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
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
|
// RUN: mlir-translate -mlir-to-llvmir %s | FileCheck %s
llvm.func @rocdl_special_regs() -> i32 {
// CHECK-LABEL: rocdl_special_regs
// CHECK: call i32 @llvm.amdgcn.workitem.id.x()
%1 = rocdl.workitem.id.x : i32
// CHECK: call i32 @llvm.amdgcn.workitem.id.y()
%2 = rocdl.workitem.id.y : i32
// CHECK: call i32 @llvm.amdgcn.workitem.id.z()
%3 = rocdl.workitem.id.z : i32
// CHECK: call i32 @llvm.amdgcn.workgroup.id.x()
%4 = rocdl.workgroup.id.x : i32
// CHECK: call i32 @llvm.amdgcn.workgroup.id.y()
%5 = rocdl.workgroup.id.y : i32
// CHECK: call i32 @llvm.amdgcn.workgroup.id.z()
%6 = rocdl.workgroup.id.z : i32
// CHECK: call i64 @__ockl_get_local_size(i32 0)
%7 = rocdl.workgroup.dim.x : i64
// CHECK: call i64 @__ockl_get_local_size(i32 1)
%8 = rocdl.workgroup.dim.y : i64
// CHECK: call i64 @__ockl_get_local_size(i32 2)
%9 = rocdl.workgroup.dim.z : i64
// CHECK: call i64 @__ockl_get_num_groups(i32 0)
%10 = rocdl.grid.dim.x : i64
// CHECK: call i64 @__ockl_get_num_groups(i32 1)
%11 = rocdl.grid.dim.y : i64
// CHECK: call i64 @__ockl_get_num_groups(i32 2)
%12 = rocdl.grid.dim.z : i64
// CHECK: call i32 @llvm.amdgcn.workitem.id.x(),{{.*}} !range ![[$RANGE:[0-9]+]]
%13 = rocdl.workitem.id.x {range = array<i32: 0, 64>} : i32
llvm.return %1 : i32
}
llvm.func @kernel_func() attributes {rocdl.kernel} {
// CHECK-LABEL: amdgpu_kernel void @kernel_func()
// CHECK: #[[$KERNEL_ATTRS:[0-9]+]]
llvm.return
}
llvm.func @kernel_func_workgroups()
attributes {rocdl.kernel, rocdl.max_flat_work_group_size = 1024 : index} {
// CHECK-LABEL: amdgpu_kernel void @kernel_func_workgroups()
// CHECK: #[[$KERNEL_WORKGROUP_ATTRS:[0-9]+]]
llvm.return
}
llvm.func @known_block_sizes()
attributes {rocdl.kernel,
rocdl.flat_work_group_size = "128,128",
rocdl.reqd_work_group_size = array<i32: 16, 4, 2>} {
// CHECK-LABEL: amdgpu_kernel void @known_block_sizes()
// CHECK: #[[$KNOWN_BLOCK_SIZE_ATTRS:[0-9]+]]
// CHECK: !reqd_work_group_size ![[$REQD_WORK_GROUP_SIZE:[0-9]+]]
llvm.return
}
llvm.func @rocdl.barrier() {
// CHECK: fence syncscope("workgroup") release
// CHECK-NEXT: call void @llvm.amdgcn.s.barrier()
// CHECK-NEXT: fence syncscope("workgroup") acquire
rocdl.barrier
llvm.return
}
llvm.func @rocdl.xdlops(%arg0 : f32, %arg1 : f32,
%arg2 : vector<32 x f32>, %arg3: i32,
%arg4 : vector<16 x f32>, %arg5 : vector<4xf32>,
%arg6 : vector<4xf16>, %arg7 : vector<32 x i32>,
%arg8 : vector<16 x i32>, %arg9 : vector<4xi32>,
%arg10 : vector<2xi16>, %arg11 : i64) -> vector<32 x f32> {
%csti32 = llvm.mlir.constant(42 : i32) : i32
// CHECK-LABEL: rocdl.xdlops
// CHECK: call <32 x float> @llvm.amdgcn.mfma.f32.32x32x1f32(float %{{.*}}, float %{{.*}}, <32 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r0 = rocdl.mfma.f32.32x32x1f32 %arg0, %arg1, %arg2, %csti32, %csti32, %csti32 :
(f32, f32, vector<32 x f32>,
i32, i32, i32) -> vector<32 x f32>
