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
|
// RUN: mlir-opt %s -split-input-file -linalg-generalize-named-ops | FileCheck %s
func.func @generalize_matmul_buffer(%A : memref<16x8xf32>, %B: memref<8x32xf32>, %C: memref<16x32xf32>) {
linalg.matmul ins(%A, %B: memref<16x8xf32>, memref<8x32xf32>)
outs(%C: memref<16x32xf32>)
return
}
// CHECK: #[[A_MAP:.+]] = affine_map<(d0, d1, d2) -> (d0, d2)>
// CHECK: #[[B_MAP:.+]] = affine_map<(d0, d1, d2) -> (d2, d1)>
// CHECK: #[[C_MAP:.+]] = affine_map<(d0, d1, d2) -> (d0, d1)>
// CHECK: func @generalize_matmul_buffer
// CHECK-SAME: %[[A:.+]]: memref<16x8xf32>
// CHECK-SAME: %[[B:.+]]: memref<8x32xf32>
// CHECK-SAME: %[[C:.+]]: memref<16x32xf32>
// CHECK: linalg.generic
// CHECK-SAME: indexing_maps = [#[[A_MAP]], #[[B_MAP]], #[[C_MAP]]]
// CHECK-SAME: iterator_types = ["parallel", "parallel", "reduction"]
// CHECK-SAME: ins(%[[A]], %[[B]]
// CHECK-SAME: outs(%[[C]]
// CHECK: ^{{.*}}(%[[A_ARG:.+]]: f32, %[[B_ARG:.+]]: f32, %[[C_ARG:.+]]: f32)
// CHECK: %[[MUL:.+]] = arith.mulf %[[A_ARG]], %[[B_ARG]] : f32
// CHECK: %[[ADD:.+]] = arith.addf %[[C_ARG]], %[[MUL]] : f32
// CHECK: linalg.yield %[[ADD]] : f32
// -----
func.func @generalize_matmul_tensor(%A : tensor<16x8xf32>, %B: tensor<8x32xf32>, %C: tensor<16x32xf32>) -> tensor<16x32xf32> {
%0 = linalg.matmul ins(%A, %B: tensor<16x8xf32>, tensor<8x32xf32>)
outs(%C: tensor<16x32xf32>) -> tensor<16x32xf32>
return %0: tensor<16x32xf32>
}
// CHECK: func @generalize_matmul_tensor
// CHECK: linalg.generic
// CHECK-SAME: ins(%{{.+}}, %{{.+}} : tensor<16x8xf32>, tensor<8x32xf32>)
// CHECK-SAME: outs(%{{.+}} : tensor<16x32xf32>)
// CHECK: ^{{.*}}(%[[A_ARG:.+]]: f32, %[[B_ARG:.+]]: f32, %[[C_ARG:.+]]: f32)
// CHECK-NEXT: %[[MUL:.+]] = arith.mulf %[[A_ARG]], %[[B_ARG]] : f32
// CHECK-NEXT: %[[ADD:.+]] = arith.addf %[[C_ARG]], %[[MUL]] : f32
// CHECK-NEXT: linalg.yield %[[ADD]] : f32
// CHECK-NEXT: -> tensor<16x32xf32>
// -----
func.func @generalize_matmul_tensor_complex(%A : tensor<16x8xcomplex<f32>>,
%B: tensor<8x32xcomplex<f32>>,
%C: tensor<16x32xcomplex<f32>>)
-> tensor<16x32xcomplex<f32>> {
%0 = linalg.matmul ins(%A, %B: tensor<16x8xcomplex<f32>>, tensor<8x32xcomplex<f32>>)
outs(%C: tensor<16x32xcomplex<f32>>) -> tensor<16x32xcomplex<f32>>
return %0: tensor<16x32xcomplex<f32>>
}
// CHECK: func @generalize_matmul_tensor_complex
// CHECK: linalg.generic
