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
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
|
/* Induction variable canonicalization and loop peeling.
Copyright (C) 2004-2017 Free Software Foundation, Inc.
This file is part of GCC.
GCC is free software; you can redistribute it and/or modify it
under the terms of the GNU General Public License as published by the
Free Software Foundation; either version 3, or (at your option) any
later version.
GCC is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
for more details.
You should have received a copy of the GNU General Public License
along with GCC; see the file COPYING3. If not see
<http://www.gnu.org/licenses/>. */
/* This pass detects the loops that iterate a constant number of times,
adds a canonical induction variable (step -1, tested against 0)
and replaces the exit test. This enables the less powerful rtl
level analysis to use this information.
This might spoil the code in some cases (by increasing register pressure).
Note that in the case the new variable is not needed, ivopts will get rid
of it, so it might only be a problem when there are no other linear induction
variables. In that case the created optimization possibilities are likely
to pay up.
We also perform
- complete unrolling (or peeling) when the loops is rolling few enough
times
- simple peeling (i.e. copying few initial iterations prior the loop)
when number of iteration estimate is known (typically by the profile
info). */
#include "config.h"
#include "system.h"
#include "coretypes.h"
#include "backend.h"
#include "tree.h"
#include "gimple.h"
#include "cfghooks.h"
#include "tree-pass.h"
#include "ssa.h"
#include "cgraph.h"
#include "gimple-pretty-print.h"
#include "fold-const.h"
#include "profile.h"
#include "gimple-fold.h"
#include "tree-eh.h"
#include "gimple-iterator.h"
#include "tree-cfg.h"
#include "tree-ssa-loop-manip.h"
#include "tree-ssa-loop-niter.h"
#include "tree-ssa-loop.h"
#include "tree-into-ssa.h"
#include "cfgloop.h"
#include "tree-chrec.h"
#include "tree-scalar-evolution.h"
#include "params.h"
#include "tree-inline.h"
#include "tree-cfgcleanup.h"
#include "builtins.h"
/* Specifies types of loops that may be unrolled. */
enum unroll_level
{
UL_SINGLE_ITER, /* Only loops that exit immediately in the first
iteration. */
UL_NO_GROWTH, /* Only loops whose unrolling will not cause increase
of code size. */
UL_ALL /* All suitable loops. */
};
/* Adds a canonical induction variable to LOOP iterating NITER times. EXIT
is the exit edge whose condition is replaced. */
static void
create_canonical_iv (struct loop *loop, edge exit, tree niter)
{
edge in;
tree type, var;
gcond *cond;
gimple_stmt_iterator incr_at;
enum tree_code cmp;
if (dump_file && (dump_flags & TDF_DETAILS))
{
fprintf (dump_file, "Added canonical iv to loop %d, ", loop->num);
print_generic_expr (dump_file, niter, TDF_SLIM);
fprintf (dump_file, " iterations.\n");
}
cond = as_a <gcond *> (last_stmt (exit->src));
in = EDGE_SUCC (exit->src, 0);
if (in == exit)
in = EDGE_SUCC (exit->src, 1);
/* Note that we do not need to worry about overflows, since
type of niter is always unsigned and all comparisons are
just for equality/nonequality -- i.e. everything works
with a modulo arithmetics. */
type = TREE_TYPE (niter);
niter = fold_build2 (PLUS_EXPR, type,
niter,
build_int_cst (type, 1));
incr_at = gsi_last_bb (in->src);
create_iv (niter,
build_int_cst (type, -1),
NULL_TREE, loop,
&incr_at, false, NULL, &var);
cmp = (exit->flags & EDGE_TRUE_VALUE) ? EQ_EXPR : NE_EXPR;
gimple_cond_set_code (cond, cmp);
gimple_cond_set_lhs (cond, var);
gimple_cond_set_rhs (cond, build_int_cst (type, 0));
update_stmt (cond);
}
/* Describe size of loop as detected by tree_estimate_loop_size. */
struct loop_size
{
/* Number of instructions in the loop. */
int overall;
/* Number of instructions that will be likely optimized out in
peeled iterations of loop (i.e. computation based on induction
variable where induction variable starts at known constant.) */
int eliminated_by_peeling;
/* Same statistics for last iteration of loop: it is smaller because
instructions after exit are not executed. */
int last_iteration;
int last_iteration_eliminated_by_peeling;
/* If some IV computation will become constant. */
bool constant_iv;
/* Number of call stmts that are not a builtin and are pure or const
present on the hot path. */
int num_pure_calls_on_hot_path;
/* Number of call stmts that are not a builtin and are not pure nor const
present on the hot path. */
int num_non_pure_calls_on_hot_path;
/* Number of statements other than calls in the loop. */
int non_call_stmts_on_hot_path;
/* Number of branches seen on the hot path. */
int num_branches_on_hot_path;
};
/* Return true if OP in STMT will be constant after peeling LOOP. */
static bool
constant_after_peeling (tree op, gimple *stmt, struct loop *loop)
{
if (is_gimple_min_invariant (op))
return true;
/* We can still fold accesses to constant arrays when index is known. */
if (TREE_CODE (op) != SSA_NAME)
{
tree base = op;
/* First make fast look if we see constant array inside. */
while (handled_component_p (base))
base = TREE_OPERAND (base, 0);
if ((DECL_P (base)
&& ctor_for_folding (base) != error_mark_node)
|| CONSTANT_CLASS_P (base))
{
/* If so, see if we understand all the indices. */
base = op;
while (handled_component_p (base))
{
if (TREE_CODE (base) == ARRAY_REF
&& !constant_after_peeling (TREE_OPERAND (base, 1), stmt, loop))
return false;
base = TREE_OPERAND (base, 0);
}
return true;
}
return false;
}
/* Induction variables are constants when defined in loop. */
if (loop_containing_stmt (stmt) != loop)
return false;
tree ev = analyze_scalar_evolution (loop, op);
if (chrec_contains_undetermined (ev)
|| chrec_contains_symbols (ev))
return false;
return true;
}
/* Computes an estimated number of insns in LOOP.