// CHECK: call <16 x float> @llvm.amdgcn.mfma.f32.16x16x1f32(float %{{.*}}, float %{{.*}}, <16 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r1 = rocdl.mfma.f32.16x16x1f32 %arg0, %arg1, %arg4, %csti32, %csti32, %csti32 :
(f32, f32, vector<16 x f32>,
i32, i32, i32) -> vector<16 x f32>
// CHECK: call <4 x float> @llvm.amdgcn.mfma.f32.16x16x4f32(float %{{.*}}, float %{{.*}}, <4 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r2 = rocdl.mfma.f32.16x16x4f32 %arg0, %arg1, %arg5, %csti32, %csti32, %csti32 :
(f32, f32, vector<4xf32>,
i32, i32, i32) -> vector<4xf32>
// CHECK: call <4 x float> @llvm.amdgcn.mfma.f32.4x4x1f32(float %{{.*}}, float %{{.*}}, <4 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r3 = rocdl.mfma.f32.4x4x1f32 %arg0, %arg1, %arg5, %csti32, %csti32, %csti32 :
(f32, f32, vector<4xf32>,
i32, i32, i32) -> vector<4xf32>
// CHECK: call <16 x float> @llvm.amdgcn.mfma.f32.32x32x2f32(float %{{.*}}, float %{{.*}}, <16 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r4= rocdl.mfma.f32.32x32x2f32 %arg0, %arg1, %arg4, %csti32, %csti32, %csti32 :
(f32, f32, vector<16 x f32>,
i32, i32, i32) -> vector<16 x f32>
// CHECK: call <32 x float> @llvm.amdgcn.mfma.f32.32x32x4f16(<4 x half> %{{.*}}, <4 x half> %{{.*}}, <32 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r5 = rocdl.mfma.f32.32x32x4f16 %arg6, %arg6, %arg2, %csti32, %csti32, %csti32 :
(vector<4xf16>, vector<4xf16>, vector<32 x f32>,
i32, i32, i32) -> vector<32 x f32>
// CHECK: call <16 x float> @llvm.amdgcn.mfma.f32.16x16x4f16(<4 x half> %{{.*}}, <4 x half> %{{.*}}, <16 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r6 = rocdl.mfma.f32.16x16x4f16 %arg6, %arg6, %arg4, %csti32, %csti32, %csti32 :
(vector<4xf16>, vector<4xf16>, vector<16 x f32>,
i32, i32, i32) -> vector<16 x f32>
// CHECK: call <4 x float> @llvm.amdgcn.mfma.f32.4x4x4f16(<4 x half> %{{.*}}, <4 x half> %{{.*}}, <4 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r7 = rocdl.mfma.f32.4x4x4f16 %arg6, %arg6, %arg5, %csti32, %csti32, %csti32 :
(vector<4xf16>, vector<4xf16>, vector<4xf32>,
i32, i32, i32) -> vector<4xf32>
// CHECK: call <16 x float> @llvm.amdgcn.mfma.f32.32x32x8f16(<4 x half> %{{.*}}, <4 x half> %{{.*}}, <16 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r8 = rocdl.mfma.f32.32x32x8f16 %arg6, %arg6, %arg4, %csti32, %csti32, %csti32 :
(vector<4xf16>, vector<4xf16>, vector<16 x f32>,
i32, i32, i32) -> vector<16 x f32>
// CHECK: call <4 x float> @llvm.amdgcn.mfma.f32.16x16x16f16(<4 x half> %{{.*}}, <4 x half> %{{.*}}, <4 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r9 = rocdl.mfma.f32.16x16x16f16 %arg6, %arg6, %arg5, %csti32, %csti32, %csti32 :
(vector<4xf16>, vector<4xf16>, vector<4xf32>,
i32, i32, i32) -> vector<4xf32>
// CHECK: call <32 x i32> @llvm.amdgcn.mfma.i32.32x32x4i8(i32 %{{.*}}, i32 %{{.*}}, <32 x i32> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r10 = rocdl.mfma.i32.32x32x4i8 %arg3, %arg3, %arg7, %csti32, %csti32, %csti32 :
