1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
|
; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=cayman -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}rcp_pat_f32:
; GCN: s_load_dword [[SRC:s[0-9]+]]
; GCN: v_rcp_f32_e32 [[RCP:v[0-9]+]], [[SRC]]
; GCN: buffer_store_dword [[RCP]]
; EG: RECIP_IEEE
define amdgpu_kernel void @rcp_pat_f32(ptr addrspace(1) %out, float %src) #0 {
%rcp = fdiv float 1.0, %src, !fpmath !0
store float %rcp, ptr addrspace(1) %out, align 4
ret void
}
; FUNC-LABEL: {{^}}rcp_ulp25_pat_f32:
; GCN: s_load_dword [[SRC:s[0-9]+]]
; GCN: v_rcp_f32_e32 [[RCP:v[0-9]+]], [[SRC]]
; GCN: buffer_store_dword [[RCP]]
; EG: RECIP_IEEE
define amdgpu_kernel void @rcp_ulp25_pat_f32(ptr addrspace(1) %out, float %src) #0 {
%rcp = fdiv float 1.0, %src, !fpmath !0
store float %rcp, ptr addrspace(1) %out, align 4
ret void
}
; FUNC-LABEL: {{^}}rcp_fast_ulp25_pat_f32:
; GCN: s_load_dword [[SRC:s[0-9]+]]
; GCN: v_rcp_f32_e32 [[RCP:v[0-9]+]], [[SRC]]
; GCN: buffer_store_dword [[RCP]]
; EG: RECIP_IEEE
define amdgpu_kernel void @rcp_fast_ulp25_pat_f32(ptr addrspace(1) %out, float %src) #0 {
%rcp = fdiv fast float 1.0, %src, !fpmath !0
store float %rcp, ptr addrspace(1) %out, align 4
ret void
}
; FUNC-LABEL: {{^}}rcp_arcp_ulp25_pat_f32:
; GCN: s_load_dword [[SRC:s[0-9]+]]
; GCN: v_rcp_f32_e32 [[RCP:v[0-9]+]], [[SRC]]
; GCN: buffer_store_dword [[RCP]]
; EG: RECIP_IEEE
define amdgpu_kernel void @rcp_arcp_ulp25_pat_f32(ptr addrspace(1) %out, float %src) #0 {
%rcp = fdiv arcp float 1.0, %src, !fpmath !0
store float %rcp, ptr addrspace(1) %out, align 4
ret void
}
; FUNC-LABEL: {{^}}rcp_global_fast_ulp25_pat_f32:
; GCN: s_load_dword [[SRC:s[0-9]+]]
; GCN: v_rcp_f32_e32 [[RCP:v[0-9]+]], [[SRC]]
; GCN: buffer_store_dword [[RCP]]
; EG: RECIP_IEEE
define amdgpu_kernel void @rcp_global_fast_ulp25_pat_f32(ptr addrspace(1) %out, float %src) #2 {
%rcp = fdiv float 1.0, %src, !fpmath !0
store float %rcp, ptr addrspace(1) %out, align 4
ret void
}
; FUNC-LABEL: {{^}}rcp_fabs_pat_f32:
; GCN: s_load_dword [[SRC:s[0-9]+]]
; GCN: v_rcp_f32_e64 [[RCP:v[0-9]+]], |[[SRC]]|
; GCN: buffer_store_dword [[RCP]]
; EG: RECIP_IEEE
define amdgpu_kernel void @rcp_fabs_pat_f32(ptr addrspace(1) %out, float %src) #0 {
%src.fabs = call float @llvm.fabs.f32(float %src)
%rcp = fdiv float 1.0, %src.fabs, !fpmath !0
store float %rcp, ptr addrspace(1) %out, align 4
ret void
}
; FUNC-LABEL: {{^}}neg_rcp_pat_f32:
; GCN: s_load_dword [[SRC:s[0-9]+]]
; GCN: v_rcp_f32_e64 [[RCP:v[0-9]+]], -[[SRC]]
; GCN: buffer_store_dword [[RCP]]
; EG: RECIP_IEEE
define amdgpu_kernel void @neg_rcp_pat_f32(ptr addrspace(1) %out, float %src) #0 {
%rcp = fdiv float -1.0, %src, !fpmath !0
