forked from swiftlang/swift
-
Notifications
You must be signed in to change notification settings - Fork 0
/
sese_loop_canonicalization.sil
363 lines (315 loc) · 17.2 KB
/
sese_loop_canonicalization.sil
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
// RUN: %target-sil-opt -tf-xla-cfg-canonicalize -tf-ensure-single-loop-exit -assume-parsing-unqualified-ownership-sil %s -o /dev/null | %FileCheck %s
import Builtin
import Swift
import TensorFlow
/*
public func loopTest(breakCount:Int32) -> Int32 {
var count:Int32 = 0
var sum:Int32 = 0
while (count < 100) {
sum += count
count += 1
if (count == breakCount) {
break;
}
}
return sum
}
*/
// CHECK-LABEL:--- XLA CFG Canonicalize: $loopWithBreak
// CHECK: [sequence
// CHECK: <while Preheader: [[PHDR:bb[0-9]+]], Header: [[HDR:bb[0-9]+]], exit: [[EXIT:bb[0-9]+]]
// CHECK: [sequence
// CHECK: {condition Header: [[ORIG_HDR:bb[0-9]+]]
// CHECK: {condition Header: [[BRK_COND:bb[0-9]+]]
// CHECK: block [[BRK_TRUE:bb[0-9]+]]
// CHECK: block [[BRK_FALSE:bb[0-9]+]]}
// CHECK: block [[ORIG_EXIT:bb[0-9]+]]}
// CHECK: block [[LATCH:bb[0-9]+]]
// CHECK: block [[EXIT]]]
// CHECK: --- XLA CFG Canonicalize end
//-- Preheader sets up the undefs, exit index, and stayInLoop flag.
// CHECK: [[PHDR]](%0 : $Builtin.Int32):
// CHECK: [[CONST_ONE:%.*]] = integer_literal $Builtin.Int32, 1
// CHECK: [[PHDR_EXIT:%.*]] = graph_op "Const"() {dtype$dtype: i32 3, value$tensor: i32 0, __device: "ALL_DEVICES"}
// CHECK: [[PHDR_FLAG:%.*]] = graph_op "Const"() {dtype$dtype: i32 10, value$tensor: i1 -1, __device: "ALL_DEVICES"}
// CHECK: br [[HDR]]([[A:%.*]] : $Builtin.Int32, [[A]] : $Builtin.Int32, [[A]] : $Builtin.Int32, [[PHDR_EXIT]] : $TensorHandle<Builtin.Int32>, [[PHDR_FLAG]] : $TensorHandle<Builtin.Int1>)
// Original header
// CHECK: [[ORIG_HDR]]:
// CHECK: [[A:%.*]] = builtin "cmp_slt_Int32"
// CHECK: cond_br [[A]], [[BRK_COND]], [[ORIG_EXIT]]
//- Sets up index to 1 and flag to false on the exit branch of the original header to latch.
// CHECK: [[ORIG_EXIT]]:
// CHECK: [[LOCAL_EXIT_INDEX:%.*]] = graph_op "Const"() {dtype$dtype: i32 3, value$tensor: i32 1, __device: "ALL_DEVICES"} : $TensorHandle<Builtin.Int32>
// CHECK: [[LOCAL_STAY_FLAG:%.*]] = graph_op "Const"() {dtype$dtype: i32 10, value$tensor: i1 0, __device: "ALL_DEVICES"} : $TensorHandle<Builtin.Int1>
// CHECK: br [[LATCH]]([[SUM_AT_HDR:%.*]] : $Builtin.Int32, [[COUNT_AT_HDR:%.*]] : $Builtin.Int32, [[SUM_AT_HDR]] : $Builtin.Int32, [[LOCAL_EXIT_INDEX]] : $TensorHandle<Builtin.Int32>, [[LOCAL_STAY_FLAG]] : $TensorHandle<Builtin.Int1>)
// CHECK: [[BRK_COND]]:
// CHECK: [[SUM_ESCAPING:%.*]] = builtin "sadd_with_overflow_Int32"([[SUM_AT_HDR]] : $Builtin.Int32, [[COUNT_AT_HDR]] : $Builtin.Int32) : $Builtin.Int32
// CHECK: [[COUNT_ESCAPING:%.*]] = builtin "sadd_with_overflow_Int32"([[COUNT_AT_HDR]] : $Builtin.Int32, [[CONST_ONE]] : $Builtin.Int32) : $Builtin.Int32
// CHECK: cond_br {{.*}}, [[BRK_TRUE]], [[BRK_FALSE]]
//- Sets up index to 0 and flag to true on the false branch of the if with break to latch.