(i32, i32, vector<32 x i32>,
i32, i32, i32) -> vector<32 x i32>
// CHECK: call <16 x i32> @llvm.amdgcn.mfma.i32.16x16x4i8(i32 %{{.*}}, i32 %{{.*}}, <16 x i32> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r11 = rocdl.mfma.i32.16x16x4i8 %arg3, %arg3, %arg8, %csti32, %csti32, %csti32 :
(i32, i32, vector<16 x i32>,
i32, i32, i32) -> vector<16 x i32>
// CHECK: call <4 x i32> @llvm.amdgcn.mfma.i32.4x4x4i8(i32 %{{.*}}, i32 %{{.*}}, <4 x i32> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r12 = rocdl.mfma.i32.4x4x4i8 %arg3, %arg3, %arg9, %csti32, %csti32, %csti32 :
(i32, i32, vector<4xi32>,
i32, i32, i32) -> vector<4xi32>
// CHECK: call <16 x i32> @llvm.amdgcn.mfma.i32.32x32x8i8(i32 %{{.*}}, i32 %{{.*}}, <16 x i32> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r13 = rocdl.mfma.i32.32x32x8i8 %arg3, %arg3, %arg8, %csti32, %csti32, %csti32 :
(i32, i32, vector<16 x i32>,
i32, i32, i32) -> vector<16 x i32>
// CHECK: call <4 x i32> @llvm.amdgcn.mfma.i32.16x16x16i8(i32 %{{.*}}, i32 %{{.*}}, <4 x i32> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r14 = rocdl.mfma.i32.16x16x16i8 %arg3, %arg3, %arg9, %csti32, %csti32, %csti32 :
(i32, i32, vector<4xi32>,
i32, i32, i32) -> vector<4xi32>
// CHECK: call <32 x float> @llvm.amdgcn.mfma.f32.32x32x2bf16(<2 x i16> %{{.*}}, <2 x i16> %{{.*}}, <32 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r15 = rocdl.mfma.f32.32x32x2bf16 %arg10, %arg10, %arg2, %csti32, %csti32, %csti32 :
(vector<2xi16>, vector<2xi16>, vector<32 x f32>,
i32, i32, i32) -> vector<32 x f32>
// CHECK: call <16 x float> @llvm.amdgcn.mfma.f32.16x16x2bf16(<2 x i16> %{{.*}}, <2 x i16> %{{.*}}, <16 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r16 = rocdl.mfma.f32.16x16x2bf16 %arg10, %arg10, %arg4, %csti32, %csti32, %csti32 :
(vector<2xi16>, vector<2xi16>, vector<16 x f32>,
i32, i32, i32) -> vector<16 x f32>
// CHECK: call <4 x float> @llvm.amdgcn.mfma.f32.4x4x2bf16(<2 x i16> %{{.*}}, <2 x i16> %{{.*}}, <4 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r17 = rocdl.mfma.f32.4x4x2bf16 %arg10, %arg10, %arg5, %csti32, %csti32, %csti32 :
(vector<2xi16>, vector<2xi16>, vector<4xf32>,
i32, i32, i32) -> vector<4xf32>
// CHECK: call <16 x float> @llvm.amdgcn.mfma.f32.32x32x4bf16(<2 x i16> %{{.*}}, <2 x i16> %{{.*}}, <16 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r18 = rocdl.mfma.f32.32x32x4bf16 %arg10, %arg10, %arg4, %csti32, %csti32, %csti32 :
(vector<2xi16>, vector<2xi16>, vector<16 x f32>,
i32, i32, i32) -> vector<16 x f32>
// CHECK: call <4 x float> @llvm.amdgcn.mfma.f32.16x16x8bf16(<2 x i16> %{{.*}}, <2 x i16> %{{.*}}, <4 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r19 = rocdl.mfma.f32.16x16x8bf16 %arg10, %arg10, %arg5, %csti32, %csti32, %csti32 :
(vector<2xi16>, vector<2xi16>, vector<4xf32>,
i32, i32, i32) -> vector<4xf32>
