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
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
|
// RUN: mlir-opt -test-linalg-decompose-ops -cse -split-input-file %s | FileCheck %s
// RUN: mlir-opt -test-linalg-decompose-ops=remove-dead-args-and-results -cse -split-input-file %s | FileCheck %s --check-prefix=CANONICALIZECHECK
func.func @simple_op(%arg0 : tensor<?x?xf32>, %arg1 : tensor<?xf32>, %arg2 : tensor<?xf32>)
-> (tensor<?x?xf32>, tensor<?x?xf32>) {
%c0 = arith.constant 0 : index
%c1 = arith.constant 1 : index
%d0 = tensor.dim %arg0, %c0 : tensor<?x?xf32>
%d1 = tensor.dim %arg0, %c1 : tensor<?x?xf32>
%init1 = tensor.empty(%d1, %d0) : tensor<?x?xf32>
%init2 = tensor.empty(%d0, %d1) : tensor<?x?xf32>
%result:2 = linalg.generic {
indexing_maps = [affine_map<(d0, d1) -> (d0, d1)>, affine_map<(d0, d1) -> (d0)>,
affine_map<(d0, d1) -> (d1)>, affine_map<(d0, d1) -> (d1, d0)>,
affine_map<(d0, d1) -> (d0, d1)>],
iterator_types = ["parallel", "parallel"]}
ins(%arg0, %arg1, %arg2 : tensor<?x?xf32>, tensor<?xf32>, tensor<?xf32>)
outs(%init1, %init2 : tensor<?x?xf32>, tensor<?x?xf32>) {
^bb0(%b0 : f32, %b1 : f32, %b2 : f32, %b3 : f32, %b4 : f32) :
%0 = arith.addf %b0, %b1 : f32
%1 = arith.mulf %0, %b2 : f32
linalg.yield %0, %1 : f32, f32
} -> (tensor<?x?xf32>, tensor<?x?xf32>)
return %result#0, %result#1 : tensor<?x?xf32>, tensor<?x?xf32>
}
// CHECK-DAG: #[[MAP0:.+]] = affine_map<(d0, d1) -> (d0, d1)>
// CHECK-DAG: #[[MAP1:.+]] = affine_map<(d0, d1) -> (d0)>
// CHECK-DAG: #[[MAP2:.+]] = affine_map<(d0, d1) -> (d1)>
// CHECK-DAG: #[[MAP3:.+]] = affine_map<(d0, d1) -> (d1, d0)>
// CHECK: func @simple_op(
// CHECK-SAME: %[[ARG0:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
// CHECK-SAME: %[[ARG1:[a-zA-Z0-9_]+]]: tensor<?xf32>
// CHECK-SAME: %[[ARG2:[a-zA-Z0-9_]+]]: tensor<?xf32>
// CHECK-DAG: %[[C0:.+]] = arith.constant 0 : index
// CHECK-DAG: %[[C1:.+]] = arith.constant 1 : index
// CHECK-DAG: %[[D0:.+]] = tensor.dim %[[ARG0]], %[[C0]]
// CHECK-DAG: %[[D1:.+]] = tensor.dim %[[ARG0]], %[[C1]]
// CHECK-DAG: %[[INIT1:.+]] = tensor.empty(%[[D1]], %[[D0]])
// CHECK-DAG: %[[INIT2:.+]] = tensor.empty(%[[D0]], %[[D1]])
// CHECK-DAG: %[[GENERIC1:.+]]:3 = linalg.generic
// CHECK-SAME: [#[[MAP0]], #[[MAP1]], #[[MAP2]], #[[MAP3]], #[[MAP0]], #[[MAP3]]]
// CHECK-SAME: ["parallel", "parallel"]
// CHECK-SAME: ins(%[[ARG0]], %[[ARG1]], %[[ARG2]] :
// CHECK-SAME: outs(%[[INIT1]], %[[INIT2]], %[[INIT1]] :
// CHECK-NEXT: ^bb0(
// CHECK-SAME: %[[B0:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B1:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B2:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B3:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B4:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B5:[a-zA-Z0-9_]+]]: f32):