// CHECK: [[BRK_FALSE]]:
// CHECK: [[LOCAL_EXIT_INDEX:%.*]] = graph_op "Const"() {dtype$dtype: i32 3, value$tensor: i32 0, __device: "ALL_DEVICES"} : $TensorHandle<Builtin.Int32>
// CHECK: [[LOCAL_STAY_FLAG:%.*]] = graph_op "Const"() {dtype$dtype: i32 10, value$tensor: i1 -1, __device: "ALL_DEVICES"} : $TensorHandle<Builtin.Int1>
// CHECK: br [[LATCH]]([[SUM_ESCAPING]] : $Builtin.Int32, [[COUNT_ESCAPING]] : $Builtin.Int32, [[SUM_ESCAPING_AT_HDR:%.*]] : $Builtin.Int32, [[LOCAL_EXIT_INDEX]] : $TensorHandle<Builtin.Int32>, [[LOCAL_STAY_FLAG]] : $TensorHandle<Builtin.Int1>)
//- Sets up index to 1 and flag to false on the true branch of the if with break to latch.
// CHECK: [[BRK_TRUE]]:
// CHECK: [[LOCAL_EXIT_INDEX:%.*]] = graph_op "Const"() {dtype$dtype: i32 3, value$tensor: i32 1, __device: "ALL_DEVICES"} : $TensorHandle<Builtin.Int32>
// CHECK: [[LOCAL_STAY_FLAG:%.*]] = graph_op "Const"() {dtype$dtype: i32 10, value$tensor: i1 0, __device: "ALL_DEVICES"} : $TensorHandle<Builtin.Int1>
// CHECK: br [[LATCH]]([[SUM_AT_HDR]] : $Builtin.Int32, [[COUNT_AT_HDR]] : $Builtin.Int32, [[SUM_ESCAPING]] : $Builtin.Int32, [[LOCAL_EXIT_INDEX]] : $TensorHandle<Builtin.Int32>, [[LOCAL_STAY_FLAG]] : $TensorHandle<Builtin.Int1>)
//-- New Header simply checks flag
// CHECK: [[HDR]]([[SUM_AT_HDR]] : $Builtin.Int32, [[COUNT_AT_HDR]] : $Builtin.Int32, [[SUM_ESCAPING_AT_HDR]] : $Builtin.Int32, [[HDR_EXIT_ARG:%.*]] : $TensorHandle<Builtin.Int32>, [[HDR_FLAG_ARG:%.*]] : $TensorHandle<Builtin.Int1>):
// CHECK: [[B:%.*]] = graph_op "tf_tensor_to_i1"([[HDR_FLAG_ARG]] : $TensorHandle<Builtin.Int1>) {{.*}} : $Builtin.Int1
// CHECK: cond_br [[B]], [[ORIG_HDR]], [[EXIT]]
sil @$loopWithBreak : $@convention(thin) (Builtin.Int32) -> Builtin.Int32 {
// %0 // user: %12
bb0(%0 : $Builtin.Int32):
%1 = integer_literal $Builtin.Int32, 0 // user: %4
%2 = integer_literal $Builtin.Int32, 1 // users: %11, %4
%3 = integer_literal $Builtin.Int32, 100 // user: %7
br bb1(%1 : $Builtin.Int32, %1 : $Builtin.Int32) // id: %4
// %5 // users: %10, %9
// %6 // users: %11, %10, %7
bb1(%5 : $Builtin.Int32, %6 : $Builtin.Int32): // Preds: bb4 bb0
%7 = builtin "cmp_slt_Int32"(%6 : $Builtin.Int32, %3 : $Builtin.Int32) : $Builtin.Int1 // user: %8
cond_br %7, bb3, bb2 // id: %8
bb2: // Preds: bb1
br bb6(%5 : $Builtin.Int32) // id: %9
bb3: // Preds: bb1
%10 = builtin "sadd_with_overflow_Int32"(%5 : $Builtin.Int32, %6 : $Builtin.Int32) : $Builtin.Int32 // users: %15, %14
%11 = builtin "sadd_with_overflow_Int32"(%6 : $Builtin.Int32, %2 : $Builtin.Int32) : $Builtin.Int32 // users: %14, %12