// CHECK: call <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.bf8.bf8(i64 %{{.*}}, i64 %{{.*}}, <4 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r20 = rocdl.mfma.f32.16x16x32.bf8.bf8 %arg11, %arg11, %arg5, %csti32, %csti32, %csti32 :
(i64, i64, vector<4xf32>,
i32, i32, i32) -> vector<4xf32>
// CHECK: call <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.bf8.fp8(i64 %{{.*}}, i64 %{{.*}}, <4 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r21 = rocdl.mfma.f32.16x16x32.bf8.fp8 %arg11, %arg11, %arg5, %csti32, %csti32, %csti32 :
(i64, i64, vector<4xf32>,
i32, i32, i32) -> vector<4xf32>
// CHECK: call <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.fp8.bf8(i64 %{{.*}}, i64 %{{.*}}, <4 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r22 = rocdl.mfma.f32.16x16x32.fp8.bf8 %arg11, %arg11, %arg5, %csti32, %csti32, %csti32 :
(i64, i64, vector<4xf32>,
i32, i32, i32) -> vector<4xf32>
// CHECK: call <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.fp8.fp8(i64 %{{.*}}, i64 %{{.*}}, <4 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r23 = rocdl.mfma.f32.16x16x32.fp8.fp8 %arg11, %arg11, %arg5, %csti32, %csti32, %csti32 :
(i64, i64, vector<4xf32>,
i32, i32, i32) -> vector<4xf32>
// CHECK: call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.bf8.bf8(i64 %{{.*}}, i64 %{{.*}}, <16 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r24 = rocdl.mfma.f32.32x32x16.bf8.bf8 %arg11, %arg11, %arg4, %csti32, %csti32, %csti32 :
(i64, i64, vector<16xf32>,
i32, i32, i32) -> vector<16xf32>
// CHECK: call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.bf8.fp8(i64 %{{.*}}, i64 %{{.*}}, <16 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r25 = rocdl.mfma.f32.32x32x16.bf8.fp8 %arg11, %arg11, %arg4, %csti32, %csti32, %csti32 :
(i64, i64, vector<16xf32>,
i32, i32, i32) -> vector<16xf32>
// CHECK: call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.fp8.bf8(i64 %{{.*}}, i64 %{{.*}}, <16 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r26 = rocdl.mfma.f32.32x32x16.fp8.bf8 %arg11, %arg11, %arg4, %csti32, %csti32, %csti32 :
(i64, i64, vector<16xf32>,
i32, i32, i32) -> vector<16xf32>
// CHECK: call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.bf8.bf8(i64 %{{.*}}, i64 %{{.*}}, <16 x float> %{{.*}}, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
%r27 = rocdl.mfma.f32.32x32x16.bf8.bf8 %arg11, %arg11, %arg4, %csti32, %csti32, %csti32 :
(i64, i64, vector<16xf32>,
i32, i32, i32) -> vector<16xf32>
llvm.return %r0 : vector<32 x f32>
}
llvm.func @rocdl.wmma(%arg0 : vector<8xf32>, %arg1 : vector<16 x f16>, %arg2 : vector<16 x i16>, %arg3 : vector<8 x i32>,
%arg4 : vector<2xi32>, %arg5 : vector<4xi32>, %arg6 : vector<4xf32>, %arg7 : vector<8xf16>, %arg8 : vector<8xi16>) -> vector<8xf32> {
%zero = llvm.mlir.constant(false) : i1
// ---- Wave32 -----
// f16 -> f32
// CHECK: call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32(<16 x half> %{{.*}}, <16 x half> %{{.*}}, <8 x float> %{{.*}})
