1; RUN: llc -march=amdgcn < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
2; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
3
4declare double @llvm.maxnum.f64(double, double) #0
5declare <2 x double> @llvm.maxnum.v2f64(<2 x double>, <2 x double>) #0
6declare <4 x double> @llvm.maxnum.v4f64(<4 x double>, <4 x double>) #0
7declare <8 x double> @llvm.maxnum.v8f64(<8 x double>, <8 x double>) #0
8declare <16 x double> @llvm.maxnum.v16f64(<16 x double>, <16 x double>) #0
9
10; FUNC-LABEL: @test_fmax_f64
11; SI: v_max_f64
12define amdgpu_kernel void @test_fmax_f64(double addrspace(1)* %out, double %a, double %b) nounwind {
13  %val = call double @llvm.maxnum.f64(double %a, double %b) #0
14  store double %val, double addrspace(1)* %out, align 8
15  ret void
16}
17
18; FUNC-LABEL: @test_fmax_v2f64
19; SI: v_max_f64
20; SI: v_max_f64
21define amdgpu_kernel void @test_fmax_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %a, <2 x double> %b) nounwind {
22  %val = call <2 x double> @llvm.maxnum.v2f64(<2 x double> %a, <2 x double> %b) #0
23  store <2 x double> %val, <2 x double> addrspace(1)* %out, align 16
24  ret void
25}
26
27; FUNC-LABEL: @test_fmax_v4f64
28; SI: v_max_f64
29; SI: v_max_f64
30; SI: v_max_f64
31; SI: v_max_f64
32define amdgpu_kernel void @test_fmax_v4f64(<4 x double> addrspace(1)* %out, <4 x double> %a, <4 x double> %b) nounwind {
33  %val = call <4 x double> @llvm.maxnum.v4f64(<4 x double> %a, <4 x double> %b) #0
34  store <4 x double> %val, <4 x double> addrspace(1)* %out, align 32
35  ret void
36}
37
38; FUNC-LABEL: @test_fmax_v8f64
39; SI: v_max_f64
40; SI: v_max_f64
41; SI: v_max_f64
42; SI: v_max_f64
43; SI: v_max_f64
44; SI: v_max_f64
45; SI: v_max_f64
46; SI: v_max_f64
47define amdgpu_kernel void @test_fmax_v8f64(<8 x double> addrspace(1)* %out, <8 x double> %a, <8 x double> %b) nounwind {
48  %val = call <8 x double> @llvm.maxnum.v8f64(<8 x double> %a, <8 x double> %b) #0
49  store <8 x double> %val, <8 x double> addrspace(1)* %out, align 64
50  ret void
51}
52
53; FUNC-LABEL: @test_fmax_v16f64
54; SI: v_max_f64
55; SI: v_max_f64
56; SI: v_max_f64
57; SI: v_max_f64
58; SI: v_max_f64
59; SI: v_max_f64
60; SI: v_max_f64
61; SI: v_max_f64
62; SI: v_max_f64
63; SI: v_max_f64
64; SI: v_max_f64
65; SI: v_max_f64
66; SI: v_max_f64
67; SI: v_max_f64
68; SI: v_max_f64
69; SI: v_max_f64
70define amdgpu_kernel void @test_fmax_v16f64(<16 x double> addrspace(1)* %out, <16 x double> %a, <16 x double> %b) nounwind {
71  %val = call <16 x double> @llvm.maxnum.v16f64(<16 x double> %a, <16 x double> %b) #0
72  store <16 x double> %val, <16 x double> addrspace(1)* %out, align 128
73  ret void
74}
75
76attributes #0 = { nounwind readnone }
77