EXIT (if non-NULL) is an exite edge that will be eliminated in all but last
iteration of the loop.
EDGE_TO_CANCEL (if non-NULL) is an non-exit edge eliminated in the last iteration
of loop.
Return results in SIZE, estimate benefits for complete unrolling exiting by EXIT.
Stop estimating after UPPER_BOUND is met. Return true in this case. */
static bool
tree_estimate_loop_size (struct loop *loop, edge exit, edge edge_to_cancel,
struct loop_size *size, int upper_bound)
{
basic_block *body = get_loop_body (loop);
gimple_stmt_iterator gsi;
unsigned int i;
bool after_exit;
vec<basic_block> path = get_loop_hot_path (loop);
size->overall = 0;
size->eliminated_by_peeling = 0;
size->last_iteration = 0;
size->last_iteration_eliminated_by_peeling = 0;
size->num_pure_calls_on_hot_path = 0;
size->num_non_pure_calls_on_hot_path = 0;
size->non_call_stmts_on_hot_path = 0;
size->num_branches_on_hot_path = 0;
size->constant_iv = 0;
if (dump_file && (dump_flags & TDF_DETAILS))
fprintf (dump_file, "Estimating sizes for loop %i\n", loop->num);
for (i = 0; i < loop->num_nodes; i++)
{
if (edge_to_cancel && body[i] != edge_to_cancel->src
&& dominated_by_p (CDI_DOMINATORS, body[i], edge_to_cancel->src))
after_exit = true;
else
after_exit = false;
if (dump_file && (dump_flags & TDF_DETAILS))
fprintf (dump_file, " BB: %i, after_exit: %i\n", body[i]->index,
after_exit);
for (gsi = gsi_start_bb (body[i]); !gsi_end_p (gsi); gsi_next (&gsi))
{
gimple *stmt = gsi_stmt (gsi);
int num = estimate_num_insns (stmt, &eni_size_weights);
bool likely_eliminated = false;
bool likely_eliminated_last = false;
bool likely_eliminated_peeled = false;
if (dump_file && (dump_flags & TDF_DETAILS))
{
fprintf (dump_file, " size: %3i ", num);
print_gimple_stmt (dump_file, gsi_stmt (gsi), 0);
}
/* Look for reasons why we might optimize this stmt away. */
if (!gimple_has_side_effects (stmt))
{
/* Exit conditional. */
if (exit && body[i] == exit->src
&& stmt == last_stmt (exit->src))
{
if (dump_file && (dump_flags & TDF_DETAILS))
fprintf (dump_file, " Exit condition will be eliminated "
"in peeled copies.\n");
likely_eliminated_peeled = true;
}
if (edge_to_cancel && body[i] == edge_to_cancel->src
&& stmt == last_stmt (edge_to_cancel->src))
{
if (dump_file && (dump_flags & TDF_DETAILS))
fprintf (dump_file, " Exit condition will be eliminated "
"in last copy.\n");
likely_eliminated_last = true;
}
/* Sets of IV variables */
if (gimple_code (stmt) == GIMPLE_ASSIGN
&& constant_after_peeling (gimple_assign_lhs (stmt), stmt, loop))
{
if (dump_file && (dump_flags & TDF_DETAILS))
fprintf (dump_file, " Induction variable computation will"
" be folded away.\n");
likely_eliminated = true;
}
/* Assignments of IV variables. */
else if (gimple_code (stmt) == GIMPLE_ASSIGN
&& TREE_CODE (gimple_assign_lhs (stmt)) == SSA_NAME
&& constant_after_peeling (gimple_assign_rhs1 (stmt),
stmt, loop)
&& (gimple_assign_rhs_class (stmt) != GIMPLE_BINARY_RHS
|| constant_after_peeling (gimple_assign_rhs2 (stmt),
stmt, loop)))
{
size->constant_iv = true;
if (dump_file && (dump_flags & TDF_DETAILS))
fprintf (dump_file,
" Constant expression will be folded away.\n");
likely_eliminated = true;
}
/* Conditionals. */
else if ((gimple_code (stmt) == GIMPLE_COND
&& constant_after_peeling (gimple_cond_lhs (stmt), stmt,
loop)
&& constant_after_peeling (gimple_cond_rhs (stmt), stmt,
loop)
/* We don't simplify all constant compares so make sure
they are not both constant already. See PR70288. */
&& (! is_gimple_min_invariant (gimple_cond_lhs (stmt))
|| ! is_gimple_min_invariant
(gimple_cond_rhs (stmt))))
|| (gimple_code (stmt) == GIMPLE_SWITCH
&& constant_after_peeling (gimple_switch_index (
as_a <gswitch *>
(stmt)),
stmt, loop)
&& ! is_gimple_min_invariant
(gimple_switch_index
(as_a <gswitch *> (stmt)))))
{
if (dump_file && (dump_flags & TDF_DETAILS))
fprintf (dump_file, " Constant conditional.\n");
likely_eliminated = true;
}
}
size->overall += num;
if (likely_eliminated || likely_eliminated_peeled)
size->eliminated_by_peeling += num;
if (!after_exit)
{
size->last_iteration += num;
if (likely_eliminated || likely_eliminated_last)
size->last_iteration_eliminated_by_peeling += num;
}
if ((size->overall * 3 / 2 - size->eliminated_by_peeling
- size->last_iteration_eliminated_by_peeling) > upper_bound)
{
free (body);
path.release ();
return true;
}
}
}
while (path.length ())
{
basic_block bb = path.pop ();
for (gsi = gsi_start_bb (bb); !gsi_end_p (gsi); gsi_next (&gsi))
{
gimple *stmt = gsi_stmt (gsi);
if (gimple_code (stmt) == GIMPLE_CALL
&& !gimple_inexpensive_call_p (as_a <gcall *> (stmt)))
{
int flags = gimple_call_flags (stmt);
if (flags & (ECF_PURE | ECF_CONST))
size->num_pure_calls_on_hot_path++;
else
size->num_non_pure_calls_on_hot_path++;
size->num_branches_on_hot_path ++;
}
/* Count inexpensive calls as non-calls, because they will likely
expand inline. */
else if (gimple_code (stmt) != GIMPLE_DEBUG)
size->non_call_stmts_on_hot_path++;
if (((gimple_code (stmt) == GIMPLE_COND
&& (!constant_after_peeling (gimple_cond_lhs (stmt), stmt, loop)
|| constant_after_peeling (gimple_cond_rhs (stmt), stmt,
loop)))
|| (gimple_code (stmt) == GIMPLE_SWITCH
&& !constant_after_peeling (gimple_switch_index (
as_a <gswitch *> (stmt)),
stmt, loop)))
&& (!exit || bb != exit->src))
size->num_branches_on_hot_path++;
}
}
path.release ();
if (dump_file && (dump_flags & TDF_DETAILS))
fprintf (dump_file, "size: %i-%i, last_iteration: %i-%i\n", size->overall,
size->eliminated_by_peeling, size->last_iteration,
size->last_iteration_eliminated_by_peeling);
free (body);
return false;
}
/* Estimate number of insns of completely unrolled loop.