%12 = builtin "cmp_eq_Int32"(%11 : $Builtin.Int32, %0 : $Builtin.Int32) : $Builtin.Int1 // user: %13
cond_br %12, bb5, bb4 // id: %13
bb4: // Preds: bb3
br bb1(%10 : $Builtin.Int32, %11 : $Builtin.Int32) // id: %14
bb5: // Preds: bb3
br bb6(%10 : $Builtin.Int32) // id: %15
// %16 // user: %17
bb6(%16 : $Builtin.Int32): // Preds: bb5 bb2
return %16 : $Builtin.Int32 // id: %17
}
/*
public func nestedLoopWithBreak(breakCount:Int32) -> Int32 {
var sum:Int32 = 0
var j:Int32 = 0
var count:Int32 = 0
while j < 100 {
while (count < 100) {
sum += count
count += 1
if (count == breakCount) {
break;
}
}
j += 1
count -= breakCount
if (sum > breakCount) {
break;
}
}
return sum
}
*/
// CHECK-LABEL: --- XLA CFG Canonicalize: $nestedLoopWithBreak
// CHECK:[sequence
// CHECK: <while Preheader: [[OPHDR:bb[0-9]+]], Header: [[OHDR:bb[0-9]+]], exit: [[OEXIT:bb[0-9]+]]
// CHECK: [sequence
// CHECK: {condition Header: [[OCOND:bb[0-9]+]]
// CHECK: [sequence
// CHECK: <while Preheader: [[IPHDR:bb[0-9]+]], Header: [[IHDR:bb[0-9]+]], exit: [[IEXIT:bb[0-9]+]]
// CHECK: [sequence
// CHECK: {condition Header: {{bb[0-9]+}}
// CHECK: {condition Header: {{bb[0-9]+}}
// CHECK: block {{bb[0-9]+}}
// CHECK: block {{bb[0-9]+}}
// CHECK: block {{bb[0-9]+}}
// CHECK: block {{bb[0-9]+}}]>
// CHECK: {condition Header: {{bb[0-9]+}}
// CHECK: block {{bb[0-9]+}}
// CHECK: block {{bb[0-9]+}}}]
// CHECK: block {{bb[0-9]+}}}
// CHECK: block {{bb[0-9]+}}]>
// CHECK: block [[OEXIT]]]
//-- Loop preheaders have appropriate number of additional arguments
//-- Outer loop
// CHECK: [[OPHDR]](%0 : $Builtin.Int32):
// CHECK: br [[OHDR]]([[A:%.*]] : $Builtin.Int32, [[A]] : $Builtin.Int32, [[A]] : $Builtin.Int32, [[A]] : $Builtin.Int32, {{.*}} : $TensorHandle<Builtin.Int32>, {{.*}} : $TensorHandle<Builtin.Int1>)
//-- Inner loop
// CHECK: [[IPHDR]]: // Preds: [[OCOND]]
// CHECK: br [[IHDR]]({{.*}} : $Builtin.Int32, {{.*}} : $Builtin.Int32, [[B:%[0-9]+]] : $Builtin.Int32, [[B]] : $Builtin.Int32, {{.*}} : $TensorHandle<Builtin.Int32>, {{.*}} : $TensorHandle<Builtin.Int1>)
sil @$nestedLoopWithBreak : $@convention(thin) (Builtin.Int32) -> Builtin.Int32 {
// %0 // users: %27, %26, %19
bb0(%0 : $Builtin.Int32):
%1 = integer_literal $Builtin.Int32, 0 // users: %4, %4, %4
%2 = integer_literal $Builtin.Int32, 100 // users: %14, %8
%3 = integer_literal $Builtin.Int32, 1 // users: %25, %18
br bb1(%1 : $Builtin.Int32, %1 : $Builtin.Int32, %1 : $Builtin.Int32) // id: %4
// %5 // users: %11, %10
