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: v_div_scale_f32
9
10; GCN: v_rcp_f32_e32
11; GCN: v_fma_f32
12; GCN: v_mul_f32_e32
13; GCN: v_div_fmas_f32
14; GCN: v_div_fixup_f32
15; GCN: v_trunc_f32_e32
16; GCN: v_mad_f32
17; GCN: s_endpgm
18define void @frem_f32(float addrspace(1)* %out, float addrspace(1)* %in1,
19                      float addrspace(1)* %in2) #0 {
20   %gep2 = getelementptr float, float addrspace(1)* %in2, i32 4
21   %r0 = load float, float addrspace(1)* %in1, align 4
22   %r1 = load float, float addrspace(1)* %gep2, align 4
23   %r2 = frem float %r0, %r1
24   store float %r2, float addrspace(1)* %out, align 4
25   ret void
26}
27
28; FUNC-LABEL: {{^}}unsafe_frem_f32:
29; GCN: buffer_load_dword [[Y:v[0-9]+]], {{.*}} offset:16
30; GCN: buffer_load_dword [[X:v[0-9]+]], {{.*}}
31; GCN: v_rcp_f32_e32 [[INVY:v[0-9]+]], [[Y]]
32; GCN: v_mul_f32_e32 [[DIV:v[0-9]+]], [[INVY]], [[X]]
33; GCN: v_trunc_f32_e32 [[TRUNC:v[0-9]+]], [[DIV]]
34; GCN: v_mad_f32 [[RESULT:v[0-9]+]], -[[TRUNC]], [[Y]], [[X]]
35; GCN: buffer_store_dword [[RESULT]]
36; GCN: s_endpgm
37define void @unsafe_frem_f32(float addrspace(1)* %out, float addrspace(1)* %in1,
38                             float addrspace(1)* %in2) #1 {
39   %gep2 = getelementptr float, float addrspace(1)* %in2, i32 4
40   %r0 = load float, float addrspace(1)* %in1, align 4
41   %r1 = load float, float addrspace(1)* %gep2, align 4
42   %r2 = frem float %r0, %r1
43   store float %r2, float addrspace(1)* %out, align 4
44   ret void
45}
46
47; FUNC-LABEL: {{^}}frem_f64:
48; GCN: buffer_load_dwordx2 [[Y:v\[[0-9]+:[0-9]+\]]], {{.*}}, 0
49; GCN: buffer_load_dwordx2 [[X:v\[[0-9]+:[0-9]+\]]], {{.*}}, 0
50; GCN-DAG: v_div_fmas_f64
51; GCN-DAG: v_div_scale_f64
52; GCN-DAG: v_mul_f64
53; CI: v_trunc_f64_e32
54; CI: v_mul_f64
55; GCN: v_add_f64
56; GCN: buffer_store_dwordx2
57; GCN: s_endpgm
58define void @frem_f64(double addrspace(1)* %out, double addrspace(1)* %in1,
59                      double addrspace(1)* %in2) #0 {
60   %r0 = load double, double addrspace(1)* %in1, align 8
61   %r1 = load double, double addrspace(1)* %in2, align 8
62   %r2 = frem double %r0, %r1
63   store double %r2, double addrspace(1)* %out, align 8
64   ret void
65}
66
67; FUNC-LABEL: {{^}}unsafe_frem_f64:
68; GCN: v_rcp_f64_e32
69; GCN: v_mul_f64
70; SI: v_bfe_u32
71; CI: v_trunc_f64_e32
72; GCN: v_fma_f64
73; GCN: s_endpgm
74define void @unsafe_frem_f64(double addrspace(1)* %out, double addrspace(1)* %in1,
75                             double addrspace(1)* %in2) #1 {
76   %r0 = load double, double addrspace(1)* %in1, align 8
77   %r1 = load double, double addrspace(1)* %in2, align 8
78   %r2 = frem double %r0, %r1
79   store double %r2, double addrspace(1)* %out, align 8
80   ret void
81}
82
83define void @frem_v2f32(<2 x float> addrspace(1)* %out, <2 x float> addrspace(1)* %in1,
84                        <2 x float> addrspace(1)* %in2) #0 {
85   %gep2 = getelementptr <2 x float>, <2 x float> addrspace(1)* %in2, i32 4
86   %r0 = load <2 x float>, <2 x float> addrspace(1)* %in1, align 8
87   %r1 = load <2 x float>, <2 x float> addrspace(1)* %gep2, align 8
88   %r2 = frem <2 x float> %r0, %r1
89   store <2 x float> %r2, <2 x float> addrspace(1)* %out, align 8
90   ret void
91}
92
93define void @frem_v4f32(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in1,
94                        <4 x float> addrspace(1)* %in2) #0 {
95   %gep2 = getelementptr <4 x float>, <4 x float> addrspace(1)* %in2, i32 4
96   %r0 = load <4 x float>, <4 x float> addrspace(1)* %in1, align 16
97   %r1 = load <4 x float>, <4 x float> addrspace(1)* %gep2, align 16
98   %r2 = frem <4 x float> %r0, %r1
99   store <4 x float> %r2, <4 x float> addrspace(1)* %out, align 16
100   ret void
101}
102
103define void @frem_v2f64(<2 x double> addrspace(1)* %out, <2 x double> addrspace(1)* %in1,
104                        <2 x double> addrspace(1)* %in2) #0 {
105   %gep2 = getelementptr <2 x double>, <2 x double> addrspace(1)* %in2, i32 4
106   %r0 = load <2 x double>, <2 x double> addrspace(1)* %in1, align 16
107   %r1 = load <2 x double>, <2 x double> addrspace(1)* %gep2, align 16
108   %r2 = frem <2 x double> %r0, %r1
109   store <2 x double> %r2, <2 x double> addrspace(1)* %out, align 16
110   ret void
111}
112
113attributes #0 = { nounwind "unsafe-fp-math"="false" }
114attributes #1 = { nounwind "unsafe-fp-math"="true" }
115