// CHECK-SAME: ins(%{{.+}}, %{{.+}} : tensor<16x8xcomplex<f32>>, tensor<8x32xcomplex<f32>>)
// CHECK-SAME: outs(%{{.+}} : tensor<16x32xcomplex<f32>>)
// CHECK: ^{{.*}}(%[[A_ARG:.+]]: complex<f32>, %[[B_ARG:.+]]: complex<f32>, %[[C_ARG:.+]]: complex<f32>)
// CHECK-NEXT: %[[MUL:.+]] = complex.mul %[[A_ARG]], %[[B_ARG]] : complex<f32>
// CHECK-NEXT: %[[ADD:.+]] = complex.add %[[C_ARG]], %[[MUL]] : complex<f32>
// CHECK-NEXT: linalg.yield %[[ADD]] : complex<f32>
// CHECK-NEXT: -> tensor<16x32xcomplex<f32>>
// -----
func.func @depthwise_conv_2d_nhwc_hwcm(%input: memref<2x4x5x2xf32>, %filter: memref<2x2x2x3xf32>, %output: memref<2x3x4x2x3xf32>) {
linalg.depthwise_conv_2d_nhwc_hwcm
{ dilations = dense<1> : tensor<2xi64>, strides = dense<1> : tensor<2xi64> }
ins(%input, %filter : memref<2x4x5x2xf32>, memref<2x2x2x3xf32>)
outs(%output : memref<2x3x4x2x3xf32>)
return
}
// CHECK-DAG: #[[MAP0:.+]] = affine_map<(d0, d1, d2, d3, d4, d5, d6) -> (d0, d1 + d5, d2 + d6, d3)>
// CHECK-DAG: #[[MAP1:.+]] = affine_map<(d0, d1, d2, d3, d4, d5, d6) -> (d5, d6, d3, d4)>
// CHECK-DAG: #[[MAP2:.+]] = affine_map<(d0, d1, d2, d3, d4, d5, d6) -> (d0, d1, d2, d3, d4)>
// CHECK: func @depthwise_conv_2d_nhwc_hwcm
// CHECK: linalg.generic
// CHECK-SAME: indexing_maps = [#[[MAP0]], #[[MAP1]], #[[MAP2]]]
// CHECK-SAME: iterator_types = ["parallel", "parallel", "parallel", "parallel", "parallel", "reduction", "reduction"]}
// CHECK-SAME: ins(%{{.+}}, %{{.+}} : memref<2x4x5x2xf32>, memref<2x2x2x3xf32>)
// CHECK-SAME: outs(%{{.+}} : memref<2x3x4x2x3xf32>)
// CHECK: ^{{.+}}(%[[BBARG0:.+]]: f32, %[[BBARG1:.+]]: f32, %[[BBARG2:.+]]: f32)
// CHECK-NEXT: %[[MUL:.+]] = arith.mulf %[[BBARG0]], %[[BBARG1]] : f32
// CHECK-NEXT: %[[ADD:.+]] = arith.addf %[[BBARG2]], %[[MUL]] : f32
// CHECK-NEXT: linalg.yield %[[ADD]] : f32
// -----
func.func @depthwise_conv_2d_nhwc_hwcm(%input: memref<2x4x5x2xf32>, %filter: memref<2x2x2x3xf32>, %output: memref<2x2x3x2x3xf32>) {
linalg.depthwise_conv_2d_nhwc_hwcm
{ dilations = dense<2> : tensor<2xi64>, strides = dense<1> : tensor<2xi64> }
ins(%input, %filter : memref<2x4x5x2xf32>, memref<2x2x2x3xf32>)
outs(%output : memref<2x2x3x2x3xf32>)
return
}
// CHECK-DAG: #[[MAP0:.+]] = affine_map<(d0, d1, d2, d3, d4, d5, d6) -> (d0, d1 + d5 * 2, d2 + d6 * 2, d3)>
// CHECK-DAG: #[[MAP1:.+]] = affine_map<(d0, d1, d2, d3, d4, d5, d6) -> (d5, d6, d3, d4)>
