1; RUN: llc -march=amdgcn -mcpu=SI -enable-misched < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=FUNC %s
2; RUN: llc -march=amdgcn -mcpu=bonaire -enable-misched < %s | FileCheck -check-prefix=CI -check-prefix=GCN -check-prefix=FUNC %s
3; RUN: llc -march=amdgcn -mcpu=tonga -enable-misched < %s | FileCheck -check-prefix=CI -check-prefix=GCN -check-prefix=FUNC %s
4
5; FUNC-LABEL: {{^}}frem_f32:
6; GCN-DAG: buffer_load_dword [[X:v[0-9]+]], {{.*$}}
7; GCN-DAG: buffer_load_dword [[Y:v[0-9]+]], {{.*}} offset:16
8; GCN-DAG: v_cmp
9; GCN-DAG: v_mul_f32
10; GCN: v_rcp_f32_e32
11; GCN: v_mul_f32_e32
12; GCN: v_mul_f32_e32
13; GCN: v_trunc_f32_e32
14; GCN: v_mad_f32
15; GCN: s_endpgm
16define void @frem_f32(float addrspace(1)* %out, float addrspace(1)* %in1,
17                      float addrspace(1)* %in2) #0 {
18   %gep2 = getelementptr float, float addrspace(1)* %in2, i32 4
19   %r0 = load float, float addrspace(1)* %in1, align 4
20   %r1 = load float, float addrspace(1)* %gep2, align 4
21   %r2 = frem float %r0, %r1
22   store float %r2, float addrspace(1)* %out, align 4
23   ret void
24}
25
26; FUNC-LABEL: {{^}}unsafe_frem_f32:
27; GCN: buffer_load_dword [[Y:v[0-9]+]], {{.*}} offset:16
28; GCN: buffer_load_dword [[X:v[0-9]+]], {{.*}}
29; GCN: v_rcp_f32_e32 [[INVY:v[0-9]+]], [[Y]]
30; GCN: v_mul_f32_e32 [[DIV:v[0-9]+]], [[INVY]], [[X]]
31; GCN: v_trunc_f32_e32 [[TRUNC:v[0-9]+]], [[DIV]]
32; GCN: v_mad_f32 [[RESULT:v[0-9]+]], -[[TRUNC]], [[Y]], [[X]]
33; GCN: buffer_store_dword [[RESULT]]
34; GCN: s_endpgm
35define void @unsafe_frem_f32(float addrspace(1)* %out, float addrspace(1)* %in1,
36                             float addrspace(1)* %in2) #1 {
37   %gep2 = getelementptr float, float addrspace(1)* %in2, i32 4
38   %r0 = load float, float addrspace(1)* %in1, align 4
39   %r1 = load float, float addrspace(1)* %gep2, align 4
40   %r2 = frem float %r0, %r1
41   store float %r2, float addrspace(1)* %out, align 4
42   ret void
43}
44
45; FUNC-LABEL: {{^}}frem_f64:
46; GCN: buffer_load_dwordx2 [[Y:v\[[0-9]+:[0-9]+\]]], {{.*}}, 0
47; GCN: buffer_load_dwordx2 [[X:v\[[0-9]+:[0-9]+\]]], {{.*}}, 0
48; GCN-DAG: v_div_fmas_f64
49; GCN-DAG: v_div_scale_f64
50; GCN-DAG: v_mul_f64
51; CI: v_trunc_f64_e32
52; CI: v_mul_f64
53; GCN: v_add_f64
54; GCN: buffer_store_dwordx2
55; GCN: s_endpgm
56define void @frem_f64(double addrspace(1)* %out, double addrspace(1)* %in1,
57                      double addrspace(1)* %in2) #0 {
58   %r0 = load double, double addrspace(1)* %in1, align 8
59   %r1 = load double, double addrspace(1)* %in2, align 8
60   %r2 = frem double %r0, %r1
61   store double %r2, double addrspace(1)* %out, align 8
62   ret void
63}
64
65; FUNC-LABEL: {{^}}unsafe_frem_f64:
66; GCN: v_rcp_f64_e32
67; GCN: v_mul_f64
68; SI: v_bfe_u32
69; CI: v_trunc_f64_e32
70; GCN: v_fma_f64
71; GCN: s_endpgm
72define void @unsafe_frem_f64(double addrspace(1)* %out, double addrspace(1)* %in1,
73                             double addrspace(1)* %in2) #1 {
74   %r0 = load double, double addrspace(1)* %in1, align 8
75   %r1 = load double, double addrspace(1)* %in2, align 8
76   %r2 = frem double %r0, %r1
77   store double %r2, double addrspace(1)* %out, align 8
78   ret void
79}
80
81define void @frem_v2f32(<2 x float> addrspace(1)* %out, <2 x float> addrspace(1)* %in1,
82                        <2 x float> addrspace(1)* %in2) #0 {
83   %gep2 = getelementptr <2 x float>, <2 x float> addrspace(1)* %in2, i32 4
84   %r0 = load <2 x float>, <2 x float> addrspace(1)* %in1, align 8
85   %r1 = load <2 x float>, <2 x float> addrspace(1)* %gep2, align 8
86   %r2 = frem <2 x float> %r0, %r1
87   store <2 x float> %r2, <2 x float> addrspace(1)* %out, align 8
88   ret void
89}
90
91define void @frem_v4f32(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in1,
92                        <4 x float> addrspace(1)* %in2) #0 {
93   %gep2 = getelementptr <4 x float>, <4 x float> addrspace(1)* %in2, i32 4
94   %r0 = load <4 x float>, <4 x float> addrspace(1)* %in1, align 16
95   %r1 = load <4 x float>, <4 x float> addrspace(1)* %gep2, align 16
96   %r2 = frem <4 x float> %r0, %r1
97   store <4 x float> %r2, <4 x float> addrspace(1)* %out, align 16
98   ret void
99}
100
101define void @frem_v2f64(<2 x double> addrspace(1)* %out, <2 x double> addrspace(1)* %in1,
102                        <2 x double> addrspace(1)* %in2) #0 {
103   %gep2 = getelementptr <2 x double>, <2 x double> addrspace(1)* %in2, i32 4
104   %r0 = load <2 x double>, <2 x double> addrspace(1)* %in1, align 16
105   %r1 = load <2 x double>, <2 x double> addrspace(1)* %gep2, align 16
106   %r2 = frem <2 x double> %r0, %r1
107   store <2 x double> %r2, <2 x double> addrspace(1)* %out, align 16
108   ret void
109}
110
111attributes #0 = { nounwind "unsafe-fp-math"="false" }
112attributes #1 = { nounwind "unsafe-fp-math"="true" }
113