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
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
| ; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN %s
; GCN-LABEL: {{^}}select_and1:
; GCN: v_cndmask_b32_e32 [[SEL:v[0-9]+]], 0, v{{[0-9]+}},
; GCN-NOT: v_and_b32
; GCN: store_dword v[{{[0-9:]+}}], [[SEL]],
define amdgpu_kernel void @select_and1(i32 addrspace(1)* %p, i32 %x, i32 %y) {
%c = icmp slt i32 %x, 11
%s = select i1 %c, i32 0, i32 -1
%a = and i32 %y, %s
store i32 %a, i32 addrspace(1)* %p, align 4
ret void
}
; GCN-LABEL: {{^}}select_and2:
; GCN: v_cndmask_b32_e32 [[SEL:v[0-9]+]], 0, v{{[0-9]+}},
; GCN-NOT: v_and_b32
; GCN: store_dword v[{{[0-9:]+}}], [[SEL]],
define amdgpu_kernel void @select_and2(i32 addrspace(1)* %p, i32 %x, i32 %y) {
%c = icmp slt i32 %x, 11
%s = select i1 %c, i32 0, i32 -1
%a = and i32 %s, %y
store i32 %a, i32 addrspace(1)* %p, align 4
ret void
}
; GCN-LABEL: {{^}}select_and3:
; GCN: v_cndmask_b32_e32 [[SEL:v[0-9]+]], 0, v{{[0-9]+}},
; GCN-NOT: v_and_b32
; GCN: store_dword v[{{[0-9:]+}}], [[SEL]],
define amdgpu_kernel void @select_and3(i32 addrspace(1)* %p, i32 %x, i32 %y) {
%c = icmp slt i32 %x, 11
%s = select i1 %c, i32 -1, i32 0
%a = and i32 %y, %s
store i32 %a, i32 addrspace(1)* %p, align 4
ret void
}
; GCN-LABEL: {{^}}select_and_v4:
; GCN: v_cndmask_b32_e32 [[SEL:v[0-9]+]], 0, v{{[0-9]+}},
; GCN: v_cndmask_b32_e32 [[SEL:v[0-9]+]], 0, v{{[0-9]+}},
; GCN: v_cndmask_b32_e32 [[SEL:v[0-9]+]], 0, v{{[0-9]+}},
; GCN: v_cndmask_b32_e32 [[SEL:v[0-9]+]], 0, v{{[0-9]+}},
; GCN-NOT: v_and_b32
; GCN: store_dword
define amdgpu_kernel void @select_and_v4(<4 x i32> addrspace(1)* %p, i32 %x, <4 x i32> %y) {
%c = icmp slt i32 %x, 11
%s = select i1 %c, <4 x i32> zeroinitializer, <4 x i32> <i32 -1, i32 -1, i32 -1, i32 -1>
%a = and <4 x i32> %s, %y
store <4 x i32> %a, <4 x i32> addrspace(1)* %p, align 32
ret void
}
; GCN-LABEL: {{^}}select_or1:
; GCN: v_cndmask_b32_e32 [[SEL:v[0-9]+]], -1, v{{[0-9]+}},
; GCN-NOT: v_or_b32
; GCN: store_dword v[{{[0-9:]+}}], [[SEL]],
define amdgpu_kernel void @select_or1(i32 addrspace(1)* %p, i32 %x, i32 %y) {
%c = icmp slt i32 %x, 11
%s = select i1 %c, i32 0, i32 -1
%a = or i32 %y, %s
store i32 %a, i32 addrspace(1)* %p, align 4
ret void
}
; GCN-LABEL: {{^}}select_or2:
; GCN: v_cndmask_b32_e32 [[SEL:v[0-9]+]], -1, v{{[0-9]+}},
; GCN-NOT: v_or_b32
; GCN: store_dword v[{{[0-9:]+}}], [[SEL]],
define amdgpu_kernel void @select_or2(i32 addrspace(1)* %p, i32 %x, i32 %y) {
%c = icmp slt i32 %x, 11
%s = select i1 %c, i32 0, i32 -1
%a = or i32 %s, %y
store i32 %a, i32 addrspace(1)* %p, align 4
ret void
}
; GCN-LABEL: {{^}}select_or3:
; GCN: v_cndmask_b32_e32 [[SEL:v[0-9]+]], -1, v{{[0-9]+}},