// CHECK-NEXT: %[[S0:.+]] = arith.addf %[[B0]], %[[B1]]
// CHECK-NEXT: linalg.yield %[[S0]], %{{[a-zA-Z0-9_]+}}, %[[S0]]
// CHECK: %[[GENERIC2:.+]]:2 = linalg.generic
// CHECK-SAME: [#[[MAP0]], #[[MAP1]], #[[MAP2]], #[[MAP3]], #[[MAP3]], #[[MAP0]]]
// CHECK-SAME: ["parallel", "parallel"]
// CHECK-SAME: ins(%[[ARG0]], %[[ARG1]], %[[ARG2]], %[[GENERIC1]]#2 :
// CHECK-SAME: outs(%[[INIT1]], %[[INIT2]] :
// CHECK-NEXT: ^bb0(
// CHECK-SAME: %[[B6:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B7:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B8:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B9:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B10:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B11:[a-zA-Z0-9_]+]]: f32):
// CHECK-NEXT: %[[S1:.+]] = arith.mulf %[[B9]], %[[B8]]
// CHECK-NEXT: linalg.yield %[[B9]], %[[S1]]
// CHECK: return %[[GENERIC1]]#0, %[[GENERIC2]]#1
// With cse + canonicalization
// CANONICALIZECHECK-DAG: #[[MAP0:.+]] = affine_map<(d0, d1) -> (d0, d1)>
// CANONICALIZECHECK-DAG: #[[MAP1:.+]] = affine_map<(d0, d1) -> (d0)>
// CANONICALIZECHECK-DAG: #[[MAP2:.+]] = affine_map<(d0, d1) -> (d1, d0)>
// CANONICALIZECHECK-DAG: #[[MAP3:.+]] = affine_map<(d0, d1) -> (d1)>
// CANONICALIZECHECK: func @simple_op(
// CANONICALIZECHECK-SAME: %[[ARG0:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
// CANONICALIZECHECK-SAME: %[[ARG1:[a-zA-Z0-9_]+]]: tensor<?xf32>
// CANONICALIZECHECK-SAME: %[[ARG2:[a-zA-Z0-9_]+]]: tensor<?xf32>
// CANONICALIZECHECK-DAG: %[[C0:.+]] = arith.constant 0 : index
// CANONICALIZECHECK-DAG: %[[C1:.+]] = arith.constant 1 : index
// CANONICALIZECHECK-DAG: %[[D0:.+]] = tensor.dim %[[ARG0]], %[[C0]]
// CANONICALIZECHECK-DAG: %[[D1:.+]] = tensor.dim %[[ARG0]], %[[C1]]
// CANONICALIZECHECK-DAG: %[[INIT1:.+]] = tensor.empty(%[[D1]], %[[D0]])
// CANONICALIZECHECK-DAG: %[[INIT2:.+]] = tensor.empty(%[[D0]], %[[D1]])
// CANONICALIZECHECK-DAG: %[[GENERIC1:.+]] = linalg.generic
// CANONICALIZECHECK-SAME: [#[[MAP0]], #[[MAP1]], #[[MAP2]]]
// CANONICALIZECHECK-SAME: ["parallel", "parallel"]
// CANONICALIZECHECK-SAME: ins(%[[ARG0]], %[[ARG1]] :
// CANONICALIZECHECK-SAME: outs(%[[INIT1]] :
// CANONICALIZECHECK-NEXT: ^bb0(
// CANONICALIZECHECK-SAME: %[[B0:[a-zA-Z0-9_]+]]: f32
// CANONICALIZECHECK-SAME: %[[B1:[a-zA-Z0-9_]+]]: f32
// CANONICALIZECHECK-SAME: %[[B2:[a-zA-Z0-9_]+]]: f32):
// CANONICALIZECHECK-NEXT: %[[S0:.+]] = arith.addf %[[B0]], %[[B1]]
// CANONICALIZECHECK-NEXT: linalg.yield %[[S0]]
// CANONICALIZECHECK: %[[GENERIC2:.+]] = linalg.generic
// CANONICALIZECHECK-SAME: [#[[MAP3]], #[[MAP2]], #[[MAP0]]]
// CANONICALIZECHECK-SAME: ["parallel", "parallel"]