%r0 = rocdl.wmma.f32.16x16x16.f16 %arg1, %arg1, %arg0 : (vector<16xf16>, vector<16xf16>, vector<8xf32>) -> vector<8xf32>
// bf16 -> f32
// CHECK: call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32(<16 x i16> %{{.*}}, <16 x i16> %{{.*}}, <8 x float> %{{.*}})
%r1 = rocdl.wmma.f32.16x16x16.bf16 %arg2, %arg2, %arg0 : (vector<16xi16>, vector<16xi16>, vector<8xf32>) -> vector<8xf32>
// f16 -> f16 (OPSEL = {0,1})
// CHECK: call <16 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v16f16(<16 x half> %{{.*}}, <16 x half> %{{.*}}, <16 x half> %{{.*}}, i1 {{.*}})
%r2 = rocdl.wmma.f16.16x16x16.f16 %arg1, %arg1, %arg1, %zero : (vector<16xf16>, vector<16xf16>, vector<16xf16>, i1) -> vector<16xf16>
// bf16 -> bf16 (OPSEL = {0,1})
// CHECK: call <16 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v16i16(<16 x i16> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> %{{.*}}, i1 {{.*}})
%r4 = rocdl.wmma.bf16.16x16x16.bf16 %arg2, %arg2, %arg2, %zero : (vector<16xi16>, vector<16xi16>, vector<16xi16>, i1) -> vector<16xi16>
// int8 -> int32 (signA = {0,1}, signB = {0,1}, clamp = {0,1})
// CHECK: call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32(i1 {{.*}}, <4 x i32> %{{.*}}, i1 {{.*}}, <4 x i32> %{{.*}}, <8 x i32> %{{.*}}, i1 {{.*}})
%r5 = rocdl.wmma.i32.16x16x16.iu8 %zero, %arg5, %zero, %arg5, %arg3, %zero : (i1, vector<4xi32>, i1, vector<4xi32>, vector<8xi32>, i1) -> vector<8xi32>
// int4 -> int32 (signA = {0,1}, signB = {0,1}, clamp = {0,1})
// CHECK: call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32(i1 {{.*}}, <2 x i32> %{{.*}}, i1 {{.*}}, <2 x i32> %{{.*}}, <8 x i32> %{{.*}}, i1 {{.*}})
%r6 = rocdl.wmma.i32.16x16x16.iu4 %zero, %arg4, %zero, %arg4, %arg3, %zero : (i1, vector<2xi32>, i1, vector<2xi32>, vector<8xi32>, i1) -> vector<8xi32>
// ---- Wave64 -----
// f16 -> f32
// CHECK: call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32(<16 x half> %{{.*}}, <16 x half> %{{.*}}, <4 x float> %{{.*}})
%r7 = rocdl.wmma.f32.16x16x16.f16 %arg1, %arg1, %arg6 : (vector<16xf16>, vector<16xf16>, vector<4xf32>) -> vector<4xf32>
// bf16 -> f32
// CHECK: call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32(<16 x i16> %{{.*}}, <16 x i16> %{{.*}}, <4 x float> %{{.*}})
%r8 = rocdl.wmma.f32.16x16x16.bf16 %arg2, %arg2, %arg6 : (vector<16xi16>, vector<16xi16>, vector<4xf32>) -> vector<4xf32>
// f16 -> f16 (OPSEL = {0,1})
// CHECK: call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16(<16 x half> %{{.*}}, <16 x half> %{{.*}}, <8 x half> %{{.*}}, i1 {{.*}})
%r9 = rocdl.wmma.f16.16x16x16.f16 %arg1, %arg1, %arg7, %zero : (vector<16xf16>, vector<16xf16>, vector<8xf16>, i1) -> vector<8xf16>
// bf16 -> bf16 (OPSEL = {0,1})
// CHECK: call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16(<16 x i16> %{{.*}}, <16 x i16> %{{.*}}, <8 x i16> %{{.*}}, i1 {{.*}})
%r11 = rocdl.wmma.bf16.16x16x16.bf16 %arg2, %arg2, %arg8, %zero : (vector<16xi16>, vector<16xi16>, vector<8xi16>, i1) -> vector<8xi16>
// int8 -> int32 (signA = {0,1}, signB = {0,1}, clamp = {0,1})
