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
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
|
; RUN: llc < %s -march=nvptx -mcpu=sm_20 | FileCheck %s
declare float @llvm.sqrt.f32(float)
declare double @llvm.sqrt.f64(double)
; CHECK-LABEL: sqrt_div(
; CHECK: sqrt.rn.f32
; CHECK: div.rn.f32
define float @sqrt_div(float %a, float %b) {
%t1 = tail call float @llvm.sqrt.f32(float %a)
%t2 = fdiv float %t1, %b
ret float %t2
}
; CHECK-LABEL: sqrt_div_fast(
; CHECK: sqrt.rn.f32
; CHECK: div.approx.f32
define float @sqrt_div_fast(float %a, float %b) #0 {
%t1 = tail call float @llvm.sqrt.f32(float %a)
%t2 = fdiv float %t1, %b
ret float %t2
}
; CHECK-LABEL: sqrt_div_fast_ninf(
; CHECK: sqrt.approx.f32
; CHECK: div.approx.f32
define float @sqrt_div_fast_ninf(float %a, float %b) #0 {
%t1 = tail call ninf afn float @llvm.sqrt.f32(float %a)
%t2 = fdiv float %t1, %b
ret float %t2
}
; CHECK-LABEL: sqrt_div_ftz(
; CHECK: sqrt.rn.ftz.f32
; CHECK: div.rn.ftz.f32
define float @sqrt_div_ftz(float %a, float %b) #1 {
%t1 = tail call float @llvm.sqrt.f32(float %a)
%t2 = fdiv float %t1, %b
ret float %t2
}
; CHECK-LABEL: sqrt_div_fast_ftz(
; CHECK: sqrt.rn.ftz.f32
; CHECK: div.approx.ftz.f32
define float @sqrt_div_fast_ftz(float %a, float %b) #0 #1 {
%t1 = tail call float @llvm.sqrt.f32(float %a)
%t2 = fdiv float %t1, %b
ret float %t2
}
; CHECK-LABEL: sqrt_div_fast_ftz_ninf(
; CHECK: sqrt.approx.ftz.f32
; CHECK: div.approx.ftz.f32
define float @sqrt_div_fast_ftz_ninf(float %a, float %b) #0 #1 {
%t1 = tail call ninf afn float @llvm.sqrt.f32(float %a)
%t2 = fdiv float %t1, %b
ret float %t2
}
; There are no fast-math or ftz versions of sqrt and div for f64. We use
; reciprocal(rsqrt(x)) for sqrt(x), and emit a vanilla divide.
; CHECK-LABEL: sqrt_div_fast_ftz_f64(
; CHECK: sqrt.rn.f64
; CHECK: div.rn.f64
define double @sqrt_div_fast_ftz_f64(double %a, double %b) #0 #1 {
%t1 = tail call double @llvm.sqrt.f64(double %a)
%t2 = fdiv double %t1, %b
ret double %t2
}
; CHECK-LABEL: sqrt_div_fast_ftz_f64_ninf(
; CHECK: rsqrt.approx.f64
; CHECK: rcp.approx.ftz.f64
; CHECK: div.rn.f64
define double @sqrt_div_fast_ftz_f64_ninf(double %a, double %b) #0 #1 {
%t1 = tail call ninf afn double @llvm.sqrt.f64(double %a)
%t2 = fdiv double %t1, %b
ret double %t2
}
; CHECK-LABEL: rsqrt(
; CHECK-NOT: rsqrt.approx
; CHECK: sqrt.rn.f32
; CHECK-NOT: rsqrt.approx
define float @rsqrt(float %a) {
%b = tail call float @llvm.sqrt.f32(float %a)
%ret = fdiv float 1.0, %b
ret float %ret
}
; CHECK-LABEL: rsqrt_fast(
; CHECK-NOT: div.
; CHECK-NOT: sqrt.
; CHECK: rsqrt.approx.f32
; CHECK-NOT: div.
; CHECK-NOT: sqrt.
define float @rsqrt_fast(float %a) #0 {
%b = tail call float @llvm.sqrt.f32(float %a)
%ret = fdiv float 1.0, %b
ret float %ret
}
; CHECK-LABEL: rsqrt_fast_ftz(
; CHECK-NOT: div.
; CHECK-NOT: sqrt.
; CHECK: rsqrt.approx.ftz.f32
; CHECK-NOT: div.