// CHECK-DAG: #[[MAP2:.+]] = affine_map<(d0, d1, d2, d3, d4, d5, d6) -> (d0, d1, d2, d3, d4)>
// CHECK: func @depthwise_conv_2d_nhwc_hwcm
// CHECK: linalg.generic
// CHECK-SAME: indexing_maps = [#[[MAP0]], #[[MAP1]], #[[MAP2]]]
// CHECK-SAME: iterator_types = ["parallel", "parallel", "parallel", "parallel", "parallel", "reduction", "reduction"]}
// CHECK-SAME: ins(%{{.+}}, %{{.+}} : memref<2x4x5x2xf32>, memref<2x2x2x3xf32>)
// CHECK-SAME: outs(%{{.+}} : memref<2x2x3x2x3xf32>)
// CHECK: ^{{.+}}(%[[BBARG0:.+]]: f32, %[[BBARG1:.+]]: f32, %[[BBARG2:.+]]: f32)
// CHECK-NEXT: %[[MUL:.+]] = arith.mulf %[[BBARG0]], %[[BBARG1]] : f32
// CHECK-NEXT: %[[ADD:.+]] = arith.addf %[[BBARG2]], %[[MUL]] : f32
// CHECK-NEXT: linalg.yield %[[ADD]] : f32
// -----
func.func @depthwise_conv_2d_nhwc_hwc(%input: memref<1x113x113x96xf32>, %filter: memref<3x3x96xf32>, %output: memref<1x56x56x96xf32>) {
linalg.depthwise_conv_2d_nhwc_hwc {dilations = dense<1> : vector<2xi64>, strides = dense<2> : vector<2xi64>}
ins(%input, %filter: memref<1x113x113x96xf32>, memref<3x3x96xf32>)
outs(%output: memref<1x56x56x96xf32>)
return
}
// CHECK-DAG: #[[MAP0:.+]] = affine_map<(d0, d1, d2, d3, d4, d5) -> (d0, d1 * 2 + d4, d2 * 2 + d5, d3)>
// CHECK-DAG: #[[MAP1:.+]] = affine_map<(d0, d1, d2, d3, d4, d5) -> (d4, d5, d3)>
// CHECK-DAG: #[[MAP2:.+]] = affine_map<(d0, d1, d2, d3, d4, d5) -> (d0, d1, d2, d3)>
// CHECK: func @depthwise_conv_2d_nhwc_hwc
// CHECK: linalg.generic
// CHECK-SAME: indexing_maps = [#[[MAP0]], #[[MAP1]], #[[MAP2]]]
// CHECK-SAME: iterator_types = ["parallel", "parallel", "parallel", "parallel", "reduction", "reduction"]}
// CHECK-SAME: ins(%{{.+}}, %{{.+}} : memref<1x113x113x96xf32>, memref<3x3x96xf32>)
// CHECK-SAME: outs(%{{.+}} : memref<1x56x56x96xf32>)
// CHECK: ^{{.+}}(%[[BBARG0:.+]]: f32, %[[BBARG1:.+]]: f32, %[[BBARG2:.+]]: f32)
// CHECK-NEXT: %[[MUL:.+]] = arith.mulf %[[BBARG0]], %[[BBARG1]] : f32
// CHECK-NEXT: %[[ADD:.+]] = arith.addf %[[BBARG2]], %[[MUL]] : f32
// CHECK-NEXT: linalg.yield %[[ADD]] : f32
// -----
func.func @conv_1d_nwc_wcf(%input: memref<?x?x?xf32>, %filter: memref<?x?x?xf32>, %output: memref<?x?x?xf32>) {
linalg.conv_1d_nwc_wcf {dilations = dense<1> : tensor<1xi64>,
strides = dense<1> : tensor<1xi64>}
ins (%input, %filter: memref<?x?x?xf32>, memref<?x?x?xf32>)
outs (%output: memref<?x?x?xf32>)
return
}
// CHECK-DAG: #[[MAP0:.+]] = affine_map<(d0, d1, d2, d3, d4) -> (d0, d1 + d3, d4)>