// CHECK: call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32(i1 {{.*}}, <4 x i32> %{{.*}}, i1 {{.*}}, <4 x i32> %{{.*}}, <4 x i32> %{{.*}}, i1 {{.*}})
%r12 = rocdl.wmma.i32.16x16x16.iu8 %zero, %arg5, %zero, %arg5, %arg5, %zero : (i1, vector<4xi32>, i1, vector<4xi32>, vector<4xi32>, i1) -> vector<4xi32>
// int4 -> int32 (signA = {0,1}, signB = {0,1}, clamp = {0,1})
// CHECK: call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32(i1 {{.*}}, <2 x i32> %{{.*}}, i1 {{.*}}, <2 x i32> %{{.*}}, <4 x i32> %{{.*}}, i1 {{.*}})
%r13 = rocdl.wmma.i32.16x16x16.iu4 %zero, %arg4, %zero, %arg4, %arg5, %zero : (i1, vector<2xi32>, i1, vector<2xi32>, vector<4xi32>, i1) -> vector<4xi32>
llvm.return %r0 : vector<8xf32>
}
llvm.func @rocdl.mubuf(%rsrc : vector<4xi32>, %vindex : i32,
%offset : i32, %vdata1 : vector<1xf32>,
%vdata2 : vector<2xf32>, %vdata4 : vector<4xf32>) {
%glc = llvm.mlir.constant(false) : i1
%slc = llvm.mlir.constant(true) : i1
// CHECK-LABEL: rocdl.mubuf
// CHECK: call <1 x float> @llvm.amdgcn.buffer.load.v1f32(<4 x i32> %{{.*}}, i32 %{{.*}}, i32 %{{.*}}, i1 {{.*}}, i1 {{.*}})
// CHECK: call <2 x float> @llvm.amdgcn.buffer.load.v2f32(<4 x i32> %{{.*}}, i32 %{{.*}}, i32 %{{.*}}, i1 {{.*}}, i1 {{.*}})
// CHECK: call <4 x float> @llvm.amdgcn.buffer.load.v4f32(<4 x i32> %{{.*}}, i32 %{{.*}}, i32 %{{.*}}, i1 {{.*}}, i1 {{.*}})
// CHECK: call void @llvm.amdgcn.buffer.store.v1f32(<1 x float> %{{.*}}, <4 x i32> %{{.*}}, i32 %{{.*}}, i32 %{{.*}}, i1 {{.*}}, i1 {{.*}})
// CHECK: call void @llvm.amdgcn.buffer.store.v2f32(<2 x float> %{{.*}}, <4 x i32> %{{.*}}, i32 %{{.*}}, i32 %{{.*}}, i1 {{.*}}, i1 {{.*}})
// CHECK: call void @llvm.amdgcn.buffer.store.v4f32(<4 x float> %{{.*}}, <4 x i32> %{{.*}}, i32 %{{.*}}, i32 %{{.*}}, i1 {{.*}}, i1 {{.*}})
%r1 = rocdl.buffer.load %rsrc, %vindex, %offset, %glc, %slc : vector<1xf32>
%r2 = rocdl.buffer.load %rsrc, %vindex, %offset, %glc, %slc : vector<2xf32>
%r4 = rocdl.buffer.load %rsrc, %vindex, %offset, %glc, %slc : vector<4xf32>
rocdl.buffer.store %vdata1, %rsrc, %vindex, %offset, %glc, %slc : vector<1xf32>
rocdl.buffer.store %vdata2, %rsrc, %vindex, %offset, %glc, %slc : vector<2xf32>
rocdl.buffer.store %vdata4, %rsrc, %vindex, %offset, %glc, %slc : vector<4xf32>
llvm.return
}
llvm.func @rocdl.raw.buffer(%rsrc : vector<4xi32>,
%offset : i32, %soffset : i32,
%vdata1 : i32,
%vdata2 : vector<2xi32>,
%vdata4 : vector<4xi32>) {
%aux = llvm.mlir.constant(0 : i32) : i32
// CHECK-LABEL: rocdl.raw.buffer
// CHECK: call i32 @llvm.amdgcn.raw.buffer.load.i32(<4 x i32> %{{.*}}, i32 %{{.*}}, i32 %{{.*}}, i32 {{.*}}
// CHECK: call <2 x i32> @llvm.amdgcn.raw.buffer.load.v2i32(<4 x i32> %{{.*}}, i32 %{{.*}}, i32 %{{.*}}, i32 {{.*}}
// CHECK: call <4 x i32> @llvm.amdgcn.raw.buffer.load.v4i32(<4 x i32> %{{.*}}, i32 %{{.*}}, i32 %{{.*}}, i32 {{.*}}
// CHECK: call void @llvm.amdgcn.raw.buffer.store.i32(i32 %{{.*}}, <4 x i32> %{{.*}}, i32 %{{.*}}, i32 %{{.*}}, i32 {{.*}}
