1; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck %s
2; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck %s
3
4; This is used to crash in LiveIntervalAnalysis via SILoadStoreOptimizer
5; while fixing up the merge of two ds_write instructions.
6
7@tess_lds = external addrspace(3) global [8192 x i32]
8
9; CHECK-LABEL: {{^}}main:
10; CHECK-DAG: ds_write_b32
11; CHECK-DAG: ds_write_b32
12; CHECK-DAG: v_mov_b32_e32 v1, v0
13; CHECK: tbuffer_store_format_xyzw v[0:3],
14define amdgpu_vs void @main(i32 inreg %arg) {
15main_body:
16  %tmp = load float, float addrspace(3)* undef, align 4
17  %tmp1 = load float, float addrspace(3)* undef, align 4
18  store float %tmp, float addrspace(3)* null, align 4
19  %tmp2 = bitcast float %tmp to i32
20  %tmp3 = add nuw nsw i32 0, 1
21  %tmp4 = zext i32 %tmp3 to i64
22  %tmp5 = getelementptr [8192 x i32], [8192 x i32] addrspace(3)* @tess_lds, i64 0, i64 %tmp4
23  %tmp6 = bitcast i32 addrspace(3)* %tmp5 to float addrspace(3)*
24  store float %tmp1, float addrspace(3)* %tmp6, align 4
25  %tmp7 = bitcast float %tmp1 to i32
26  %tmp8 = insertelement <4 x i32> undef, i32 %tmp2, i32 0
27  %tmp9 = insertelement <4 x i32> %tmp8, i32 %tmp7, i32 1
28  %tmp10 = insertelement <4 x i32> %tmp9, i32 undef, i32 2
29  %tmp11 = insertelement <4 x i32> %tmp10, i32 undef, i32 3
30  call void @llvm.amdgcn.struct.tbuffer.store.v4i32(<4 x i32> %tmp11, <4 x i32> undef, i32 0, i32 0, i32 %arg, i32 78, i32 3) #2
31  ret void
32}
33
34declare void @llvm.amdgcn.struct.tbuffer.store.v4i32(<4 x i32>, <4 x i32>, i32, i32, i32, i32 immarg, i32 immarg) #0
35
36attributes #0 = { nounwind willreturn writeonly }
37