// CHECK-DAG: #[[MAP1:.+]] = affine_map<(d0, d1, d2, d3, d4) -> (d3, d4, d2)>
// CHECK-DAG: #[[MAP2:.+]] = affine_map<(d0, d1, d2, d3, d4) -> (d0, d1, d2)>
// CHECK: func @conv_1d_nwc_wcf
// CHECK: linalg.generic
// CHECK-SAME: indexing_maps = [#[[MAP0]], #[[MAP1]], #[[MAP2]]]
// CHECK-SAME: iterator_types = ["parallel", "parallel", "parallel", "reduction", "reduction"]}
// CHECK-SAME: ins(%{{.+}}, %{{.+}} : memref<?x?x?xf32>, memref<?x?x?xf32>)
// CHECK-SAME: outs(%{{.+}} : memref<?x?x?xf32>)
// CHECK: ^{{.+}}(%[[BBARG0:.+]]: f32, %[[BBARG1:.+]]: f32, %[[BBARG2:.+]]: f32)
// CHECK-NEXT: %[[MUL:.+]] = arith.mulf %[[BBARG0]], %[[BBARG1]] : f32
// CHECK-NEXT: %[[ADD:.+]] = arith.addf %[[BBARG2]], %[[MUL]] : f32
// CHECK-NEXT: linalg.yield %[[ADD]] : f32
// -----
func.func @conv_1d_ncw_fcw(%input: memref<?x?x?xf32>, %filter: memref<?x?x?xf32>, %output: memref<?x?x?xf32>) {
linalg.conv_1d_ncw_fcw {dilations = dense<1> : tensor<1xi64>,
strides = dense<1> : tensor<1xi64>}
ins (%input, %filter: memref<?x?x?xf32>, memref<?x?x?xf32>)
outs (%output: memref<?x?x?xf32>)
return
}
// CHECK-DAG: #[[MAP0:.+]] = affine_map<(d0, d1, d2, d3, d4) -> (d0, d3, d2 + d4)>
// CHECK-DAG: #[[MAP1:.+]] = affine_map<(d0, d1, d2, d3, d4) -> (d1, d3, d4)>
// CHECK-DAG: #[[MAP2:.+]] = affine_map<(d0, d1, d2, d3, d4) -> (d0, d1, d2)>
// CHECK: func @conv_1d_ncw_fcw
// CHECK: linalg.generic
// CHECK-SAME: indexing_maps = [#[[MAP0]], #[[MAP1]], #[[MAP2]]]
// CHECK-SAME: iterator_types = ["parallel", "parallel", "parallel", "reduction", "reduction"]}
// CHECK-SAME: ins(%{{.+}}, %{{.+}} : memref<?x?x?xf32>, memref<?x?x?xf32>)
// CHECK-SAME: outs(%{{.+}} : memref<?x?x?xf32>)
// CHECK: ^{{.+}}(%[[BBARG0:.+]]: f32, %[[BBARG1:.+]]: f32, %[[BBARG2:.+]]: f32)
// CHECK-NEXT: %[[MUL:.+]] = arith.mulf %[[BBARG0]], %[[BBARG1]] : f32
// CHECK-NEXT: %[[ADD:.+]] = arith.addf %[[BBARG2]], %[[MUL]] : f32
// CHECK-NEXT: linalg.yield %[[ADD]] : f32
// -----
func.func @generalize_fill(%output: memref<?x?xf32>, %value : f32) {
linalg.fill ins(%value : f32) outs(%output : memref<?x?xf32>)
return
}
// CHECK-DAG: #[[MAP0:.+]] = affine_map<(d0, d1) -> ()>
// CHECK-DAG: #[[MAP1:.+]] = affine_map<(d0, d1) -> (d0, d1)>
// CHECK: func @generalize_fill
// CHECK-SAME: (%[[ARG0:.+]]: memref<?x?xf32>, %[[VAL:.+]]: f32)
// CHECK: linalg.generic
// CHECK-SAME: indexing_maps = [#[[MAP0]], #[[MAP1]]]