; GCN-NOT: v_or_b32
; GCN: store_dword v[{{[0-9:]+}}], [[SEL]],
define amdgpu_kernel void @select_or3(i32 addrspace(1)* %p, i32 %x, i32 %y) {
%c = icmp slt i32 %x, 11
%s = select i1 %c, i32 -1, i32 0
%a = or i32 %y, %s
store i32 %a, i32 addrspace(1)* %p, align 4
ret void
}
; GCN-LABEL: {{^}}select_or_v4:
; GCN: v_cndmask_b32_e32 [[SEL:v[0-9]+]], -1, v{{[0-9]+}},
; GCN: v_cndmask_b32_e32 [[SEL:v[0-9]+]], -1, v{{[0-9]+}},
; GCN: v_cndmask_b32_e32 [[SEL:v[0-9]+]], -1, v{{[0-9]+}},
; GCN: v_cndmask_b32_e32 [[SEL:v[0-9]+]], -1, v{{[0-9]+}},
; GCN-NOT: v_or_b32
; GCN: store_dword
define amdgpu_kernel void @select_or_v4(<4 x i32> addrspace(1)* %p, i32 %x, <4 x i32> %y) {
%c = icmp slt i32 %x, 11
%s = select i1 %c, <4 x i32> zeroinitializer, <4 x i32> <i32 -1, i32 -1, i32 -1, i32 -1>
%a = or <4 x i32> %s, %y
store <4 x i32> %a, <4 x i32> addrspace(1)* %p, align 32
ret void
}
; GCN-LABEL: {{^}}sel_constants_sub_constant_sel_constants:
; GCN: v_cndmask_b32_e64 v{{[0-9]+}}, 2, 9,
define amdgpu_kernel void @sel_constants_sub_constant_sel_constants(i32 addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, i32 -4, i32 3
%bo = sub i32 5, %sel
store i32 %bo, i32 addrspace(1)* %p, align 4
ret void
}
; GCN-LABEL: {{^}}sel_constants_sub_constant_sel_constants_i16:
; TODO: shrink i16 constant. This is correct but suboptimal.
; GCN: v_mov_b32_e32 [[T:v[0-9]+]], 0xffff0009
; GCN: v_cndmask_b32_e32 v{{[0-9]+}}, 2, [[T]],
define amdgpu_kernel void @sel_constants_sub_constant_sel_constants_i16(i16 addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, i16 -4, i16 3
%bo = sub i16 5, %sel
store i16 %bo, i16 addrspace(1)* %p, align 2
ret void
}
; GCN-LABEL: {{^}}sel_constants_sub_constant_sel_constants_i16_neg:
; GCN: v_mov_b32_e32 [[F:v[0-9]+]], 0xfffff449
; GCN: v_cndmask_b32_e64 v{{[0-9]+}}, [[F]], -3,
define amdgpu_kernel void @sel_constants_sub_constant_sel_constants_i16_neg(i16 addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, i16 4, i16 3000
%bo = sub i16 1, %sel
store i16 %bo, i16 addrspace(1)* %p, align 2
ret void
}
; GCN-LABEL: {{^}}sel_constants_sub_constant_sel_constants_v2i16:
; GCN-DAG: v_mov_b32_e32 [[F:v[0-9]+]], 0x60002
; GCN-DAG: v_mov_b32_e32 [[T:v[0-9]+]], 0x50009
; GCN: v_cndmask_b32_e32 v{{[0-9]+}}, [[F]], [[T]],
define amdgpu_kernel void @sel_constants_sub_constant_sel_constants_v2i16(<2 x i16> addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, <2 x i16> <i16 -4, i16 2>, <2 x i16> <i16 3, i16 1>
%bo = sub <2 x i16> <i16 5, i16 7>, %sel
store <2 x i16> %bo, <2 x i16> addrspace(1)* %p, align 4
ret void
}
; GCN-LABEL: {{^}}sel_constants_sub_constant_sel_constants_v4i32:
; GCN-DAG: v_cndmask_b32_e64 v{{[0-9]+}}, 2, 9,
; GCN-DAG: v_cndmask_b32_e64 v{{[0-9]+}}, 6, 5,