// CHECK: call void @llvm.amdgcn.raw.buffer.store.v2i32(<2 x i32> %{{.*}}, <4 x i32> %{{.*}}, i32 %{{.*}}, i32 %{{.*}}, i32 {{.*}}
// CHECK: call void @llvm.amdgcn.raw.buffer.store.v4i32(<4 x i32> %{{.*}}, <4 x i32> %{{.*}}, i32 %{{.*}}, i32 %{{.*}}, i32 {{.*}}
%r1 = rocdl.raw.buffer.load %rsrc, %offset, %soffset, %aux : i32
%r2 = rocdl.raw.buffer.load %rsrc, %offset, %soffset, %aux : vector<2xi32>
%r4 = rocdl.raw.buffer.load %rsrc, %offset, %soffset, %aux : vector<4xi32>
rocdl.raw.buffer.store %vdata1, %rsrc, %offset, %soffset, %aux : i32
rocdl.raw.buffer.store %vdata2, %rsrc, %offset, %soffset, %aux : vector<2xi32>
rocdl.raw.buffer.store %vdata4, %rsrc, %offset, %soffset, %aux : vector<4xi32>
llvm.return
}
llvm.func @rocdl.raw.buffer.atomic.f32(%rsrc : vector<4xi32>,
%offset : i32, %soffset : i32,
%vdata1 : f32) {
%aux = llvm.mlir.constant(0 : i32) : i32
// CHECK-LABEL: rocdl.raw.buffer.atomic.f32
// CHECK: call float @llvm.amdgcn.raw.buffer.atomic.fadd.f32(float %{{.*}}, <4 x i32> %{{.*}}, i32 %{{.*}}, i32 %{{.*}}, i32 {{.*}}
// CHECK: call float @llvm.amdgcn.raw.buffer.atomic.fmax.f32(float %{{.*}}, <4 x i32> %{{.*}}, i32 %{{.*}}, i32 %{{.*}}, i32 {{.*}}
rocdl.raw.buffer.atomic.fadd %vdata1, %rsrc, %offset, %soffset, %aux : f32
rocdl.raw.buffer.atomic.fmax %vdata1, %rsrc, %offset, %soffset, %aux : f32
llvm.return
}
llvm.func @rocdl.raw.buffer.atomic.i32(%rsrc : vector<4xi32>,
%offset : i32, %soffset : i32,
%vdata1 : i32) {
%aux = llvm.mlir.constant(0 : i32) : i32
// CHECK-LABEL: rocdl.raw.buffer.atomic.i32
// CHECK: call i32 @llvm.amdgcn.raw.buffer.atomic.smax.i32(i32 %{{.*}}, <4 x i32> %{{.*}}, i32 %{{.*}}, i32 %{{.*}}, i32 {{.*}}
// CHECK: call i32 @llvm.amdgcn.raw.buffer.atomic.umin.i32(i32 %{{.*}}, <4 x i32> %{{.*}}, i32 %{{.*}}, i32 %{{.*}}, i32 {{.*}}
rocdl.raw.buffer.atomic.smax %vdata1, %rsrc, %offset, %soffset, %aux : i32
rocdl.raw.buffer.atomic.umin %vdata1, %rsrc, %offset, %soffset, %aux : i32
llvm.return
}
llvm.func @rocdl.raw.buffer.atomic.cmpswap(%rsrc : vector<4xi32>,
%offset : i32, %soffset : i32,
%src : i32, %cmp : i32) -> i32 {
%aux = llvm.mlir.constant(0 : i32) : i32
// CHECK-LABEL: rocdl.raw.buffer.atomic.cmpswap
// CHECK: [[val:%.+]] = call i32 @llvm.amdgcn.raw.buffer.atomic.cmpswap.i32(i32 %{{.*}}, i32 %{{.*}}, <4 x i32> %{{.*}}, i32 %{{.*}}, i32 %{{.*}}, i32 {{.*}}
// CHECK: ret i32 [[val]]
%val = rocdl.raw.buffer.atomic.cmpswap(%src, %cmp, %rsrc, %offset, %soffset, %aux) : i32, vector<4xi32>
llvm.return %val : i32
}
// CHECK-DAG: attributes #[[$KERNEL_ATTRS]] = { "amdgpu-flat-work-group-size"="1,256" "amdgpu-implicitarg-num-bytes"="56" }
// CHECK-DAG: attributes #[[$KERNEL_WORKGROUP_ATTRS]] = { "amdgpu-flat-work-group-size"="1,1024"
// CHECK-DAG: attributes #[[$KNOWN_BLOCK_SIZE_ATTRS]] = { "amdgpu-flat-work-group-size"="128,128"
// CHECK-DAG: ![[$RANGE]] = !{i32 0, i32 64}
// CHECK-DAG: ![[$REQD_WORK_GROUP_SIZE]] = !{i32 16, i32 4, i32 2}
|