It is (NUNROLL + 1) * size of loop body with taking into account
the fact that in last copy everything after exit conditional
is dead and that some instructions will be eliminated after
peeling.
Loop body is likely going to simplify further, this is difficult
to guess, we just decrease the result by 1/3. */
static unsigned HOST_WIDE_INT
estimated_unrolled_size (struct loop_size *size,
unsigned HOST_WIDE_INT nunroll)
{
HOST_WIDE_INT unr_insns = ((nunroll)
* (HOST_WIDE_INT) (size->overall
- size->eliminated_by_peeling));
if (!nunroll)
unr_insns = 0;
unr_insns += size->last_iteration - size->last_iteration_eliminated_by_peeling;
unr_insns = unr_insns * 2 / 3;
if (unr_insns <= 0)
unr_insns = 1;
return unr_insns;
}
/* Loop LOOP is known to not loop. See if there is an edge in the loop
body that can be remove to make the loop to always exit and at
the same time it does not make any code potentially executed
during the last iteration dead.
After complete unrolling we still may get rid of the conditional
on the exit in the last copy even if we have no idea what it does.
This is quite common case for loops of form
int a[5];
for (i=0;i<b;i++)
a[i]=0;
Here we prove the loop to iterate 5 times but we do not know
it from induction variable.
For now we handle only simple case where there is exit condition
just before the latch block and the latch block contains no statements
with side effect that may otherwise terminate the execution of loop
(such as by EH or by terminating the program or longjmp).
In the general case we may want to cancel the paths leading to statements
loop-niter identified as having undefined effect in the last iteration.
The other cases are hopefully rare and will be cleaned up later. */
static edge
loop_edge_to_cancel (struct loop *loop)
{
vec<edge> exits;
unsigned i;
edge edge_to_cancel;
gimple_stmt_iterator gsi;
/* We want only one predecestor of the loop. */
if (EDGE_COUNT (loop->latch->preds) > 1)
return NULL;
exits = get_loop_exit_edges (loop);
FOR_EACH_VEC_ELT (exits, i, edge_to_cancel)
{
/* Find the other edge than the loop exit
leaving the conditoinal. */
if (EDGE_COUNT (edge_to_cancel->src->succs) != 2)
continue;
if (EDGE_SUCC (edge_to_cancel->src, 0) == edge_to_cancel)
edge_to_cancel = EDGE_SUCC (edge_to_cancel->src, 1);
else
edge_to_cancel = EDGE_SUCC (edge_to_cancel->src, 0);
/* We only can handle conditionals. */
if (!(edge_to_cancel->flags & (EDGE_TRUE_VALUE | EDGE_FALSE_VALUE)))
continue;
/* We should never have conditionals in the loop latch. */
gcc_assert (edge_to_cancel->dest != loop->header);
/* Check that it leads to loop latch. */
if (edge_to_cancel->dest != loop->latch)
continue;
exits.release ();
/* Verify that the code in loop latch does nothing that may end program
execution without really reaching the exit. This may include
non-pure/const function calls, EH statements, volatile ASMs etc. */
for (gsi = gsi_start_bb (loop->latch); !gsi_end_p (gsi); gsi_next (&gsi))
if (gimple_has_side_effects (gsi_stmt (gsi)))
return NULL;
return edge_to_cancel;
}
exits.release ();
return NULL;
}
/* Remove all tests for exits that are known to be taken after LOOP was
peeled NPEELED times. Put gcc_unreachable before every statement
known to not be executed. */
static bool
remove_exits_and_undefined_stmts (struct loop *loop, unsigned int npeeled)
{
struct nb_iter_bound *elt;
bool changed = false;
for (elt = loop->bounds; elt; elt = elt->next)
{
/* If statement is known to be undefined after peeling, turn it
into unreachable (or trap when debugging experience is supposed
to be good). */
if (!elt->is_exit
&& wi::ltu_p (elt->bound, npeeled))
{
gimple_stmt_iterator gsi = gsi_for_stmt (elt->stmt);
gcall *stmt = gimple_build_call
(builtin_decl_implicit (BUILT_IN_UNREACHABLE), 0);
gimple_set_location (stmt, gimple_location (elt->stmt));
gsi_insert_before (&gsi, stmt, GSI_NEW_STMT);
split_block (gimple_bb (stmt), stmt);
changed = true;
if (dump_file && (dump_flags & TDF_DETAILS))
{
fprintf (dump_file, "Forced statement unreachable: ");
print_gimple_stmt (dump_file, elt->stmt, 0);
}
}
/* If we know the exit will be taken after peeling, update. */
else if (elt->is_exit
&& wi::leu_p (elt->bound, npeeled))
{
basic_block bb = gimple_bb (elt->stmt);
edge exit_edge = EDGE_SUCC (bb, 0);
if (dump_file && (dump_flags & TDF_DETAILS))
{