// CANONICALIZECHECK-SAME: ins(%[[ARG2]], %[[GENERIC1]] :
// CANONICALIZECHECK-SAME: outs(%[[INIT2]] :
// CANONICALIZECHECK-NEXT: ^bb0(
// CANONICALIZECHECK-SAME: %[[B3:[a-zA-Z0-9_]+]]: f32
// CANONICALIZECHECK-SAME: %[[B4:[a-zA-Z0-9_]+]]: f32
// CANONICALIZECHECK-SAME: %[[B5:[a-zA-Z0-9_]+]]: f32):
// CANONICALIZECHECK-NEXT: %[[S1:.+]] = arith.mulf %[[B4]], %[[B3]]
// CANONICALIZECHECK-NEXT: linalg.yield %[[S1]]
// CANONICALIZECHECK: return %[[GENERIC1]], %[[GENERIC2]]
// -----
func.func @simple_op_permuted_outputs(%arg0 : tensor<?x?xf32>, %arg1 : tensor<?xf32>, %arg2 : tensor<?xf32>)
-> (tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>) {
%c0 = arith.constant 0 : index
%c1 = arith.constant 1 : index
%d0 = tensor.dim %arg0, %c0 : tensor<?x?xf32>
%d1 = tensor.dim %arg0, %c1 : tensor<?x?xf32>
%init1 = tensor.empty(%d1, %d0) : tensor<?x?xf32>
%init2 = tensor.empty(%d0, %d1) : tensor<?x?xf32>
%result:3 = linalg.generic {
indexing_maps = [affine_map<(d0, d1) -> (d0, d1)>, affine_map<(d0, d1) -> (d0)>,
affine_map<(d0, d1) -> (d1)>, affine_map<(d0, d1) -> (d1, d0)>,
affine_map<(d0, d1) -> (d0, d1)>, affine_map<(d0, d1) -> (d0, d1)>],
iterator_types = ["parallel", "parallel"]}
ins(%arg0, %arg1, %arg2 : tensor<?x?xf32>, tensor<?xf32>, tensor<?xf32>)
outs(%init1, %init2, %init2 : tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>) {
^bb0(%b0 : f32, %b1 : f32, %b2 : f32, %b3 : f32, %b4 : f32, %b5 : f32) :
%0 = arith.addf %b0, %b1 : f32
%1 = arith.mulf %0, %b2 : f32
linalg.yield %0, %1, %0 : f32, f32, f32
} -> (tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>)
return %result#0, %result#1, %result#2 : tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>
}
// CHECK-DAG: #[[MAP0:.+]] = affine_map<(d0, d1) -> (d0, d1)>
// CHECK-DAG: #[[MAP1:.+]] = affine_map<(d0, d1) -> (d0)>
// CHECK-DAG: #[[MAP2:.+]] = affine_map<(d0, d1) -> (d1)>
// CHECK-DAG: #[[MAP3:.+]] = affine_map<(d0, d1) -> (d1, d0)>
// CHECK: func @simple_op_permuted_outputs(
// CHECK-SAME: %[[ARG0:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
// CHECK-SAME: %[[ARG1:[a-zA-Z0-9_]+]]: tensor<?xf32>
// CHECK-SAME: %[[ARG2:[a-zA-Z0-9_]+]]: tensor<?xf32>
// CHECK-DAG: %[[C0:.+]] = arith.constant 0 : index
// CHECK-DAG: %[[C1:.+]] = arith.constant 1 : index
// CHECK-DAG: %[[D0:.+]] = tensor.dim %[[ARG0]], %[[C0]]
// CHECK-DAG: %[[D1:.+]] = tensor.dim %[[ARG0]], %[[C1]]
// CHECK-DAG: %[[INIT1:.+]] = tensor.empty(%[[D1]], %[[D0]])
// CHECK-DAG: %[[INIT2:.+]] = tensor.empty(%[[D0]], %[[D1]])
// CHECK-DAG: %[[GENERIC1:.+]]:4 = linalg.generic
// CHECK-SAME: [#[[MAP0]], #[[MAP1]], #[[MAP2]], #[[MAP3]], #[[MAP0]], #[[MAP0]], #[[MAP3]]]
// CHECK-SAME: ["parallel", "parallel"]