// %6 // users: %25, %8
// %7 // user: %11
bb1(%5 : $Builtin.Int32, %6 : $Builtin.Int32, %7 : $Builtin.Int32): // Preds: bb11 bb0
%8 = builtin "cmp_slt_Int32"(%6 : $Builtin.Int32, %2 : $Builtin.Int32) : $Builtin.Int1 // user: %9
cond_br %8, bb3, bb2 // id: %9
bb2: // Preds: bb1
br bb12(%5 : $Builtin.Int32) // id: %10
bb3: // Preds: bb1
br bb4(%5 : $Builtin.Int32, %7 : $Builtin.Int32) // id: %11
// %12 // users: %17, %16
// %13 // users: %18, %17, %16, %14
bb4(%12 : $Builtin.Int32, %13 : $Builtin.Int32): // Preds: bb7 bb3
%14 = builtin "cmp_slt_Int32"(%13 : $Builtin.Int32, %2 : $Builtin.Int32) : $Builtin.Int1 // user: %15
cond_br %14, bb6, bb5 // id: %15
bb5: // Preds: bb4
br bb9(%12 : $Builtin.Int32, %13 : $Builtin.Int32) // id: %16
bb6: // Preds: bb4
%17 = builtin "sadd_with_overflow_Int32"(%12 : $Builtin.Int32, %13 : $Builtin.Int32) : $Builtin.Int32 // users: %22, %21
%18 = builtin "sadd_with_overflow_Int32"(%13 : $Builtin.Int32, %3 : $Builtin.Int32) : $Builtin.Int32 // users: %22, %21, %19
%19 = builtin "cmp_eq_Int32"(%18 : $Builtin.Int32, %0 : $Builtin.Int32) : $Builtin.Int1 // user: %20
cond_br %19, bb8, bb7 // id: %20
bb7: // Preds: bb6
br bb4(%17 : $Builtin.Int32, %18 : $Builtin.Int32) // id: %21
bb8: // Preds: bb6
br bb9(%17 : $Builtin.Int32, %18 : $Builtin.Int32) // id: %22
// %23 // users: %30, %29, %27
// %24 // user: %26
bb9(%23 : $Builtin.Int32, %24 : $Builtin.Int32): // Preds: bb8 bb5
%25 = builtin "sadd_with_overflow_Int32"(%6 : $Builtin.Int32, %3 : $Builtin.Int32) : $Builtin.Int32 // user: %30
%26 = builtin "ssub_with_overflow_Int32"(%24 : $Builtin.Int32, %0 : $Builtin.Int32) : $Builtin.Int32 // user: %30
%27 = builtin "cmp_slt_Int32"(%0 : $Builtin.Int32, %23 : $Builtin.Int32) : $Builtin.Int1 // user: %28
cond_br %27, bb10, bb11 // id: %28
bb10: // Preds: bb9
br bb12(%23 : $Builtin.Int32) // id: %29
bb11: // Preds: bb9
br bb1(%23 : $Builtin.Int32, %25 : $Builtin.Int32, %26 : $Builtin.Int32) // id: %30
// %31 // user: %32
bb12(%31 : $Builtin.Int32): // Preds: bb10 bb2
return %31 : $Builtin.Int32 // id: %32
} // end sil function '$S20canonicalize_cfg_xla10nestedLoop10breakCounts5Int32VAE_tF'
/* do while loop of the following form
public func nestedLoopWithBreak(breakCount:Int32) -> Int32 {
var j:Int32 = 0
do {
j = j + 1
if j == breakCount { break; }
j = j + 1
} while (j < 100)
return j
}
*/
// CHECK-LABEL: --- XLA CFG Canonicalize: $doWhileLoop
// CHECK: [sequence
// CHECK: {condition Header: {{bb[0-9]+}}
// CHECK: {condition Header: {{bb[0-9]+}}
