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