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
|
// RUN: mlir-opt -allow-unregistered-dialect %s -pass-pipeline='builtin.module(func.func(affine-loop-fusion{mode=producer}))' -split-input-file | FileCheck %s --check-prefix=PRODUCER-CONSUMER
// RUN: mlir-opt -allow-unregistered-dialect %s -pass-pipeline='builtin.module(func.func(affine-loop-fusion{fusion-maximal mode=sibling}))' -split-input-file | FileCheck %s --check-prefix=SIBLING-MAXIMAL
// Part I of fusion tests in mlir/test/Transforms/loop-fusion.mlir.
// Part II of fusion tests in mlir/test/Transforms/loop-fusion-2.mlir
// Part III of fusion tests in mlir/test/Transforms/loop-fusion-3.mlir
// Expects fusion of producer into consumer at depth 4 and subsequent removal of
// source loop.
// PRODUCER-CONSUMER-LABEL: func @unflatten4d
func.func @unflatten4d(%arg1: memref<7x8x9x10xf32>) {
%m = memref.alloc() : memref<5040xf32>
%cf7 = arith.constant 7.0 : f32
affine.for %i0 = 0 to 7 {
affine.for %i1 = 0 to 8 {
affine.for %i2 = 0 to 9 {
affine.for %i3 = 0 to 10 {
affine.store %cf7, %m[720 * %i0 + 90 * %i1 + 10 * %i2 + %i3] : memref<5040xf32>
}
}
}
}
affine.for %i0 = 0 to 7 {
affine.for %i1 = 0 to 8 {
affine.for %i2 = 0 to 9 {
affine.for %i3 = 0 to 10 {
%v0 = affine.load %m[720 * %i0 + 90 * %i1 + 10 * %i2 + %i3] : memref<5040xf32>
affine.store %v0, %arg1[%i0, %i1, %i2, %i3] : memref<7x8x9x10xf32>
}
}
}
}
return
}
// PRODUCER-CONSUMER: affine.for
// PRODUCER-CONSUMER-NEXT: affine.for
// PRODUCER-CONSUMER-NEXT: affine.for
// PRODUCER-CONSUMER-NEXT: affine.for
// PRODUCER-CONSUMER-NOT: affine.for
// PRODUCER-CONSUMER: return
// -----
// Expects fusion of producer into consumer at depth 2 and subsequent removal of
// source loop.
// PRODUCER-CONSUMER-LABEL: func @unflatten2d_with_transpose
func.func @unflatten2d_with_transpose(%arg1: memref<8x7xf32>) {
%m = memref.alloc() : memref<56xf32>
%cf7 = arith.constant 7.0 : f32
affine.for %i0 = 0 to 7 {
affine.for %i1 = 0 to 8 {
affine.store %cf7, %m[8 * %i0 + %i1] : memref<56xf32>
}
}
affine.for %i0 = 0 to 8 {
affine.for %i1 = 0 to 7 {
%v0 = affine.load %m[%i0 + 8 * %i1] : memref<56xf32>
affine.store %v0, %arg1[%i0, %i1] : memref<8x7xf32>
}
}
return
}
// PRODUCER-CONSUMER: affine.for
// PRODUCER-CONSUMER-NEXT: affine.for
// PRODUCER-CONSUMER-NOT: affine.for
// PRODUCER-CONSUMER: return
// -----
// Expects fusion of producer into consumer at depth 1 and source loop to not
// be removed due to difference in loop steps.
// PRODUCER-CONSUMER-LABEL: func @check_src_dst_step
func.func @check_src_dst_step(%m : memref<100xf32>,
%src: memref<100xf32>,
%out: memref<100xf32>) {
affine.for %i0 = 0 to 100 {
%r1 = affine.load %src[%i0]: memref<100xf32>
affine.store %r1, %m[%i0] : memref<100xf32>
}
affine.for %i2 = 0 to 100 step 2 {
%r2 = affine.load %m[%i2] : memref<100xf32>
affine.store %r2, %out[%i2] : memref<100xf32>
}
return
}
// Check if the fusion did take place as well as that the source loop was
// not removed. To check if fusion took place, the read instruction from the
// original source loop is checked to be in the fused loop.
//
// PRODUCER-CONSUMER: affine.for %[[idx_0:.*]] = 0 to 100 {
// PRODUCER-CONSUMER-NEXT: %[[result_0:.*]] = affine.load %[[arr1:.*]][%[[idx_0]]] : memref<100xf32>
// PRODUCER-CONSUMER-NEXT: affine.store %[[result_0]], %{{.*}}[%[[idx_0]]] : memref<100xf32>
// PRODUCER-CONSUMER-NEXT: }
// PRODUCER-CONSUMER: affine.for %[[idx_1:.*]] = 0 to 100 step 2 {
// PRODUCER-CONSUMER: affine.load %[[arr1]][%[[idx_1]]] : memref<100xf32>
// PRODUCER-CONSUMER: }
// PRODUCER-CONSUMER: return
// -----
// SIBLING-MAXIMAL-LABEL: func @reduce_add_non_maximal_f32_f32(
func.func @reduce_add_non_maximal_f32_f32(%arg0: memref<64x64xf32, 1>, %arg1 : memref<1x64xf32, 1>, %arg2 : memref<1x64xf32, 1>) {
%cst_0 = arith.constant 0.000000e+00 : f32
%cst_1 = arith.constant 1.000000e+00 : f32
affine.for %arg3 = 0 to 1 {
affine.for %arg4 = 0 to 64 {
%accum = affine.for %arg5 = 0 to 64 iter_args (%prevAccum = %cst_0) -> f32 {
%4 = affine.load %arg0[%arg5, %arg4] : memref<64x64xf32, 1>
%5 = arith.addf %prevAccum, %4 : f32
affine.yield %5 : f32
}
%accum_dbl = arith.addf %accum, %accum : f32
affine.store %accum_dbl, %arg1[%arg3, %arg4] : memref<1x64xf32, 1>
}
}
affine.for %arg3 = 0 to 1 {
affine.for %arg4 = 0 to 64 {
// Following loop trip count does not match the corresponding source trip count.
