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
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 %s
; Run with unsafe-fp-math to make sure nothing tries to turn this into 1 / rsqrt(x)
define amdgpu_kernel void @v_safe_fsqrt_f32(ptr addrspace(1) %out, ptr addrspace(1) %in) {
; R600-LABEL: v_safe_fsqrt_f32:
; R600: ; %bb.0:
; R600-NEXT: ALU 0, @8, KC0[CB0:0-32], KC1[]
; R600-NEXT: TEX 0 @6
; R600-NEXT: ALU 3, @9, KC0[CB0:0-32], KC1[]
; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
; R600-NEXT: CF_END
; R600-NEXT: PAD
; R600-NEXT: Fetch clause starting at 6:
; R600-NEXT: VTX_READ_32 T0.X, T0.X, 0, #1
; R600-NEXT: ALU clause starting at 8:
; R600-NEXT: MOV * T0.X, KC0[2].Z,
; R600-NEXT: ALU clause starting at 9:
; R600-NEXT: RECIPSQRT_IEEE * T0.X, T0.X,
; R600-NEXT: LSHR T1.X, KC0[2].Y, literal.x,
; R600-NEXT: RECIP_IEEE * T0.X, PS,
; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%r0 = load float, ptr addrspace(1) %in
%r1 = call float @llvm.sqrt.f32(float %r0)
store float %r1, ptr addrspace(1) %out
ret void
}
define amdgpu_kernel void @v_unsafe_fsqrt_f32(ptr addrspace(1) %out, ptr addrspace(1) %in) #1 {
; R600-LABEL: v_unsafe_fsqrt_f32:
; R600: ; %bb.0:
; R600-NEXT: ALU 0, @8, KC0[CB0:0-32], KC1[]
; R600-NEXT: TEX 0 @6
; R600-NEXT: ALU 3, @9, KC0[CB0:0-32], KC1[]
; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
; R600-NEXT: CF_END
; R600-NEXT: PAD
; R600-NEXT: Fetch clause starting at 6:
; R600-NEXT: VTX_READ_32 T0.X, T0.X, 0, #1
; R600-NEXT: ALU clause starting at 8:
; R600-NEXT: MOV * T0.X, KC0[2].Z,
; R600-NEXT: ALU clause starting at 9:
; R600-NEXT: RECIPSQRT_IEEE * T0.X, T0.X,
; R600-NEXT: LSHR T1.X, KC0[2].Y, literal.x,
; R600-NEXT: RECIP_IEEE * T0.X, PS,
; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%r0 = load float, ptr addrspace(1) %in
%r1 = call float @llvm.sqrt.f32(float %r0)
store float %r1, ptr addrspace(1) %out
ret void
}
define amdgpu_kernel void @s_sqrt_f32(ptr addrspace(1) %out, float %in) {
; R600-LABEL: s_sqrt_f32:
; R600: ; %bb.0: ; %entry
; R600-NEXT: ALU 3, @4, KC0[CB0:0-32], KC1[]
; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
; R600-NEXT: CF_END
; R600-NEXT: PAD
; R600-NEXT: ALU clause starting at 4:
; R600-NEXT: RECIPSQRT_IEEE * T0.X, KC0[2].Z,
; R600-NEXT: LSHR T1.X, KC0[2].Y, literal.x,
; R600-NEXT: RECIP_IEEE * T0.X, PS,
; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
entry:
%fdiv = call float @llvm.sqrt.f32(float %in)
store float %fdiv, ptr addrspace(1) %out
ret void
}
define amdgpu_kernel void @s_sqrt_v2f32(ptr addrspace(1) %out, <2 x float> %in) {
; R600-LABEL: s_sqrt_v2f32:
; R600: ; %bb.0: ; %entry
; R600-NEXT: ALU 5, @4, KC0[CB0:0-32], KC1[]
; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
; R600-NEXT: CF_END
; R600-NEXT: PAD
; R600-NEXT: ALU clause starting at 4:
; R600-NEXT: RECIPSQRT_IEEE * T0.X, KC0[2].W,
; R600-NEXT: RECIPSQRT_IEEE * T0.Y, KC0[3].X,
; R600-NEXT: RECIP_IEEE * T0.Y, PS,