fprintf (dump_file, "Forced exit to be taken: ");
print_gimple_stmt (dump_file, elt->stmt, 0);
}
if (!loop_exit_edge_p (loop, exit_edge))
exit_edge = EDGE_SUCC (bb, 1);
exit_edge->probability = REG_BR_PROB_BASE;
exit_edge->count = exit_edge->src->count;
gcc_checking_assert (loop_exit_edge_p (loop, exit_edge));
gcond *cond_stmt = as_a <gcond *> (elt->stmt);
if (exit_edge->flags & EDGE_TRUE_VALUE)
gimple_cond_make_true (cond_stmt);
else
gimple_cond_make_false (cond_stmt);
update_stmt (cond_stmt);
changed = true;
}
}
return changed;
}
/* Remove all exits that are known to be never taken because of the loop bound
discovered. */
static bool
remove_redundant_iv_tests (struct loop *loop)
{
struct nb_iter_bound *elt;
bool changed = false;
if (!loop->any_upper_bound)
return false;
for (elt = loop->bounds; elt; elt = elt->next)
{
/* Exit is pointless if it won't be taken before loop reaches
upper bound. */
if (elt->is_exit && loop->any_upper_bound
&& wi::ltu_p (loop->nb_iterations_upper_bound, elt->bound))
{
basic_block bb = gimple_bb (elt->stmt);
edge exit_edge = EDGE_SUCC (bb, 0);
struct tree_niter_desc niter;
if (!loop_exit_edge_p (loop, exit_edge))
exit_edge = EDGE_SUCC (bb, 1);
/* Only when we know the actual number of iterations, not
just a bound, we can remove the exit. */
if (!number_of_iterations_exit (loop, exit_edge,
&niter, false, false)
|| !integer_onep (niter.assumptions)
|| !integer_zerop (niter.may_be_zero)
|| !niter.niter
|| TREE_CODE (niter.niter) != INTEGER_CST
|| !wi::ltu_p (loop->nb_iterations_upper_bound,
wi::to_widest (niter.niter)))
continue;
if (dump_file && (dump_flags & TDF_DETAILS))
{
fprintf (dump_file, "Removed pointless exit: ");
print_gimple_stmt (dump_file, elt->stmt, 0);
}
gcond *cond_stmt = as_a <gcond *> (elt->stmt);
if (exit_edge->flags & EDGE_TRUE_VALUE)
gimple_cond_make_false (cond_stmt);
else
gimple_cond_make_true (cond_stmt);
update_stmt (cond_stmt);
changed = true;
}
}
return changed;
}
/* Stores loops that will be unlooped and edges that will be removed
after we process whole loop tree. */
static vec<loop_p> loops_to_unloop;
static vec<int> loops_to_unloop_nunroll;
static vec<edge> edges_to_remove;
/* Stores loops that has been peeled. */
static bitmap peeled_loops;
/* Cancel all fully unrolled loops by putting __builtin_unreachable
on the latch edge.
We do it after all unrolling since unlooping moves basic blocks
across loop boundaries trashing loop closed SSA form as well
as SCEV info needed to be intact during unrolling.
IRRED_INVALIDATED is used to bookkeep if information about
irreducible regions may become invalid as a result
of the transformation.
LOOP_CLOSED_SSA_INVALIDATED is used to bookkepp the case
when we need to go into loop closed SSA form. */
static void
unloop_loops (bitmap loop_closed_ssa_invalidated,
bool *irred_invalidated)
{
while (loops_to_unloop.length ())
{
struct loop *loop = loops_to_unloop.pop ();
int n_unroll = loops_to_unloop_nunroll.pop ();
basic_block latch = loop->latch;
edge latch_edge = loop_latch_edge (loop);
int flags = latch_edge->flags;
location_t locus = latch_edge->goto_locus;
gcall *stmt;
gimple_stmt_iterator gsi;
remove_exits_and_undefined_stmts (loop, n_unroll);
/* Unloop destroys the latch edge. */
unloop (loop, irred_invalidated, loop_closed_ssa_invalidated);
/* Create new basic block for the latch edge destination and wire
it in. */
stmt = gimple_build_call (builtin_decl_implicit (BUILT_IN_UNREACHABLE), 0);
latch_edge = make_edge (latch, create_basic_block (NULL, NULL, latch), flags);
latch_edge->probability = 0;
latch_edge->count = profile_count::zero ();
latch_edge->flags |= flags;
latch_edge->goto_locus = locus;
add_bb_to_loop (latch_edge->dest, current_loops->tree_root);
latch_edge->dest->count = profile_count::zero ();
latch_edge->dest->frequency = 0;
set_immediate_dominator (CDI_DOMINATORS, latch_edge->dest, latch_edge->src);
gsi = gsi_start_bb (latch_edge->dest);
gsi_insert_after (&gsi, stmt, GSI_NEW_STMT);
}
loops_to_unloop.release ();
loops_to_unloop_nunroll.release ();
/* Remove edges in peeled copies. */
unsigned i;
edge e;
FOR_EACH_VEC_ELT (edges_to_remove, i, e)
{
bool ok = remove_path (e, irred_invalidated, loop_closed_ssa_invalidated);
gcc_assert (ok);
}
edges_to_remove.release ();
}
/* Tries to unroll LOOP completely, i.e. NITER times.