// CHECK: block {{bb[0-9]+}}
// CHECK: block {{bb[0-9]+}}}
// CHECK: block {{bb[0-9]+}}}
// CHECK: <while Preheader: [[PHDR:bb[0-9]+]], Header: [[HDR:bb[0-9]+]], exit: [[EXIT:bb[0-9]+]]
// CHECK: [sequence
// CHECK: {condition Header: {{bb[0-9]+}}
// CHECK: {condition Header: {{bb[0-9]+}}
// CHECK: block {{bb[0-9]+}}
// CHECK: block {{bb[0-9]+}}}
// CHECK: block {{bb[0-9]+}}}
// CHECK: block {{bb[0-9]+}}]>
// CHECK: block {{bb[0-9]+}}]
// Make sure undef is still left in this case for now.
// CHECK: sil @$doWhileLoop : {{.*}} (Builtin.Int32) -> Builtin.Int32 {
// CHECK: [[PHDR]]({{.*}} : $TensorHandle<Builtin.Int1>):
// CHECK: br [[HDR]]({{.*}} : $Builtin.Int32, {{.*}} : $Builtin.Int32, {{.*}} : $TensorHandle<Builtin.Int32>, {{.*}} : $TensorHandle<Builtin.Int1>)
sil @$doWhileLoop : $@convention(thin) (Builtin.Int32) -> Builtin.Int32 {
bb0(%0 : $Builtin.Int32):
%1 = integer_literal $Builtin.Int32, 0
%2 = integer_literal $Builtin.Int32, 1
%3 = integer_literal $Builtin.Int32, 100
br bb1 (%3 : $Builtin.Int32)
bb1(%4: $Builtin.Int32):
%5 = builtin "sadd_with_overflow_Int32"(%4 : $Builtin.Int32, %2 : $Builtin.Int32) : $Builtin.Int32
%6 = builtin "cmp_eq_Int32"(%5 : $Builtin.Int32, %0 : $Builtin.Int32) : $Builtin.Int1
cond_br %6, bb2, bb3(%5 : $Builtin.Int32)
bb2:
%7 = builtin "sadd_with_overflow_Int32"(%5 : $Builtin.Int32, %2 : $Builtin.Int32) : $Builtin.Int32
%8 = builtin "cmp_slt_Int32"(%7 : $Builtin.Int32, %3 : $Builtin.Int32) : $Builtin.Int1
cond_br %8, bb1(%7 : $Builtin.Int32), bb3(%5 : $Builtin.Int32)
bb3 (%9 : $Builtin.Int32):
return %9 : $Builtin.Int32
}
// The SIL code is a vastly simplified version of the SIL obtained from swift
// for the following function in -O mode:
// public func natSumWithBreak(_ breakIndex: Int32) -> Tensor<Int32>{
// var i: Int32 = 1
// var sum = Tensor<Int32>(0)
// let maxCount: Int32 = 100
// while i <= maxCount {
// sum += i
// i += 1
// var j = Tensor<Int32>(breakIndex * i)
// if (i == breakIndex) {
// sum -= j
// break
// }
// j -= j
// sum += j
// }
// sum += sum
// return sum
// }
//
// This is an example where the nodes that lead to the common post dominator of all exit nodes
// is also reachable from nodes outside of the loop. The key issue is that bb9 is the common
// post dominator of the exit blocks bb5 and bb8 of the loop. As a first step, we attempt to
// make the common post dominator the single exit block. In order to do that, we need to
// move the blocks reachable from exit blocks bb5 and bb8 up to, but excluding bb9, into the loop.
