1; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
2; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
3; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
4
5
6; FUNC-LABEL: {{^}}v_fsub_f32:
7; SI: v_subrev_f32_e32 {{v[0-9]+}}, {{v[0-9]+}}, {{v[0-9]+}}
8define void @v_fsub_f32(float addrspace(1)* %out, float addrspace(1)* %in) {
9  %b_ptr = getelementptr float addrspace(1)* %in, i32 1
10  %a = load float addrspace(1)* %in, align 4
11  %b = load float addrspace(1)* %b_ptr, align 4
12  %result = fsub float %a, %b
13  store float %result, float addrspace(1)* %out, align 4
14  ret void
15}
16
17; FUNC-LABEL: {{^}}s_fsub_f32:
18; R600: ADD {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[2].Z, -KC0[2].W
19
20; SI: v_sub_f32_e32 {{v[0-9]+}}, {{s[0-9]+}}, {{v[0-9]+}}
21define void @s_fsub_f32(float addrspace(1)* %out, float %a, float %b) {
22  %sub = fsub float %a, %b
23  store float %sub, float addrspace(1)* %out, align 4
24  ret void
25}
26
27declare float @llvm.R600.load.input(i32) readnone
28
29declare void @llvm.AMDGPU.store.output(float, i32)
30
31; FUNC-LABEL: {{^}}fsub_v2f32:
32; R600-DAG: ADD {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[3].X, -KC0[3].Z
33; R600-DAG: ADD {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[2].W, -KC0[3].Y
34
35; FIXME: Should be using SGPR directly for first operand
36; SI: v_subrev_f32_e32 {{v[0-9]+}}, {{v[0-9]+}}, {{v[0-9]+}}
37; SI: v_subrev_f32_e32 {{v[0-9]+}}, {{v[0-9]+}}, {{v[0-9]+}}
38define void @fsub_v2f32(<2 x float> addrspace(1)* %out, <2 x float> %a, <2 x float> %b) {
39  %sub = fsub <2 x float> %a, %b
40  store <2 x float> %sub, <2 x float> addrspace(1)* %out, align 8
41  ret void
42}
43
44; FUNC-LABEL: {{^}}v_fsub_v4f32:
45; R600: ADD {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], -T[0-9]+\.[XYZW]}}
46; R600: ADD {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], -T[0-9]+\.[XYZW]}}
47; R600: ADD {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], -T[0-9]+\.[XYZW]}}
48; R600: ADD {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], -T[0-9]+\.[XYZW]}}
49
50; SI: v_subrev_f32_e32 {{v[0-9]+}}, {{v[0-9]+}}, {{v[0-9]+}}
51; SI: v_subrev_f32_e32 {{v[0-9]+}}, {{v[0-9]+}}, {{v[0-9]+}}
52; SI: v_subrev_f32_e32 {{v[0-9]+}}, {{v[0-9]+}}, {{v[0-9]+}}
53; SI: v_subrev_f32_e32 {{v[0-9]+}}, {{v[0-9]+}}, {{v[0-9]+}}
54define void @v_fsub_v4f32(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in) {
55  %b_ptr = getelementptr <4 x float> addrspace(1)* %in, i32 1
56  %a = load <4 x float> addrspace(1)* %in, align 16
57  %b = load <4 x float> addrspace(1)* %b_ptr, align 16
58  %result = fsub <4 x float> %a, %b
59  store <4 x float> %result, <4 x float> addrspace(1)* %out, align 16
60  ret void
61}
62
63; FIXME: Should be using SGPR directly for first operand
64
65; FUNC-LABEL: {{^}}s_fsub_v4f32:
66; SI: v_subrev_f32_e32 {{v[0-9]+}}, {{v[0-9]+}}, {{v[0-9]+}}
67; SI: v_subrev_f32_e32 {{v[0-9]+}}, {{v[0-9]+}}, {{v[0-9]+}}
68; SI: v_subrev_f32_e32 {{v[0-9]+}}, {{v[0-9]+}}, {{v[0-9]+}}
69; SI: v_subrev_f32_e32 {{v[0-9]+}}, {{v[0-9]+}}, {{v[0-9]+}}
70; SI: s_endpgm
71define void @s_fsub_v4f32(<4 x float> addrspace(1)* %out, <4 x float> %a, <4 x float> %b) {
72  %result = fsub <4 x float> %a, %b
73  store <4 x float> %result, <4 x float> addrspace(1)* %out, align 16
74  ret void
75}
76