%accum = affine.for %arg5 = 0 to 32 iter_args (%prevAccum = %cst_1) -> f32 {
%4 = affine.load %arg0[%arg5, %arg4] : memref<64x64xf32, 1>
%5 = arith.mulf %prevAccum, %4 : f32
affine.yield %5 : f32
}
%accum_sqr = arith.mulf %accum, %accum : f32
affine.store %accum_sqr, %arg2[%arg3, %arg4] : memref<1x64xf32, 1>
}
}
return
}
// Test checks the loop structure is preserved after sibling fusion
// since the destination loop and source loop trip counts do not
// match.
// SIBLING-MAXIMAL: %[[cst_0:.*]] = arith.constant 0.000000e+00 : f32
// SIBLING-MAXIMAL-NEXT: %[[cst_1:.*]] = arith.constant 1.000000e+00 : f32
// SIBLING-MAXIMAL-NEXT: affine.for %[[idx_0:.*]]= 0 to 1 {
// SIBLING-MAXIMAL-NEXT: affine.for %[[idx_1:.*]] = 0 to 64 {
// SIBLING-MAXIMAL-NEXT: %[[result_1:.*]] = affine.for %[[idx_2:.*]] = 0 to 32 iter_args(%[[iter_0:.*]] = %[[cst_1]]) -> (f32) {
// SIBLING-MAXIMAL-NEXT: %[[result_0:.*]] = affine.for %[[idx_3:.*]] = 0 to 64 iter_args(%[[iter_1:.*]] = %[[cst_0]]) -> (f32) {
// -----
// SIBLING-MAXIMAL-LABEL: func @sibling_load_only
func.func @sibling_load_only(%arg0: memref<10xf32>) {
affine.for %arg1 = 0 to 10 {
%0 = affine.load %arg0[%arg1] : memref<10xf32>
}
affine.for %arg1 = 0 to 10 {
%0 = affine.load %arg0[%arg1] : memref<10xf32>
}
// SIBLING-MAXIMAL-NEXT: affine.for
// SIBLING-MAXIMAL-NEXT: affine.load
// SIBLING-MAXIMAL-NEXT: affine.load
return
}
// -----
// PRODUCER-CONSUMER-LABEL: func @fusion_for_multiple_blocks() {
func.func @fusion_for_multiple_blocks() {
^bb0:
%m = memref.alloc() : memref<10xf32>
%cf7 = arith.constant 7.0 : f32
affine.for %i0 = 0 to 10 {
affine.store %cf7, %m[%i0] : memref<10xf32>
}
affine.for %i1 = 0 to 10 {
%v0 = affine.load %m[%i1] : memref<10xf32>
}
// PRODUCER-CONSUMER: affine.for %{{.*}} = 0 to 10 {
// PRODUCER-CONSUMER-NEXT: affine.store %{{.*}}, %{{.*}}[0] : memref<1xf32>
// PRODUCER-CONSUMER-NEXT: affine.load %{{.*}}[0] : memref<1xf32>
// PRODUCER-CONSUMER-NEXT: }
cf.br ^bb1
^bb1:
affine.for %i0 = 0 to 10 {
affine.store %cf7, %m[%i0] : memref<10xf32>
}
affine.for %i1 = 0 to 10 {
%v0 = affine.load %m[%i1] : memref<10xf32>
}
// PRODUCER-CONSUMER: affine.for %{{.*}} = 0 to 10 {
// PRODUCER-CONSUMER-NEXT: affine.store %{{.*}}, %{{.*}}[0] : memref<1xf32>
// PRODUCER-CONSUMER-NEXT: affine.load %{{.*}}[0] : memref<1xf32>
// PRODUCER-CONSUMER-NEXT: }
return
}
// -----
// PRODUCER-CONSUMER-LABEL: @fuse_higher_dim_nest_into_lower_dim_nest
func.func @fuse_higher_dim_nest_into_lower_dim_nest() {
%A = memref.alloc() : memref<8x12x128x64xf32>
%B = memref.alloc() : memref<8x128x12x64xf32>
affine.for %arg205 = 0 to 8 {
affine.for %arg206 = 0 to 128 {
affine.for %arg207 = 0 to 12 {
affine.for %arg208 = 0 to 64 {
%a = affine.load %A[%arg205, %arg207, %arg206, %arg208] : memref<8x12x128x64xf32>
affine.store %a, %B[%arg205, %arg206, %arg207, %arg208] : memref<8x128x12x64xf32>
}
}
}
}
%C = memref.alloc() : memref<8x128x768xf16>
affine.for %arg205 = 0 to 8 {
affine.for %arg206 = 0 to 128 {
affine.for %arg207 = 0 to 768 {
%b = affine.load %B[%arg205, %arg206, %arg207 floordiv 64, %arg207 mod 64] : memref<8x128x12x64xf32>
%c = arith.truncf %b : f32 to f16
affine.store %c, %C[%arg205, %arg206, %arg207] : memref<8x128x768xf16>
}
}
}
// Check that fusion happens into the innermost loop of the consumer.
// PRODUCER-CONSUMER: affine.for
// PRODUCER-CONSUMER-NEXT: affine.for %{{.*}} = 0 to 128
// PRODUCER-CONSUMER-NEXT: affine.for %{{.*}} = 0 to 768
// PRODUCER-CONSUMER-NOT: affine.for
// PRODUCER-CONSUMER: return
return
}
|
