1; RUN: llc < %s | FileCheck %s
2; RUN: %if ptxas %{ llc < %s | %ptxas-verify %}
3
4target triple = "nvptx64-nvidia-cuda"
5
6; CHECK-LABEL: test1
7; CHECK: ld.global.v2.f32
8; CHECK: ld.global.v2.f32
9; CHECK: st.global.v2.f32
10; CHECK: st.global.v2.f32
11define void @test1(float addrspace(1)* noalias align 8 %in, float addrspace(1)* noalias align 8 %out) {
12  %in.1 = getelementptr float, float addrspace(1)* %in, i32 1
13  %in.2 = getelementptr float, float addrspace(1)* %in, i32 2
14  %in.3 = getelementptr float, float addrspace(1)* %in, i32 3
15  %v0 = load float, float addrspace(1)* %in, align 8
16  %v1 = load float, float addrspace(1)* %in.1, align 4
17  %v2 = load float, float addrspace(1)* %in.2, align 8
18  %v3 = load float, float addrspace(1)* %in.3, align 4
19  %sum0 = fadd float %v0, %v1
20  %sum1 = fadd float %v1, %v2
21  %sum2 = fadd float %v3, %v1
22  %sum3 = fadd float %v2, %v3
23  %out.1 = getelementptr float, float addrspace(1)* %out, i32 1
24  %out.2 = getelementptr float, float addrspace(1)* %out, i32 2
25  %out.3 = getelementptr float, float addrspace(1)* %out, i32 3
26  store float %sum0, float addrspace(1)* %out, align 8
27  store float %sum1, float addrspace(1)* %out.1, align 4
28  store float %sum2, float addrspace(1)* %out.2, align 8
29  store float %sum3, float addrspace(1)* %out.3, align 4
30  ret void
31}
32