UL determines which loops we are allowed to unroll.
EXIT is the exit of the loop that should be eliminated.
MAXITER specfy bound on number of iterations, -1 if it is
not known or too large for HOST_WIDE_INT. The location
LOCUS corresponding to the loop is used when emitting
a summary of the unroll to the dump file. */
static bool
try_unroll_loop_completely (struct loop *loop,
edge exit, tree niter,
enum unroll_level ul,
HOST_WIDE_INT maxiter,
location_t locus)
{
unsigned HOST_WIDE_INT n_unroll = 0, ninsns, unr_insns;
struct loop_size size;
bool n_unroll_found = false;
edge edge_to_cancel = NULL;
dump_flags_t report_flags = MSG_OPTIMIZED_LOCATIONS | TDF_DETAILS;
/* See if we proved number of iterations to be low constant.
EXIT is an edge that will be removed in all but last iteration of
the loop.
EDGE_TO_CACNEL is an edge that will be removed from the last iteration
of the unrolled sequence and is expected to make the final loop not
rolling.
If the number of execution of loop is determined by standard induction
variable test, then EXIT and EDGE_TO_CANCEL are the two edges leaving
from the iv test. */
if (tree_fits_uhwi_p (niter))
{
n_unroll = tree_to_uhwi (niter);
n_unroll_found = true;
edge_to_cancel = EDGE_SUCC (exit->src, 0);
if (edge_to_cancel == exit)
edge_to_cancel = EDGE_SUCC (exit->src, 1);
}
/* We do not know the number of iterations and thus we can not eliminate
the EXIT edge. */
else
exit = NULL;
/* See if we can improve our estimate by using recorded loop bounds. */
if (maxiter >= 0
&& (!n_unroll_found || (unsigned HOST_WIDE_INT)maxiter < n_unroll))
{
n_unroll = maxiter;
n_unroll_found = true;
/* Loop terminates before the IV variable test, so we can not
remove it in the last iteration. */
edge_to_cancel = NULL;
}
if (!n_unroll_found)
return false;
if (n_unroll > (unsigned) PARAM_VALUE (PARAM_MAX_COMPLETELY_PEEL_TIMES))
{
if (dump_file && (dump_flags & TDF_DETAILS))
fprintf (dump_file, "Not unrolling loop %d "
"(--param max-completely-peel-times limit reached).\n",
loop->num);
return false;
}
if (!edge_to_cancel)
edge_to_cancel = loop_edge_to_cancel (loop);
if (n_unroll)
{
bool large;
if (ul == UL_SINGLE_ITER)
return false;
/* EXIT can be removed only if we are sure it passes first N_UNROLL
iterations. */
bool remove_exit = (exit && niter
&& TREE_CODE (niter) == INTEGER_CST
&& wi::leu_p (n_unroll, wi::to_widest (niter)));
large = tree_estimate_loop_size
(loop, remove_exit ? exit : NULL, edge_to_cancel, &size,
PARAM_VALUE (PARAM_MAX_COMPLETELY_PEELED_INSNS));
ninsns = size.overall;
if (large)
{
if (dump_file && (dump_flags & TDF_DETAILS))
fprintf (dump_file, "Not unrolling loop %d: it is too large.\n",
loop->num);
return false;
}
unr_insns = estimated_unrolled_size (&size, n_unroll);
if (dump_file && (dump_flags & TDF_DETAILS))
{
fprintf (dump_file, " Loop size: %d\n", (int) ninsns);
fprintf (dump_file, " Estimated size after unrolling: %d\n",
(int) unr_insns);
}
/* If the code is going to shrink, we don't need to be extra cautious
on guessing if the unrolling is going to be profitable. */
if (unr_insns
/* If there is IV variable that will become constant, we save
one instruction in the loop prologue we do not account
otherwise. */
<= ninsns + (size.constant_iv != false))
;
/* We unroll only inner loops, because we do not consider it profitable
otheriwse. We still can cancel loopback edge of not rolling loop;
this is always a good idea. */
else if (ul == UL_NO_GROWTH)
{
if (dump_file && (dump_flags & TDF_DETAILS))
fprintf (dump_file, "Not unrolling loop %d: size would grow.\n",
loop->num);
return false;
}
/* Outer loops tend to be less interesting candidates for complete
unrolling unless we can do a lot of propagation into the inner loop
body. For now we disable outer loop unrolling when the code would
grow. */
else if (loop->inner)
{
if (dump_file && (dump_flags & TDF_DETAILS))
fprintf (dump_file, "Not unrolling loop %d: "
"it is not innermost and code would grow.\n",
loop->num);
return false;
}
/* If there is call on a hot path through the loop, then
there is most probably not much to optimize. */
else if (size.num_non_pure_calls_on_hot_path)
{
if (dump_file && (dump_flags & TDF_DETAILS))
fprintf (dump_file, "Not unrolling loop %d: "
"contains call and code would grow.\n",
loop->num);
return false;
}
/* If there is pure/const call in the function, then we
can still optimize the unrolled loop body if it contains
some other interesting code than the calls and code
storing or cumulating the return value. */
else if (size.num_pure_calls_on_hot_path
/* One IV increment, one test, one ivtmp store
and one useful stmt. That is about minimal loop
doing pure call. */
&& (size.non_call_stmts_on_hot_path
<= 3 + size.num_pure_calls_on_hot_path))
{
if (dump_file && (dump_flags & TDF_DETAILS))
fprintf (dump_file, "Not unrolling loop %d: "
"contains just pure calls and code would grow.\n",
loop->num);
return false;
}
/* Complete unrolling is a major win when control flow is removed and
one big basic block is created. If the loop contains control flow
the optimization may still be a win because of eliminating the loop
overhead but it also may blow the branch predictor tables.