; GCN-DAG: v_cndmask_b32_e64 v{{[0-9]+}}, 10, 6,
; GCN-DAG: v_cndmask_b32_e64 v{{[0-9]+}}, 14, 7,
define amdgpu_kernel void @sel_constants_sub_constant_sel_constants_v4i32(<4 x i32> addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, <4 x i32> <i32 -4, i32 2, i32 3, i32 4>, <4 x i32> <i32 3, i32 1, i32 -1, i32 -3>
%bo = sub <4 x i32> <i32 5, i32 7, i32 9, i32 11>, %sel
store <4 x i32> %bo, <4 x i32> addrspace(1)* %p, align 32
ret void
}
; GCN-LABEL: {{^}}sdiv_constant_sel_constants:
; GCN: v_cndmask_b32_e64 v{{[0-9]+}}, 5, 0,
define amdgpu_kernel void @sdiv_constant_sel_constants(i64 addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, i64 121, i64 23
%bo = sdiv i64 120, %sel
store i64 %bo, i64 addrspace(1)* %p, align 8
ret void
}
; GCN-LABEL: {{^}}udiv_constant_sel_constants:
; GCN: v_cndmask_b32_e64 v{{[0-9]+}}, 5, 0,
define amdgpu_kernel void @udiv_constant_sel_constants(i64 addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, i64 -4, i64 23
%bo = udiv i64 120, %sel
store i64 %bo, i64 addrspace(1)* %p, align 8
ret void
}
; GCN-LABEL: {{^}}srem_constant_sel_constants:
; GCN: v_cndmask_b32_e64 v{{[0-9]+}}, 3, 33,
define amdgpu_kernel void @srem_constant_sel_constants(i64 addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, i64 34, i64 15
%bo = srem i64 33, %sel
store i64 %bo, i64 addrspace(1)* %p, align 8
ret void
}
; GCN-LABEL: {{^}}urem_constant_sel_constants:
; GCN: v_cndmask_b32_e64 v{{[0-9]+}}, 3, 33,
define amdgpu_kernel void @urem_constant_sel_constants(i64 addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, i64 34, i64 15
%bo = urem i64 33, %sel
store i64 %bo, i64 addrspace(1)* %p, align 8
ret void
}
; GCN-LABEL: {{^}}shl_constant_sel_constants:
; GCN: v_cndmask_b32_e64 v{{[0-9]+}}, 8, 4,
define amdgpu_kernel void @shl_constant_sel_constants(i32 addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, i32 2, i32 3
%bo = shl i32 1, %sel
store i32 %bo, i32 addrspace(1)* %p, align 4
ret void
}
; GCN-LABEL: {{^}}lshr_constant_sel_constants:
; GCN: v_cndmask_b32_e64 v{{[0-9]+}}, 8, 16,
define amdgpu_kernel void @lshr_constant_sel_constants(i32 addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, i32 2, i32 3
%bo = lshr i32 64, %sel
store i32 %bo, i32 addrspace(1)* %p, align 4
ret void
}
; GCN-LABEL: {{^}}ashr_constant_sel_constants:
; GCN: v_cndmask_b32_e64 v{{[0-9]+}}, 16, 32,
define amdgpu_kernel void @ashr_constant_sel_constants(i32 addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, i32 2, i32 3
%bo = ashr i32 128, %sel
store i32 %bo, i32 addrspace(1)* %p, align 4
ret void
}
; GCN-LABEL: {{^}}fsub_constant_sel_constants:
; GCN: v_cndmask_b32_e64 v{{[0-9]+}}, -4.0, 1.0,