// CHECK-SAME: iterator_types = ["parallel", "parallel"]}
// CHECK-SAME: ins(%[[VAL]] : f32)
// CHECK-SAME: outs(%{{.+}} : memref<?x?xf32>)
// CHECK: ^{{.+}}(%[[BBARG0:.+]]: f32, %[[BBARG1:.+]]: f32)
// CHECK-NEXT: linalg.yield %[[BBARG0]] : f32
// -----
func.func @generalize_batch_matm_vec(%lhs : memref<?x?x?xi8>, %rhs: memref<?x?xi8>, %out: memref<?x?xf32>) {
linalg.batch_matvec ins(%lhs, %rhs: memref<?x?x?xi8>, memref<?x?xi8>)
outs(%out: memref<?x?xf32>)
return
}
// CHECK: #[[MAP0:.+]] = affine_map<(d0, d1, d2) -> (d0, d1, d2)>
// CHECK: #[[MAP1:.+]] = affine_map<(d0, d1, d2) -> (d0, d2)>
// CHECK: #[[MAP2:.+]] = affine_map<(d0, d1, d2) -> (d0, d1)>
// CHECK: @generalize_batch_matm_vec
// CHECK: linalg.generic
// CHECK-SAME: indexing_maps = [#[[MAP0]], #[[MAP1]], #[[MAP2]]]
// CHECK-SAME: iterator_types = ["parallel", "parallel", "reduction"]}
// CHECK-SAME: ins(%{{.+}}, %{{.+}} : memref<?x?x?xi8>, memref<?x?xi8>)
// CHECK-SAME: outs(%{{.+}} : memref<?x?xf32>)
// CHECK: ^{{.+}}(%[[BBARG0:.+]]: i8, %[[BBARG1:.+]]: i8, %[[BBARG2:.+]]: f32)
// CHECK: %[[BBARG0_F32:.+]] = arith.sitofp %[[BBARG0]] : i8 to f32
// CHECK: %[[BBARG1_F32:.+]] = arith.sitofp %[[BBARG1]] : i8 to f32
// CHECK: %[[MUL:.+]] = arith.mulf %[[BBARG0_F32]], %[[BBARG1_F32]]
// CHECK: %[[ADD:.+]] = arith.addf %[[BBARG2]], %[[MUL]]
// CHECK: linalg.yield %[[ADD]] : f32
// -----
func.func @batch_reduce_gemm(%lhs: memref<7x8x9xf32>, %rhs: memref<7x9x8xf32>, %out: memref<8x8xf32>) {
linalg.batch_reduce_matmul ins(%lhs, %rhs: memref<7x8x9xf32>, memref<7x9x8xf32>)
outs(%out: memref<8x8xf32>)
return
}
// CHECK-DAG: #[[MAP0:.+]] = affine_map<(d0, d1, d2, d3) -> (d0, d1, d3)>
// CHECK-DAG: #[[MAP1:.+]] = affine_map<(d0, d1, d2, d3) -> (d0, d3, d2)>
// CHECK-DAG: #[[MAP2:.+]] = affine_map<(d0, d1, d2, d3) -> (d1, d2)>
// CHECK: @batch_reduce_gemm
// CHECK: linalg.generic
// CHECK-SAME: indexing_maps = [#[[MAP0]], #[[MAP1]], #[[MAP2]]]
// CHECK-SAME: iterator_types = ["reduction", "parallel", "parallel", "reduction"]}
// CHECK-SAME: ins(%{{.+}}, %{{.+}} : memref<7x8x9xf32>, memref<7x9x8xf32>)
// CHECK-SAME: outs(%{{.+}} : memref<8x8xf32>
// CHECK: ^{{.+}}(%[[BBARG0:.+]]: f32, %[[BBARG1:.+]]: f32, %[[BBARG2:.+]]: f32)
// CHECK: %[[MUL:.+]] = arith.mulf %[[BBARG0]], %[[BBARG1]] : f32
// CHECK: %[[ADD:.+]] = arith.addf %[[BBARG2]], %[[MUL]] : f32
// CHECK: linalg.yield %[[ADD]] : f32
|