// CHECK-SAME: ins(%[[ARG0]], %[[ARG1]], %[[ARG2]] :
// CHECK-SAME: outs(%[[INIT1]], %[[INIT2]], %[[INIT2]], %[[INIT1]] :
// CHECK-NEXT: ^bb0(
// CHECK-SAME: %[[B0:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B1:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B2:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B3:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B4:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B5:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B6:[a-zA-Z0-9_]+]]: f32):
// CHECK-NEXT: %[[S0:.+]] = arith.addf %[[B0]], %[[B1]]
// CHECK-NEXT: linalg.yield %[[S0]], %{{[a-zA-Z0-9_]+}}, %[[S0]]
// CHECK: %[[GENERIC2:.+]]:3 = linalg.generic
// CHECK-SAME: [#[[MAP0]], #[[MAP1]], #[[MAP2]], #[[MAP3]], #[[MAP3]], #[[MAP0]], #[[MAP0]]]
// CHECK-SAME: ["parallel", "parallel"]
// CHECK-SAME: ins(%[[ARG0]], %[[ARG1]], %[[ARG2]], %[[GENERIC1]]#3 :
// CHECK-SAME: outs(%[[INIT1]], %[[INIT2]], %[[INIT2]] :
// CHECK-NEXT: ^bb0(
// CHECK-SAME: %[[B7:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B8:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B9:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B10:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B11:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[B12:[a-zA-Z0-9_]+]]: f32):
// CHECK-NEXT: %[[S1:.+]] = arith.mulf %[[B10]], %[[B9]]
// CHECK-NEXT: linalg.yield %[[B10]], %[[S1]], %[[B10]]
// CHECK: return %[[GENERIC1]]#0, %[[GENERIC2]]#1, %[[GENERIC1]]#2
// CANONICALIZECHECK-DAG: #[[MAP0:.+]] = affine_map<(d0, d1) -> (d0, d1)>
// CANONICALIZECHECK-DAG: #[[MAP1:.+]] = affine_map<(d0, d1) -> (d0)>
// CANONICALIZECHECK-DAG: #[[MAP2:.+]] = affine_map<(d0, d1) -> (d1, d0)>
// CANONICALIZECHECK-DAG: #[[MAP3:.+]] = affine_map<(d0, d1) -> (d1)>
// CANONICALIZECHECK: func @simple_op_permuted_outputs(
// CANONICALIZECHECK-SAME: %[[ARG0:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
// CANONICALIZECHECK-SAME: %[[ARG1:[a-zA-Z0-9_]+]]: tensor<?xf32>
// CANONICALIZECHECK-SAME: %[[ARG2:[a-zA-Z0-9_]+]]: tensor<?xf32>
// CANONICALIZECHECK-DAG: %[[C0:.+]] = arith.constant 0 : index
// CANONICALIZECHECK-DAG: %[[C1:.+]] = arith.constant 1 : index
// CANONICALIZECHECK-DAG: %[[D0:.+]] = tensor.dim %[[ARG0]], %[[C0]]
// CANONICALIZECHECK-DAG: %[[D1:.+]] = tensor.dim %[[ARG0]], %[[C1]]
// CANONICALIZECHECK-DAG: %[[INIT1:.+]] = tensor.empty(%[[D1]], %[[D0]])
// CANONICALIZECHECK-DAG: %[[INIT2:.+]] = tensor.empty(%[[D0]], %[[D1]])
// CANONICALIZECHECK-DAG: %[[GENERIC1:.+]]:2 = linalg.generic
// CANONICALIZECHECK-SAME: [#[[MAP0]], #[[MAP1]], #[[MAP2]], #[[MAP0]]]
// CANONICALIZECHECK-SAME: ["parallel", "parallel"]
// CANONICALIZECHECK-SAME: ins(%[[ARG0]], %[[ARG1]] :