; R600-NEXT: LSHR T1.X, KC0[2].Y, literal.x,
; R600-NEXT: RECIP_IEEE * T0.X, T0.X,
; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
entry:
%fdiv = call <2 x float> @llvm.sqrt.v2f32(<2 x float> %in)
store <2 x float> %fdiv, ptr addrspace(1) %out
ret void
}
define amdgpu_kernel void @s_sqrt_v4f32(ptr addrspace(1) %out, <4 x float> %in) {
; R600-LABEL: s_sqrt_v4f32:
; R600: ; %bb.0: ; %entry
; R600-NEXT: ALU 9, @4, KC0[CB0:0-32], KC1[]
; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T1.X, 1
; R600-NEXT: CF_END
; R600-NEXT: PAD
; R600-NEXT: ALU clause starting at 4:
; R600-NEXT: RECIPSQRT_IEEE * T0.X, KC0[3].Y,
; R600-NEXT: RECIPSQRT_IEEE * T0.Y, KC0[3].Z,
; R600-NEXT: RECIPSQRT_IEEE * T0.Z, KC0[3].W,
; R600-NEXT: RECIPSQRT_IEEE * T0.W, KC0[4].X,
; R600-NEXT: RECIP_IEEE * T0.W, PS,
; R600-NEXT: RECIP_IEEE * T0.Z, T0.Z,
; R600-NEXT: RECIP_IEEE * T0.Y, T0.Y,
; R600-NEXT: LSHR T1.X, KC0[2].Y, literal.x,
; R600-NEXT: RECIP_IEEE * T0.X, T0.X,
; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
entry:
%fdiv = call <4 x float> @llvm.sqrt.v4f32(<4 x float> %in)
store <4 x float> %fdiv, ptr addrspace(1) %out
ret void
}
define amdgpu_kernel void @elim_redun_check_neg0(ptr addrspace(1) %out, float %in) {
; R600-LABEL: elim_redun_check_neg0:
; R600: ; %bb.0: ; %entry
; R600-NEXT: ALU 3, @4, KC0[CB0:0-32], KC1[]
; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
; R600-NEXT: CF_END
; R600-NEXT: PAD
; R600-NEXT: ALU clause starting at 4:
; R600-NEXT: RECIPSQRT_IEEE * T0.X, KC0[2].Z,
; R600-NEXT: LSHR T1.X, KC0[2].Y, literal.x,
; R600-NEXT: RECIP_IEEE * T0.X, PS,
; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
entry:
%sqrt = call float @llvm.sqrt.f32(float %in)
%cmp = fcmp olt float %in, -0.000000e+00
%res = select i1 %cmp, float 0x7FF8000000000000, float %sqrt
store float %res, ptr addrspace(1) %out
ret void
}
define amdgpu_kernel void @elim_redun_check_pos0(ptr addrspace(1) %out, float %in) {
; R600-LABEL: elim_redun_check_pos0:
; R600: ; %bb.0: ; %entry
; R600-NEXT: ALU 3, @4, KC0[CB0:0-32], KC1[]
; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
; R600-NEXT: CF_END
; R600-NEXT: PAD
; R600-NEXT: ALU clause starting at 4:
; R600-NEXT: RECIPSQRT_IEEE * T0.X, KC0[2].Z,
; R600-NEXT: LSHR T1.X, KC0[2].Y, literal.x,
; R600-NEXT: RECIP_IEEE * T0.X, PS,
; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
entry:
%sqrt = call float @llvm.sqrt.f32(float %in)
%cmp = fcmp olt float %in, 0.000000e+00
%res = select i1 %cmp, float 0x7FF8000000000000, float %sqrt
store float %res, ptr addrspace(1) %out
ret void
}
define amdgpu_kernel void @elim_redun_check_ult(ptr addrspace(1) %out, float %in) {
; R600-LABEL: elim_redun_check_ult:
; R600: ; %bb.0: ; %entry
; R600-NEXT: ALU 3, @4, KC0[CB0:0-32], KC1[]
; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
; R600-NEXT: CF_END
; R600-NEXT: PAD
; R600-NEXT: ALU clause starting at 4:
; R600-NEXT: RECIPSQRT_IEEE * T0.X, KC0[2].Z,
; R600-NEXT: LSHR T1.X, KC0[2].Y, literal.x,
