1; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
2; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
3; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
4
5; FUNC-LABEL: {{^}}fmul_f32:
6; GCN: v_mul_f32
7
8; R600: MUL_IEEE {{\** *}}{{T[0-9]+\.[XYZW]}}, KC0[2].Z, KC0[2].W
9define void @fmul_f32(float addrspace(1)* %out, float %a, float %b) {
10entry:
11  %0 = fmul float %a, %b
12  store float %0, float addrspace(1)* %out
13  ret void
14}
15
16declare float @llvm.R600.load.input(i32) readnone
17
18declare void @llvm.AMDGPU.store.output(float, i32)
19
20; FUNC-LABEL: {{^}}fmul_v2f32:
21; GCN: v_mul_f32
22; GCN: v_mul_f32
23
24; R600: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW]}}
25; R600: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW]}}
26define void @fmul_v2f32(<2 x float> addrspace(1)* %out, <2 x float> %a, <2 x float> %b) {
27entry:
28  %0 = fmul <2 x float> %a, %b
29  store <2 x float> %0, <2 x float> addrspace(1)* %out
30  ret void
31}
32
33; FUNC-LABEL: {{^}}fmul_v4f32:
34; GCN: v_mul_f32
35; GCN: v_mul_f32
36; GCN: v_mul_f32
37; GCN: v_mul_f32
38
39; R600: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
40; R600: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
41; R600: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
42; R600: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
43define void @fmul_v4f32(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in) {
44  %b_ptr = getelementptr <4 x float>, <4 x float> addrspace(1)* %in, i32 1
45  %a = load <4 x float>, <4 x float> addrspace(1) * %in
46  %b = load <4 x float>, <4 x float> addrspace(1) * %b_ptr
47  %result = fmul <4 x float> %a, %b
48  store <4 x float> %result, <4 x float> addrspace(1)* %out
49  ret void
50}
51
52; FUNC-LABEL: {{^}}test_mul_2_k:
53; GCN: v_mul_f32
54; GCN-NOT: v_mul_f32
55; GCN: s_endpgm
56define void @test_mul_2_k(float addrspace(1)* %out, float %x) #0 {
57  %y = fmul float %x, 2.0
58  %z = fmul float %y, 3.0
59  store float %z, float addrspace(1)* %out
60  ret void
61}
62
63; FUNC-LABEL: {{^}}test_mul_2_k_inv:
64; GCN: v_mul_f32
65; GCN-NOT: v_mul_f32
66; GCN-NOT: v_mad_f32
67; GCN: s_endpgm
68define void @test_mul_2_k_inv(float addrspace(1)* %out, float %x) #0 {
69  %y = fmul float %x, 3.0
70  %z = fmul float %y, 2.0
71  store float %z, float addrspace(1)* %out
72  ret void
73}
74
75; There should be three multiplies here; %a should be used twice (once
76; negated), not duplicated into mul x, 5.0 and mul x, -5.0.
77; FUNC-LABEL: {{^}}test_mul_twouse:
78; GCN: v_mul_f32
79; GCN: v_mul_f32
80; GCN: v_mul_f32
81; GCN-NOT: v_mul_f32
82define void @test_mul_twouse(float addrspace(1)* %out, float %x, float %y) #0 {
83  %a = fmul float %x, 5.0
84  %b = fsub float -0.0, %a
85  %c = fmul float %b, %y
86  %d = fmul float %c, %a
87  store float %d, float addrspace(1)* %out
88  ret void
89}
90
91attributes #0 = { nounwind }
92