// CANONICALIZECHECK-SAME: outs(%[[INIT1]], %[[INIT2]] :
// CANONICALIZECHECK-NEXT: ^bb0(
// CANONICALIZECHECK-SAME: %[[B0:[a-zA-Z0-9_]+]]: f32
// CANONICALIZECHECK-SAME: %[[B1:[a-zA-Z0-9_]+]]: f32
// CANONICALIZECHECK-SAME: %[[B2:[a-zA-Z0-9_]+]]: f32):
// CANONICALIZECHECK-NEXT: %[[S0:.+]] = arith.addf %[[B0]], %[[B1]]
// CANONICALIZECHECK-NEXT: linalg.yield %[[S0]], %[[S0]]
// CANONICALIZECHECK: %[[GENERIC2:.+]] = linalg.generic
// CANONICALIZECHECK-SAME: [#[[MAP3]], #[[MAP2]], #[[MAP0]]]
// CANONICALIZECHECK-SAME: ["parallel", "parallel"]
// CANONICALIZECHECK-SAME: ins(%[[ARG2]], %[[GENERIC1]]#0 :
// CANONICALIZECHECK-SAME: outs(%[[INIT2]] :
// CANONICALIZECHECK-NEXT: ^bb0(
// CANONICALIZECHECK-SAME: %[[B4:[a-zA-Z0-9_]+]]: f32
// CANONICALIZECHECK-SAME: %[[B5:[a-zA-Z0-9_]+]]: f32
// CANONICALIZECHECK-SAME: %[[B6:[a-zA-Z0-9_]+]]: f32):
// CANONICALIZECHECK-NEXT: %[[S1:.+]] = arith.mulf %[[B5]], %[[B4]]
// CANONICALIZECHECK-NEXT: linalg.yield %[[S1]]
// CANONICALIZECHECK: return %[[GENERIC1]]#0, %[[GENERIC2]], %[[GENERIC1]]#1
// -----
#map0 = affine_map<(d0, d1) -> (d0, d1)>
#map1 = affine_map<(d0, d1) -> (d0)>
#map2 = affine_map<(d0, d1) -> (d1, d0)>
func.func @multi_statement(%arg0 : tensor<10x20xf32>, %arg1 : tensor<10xi32>) -> tensor<20x10xf64> {
%init = tensor.empty() : tensor<20x10xf64>
%0 = linalg.generic {
indexing_maps = [#map0, #map1, #map2],
iterator_types = ["parallel", "parallel"]}
ins(%arg0, %arg1 : tensor<10x20xf32>, tensor<10xi32>)
outs(%init : tensor<20x10xf64>) {
^bb0(%b0 : f32, %b1 : i32, %b2 : f64):
%1 = arith.sitofp %b1 : i32 to f64
%2 = arith.extf %b0 : f32 to f64
%3 = arith.addf %1, %2 : f64
linalg.yield %3 : f64
} -> tensor<20x10xf64>
return %0 : tensor<20x10xf64>
}
// CHECK-DAG: #[[MAP0:.+]] = affine_map<(d0, d1) -> (d0, d1)>
// CHECK-DAG: #[[MAP1:.+]] = affine_map<(d0, d1) -> (d0)>
// CHECK-DAG: #[[MAP2:.+]] = affine_map<(d0, d1) -> (d1, d0)>
// CHECK: func @multi_statement(
// CHECK-SAME: %[[ARG0:.+]]: tensor<10x20xf32>
// CHECK-SAME: %[[ARG1:.+]]: tensor<10xi32>)
// CHECK-DAG: %[[INIT0:.+]] = tensor.empty() : tensor<20x10xf64>
// CHECK-DAG: %[[INIT1:.+]] = tensor.empty() : tensor<10x20xf64>
// CHECK: %[[GENERIC0:.+]]:2 = linalg.generic
// CHECK-SAME: indexing_maps = [#[[MAP0]], #[[MAP1]], #[[MAP2]], #[[MAP0]]]
// CHECK-SAME: iterator_types = ["parallel", "parallel"]
// CHECK-SAME: ins(%[[ARG0]], %[[ARG1]] :
// CHECK-SAME: outs(%[[INIT0]], %[[INIT1]] :
// CHECK-NEXT: ^bb0(
// CHECK-SAME: %[[B0:.+]]: f32
// CHECK-SAME: %[[B1:.+]]: i32
// CHECK-SAME: %[[B2:[a-zA-Z0-9_]+]]: f64
// CHECK-SAME: %[[B3:.+]]: f64
// CHECK-NEXT: %[[S0:.+]] = arith.sitofp %[[B1]] : i32 to f64
// CHECK-NEXT: linalg.yield %{{.+}}, %[[S0]]
// CHECK: %[[GENERIC1:.+]]:2 = linalg.generic