Limit number of branches on the hot path through the peeled
sequence. */
else if (size.num_branches_on_hot_path * (int)n_unroll
> PARAM_VALUE (PARAM_MAX_PEEL_BRANCHES))
{
if (dump_file && (dump_flags & TDF_DETAILS))
fprintf (dump_file, "Not unrolling loop %d: "
" number of branches on hot path in the unrolled sequence"
" reach --param max-peel-branches limit.\n",
loop->num);
return false;
}
else if (unr_insns
> (unsigned) PARAM_VALUE (PARAM_MAX_COMPLETELY_PEELED_INSNS))
{
if (dump_file && (dump_flags & TDF_DETAILS))
fprintf (dump_file, "Not unrolling loop %d: "
"(--param max-completely-peeled-insns limit reached).\n",
loop->num);
return false;
}
if (!n_unroll)
dump_printf_loc (report_flags, locus,
"loop turned into non-loop; it never loops.\n");
initialize_original_copy_tables ();
auto_sbitmap wont_exit (n_unroll + 1);
if (exit && niter
&& TREE_CODE (niter) == INTEGER_CST
&& wi::leu_p (n_unroll, wi::to_widest (niter)))
{
bitmap_ones (wont_exit);
if (wi::eq_p (wi::to_widest (niter), n_unroll)
|| edge_to_cancel)
bitmap_clear_bit (wont_exit, 0);
}
else
{
exit = NULL;
bitmap_clear (wont_exit);
}
if (!gimple_duplicate_loop_to_header_edge (loop, loop_preheader_edge (loop),
n_unroll, wont_exit,
exit, &edges_to_remove,
DLTHE_FLAG_UPDATE_FREQ
| DLTHE_FLAG_COMPLETTE_PEEL))
{
free_original_copy_tables ();
if (dump_file && (dump_flags & TDF_DETAILS))
fprintf (dump_file, "Failed to duplicate the loop\n");
return false;
}
free_original_copy_tables ();
}
/* Remove the conditional from the last copy of the loop. */
if (edge_to_cancel)
{
gcond *cond = as_a <gcond *> (last_stmt (edge_to_cancel->src));
force_edge_cold (edge_to_cancel, true);
if (edge_to_cancel->flags & EDGE_TRUE_VALUE)
gimple_cond_make_false (cond);
else
gimple_cond_make_true (cond);
update_stmt (cond);
/* Do not remove the path. Doing so may remove outer loop
and confuse bookkeeping code in tree_unroll_loops_completelly. */
}
/* Store the loop for later unlooping and exit removal. */
loops_to_unloop.safe_push (loop);
loops_to_unloop_nunroll.safe_push (n_unroll);
if (dump_enabled_p ())
{
if (!n_unroll)
dump_printf_loc (MSG_OPTIMIZED_LOCATIONS | TDF_DETAILS, locus,
"loop turned into non-loop; it never loops\n");
else
{
dump_printf_loc (MSG_OPTIMIZED_LOCATIONS | TDF_DETAILS, locus,
"loop with %d iterations completely unrolled",
(int) (n_unroll + 1));
if (loop->header->count.initialized_p ())
dump_printf (MSG_OPTIMIZED_LOCATIONS | TDF_DETAILS,
" (header execution count %d)",
(int)loop->header->count.to_gcov_type ());
dump_printf (MSG_OPTIMIZED_LOCATIONS | TDF_DETAILS, "\n");
}
}
if (dump_file && (dump_flags & TDF_DETAILS))
{
if (exit)
fprintf (dump_file, "Exit condition of peeled iterations was "
"eliminated.\n");
if (edge_to_cancel)
fprintf (dump_file, "Last iteration exit edge was proved true.\n");
else
fprintf (dump_file, "Latch of last iteration was marked by "
"__builtin_unreachable ().\n");
}
return true;
}
/* Return number of instructions after peeling. */
static unsigned HOST_WIDE_INT
estimated_peeled_sequence_size (struct loop_size *size,
unsigned HOST_WIDE_INT npeel)
{
return MAX (npeel * (HOST_WIDE_INT) (size->overall
- size->eliminated_by_peeling), 1);
}
/* If the loop is expected to iterate N times and is
small enough, duplicate the loop body N+1 times before
the loop itself. This way the hot path will never
enter the loop.
Parameters are the same as for try_unroll_loops_completely */
static bool
try_peel_loop (struct loop *loop,
edge exit, tree niter,
HOST_WIDE_INT maxiter)
{
HOST_WIDE_INT npeel;
struct loop_size size;
int peeled_size;
if (!flag_peel_loops || PARAM_VALUE (PARAM_MAX_PEEL_TIMES) <= 0
|| !peeled_loops)
return false;
if (bitmap_bit_p (peeled_loops, loop->num))
{
if (dump_file)
fprintf (dump_file, "Not peeling: loop is already peeled\n");
return false;
}
/* Peel only innermost loops.