; R600-NEXT: RECIP_IEEE * T0.X, PS,
; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
entry:
%sqrt = call float @llvm.sqrt.f32(float %in)
%cmp = fcmp ult float %in, -0.000000e+00
%res = select i1 %cmp, float 0x7FF8000000000000, float %sqrt
store float %res, ptr addrspace(1) %out
ret void
}
define amdgpu_kernel void @elim_redun_check_v2(ptr addrspace(1) %out, <2 x float> %in) {
; R600-LABEL: elim_redun_check_v2:
; R600: ; %bb.0: ; %entry
; R600-NEXT: ALU 5, @4, KC0[CB0:0-32], KC1[]
; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
; R600-NEXT: CF_END
; R600-NEXT: PAD
; R600-NEXT: ALU clause starting at 4:
; R600-NEXT: RECIPSQRT_IEEE * T0.X, KC0[2].W,
; R600-NEXT: RECIPSQRT_IEEE * T0.Y, KC0[3].X,
; R600-NEXT: RECIP_IEEE * T0.Y, PS,
; R600-NEXT: LSHR T1.X, KC0[2].Y, literal.x,
; R600-NEXT: RECIP_IEEE * T0.X, T0.X,
; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
entry:
%sqrt = call <2 x float> @llvm.sqrt.v2f32(<2 x float> %in)
%cmp = fcmp olt <2 x float> %in, <float -0.000000e+00, float -0.000000e+00>
%res = select <2 x i1> %cmp, <2 x float> <float 0x7FF8000000000000, float 0x7FF8000000000000>, <2 x float> %sqrt
store <2 x float> %res, ptr addrspace(1) %out
ret void
}
define amdgpu_kernel void @elim_redun_check_v2_ult(ptr addrspace(1) %out, <2 x float> %in) {
; R600-LABEL: elim_redun_check_v2_ult:
; R600: ; %bb.0: ; %entry
; R600-NEXT: ALU 5, @4, KC0[CB0:0-32], KC1[]
; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
; R600-NEXT: CF_END
; R600-NEXT: PAD
; R600-NEXT: ALU clause starting at 4:
; R600-NEXT: RECIPSQRT_IEEE * T0.X, KC0[2].W,
; R600-NEXT: RECIPSQRT_IEEE * T0.Y, KC0[3].X,
; R600-NEXT: RECIP_IEEE * T0.Y, PS,
; R600-NEXT: LSHR T1.X, KC0[2].Y, literal.x,
; R600-NEXT: RECIP_IEEE * T0.X, T0.X,
; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
entry:
%sqrt = call <2 x float> @llvm.sqrt.v2f32(<2 x float> %in)
%cmp = fcmp ult <2 x float> %in, <float -0.000000e+00, float -0.000000e+00>
%res = select <2 x i1> %cmp, <2 x float> <float 0x7FF8000000000000, float 0x7FF8000000000000>, <2 x float> %sqrt
store <2 x float> %res, ptr addrspace(1) %out
ret void
}
define amdgpu_kernel void @recip_sqrt(ptr addrspace(1) %out, float %src) nounwind {
; R600-LABEL: recip_sqrt:
; R600: ; %bb.0:
; R600-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
; R600-NEXT: CF_END
; R600-NEXT: PAD
; R600-NEXT: ALU clause starting at 4:
; R600-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
; R600-NEXT: RECIPSQRT_IEEE * T1.X, KC0[2].Z,
; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%sqrt = call float @llvm.sqrt.f32(float %src)
%recipsqrt = fdiv fast float 1.0, %sqrt
store float %recipsqrt, ptr addrspace(1) %out, align 4
ret void
}
declare float @llvm.sqrt.f32(float %in) #0
declare <2 x float> @llvm.sqrt.v2f32(<2 x float> %in) #0
declare <4 x float> @llvm.sqrt.v4f32(<4 x float> %in) #0
attributes #0 = { nounwind readnone }
attributes #1 = { nounwind "unsafe-fp-math"="true" }
|