// In this example, bb3 needs to be moved into the loop, but is reachable from from bb1 (outside
// the loop) and bb8 (which will be moved inside the loop). To deal with this case, we simply clone
// bb3 before moving.
//
// CHECK-LABEL: --- XLA CFG Canonicalize: $loopThatRequiresNodeCloning
// CHECK: [sequence
// CHECK: {condition Header: {{bb[0-9]+}}
// CHECK: block {{bb[0-9]+}}
// CHECK: [sequence
// CHECK: <while Preheader: {{bb[0-9]+}}, Header: {{bb[0-9]+}}, exit: {{bb[0-9]+}}
// CHECK: [sequence
// CHECK: {condition Header: {{bb[0-9]+}}
// CHECK: block {{bb[0-9]+}}
// CHECK: {condition Header: {{bb[0-9]+}}
// CHECK: block {{bb[0-9]+}}
// CHECK: block {{bb[0-9]+}}}}
// CHECK: block {{bb[0-9]+}}]>
// CHECK: block {{bb[0-9]+}}]}
// CHECK: block {{bb[0-9]+}}]
// CHECK: --- XLA CFG Canonicalize end
sil @$loopThatRequiresNodeCloning : $@convention(thin) (Builtin.Int32, Builtin.Int32) -> Builtin.Int32 {
bb0(%0 : $Builtin.Int32, %1 : $Builtin.Int32):
%2 = integer_literal $Builtin.Int32, 1
%3 = integer_literal $Builtin.Int32, 100
%4 = builtin "sadd_with_overflow_Int32"(%0 : $Builtin.Int32, %2 : $Builtin.Int32) : $Builtin.Int32
%5 = builtin "cmp_slt_Int32"(%4 : $Builtin.Int32, %3 : $Builtin.Int32) : $Builtin.Int1
cond_br %5, bb1, bb2
bb1:
// First arg is sum, the second arg is loop counter.
br bb3(%4 : $Builtin.Int32, %0 : $Builtin.Int32)
bb2:
br bb4(%4 : $Builtin.Int32, %0 : $Builtin.Int32)
bb3(%6 : $Builtin.Int32, %7 : $Builtin.Int32):
%8 = builtin "ssub_with_overflow_Int32"(%6 : $Builtin.Int32, %7 : $Builtin.Int32) : $Builtin.Int32
br bb9(%8 : $Builtin.Int32)
bb4(%9 : $Builtin.Int32, %10 : $Builtin.Int32):
%11 = builtin "sadd_with_overflow_Int32"(%10 : $Builtin.Int32, %2 : $Builtin.Int32) : $Builtin.Int32
%12 = builtin "sadd_with_overflow_Int32"(%9 : $Builtin.Int32, %10 : $Builtin.Int32) : $Builtin.Int32
%13 = builtin "cmp_slt_Int32"(%11 : $Builtin.Int32, %1 : $Builtin.Int32) : $Builtin.Int1
cond_br %13, bb5, bb6
bb5:
br bb9(%12 : $Builtin.Int32)
bb6:
%14 = builtin "sadd_with_overflow_Int32"(%11 : $Builtin.Int32, %2 : $Builtin.Int32) : $Builtin.Int32
%15 = builtin "sadd_with_overflow_Int32"(%12 : $Builtin.Int32, %14 : $Builtin.Int32) : $Builtin.Int32
%16 = builtin "cmp_slt_Int32"(%14 : $Builtin.Int32, %1 : $Builtin.Int32) : $Builtin.Int1
cond_br %16, bb8, bb7
bb7:
br bb4(%15 : $Builtin.Int32, %14 : $Builtin.Int32)
bb8:
br bb3(%15 : $Builtin.Int32, %14 : $Builtin.Int32)
bb9(%18 : $Builtin.Int32):
return %18 : $Builtin.Int32
}