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
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx1010 -verify-machineinstrs %s -o - | FileCheck -check-prefix=GCN %s
define amdgpu_cs float @div_sqrt(float inreg %arg1) {
; GCN-LABEL: div_sqrt:
; GCN: ; %bb.0: ; %.entry
; GCN-NEXT: v_mul_f32_e64 v0, 0x4f800000, s0
; GCN-NEXT: v_cmp_gt_f32_e64 vcc_lo, 0xf800000, s0
; GCN-NEXT: v_cndmask_b32_e32 v0, s0, v0, vcc_lo
; GCN-NEXT: v_sqrt_f32_e32 v1, v0
; GCN-NEXT: v_add_nc_u32_e32 v2, -1, v1
; GCN-NEXT: v_add_nc_u32_e32 v3, 1, v1
; GCN-NEXT: v_fma_f32 v4, -v2, v1, v0
; GCN-NEXT: v_fma_f32 v5, -v3, v1, v0
; GCN-NEXT: v_cmp_ge_f32_e64 s0, 0, v4
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, v2, s0
; GCN-NEXT: v_cmp_lt_f32_e64 s0, 0, v5
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, v3, s0
; GCN-NEXT: v_mul_f32_e32 v2, 0x37800000, v1
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
; GCN-NEXT: v_cmp_class_f32_e64 vcc_lo, v0, 0x260
; GCN-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo
; GCN-NEXT: v_rcp_f32_e32 v0, v0
; GCN-NEXT: ; return to shader part epilog
.entry:
%a = call float @llvm.sqrt.f32(float %arg1)
%b = fdiv afn float 1.000000e+00, %a
ret float %b
}
define amdgpu_cs float @sqrt_div(float inreg %arg1) {
; GCN-LABEL: sqrt_div:
; GCN: ; %bb.0: ; %.entry
; GCN-NEXT: v_rcp_f32_e32 v0, s0
; GCN-NEXT: v_mul_f32_e32 v1, 0x4f800000, v0
; GCN-NEXT: v_cmp_gt_f32_e32 vcc_lo, 0xf800000, v0
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GCN-NEXT: v_sqrt_f32_e32 v1, v0
; GCN-NEXT: v_add_nc_u32_e32 v2, -1, v1
; GCN-NEXT: v_add_nc_u32_e32 v3, 1, v1
; GCN-NEXT: v_fma_f32 v4, -v2, v1, v0
; GCN-NEXT: v_fma_f32 v5, -v3, v1, v0
; GCN-NEXT: v_cmp_ge_f32_e64 s0, 0, v4
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, v2, s0
; GCN-NEXT: v_cmp_lt_f32_e64 s0, 0, v5
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, v3, s0
; GCN-NEXT: v_mul_f32_e32 v2, 0x37800000, v1
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
; GCN-NEXT: v_cmp_class_f32_e64 vcc_lo, v0, 0x260
; GCN-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo
; GCN-NEXT: ; return to shader part epilog
.entry:
%a = fdiv afn float 1.000000e+00, %arg1
%b = call float @llvm.sqrt.f32(float %a)
ret float %b
}
define amdgpu_cs float @rcp_sqrt(float inreg %arg1) {
; GCN-LABEL: rcp_sqrt:
; GCN: ; %bb.0: ; %.entry
; GCN-NEXT: v_mul_f32_e64 v0, 0x4f800000, s0
; GCN-NEXT: v_cmp_gt_f32_e64 vcc_lo, 0xf800000, s0
; GCN-NEXT: v_cndmask_b32_e32 v0, s0, v0, vcc_lo
; GCN-NEXT: v_sqrt_f32_e32 v1, v0
; GCN-NEXT: v_add_nc_u32_e32 v2, -1, v1
; GCN-NEXT: v_add_nc_u32_e32 v3, 1, v1
; GCN-NEXT: v_fma_f32 v4, -v2, v1, v0
; GCN-NEXT: v_fma_f32 v5, -v3, v1, v0
; GCN-NEXT: v_cmp_ge_f32_e64 s0, 0, v4
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, v2, s0
; GCN-NEXT: v_cmp_lt_f32_e64 s0, 0, v5
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, v3, s0