// CHECK-SAME: indexing_maps = [#[[MAP0]], #[[MAP1]], #[[MAP0]], #[[MAP2]], #[[MAP0]]]
// CHECK-SAME: iterator_types = ["parallel", "parallel"]
// CHECK-SAME: ins(%[[ARG0]], %[[ARG1]], %[[GENERIC0]]#1 :
// CHECK-SAME: outs(%[[INIT0]], %[[INIT1]] :
// CHECK-NEXT: ^bb0(
// CHECK-SAME: %[[B4:.+]]: f32
// CHECK-SAME: %[[B5:.+]]: i32
// CHECK-SAME: %[[B6:[a-zA-Z0-9_]+]]: f64
// CHECK-SAME: %[[B7:[a-zA-Z0-9_]+]]: f64
// CHECK-SAME: %[[B8:.+]]: f64
// CHECK-NEXT: %[[S1:.+]] = arith.extf %[[B4]] : f32 to f64
// CHECK-NEXT: linalg.yield %{{.+}}, %[[S1]]
// CHECK: %[[GENERIC2:.+]] = linalg.generic
// CHECK-SAME: indexing_maps = [#[[MAP0]], #[[MAP1]], #[[MAP0]], #[[MAP0]], #[[MAP2]]]
// CHECK-SAME: iterator_types = ["parallel", "parallel"]
// CHECK-SAME: ins(%[[ARG0]], %[[ARG1]], %[[GENERIC0]]#1, %[[GENERIC1]]#1 :
// CHECK-SAME: outs(%[[INIT0]] :
// CHECK-NEXT: ^bb0(
// CHECK-SAME: %[[B9:.+]]: f32
// CHECK-SAME: %[[B10:.+]]: i32
// CHECK-SAME: %[[B11:[a-zA-Z0-9_]+]]: f64
// CHECK-SAME: %[[B12:[a-zA-Z0-9_]+]]: f64
// CHECK-SAME: %[[B13:.+]]: f64
// CHECK-NEXT: %[[S2:.+]] = arith.addf %[[B11]], %[[B12]] : f64
// CHECK-NEXT: linalg.yield %[[S2]]
// CHECK: return %[[GENERIC2]]
// CANONICALIZECHECK-DAG: #[[MAP0:.+]] = affine_map<(d0, d1) -> (d0)>
// CANONICALIZECHECK-DAG: #[[MAP1:.+]] = affine_map<(d0, d1) -> (d0, d1)>
// CANONICALIZECHECK-DAG: #[[MAP2:.+]] = affine_map<(d0, d1) -> (d1, d0)>
// CANONICALIZECHECK: func @multi_statement(
// CANONICALIZECHECK-SAME: %[[ARG0:.+]]: tensor<10x20xf32>
// CANONICALIZECHECK-SAME: %[[ARG1:.+]]: tensor<10xi32>)
// CANONICALIZECHECK-DAG: %[[INIT0:.+]] = tensor.empty() : tensor<20x10xf64>
// CANONICALIZECHECK-DAG: %[[INIT1:.+]] = tensor.empty() : tensor<10x20xf64>
// CANONICALIZECHECK: %[[GENERIC0:.+]] = linalg.generic
// CANONICALIZECHECK-SAME: indexing_maps = [#[[MAP0]], #[[MAP1]]]
// CANONICALIZECHECK-SAME: iterator_types = ["parallel", "parallel"]
// CANONICALIZECHECK-SAME: ins(%[[ARG1]] :
// CANONICALIZECHECK-SAME: outs(%[[INIT1]] :
// CANONICALIZECHECK-NEXT: ^bb0(
// CANONICALIZECHECK-SAME: %[[B0:.+]]: i32
// CANONICALIZECHECK-SAME: %[[B1:.+]]: f64
// CANONICALIZECHECK-NEXT: %[[S0:.+]] = arith.sitofp %[[B0]] : i32 to f64
// CANONICALIZECHECK-NEXT: linalg.yield %[[S0]]
// CANONICALIZECHECK: %[[GENERIC1:.+]] = linalg.generic
// CANONICALIZECHECK-SAME: indexing_maps = [#[[MAP1]], #[[MAP1]]]
// CANONICALIZECHECK-SAME: iterator_types = ["parallel", "parallel"]
// CANONICALIZECHECK-SAME: ins(%[[ARG0]] :
// CANONICALIZECHECK-SAME: outs(%[[INIT1]] :
// CANONICALIZECHECK-NEXT: ^bb0(
// CANONICALIZECHECK-SAME: %[[B2:.+]]: f32
// CANONICALIZECHECK-SAME: %[[B3:.+]]: f64
// CANONICALIZECHECK-NEXT: %[[S1:.+]] = arith.extf %[[B2]] : f32 to f64