; CHECK-NOT: sqrt.
define float @rsqrt_fast_ftz(float %a) #0 #1 {
%b = tail call float @llvm.sqrt.f32(float %a)
%ret = fdiv float 1.0, %b
ret float %ret
}
; CHECK-LABEL: fadd
; CHECK: add.rn.f32
define float @fadd(float %a, float %b) {
%t1 = fadd float %a, %b
ret float %t1
}
; CHECK-LABEL: fadd_ftz
; CHECK: add.rn.ftz.f32
define float @fadd_ftz(float %a, float %b) #1 {
%t1 = fadd float %a, %b
ret float %t1
}
declare float @llvm.sin.f32(float)
declare float @llvm.cos.f32(float)
; CHECK-LABEL: fsin_approx
; CHECK: sin.approx.f32
define float @fsin_approx(float %a) #0 {
%r = tail call float @llvm.sin.f32(float %a)
ret float %r
}
; CHECK-LABEL: fcos_approx
; CHECK: cos.approx.f32
define float @fcos_approx(float %a) #0 {
%r = tail call float @llvm.cos.f32(float %a)
ret float %r
}
; CHECK-LABEL: repeated_div_recip_allowed
define float @repeated_div_recip_allowed(i1 %pred, float %a, float %b, float %divisor) {
; CHECK: rcp.rn.f32
; CHECK: mul.rn.f32
; CHECK: mul.rn.f32
; CHECK: mul.rn.f32
; CHECK: selp.f32
%x = fdiv arcp float %a, %divisor
%y = fdiv arcp float %b, %divisor
%z = fmul float %x, %y
%w = select i1 %pred, float %z, float %y
ret float %w
}
; CHECK-LABEL: repeated_div_recip_allowed_sel
define float @repeated_div_recip_allowed_sel(i1 %pred, float %a, float %b, float %divisor) {
; CHECK: selp.f32
; CHECK: div.rn.f32
%x = fdiv arcp float %a, %divisor
%y = fdiv arcp float %b, %divisor
%w = select i1 %pred, float %x, float %y
ret float %w
}
; CHECK-LABEL: repeated_div_recip_allowed_ftz
define float @repeated_div_recip_allowed_ftz(i1 %pred, float %a, float %b, float %divisor) #1 {
; CHECK: rcp.rn.ftz.f32
; CHECK: mul.rn.ftz.f32
; CHECK: mul.rn.ftz.f32
; CHECK: mul.rn.ftz.f32
; CHECK: selp.f32
%x = fdiv arcp float %a, %divisor
%y = fdiv arcp float %b, %divisor
%z = fmul float %x, %y
%w = select i1 %pred, float %z, float %y
ret float %w
}
; CHECK-LABEL: repeated_div_recip_allowed_ftz_sel
define float @repeated_div_recip_allowed_ftz_sel(i1 %pred, float %a, float %b, float %divisor) #1 {
; CHECK: selp.f32
; CHECK: div.rn.ftz.f32
%x = fdiv arcp float %a, %divisor
%y = fdiv arcp float %b, %divisor
%w = select i1 %pred, float %x, float %y
ret float %w
}
; CHECK-LABEL: repeated_div_fast
define float @repeated_div_fast(i1 %pred, float %a, float %b, float %divisor) #0 {
; CHECK: rcp.approx.f32
; CHECK: mul.f32
; CHECK: mul.f32
; CHECK: mul.f32
; CHECK: selp.f32
%x = fdiv float %a, %divisor
%y = fdiv float %b, %divisor
%z = fmul float %x, %y
%w = select i1 %pred, float %z, float %y
ret float %w
}
; CHECK-LABEL: repeated_div_fast_sel
define float @repeated_div_fast_sel(i1 %pred, float %a, float %b, float %divisor) #0 {
; CHECK: selp.f32
; CHECK: div.approx.f32
%x = fdiv float %a, %divisor
%y = fdiv float %b, %divisor
%w = select i1 %pred, float %x, float %y
ret float %w
}
; CHECK-LABEL: repeated_div_fast_ftz
define float @repeated_div_fast_ftz(i1 %pred, float %a, float %b, float %divisor) #0 #1 {
; CHECK: rcp.approx.ftz.f32
; CHECK: mul.ftz.f32
; CHECK: mul.ftz.f32
; CHECK: mul.ftz.f32
; CHECK: selp.f32
%x = fdiv float %a, %divisor
%y = fdiv float %b, %divisor
%z = fmul float %x, %y
%w = select i1 %pred, float %z, float %y
ret float %w
}
; CHECK-LABEL: repeated_div_fast_ftz_sel
define float @repeated_div_fast_ftz_sel(i1 %pred, float %a, float %b, float %divisor) #0 #1 {
; CHECK: selp.f32
; CHECK: div.approx.ftz.f32
%x = fdiv float %a, %divisor
%y = fdiv float %b, %divisor
%w = select i1 %pred, float %x, float %y
ret float %w
}
attributes #0 = { "unsafe-fp-math" = "true" }
attributes #1 = { "denormal-fp-math-f32" = "preserve-sign" }
|