; GCN-NEXT: v_mul_f32_e32 v2, 0x37800000, v1
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
; GCN-NEXT: v_cmp_class_f32_e64 vcc_lo, v0, 0x260
; GCN-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo
; GCN-NEXT: v_rcp_f32_e32 v0, v0
; GCN-NEXT: ; return to shader part epilog
.entry:
%a = call float @llvm.sqrt.f32(float %arg1)
%b = call float @llvm.amdgcn.rcp.f32(float %a)
ret float %b
}
define amdgpu_cs float @sqrt_rcp(float inreg %arg1) {
; GCN-LABEL: sqrt_rcp:
; GCN: ; %bb.0: ; %.entry
; GCN-NEXT: v_rcp_f32_e32 v0, s0
; GCN-NEXT: v_mul_f32_e32 v1, 0x4f800000, v0
; GCN-NEXT: v_cmp_gt_f32_e32 vcc_lo, 0xf800000, v0
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GCN-NEXT: v_sqrt_f32_e32 v1, v0
; GCN-NEXT: v_add_nc_u32_e32 v2, -1, v1
; GCN-NEXT: v_add_nc_u32_e32 v3, 1, v1
; GCN-NEXT: v_fma_f32 v4, -v2, v1, v0
; GCN-NEXT: v_fma_f32 v5, -v3, v1, v0
; GCN-NEXT: v_cmp_ge_f32_e64 s0, 0, v4
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, v2, s0
; GCN-NEXT: v_cmp_lt_f32_e64 s0, 0, v5
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, v3, s0
; GCN-NEXT: v_mul_f32_e32 v2, 0x37800000, v1
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
; GCN-NEXT: v_cmp_class_f32_e64 vcc_lo, v0, 0x260
; GCN-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo
; GCN-NEXT: ; return to shader part epilog
.entry:
%a = call float @llvm.amdgcn.rcp.f32(float %arg1)
%b = call float @llvm.sqrt.f32(float %a)
ret float %b
}
define amdgpu_cs float @div_sqrt_contract(float inreg %arg1) {
; GCN-LABEL: div_sqrt_contract:
; GCN: ; %bb.0: ; %.entry
; GCN-NEXT: v_mul_f32_e64 v0, 0x4f800000, s0
; GCN-NEXT: v_cmp_gt_f32_e64 vcc_lo, 0xf800000, s0
; GCN-NEXT: v_cndmask_b32_e32 v0, s0, v0, vcc_lo
; GCN-NEXT: v_sqrt_f32_e32 v1, v0
; GCN-NEXT: v_add_nc_u32_e32 v2, -1, v1
; GCN-NEXT: v_add_nc_u32_e32 v3, 1, v1
; GCN-NEXT: v_fma_f32 v4, -v2, v1, v0
; GCN-NEXT: v_fma_f32 v5, -v3, v1, v0
; GCN-NEXT: v_cmp_ge_f32_e64 s0, 0, v4
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, v2, s0
; GCN-NEXT: v_cmp_lt_f32_e64 s0, 0, v5
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, v3, s0
; GCN-NEXT: v_mul_f32_e32 v2, 0x37800000, v1
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
; GCN-NEXT: v_cmp_class_f32_e64 vcc_lo, v0, 0x260
; GCN-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo
; GCN-NEXT: v_rcp_f32_e32 v0, v0
; GCN-NEXT: ; return to shader part epilog
.entry:
%a = call contract float @llvm.sqrt.f32(float %arg1)
%b = fdiv afn contract float 1.000000e+00, %a
ret float %b
}
define amdgpu_cs float @sqrt_div_contract(float inreg %arg1) {
; GCN-LABEL: sqrt_div_contract:
; GCN: ; %bb.0: ; %.entry
; GCN-NEXT: v_rcp_f32_e32 v0, s0
; GCN-NEXT: v_mul_f32_e32 v1, 0x4f800000, v0
; GCN-NEXT: v_cmp_gt_f32_e32 vcc_lo, 0xf800000, v0
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GCN-NEXT: v_sqrt_f32_e32 v1, v0
; GCN-NEXT: v_add_nc_u32_e32 v2, -1, v1
; GCN-NEXT: v_add_nc_u32_e32 v3, 1, v1