While the code is perfectly capable of peeling non-innermost loops,
the heuristics would probably need some improvements. */
if (loop->inner)
{
if (dump_file)
fprintf (dump_file, "Not peeling: outer loop\n");
return false;
}
if (!optimize_loop_for_speed_p (loop))
{
if (dump_file)
fprintf (dump_file, "Not peeling: cold loop\n");
return false;
}
/* Check if there is an estimate on the number of iterations. */
npeel = estimated_loop_iterations_int (loop);
if (npeel < 0)
npeel = likely_max_loop_iterations_int (loop);
if (npeel < 0)
{
if (dump_file)
fprintf (dump_file, "Not peeling: number of iterations is not "
"estimated\n");
return false;
}
if (maxiter >= 0 && maxiter <= npeel)
{
if (dump_file)
fprintf (dump_file, "Not peeling: upper bound is known so can "
"unroll completely\n");
return false;
}
/* We want to peel estimated number of iterations + 1 (so we never
enter the loop on quick path). Check against PARAM_MAX_PEEL_TIMES
and be sure to avoid overflows. */
if (npeel > PARAM_VALUE (PARAM_MAX_PEEL_TIMES) - 1)
{
if (dump_file)
fprintf (dump_file, "Not peeling: rolls too much "
"(%i + 1 > --param max-peel-times)\n", (int) npeel);
return false;
}
npeel++;
/* Check peeled loops size. */
tree_estimate_loop_size (loop, exit, NULL, &size,
PARAM_VALUE (PARAM_MAX_PEELED_INSNS));
if ((peeled_size = estimated_peeled_sequence_size (&size, (int) npeel))
> PARAM_VALUE (PARAM_MAX_PEELED_INSNS))
{
if (dump_file)
fprintf (dump_file, "Not peeling: peeled sequence size is too large "
"(%i insns > --param max-peel-insns)", peeled_size);
return false;
}
/* Duplicate possibly eliminating the exits. */
initialize_original_copy_tables ();
auto_sbitmap wont_exit (npeel + 1);
if (exit && niter
&& TREE_CODE (niter) == INTEGER_CST
&& wi::leu_p (npeel, wi::to_widest (niter)))
{
bitmap_ones (wont_exit);
bitmap_clear_bit (wont_exit, 0);
}
else
{
exit = NULL;
bitmap_clear (wont_exit);
}
if (!gimple_duplicate_loop_to_header_edge (loop, loop_preheader_edge (loop),
npeel, wont_exit,
exit, &edges_to_remove,
DLTHE_FLAG_UPDATE_FREQ))
{
free_original_copy_tables ();
return false;
}
free_original_copy_tables ();
if (dump_file && (dump_flags & TDF_DETAILS))
{
fprintf (dump_file, "Peeled loop %d, %i times.\n",
loop->num, (int) npeel);
}
if (loop->any_estimate)
{
if (wi::ltu_p (npeel, loop->nb_iterations_estimate))
loop->nb_iterations_estimate -= npeel;
else
loop->nb_iterations_estimate = 0;
}
if (loop->any_upper_bound)
{
if (wi::ltu_p (npeel, loop->nb_iterations_upper_bound))
loop->nb_iterations_upper_bound -= npeel;
else
loop->nb_iterations_upper_bound = 0;
}
if (loop->any_likely_upper_bound)
{
if (wi::ltu_p (npeel, loop->nb_iterations_likely_upper_bound))
loop->nb_iterations_likely_upper_bound -= npeel;
else
{
loop->any_estimate = true;
loop->nb_iterations_estimate = 0;
loop->nb_iterations_likely_upper_bound = 0;
}
}
profile_count entry_count = profile_count::zero ();
int entry_freq = 0;
edge e;
edge_iterator ei;
FOR_EACH_EDGE (e, ei, loop->header->preds)
if (e->src != loop->latch)
{
if (e->src->count.initialized_p ())
entry_count = e->src->count + e->src->count;
entry_freq += e->src->frequency;
gcc_assert (!flow_bb_inside_loop_p (loop, e->src));
}
int scale = 1;
if (loop->header->count > 0)
scale = entry_count.probability_in (loop->header->count);
else if (loop->header->frequency)
scale = RDIV (entry_freq * REG_BR_PROB_BASE, loop->header->frequency);
scale_loop_profile (loop, scale, 0);
bitmap_set_bit (peeled_loops, loop->num);
return true;
}
/* Adds a canonical induction variable to LOOP if suitable.
CREATE_IV is true if we may create a new iv. UL determines
which loops we are allowed to completely unroll. If TRY_EVAL is true, we try
to determine the number of iterations of a loop by direct evaluation.
Returns true if cfg is changed. */
static bool
canonicalize_loop_induction_variables (struct loop *loop,
bool create_iv, enum unroll_level ul,
bool try_eval)
{
edge exit = NULL;
tree niter;
HOST_WIDE_INT maxiter;
bool modified = false;
location_t locus = UNKNOWN_LOCATION;
niter = number_of_latch_executions (loop);
exit = single_exit (loop);
if (TREE_CODE (niter) == INTEGER_CST)
locus = gimple_location (last_stmt (exit->src));
else
{
/* If the loop has more than one exit, try checking all of them
for # of iterations determinable through scev. */
if (!exit)
niter = find_loop_niter (loop, &exit);
/* Finally if everything else fails, try brute force evaluation. */
if (try_eval
&& (chrec_contains_undetermined (niter)
|| TREE_CODE (niter) != INTEGER_CST))
niter = find_loop_niter_by_eval (loop, &exit);
if (exit)
locus = gimple_location (last_stmt (exit->src));
if (TREE_CODE (niter) != INTEGER_CST)
exit = NULL;
}
/* We work exceptionally hard here to estimate the bound
by find_loop_niter_by_eval. Be sure to keep it for future. */
if (niter && TREE_CODE (niter) == INTEGER_CST)
{
record_niter_bound (loop, wi::to_widest (niter),
exit == single_likely_exit (loop), true);
}
/* Force re-computation of loop bounds so we can remove redundant exits. */
maxiter = max_loop_iterations_int (loop);
if (dump_file && (dump_flags & TDF_DETAILS)
&& TREE_CODE (niter) == INTEGER_CST)
{
fprintf (dump_file, "Loop %d iterates ", loop->num);
print_generic_expr (dump_file, niter, TDF_SLIM);
fprintf (dump_file, " times.\n");
}
if (dump_file && (dump_flags & TDF_DETAILS)
&& maxiter >= 0)
{
fprintf (dump_file, "Loop %d iterates at most %i times.\n", loop->num,
(int)maxiter);
}
if (dump_file && (dump_flags & TDF_DETAILS)
&& likely_max_loop_iterations_int (loop) >= 0)
{
fprintf (dump_file, "Loop %d likely iterates at most %i times.\n",
loop->num, (int)likely_max_loop_iterations_int (loop));
}
/* Remove exits that are known to be never taken based on loop bound.