// CANONICALIZECHECK-NEXT: linalg.yield %[[S1]]
// CANONICALIZECHECK: %[[GENERIC2:.+]] = linalg.generic
// CANONICALIZECHECK-SAME: indexing_maps = [#[[MAP1]], #[[MAP1]], #[[MAP2]]]
// CANONICALIZECHECK-SAME: iterator_types = ["parallel", "parallel"]
// CANONICALIZECHECK-SAME: ins(%[[GENERIC0]], %[[GENERIC1]] :
// CANONICALIZECHECK-SAME: outs(%[[INIT0]] :
// CANONICALIZECHECK-NEXT: ^bb0(
// CANONICALIZECHECK-SAME: %[[B4:[a-zA-Z0-9_]+]]: f64
// CANONICALIZECHECK-SAME: %[[B5:[a-zA-Z0-9_]+]]: f64
// CANONICALIZECHECK-SAME: %[[B6:.+]]: f64
// CANONICALIZECHECK-NEXT: %[[S2:.+]] = arith.addf %[[B4]], %[[B5]] : f64
// CANONICALIZECHECK-NEXT: linalg.yield %[[S2]]
// CANONICALIZECHECK: return %[[GENERIC2]]
// -----
#map0 = affine_map<(d0, d1) -> (d0)>
#map1 = affine_map<(d0, d1) -> (d1)>
#map2 = affine_map<(d0, d1) -> (d0, d1)>
#map3 = affine_map<(d0, d1) -> (d1, d0)>
func.func @destination_passing_style(
%arg0 : tensor<?xf32>, %arg1 : tensor<?xf32>,
%arg2 : tensor<?x?xf32>, %arg3 : tensor<?x?xf32>)
-> (tensor<?x?xf32>, tensor<?x?xf32>) {
%0:2 = linalg.generic {
indexing_maps = [#map0, #map1, #map2, #map3],
iterator_types = ["parallel", "parallel"]}
ins(%arg0, %arg1 : tensor<?xf32>, tensor<?xf32>)
outs(%arg2, %arg3 : tensor<?x?xf32>, tensor<?x?xf32>) {
^bb0(%b0 : f32, %b1 : f32, %b2 : f32, %b3 : f32) :
%1 = arith.addf %b0, %b2 : f32
%2 = arith.mulf %b1, %b3 : f32
linalg.yield %1, %2 : f32, f32
} -> (tensor<?x?xf32>, tensor<?x?xf32>)
return %0#0, %0#1 : tensor<?x?xf32>, tensor<?x?xf32>
}
// CHECK-DAG: #[[MAP0:.+]] = affine_map<(d0, d1) -> (d0)>
// CHECK-DAG: #[[MAP1:.+]] = affine_map<(d0, d1) -> (d1)>
// CHECK-DAG: #[[MAP2:.+]] = affine_map<(d0, d1) -> (d0, d1)>
// CHECK-DAG: #[[MAP3:.+]] = affine_map<(d0, d1) -> (d1, d0)>
// CHECK: func.func @destination_passing_style(
// CHECK-SAME: %[[ARG0:[a-zA-Z0-9_]+]]: tensor<?xf32>
// CHECK-SAME: %[[ARG1:[a-zA-Z0-9_]+]]: tensor<?xf32>
// CHECK-SAME: %[[ARG2:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
// CHECK-SAME: %[[ARG3:[a-zA-Z0-9_]+]]: tensor<?x?xf32>)
// CHECK: %[[GENERIC1:.+]]:3 = linalg.generic
// CHECK-SAME: indexing_maps = [#[[MAP0]], #[[MAP1]], #[[MAP2]], #[[MAP3]], #[[MAP2]]]
// CHECK-SAME: iterator_types = ["parallel", "parallel"]
// CHECK-SAME: ins(%[[ARG0]], %[[ARG1]] :
// CHECK-SAME: outs(%[[ARG2]], %[[ARG3]], %[[ARG2]] :
// CHECK-NEXT: ^bb0(
// CHECK-SAME: %[[ARG4:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[ARG5:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[ARG6:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[ARG7:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[ARG8:[a-zA-Z0-9_]+]]: f32
// CHECK-NEXT: %[[S1:.+]] = arith.addf %[[ARG4]], %[[ARG6]]
// CHECK-NEXT: linalg.yield %[[S1]], %{{.+}}, %[[S1]]