store float %rcp, ptr addrspace(1) %out, align 4
ret void
}
; FUNC-LABEL: {{^}}rcp_fabs_fneg_pat_f32:
; GCN: s_load_dword [[SRC:s[0-9]+]]
; GCN: v_rcp_f32_e64 [[RCP:v[0-9]+]], -|[[SRC]]|
; GCN: buffer_store_dword [[RCP]]
define amdgpu_kernel void @rcp_fabs_fneg_pat_f32(ptr addrspace(1) %out, float %src) #0 {
%src.fabs = call float @llvm.fabs.f32(float %src)
%src.fabs.fneg = fneg float %src.fabs
%rcp = fdiv float 1.0, %src.fabs.fneg, !fpmath !0
store float %rcp, ptr addrspace(1) %out, align 4
ret void
}
; FUNC-LABEL: {{^}}rcp_fabs_fneg_pat_multi_use_f32:
; GCN: s_load_dword [[SRC:s[0-9]+]]
; GCN: v_rcp_f32_e64 [[RCP:v[0-9]+]], -|[[SRC]]|
; GCN: v_mul_f32_e64 [[MUL:v[0-9]+]], [[SRC]], -|[[SRC]]|
; GCN: buffer_store_dword [[RCP]]
; GCN: buffer_store_dword [[MUL]]
define amdgpu_kernel void @rcp_fabs_fneg_pat_multi_use_f32(ptr addrspace(1) %out, float %src) #0 {
%src.fabs = call float @llvm.fabs.f32(float %src)
%src.fabs.fneg = fneg float %src.fabs
%rcp = fdiv float 1.0, %src.fabs.fneg, !fpmath !0
store volatile float %rcp, ptr addrspace(1) %out, align 4
%other = fmul float %src, %src.fabs.fneg
store volatile float %other, ptr addrspace(1) %out, align 4
ret void
}
; FUNC-LABEL: {{^}}div_arcp_2_x_pat_f32:
; GCN: v_mul_f32_e64 [[MUL:v[0-9]+]], s{{[0-9]+}}, 0.5
; GCN: buffer_store_dword [[MUL]]
define amdgpu_kernel void @div_arcp_2_x_pat_f32(ptr addrspace(1) %out) #0 {
%x = load float, ptr addrspace(1) undef
%rcp = fdiv arcp float %x, 2.0
store float %rcp, ptr addrspace(1) %out, align 4
ret void
}
; FUNC-LABEL: {{^}}div_arcp_k_x_pat_f32:
; GCN: v_mov_b32_e32 [[V:v[0-9]+]], 0x3dcccccd
; GCN: v_mul_f32_e32 [[MUL:v[0-9]+]], s{{[0-9]+}}, [[V]]
; GCN: buffer_store_dword [[MUL]]
define amdgpu_kernel void @div_arcp_k_x_pat_f32(ptr addrspace(1) %out) #0 {
%x = load float, ptr addrspace(1) undef
%rcp = fdiv arcp float %x, 10.0
store float %rcp, ptr addrspace(1) %out, align 4
ret void
}
; FUNC-LABEL: {{^}}div_arcp_neg_k_x_pat_f32:
; GCN: v_mov_b32_e32 [[V:v[0-9]+]], 0xbdcccccd
; GCN: v_mul_f32_e32 [[MUL:v[0-9]+]], s{{[0-9]+}}, [[V]]
; GCN: buffer_store_dword [[MUL]]
define amdgpu_kernel void @div_arcp_neg_k_x_pat_f32(ptr addrspace(1) %out) #0 {
%x = load float, ptr addrspace(1) undef
%rcp = fdiv arcp float %x, -10.0
store float %rcp, ptr addrspace(1) %out, align 4
ret void
}
declare float @llvm.fabs.f32(float) #1
declare float @llvm.sqrt.f32(float) #1
attributes #0 = { nounwind "unsafe-fp-math"="false" "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
attributes #1 = { nounwind readnone }
attributes #2 = { nounwind "unsafe-fp-math"="true" "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
!0 = !{float 2.500000e+00}
|
