amdgpu-flat-work-group-size"="1,256" }