define amdgpu_kernel void @fsub_constant_sel_constants(float addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, float -2.0, float 3.0
%bo = fsub float -1.0, %sel
store float %bo, float addrspace(1)* %p, align 4
ret void
}
; GCN-LABEL: {{^}}fsub_constant_sel_constants_f16:
; TODO: it shall be possible to fold constants with OpSel
; GCN-DAG: v_mov_b32_e32 [[T:v[0-9]+]], 0x3c00
; GCN-DAG: v_mov_b32_e32 [[F:v[0-9]+]], 0xc400
; GCN: v_cndmask_b32_e32 v{{[0-9]+}}, [[F]], [[T]],
define amdgpu_kernel void @fsub_constant_sel_constants_f16(half addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, half -2.0, half 3.0
%bo = fsub half -1.0, %sel
store half %bo, half addrspace(1)* %p, align 2
ret void
}
; GCN-LABEL: {{^}}fsub_constant_sel_constants_v2f16:
; GCN-DAG: v_mov_b32_e32 [[T:v[0-9]+]], 0x45003c00
; GCN: v_cndmask_b32_e32 v{{[0-9]+}}, -2.0, [[T]],
define amdgpu_kernel void @fsub_constant_sel_constants_v2f16(<2 x half> addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, <2 x half> <half -2.0, half -3.0>, <2 x half> <half -1.0, half 4.0>
%bo = fsub <2 x half> <half -1.0, half 2.0>, %sel
store <2 x half> %bo, <2 x half> addrspace(1)* %p, align 4
ret void
}
; GCN-LABEL: {{^}}fsub_constant_sel_constants_v4f32:
; GCN-DAG: v_mov_b32_e32 [[T2:v[0-9]+]], 0x40a00000
; GCN-DAG: v_mov_b32_e32 [[T3:v[0-9]+]], 0x41100000
; GCN-DAG: v_mov_b32_e32 [[T4:v[0-9]+]], 0x41500000
; GCN-DAG: v_mov_b32_e32 [[F4:v[0-9]+]], 0x40c00000
; GCN-DAG: v_cndmask_b32_e64 v{{[0-9]+}}, 0, 1.0,
; GCN-DAG: v_cndmask_b32_e32 v{{[0-9]+}}, 2.0, [[T2]],
; GCN-DAG: v_cndmask_b32_e32 v{{[0-9]+}}, 4.0, [[T3]],
; GCN-DAG: v_cndmask_b32_e32 v{{[0-9]+}}, [[F4]], [[T4]],
define amdgpu_kernel void @fsub_constant_sel_constants_v4f32(<4 x float> addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, <4 x float> <float -2.0, float -3.0, float -4.0, float -5.0>, <4 x float> <float -1.0, float 0.0, float 1.0, float 2.0>
%bo = fsub <4 x float> <float -1.0, float 2.0, float 5.0, float 8.0>, %sel
store <4 x float> %bo, <4 x float> addrspace(1)* %p, align 32
ret void
}
; GCN-LABEL: {{^}}fdiv_constant_sel_constants:
; GCN: v_cndmask_b32_e64 v{{[0-9]+}}, 4.0, -2.0,
define amdgpu_kernel void @fdiv_constant_sel_constants(float addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, float -4.0, float 2.0
%bo = fdiv float 8.0, %sel
store float %bo, float addrspace(1)* %p, align 4
ret void
}
; GCN-LABEL: {{^}}frem_constant_sel_constants:
; GCN: v_cndmask_b32_e64 v{{[0-9]+}}, 2.0, 1.0,
define amdgpu_kernel void @frem_constant_sel_constants(float addrspace(1)* %p, i1 %cond) {
%sel = select i1 %cond, float -4.0, float 3.0
%bo = frem float 5.0, %sel
store float %bo, float addrspace(1)* %p, align 4
ret void
}
|