; GCN-NEXT: v_fma_f32 v4, -v2, v1, v0
; GCN-NEXT: v_fma_f32 v5, -v3, v1, v0
; GCN-NEXT: v_cmp_ge_f32_e64 s0, 0, v4
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, v2, s0
; GCN-NEXT: v_cmp_lt_f32_e64 s0, 0, v5
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, v3, s0
; GCN-NEXT: v_mul_f32_e32 v2, 0x37800000, v1
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
; GCN-NEXT: v_cmp_class_f32_e64 vcc_lo, v0, 0x260
; GCN-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo
; GCN-NEXT: ; return to shader part epilog
.entry:
%a = fdiv afn contract float 1.000000e+00, %arg1
%b = call contract float @llvm.sqrt.f32(float %a)
ret float %b
}
define amdgpu_cs float @rcp_sqrt_contract(float inreg %arg1) {
; GCN-LABEL: rcp_sqrt_contract:
; GCN: ; %bb.0: ; %.entry
; GCN-NEXT: v_mul_f32_e64 v0, 0x4f800000, s0
; GCN-NEXT: v_cmp_gt_f32_e64 vcc_lo, 0xf800000, s0
; GCN-NEXT: v_cndmask_b32_e32 v0, s0, v0, vcc_lo
; GCN-NEXT: v_sqrt_f32_e32 v1, v0
; GCN-NEXT: v_add_nc_u32_e32 v2, -1, v1
; GCN-NEXT: v_add_nc_u32_e32 v3, 1, v1
; GCN-NEXT: v_fma_f32 v4, -v2, v1, v0
; GCN-NEXT: v_fma_f32 v5, -v3, v1, v0
; GCN-NEXT: v_cmp_ge_f32_e64 s0, 0, v4
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, v2, s0
; GCN-NEXT: v_cmp_lt_f32_e64 s0, 0, v5
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, v3, s0
; GCN-NEXT: v_mul_f32_e32 v2, 0x37800000, v1
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
; GCN-NEXT: v_cmp_class_f32_e64 vcc_lo, v0, 0x260
; GCN-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo
; GCN-NEXT: v_rcp_f32_e32 v0, v0
; GCN-NEXT: ; return to shader part epilog
.entry:
%a = call contract float @llvm.sqrt.f32(float %arg1)
%b = call contract float @llvm.amdgcn.rcp.f32(float %a)
ret float %b
}
define amdgpu_cs float @sqrt_rcp_contract(float inreg %arg1) {
; GCN-LABEL: sqrt_rcp_contract:
; GCN: ; %bb.0: ; %.entry
; GCN-NEXT: v_rcp_f32_e32 v0, s0
; GCN-NEXT: v_mul_f32_e32 v1, 0x4f800000, v0
; GCN-NEXT: v_cmp_gt_f32_e32 vcc_lo, 0xf800000, v0
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GCN-NEXT: v_sqrt_f32_e32 v1, v0
; GCN-NEXT: v_add_nc_u32_e32 v2, -1, v1
; GCN-NEXT: v_add_nc_u32_e32 v3, 1, v1
; GCN-NEXT: v_fma_f32 v4, -v2, v1, v0
; GCN-NEXT: v_fma_f32 v5, -v3, v1, v0
; GCN-NEXT: v_cmp_ge_f32_e64 s0, 0, v4
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, v2, s0
; GCN-NEXT: v_cmp_lt_f32_e64 s0, 0, v5
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, v3, s0
; GCN-NEXT: v_mul_f32_e32 v2, 0x37800000, v1
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
; GCN-NEXT: v_cmp_class_f32_e64 vcc_lo, v0, 0x260
; GCN-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo
; GCN-NEXT: ; return to shader part epilog
.entry:
%a = call contract float @llvm.amdgcn.rcp.f32(float %arg1)
%b = call contract float @llvm.sqrt.f32(float %a)
ret float %b
}
declare float @llvm.sqrt.f32(float)
declare float @llvm.amdgcn.rcp.f32(float)
|