// CHECK: %[[GENERIC2:.+]]:2 = linalg.generic
// CHECK-SAME: indexing_maps = [#[[MAP0]], #[[MAP1]], #[[MAP2]], #[[MAP2]], #[[MAP3]]]
// CHECK-SAME: iterator_types = ["parallel", "parallel"]
// CHECK-SAME: ins(%[[ARG0]], %[[ARG1]], %[[GENERIC1]]#2 :
// CHECK-SAME: outs(%[[ARG2]], %[[ARG3]] :
// CHECK-NEXT: ^bb0(
// CHECK-SAME: %[[ARG9:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[ARG10:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[ARG11:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[ARG12:[a-zA-Z0-9_]+]]: f32
// CHECK-SAME: %[[ARG13:[a-zA-Z0-9_]+]]: f32
// CHECK-NEXT: %[[S2:.+]] = arith.mulf %[[ARG10]], %[[ARG12]]
// CHECK-NEXT: linalg.yield %[[ARG11]], %[[S2]]
// CHECK: return %[[GENERIC1]]#0, %[[GENERIC2]]#1
// CANONICALIZECHECK-DAG: #[[MAP0:.+]] = affine_map<(d0, d1) -> (d0)>
// CANONICALIZECHECK-DAG: #[[MAP1:.+]] = affine_map<(d0, d1) -> (d0, d1)>
// CANONICALIZECHECK-DAG: #[[MAP2:.+]] = affine_map<(d0, d1) -> (d1)>
// CANONICALIZECHECK-DAG: #[[MAP3:.+]] = affine_map<(d0, d1) -> (d1, d0)>
// CANONICALIZECHECK: func.func @destination_passing_style(
// CANONICALIZECHECK-SAME: %[[ARG0:[a-zA-Z0-9_]+]]: tensor<?xf32>
// CANONICALIZECHECK-SAME: %[[ARG1:[a-zA-Z0-9_]+]]: tensor<?xf32>
// CANONICALIZECHECK-SAME: %[[ARG2:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
// CANONICALIZECHECK-SAME: %[[ARG3:[a-zA-Z0-9_]+]]: tensor<?x?xf32>)
// CANONICALIZECHECK: %[[GENERIC1:.+]] = linalg.generic
// CANONICALIZECHECK-SAME: indexing_maps = [#[[MAP0]], #[[MAP1]]]
// CANONICALIZECHECK-SAME: iterator_types = ["parallel", "parallel"]
// CANONICALIZECHECK-SAME: ins(%[[ARG0]] :
// CANONICALIZECHECK-SAME: outs(%[[ARG2]] :
// CANONICALIZECHECK-NEXT: ^bb0(
// CANONICALIZECHECK-SAME: %[[ARG4:[a-zA-Z0-9_]+]]: f32
// CANONICALIZECHECK-SAME: %[[ARG5:[a-zA-Z0-9_]+]]: f32
// CANONICALIZECHECK-NEXT: %[[S1:.+]] = arith.addf %[[ARG4]], %[[ARG5]]
// CANONICALIZECHECK-NEXT: linalg.yield %[[S1]]
// CANONICALIZECHECK: %[[GENERIC2:.+]]:2 = linalg.generic
// CANONICALIZECHECK-SAME: indexing_maps = [#[[MAP2]], #[[MAP1]], #[[MAP1]], #[[MAP3]]]
// CANONICALIZECHECK-SAME: iterator_types = ["parallel", "parallel"]
// CANONICALIZECHECK-SAME: ins(%[[ARG1]], %[[GENERIC1]] :
// CANONICALIZECHECK-SAME: outs(%[[ARG2]], %[[ARG3]] :
// CANONICALIZECHECK-NEXT: ^bb0(
// CANONICALIZECHECK-SAME: %[[ARG4:[a-zA-Z0-9_]+]]: f32
// CANONICALIZECHECK-SAME: %[[ARG5:[a-zA-Z0-9_]+]]: f32
// CANONICALIZECHECK-SAME: %[[ARG6:[a-zA-Z0-9_]+]]: f32
// CANONICALIZECHECK-SAME: %[[ARG7:[a-zA-Z0-9_]+]]: f32
// CANONICALIZECHECK-NEXT: %[[S2:.+]] = arith.mulf %[[ARG4]], %[[ARG6]]
// CANONICALIZECHECK-NEXT: linalg.yield %[[ARG5]], %[[S2]]
// CANONICALIZECHECK: return %[[GENERIC1]], %[[GENERIC2]]#1
|