Needs to be called after compilation of max_loop_iterations_int that
populates the loop bounds. */
modified |= remove_redundant_iv_tests (loop);
if (try_unroll_loop_completely (loop, exit, niter, ul, maxiter, locus))
return true;
if (create_iv
&& niter && !chrec_contains_undetermined (niter)
&& exit && just_once_each_iteration_p (loop, exit->src))
create_canonical_iv (loop, exit, niter);
if (ul == UL_ALL)
modified |= try_peel_loop (loop, exit, niter, maxiter);
return modified;
}
/* The main entry point of the pass. Adds canonical induction variables
to the suitable loops. */
unsigned int
canonicalize_induction_variables (void)
{
struct loop *loop;
bool changed = false;
bool irred_invalidated = false;
bitmap loop_closed_ssa_invalidated = BITMAP_ALLOC (NULL);
estimate_numbers_of_iterations ();
FOR_EACH_LOOP (loop, LI_FROM_INNERMOST)
{
changed |= canonicalize_loop_induction_variables (loop,
true, UL_SINGLE_ITER,
true);
}
gcc_assert (!need_ssa_update_p (cfun));
unloop_loops (loop_closed_ssa_invalidated, &irred_invalidated);
if (irred_invalidated
&& loops_state_satisfies_p (LOOPS_HAVE_MARKED_IRREDUCIBLE_REGIONS))
mark_irreducible_loops ();
/* Clean up the information about numbers of iterations, since brute force
evaluation could reveal new information. */
free_numbers_of_iterations_estimates (cfun);
scev_reset ();
if (!bitmap_empty_p (loop_closed_ssa_invalidated))
{
gcc_checking_assert (loops_state_satisfies_p (LOOP_CLOSED_SSA));
rewrite_into_loop_closed_ssa (NULL, TODO_update_ssa);
}
BITMAP_FREE (loop_closed_ssa_invalidated);
if (changed)
return TODO_cleanup_cfg;
return 0;
}
/* Propagate constant SSA_NAMEs defined in basic block BB. */
static void
propagate_constants_for_unrolling (basic_block bb)
{
/* Look for degenerate PHI nodes with constant argument. */
for (gphi_iterator gsi = gsi_start_phis (bb); !gsi_end_p (gsi); )
{
gphi *phi = gsi.phi ();
tree result = gimple_phi_result (phi);
tree arg = gimple_phi_arg_def (phi, 0);
if (! SSA_NAME_OCCURS_IN_ABNORMAL_PHI (result)
&& gimple_phi_num_args (phi) == 1
&& CONSTANT_CLASS_P (arg))
{
replace_uses_by (result, arg);
gsi_remove (&gsi, true);
release_ssa_name (result);
}
else
gsi_next (&gsi);
}
/* Look for assignments to SSA names with constant RHS. */
for (gimple_stmt_iterator gsi = gsi_start_bb (bb); !gsi_end_p (gsi); )
{
gimple *stmt = gsi_stmt (gsi);
tree lhs;
if (is_gimple_assign (stmt)
&& TREE_CODE_CLASS (gimple_assign_rhs_code (stmt)) == tcc_constant
&& (lhs = gimple_assign_lhs (stmt), TREE_CODE (lhs) == SSA_NAME)
&& !SSA_NAME_OCCURS_IN_ABNORMAL_PHI (lhs))
{
replace_uses_by (lhs, gimple_assign_rhs1 (stmt));
gsi_remove (&gsi, true);
release_ssa_name (lhs);
}
else
gsi_next (&gsi);
}
}
/* Process loops from innermost to outer, stopping at the innermost
loop we unrolled. */
static bool
tree_unroll_loops_completely_1 (bool may_increase_size, bool unroll_outer,
bitmap father_bbs, struct loop *loop)
{
struct loop *loop_father;
bool changed = false;
struct loop *inner;
enum unroll_level ul;
/* Process inner loops first. */
for (inner = loop->inner; inner != NULL; inner = inner->next)
changed |= tree_unroll_loops_completely_1 (may_increase_size,
unroll_outer, father_bbs,
inner);
/* If we changed an inner loop we cannot process outer loops in this
iteration because SSA form is not up-to-date. Continue with
siblings of outer loops instead. */
if (changed)
return true;
/* Don't unroll #pragma omp simd loops until the vectorizer
attempts to vectorize those. */
if (loop->force_vectorize)
return false;
/* Try to unroll this loop. */
loop_father = loop_outer (loop);
if (!loop_father)
return false;
if (may_increase_size && optimize_loop_nest_for_speed_p (loop)
/* Unroll outermost loops only if asked to do so or they do
not cause code growth. */
&& (unroll_outer || loop_outer (loop_father)))
ul = UL_ALL;
else
ul = UL_NO_GROWTH;
if (canonicalize_loop_induction_variables
(loop, false, ul, !flag_tree_loop_ivcanon))
{
/* If we'll continue unrolling, we need to propagate constants
within the new basic blocks to fold away induction variable
computations; otherwise, the size might blow up before the
iteration is complete and the IR eventually cleaned up. */
if (loop_outer (loop_father))
bitmap_set_bit (father_bbs, loop_father->header->index);
return true;
}
return false;
}
/* Unroll LOOPS completely if they iterate just few times. Unless
MAY_INCREASE_SIZE is true, perform the unrolling only if the
size of the code does not increase. */
unsigned int
tree_unroll_loops_completely (bool may_increase_size, bool unroll_outer)
{
bitmap father_bbs = BITMAP_ALLOC (NULL);
bool changed;
int iteration = 0;
bool irred_invalidated = false;
do
{
changed = false;
bitmap loop_closed_ssa_invalidated = NULL;
if (loops_state_satisfies_p (LOOP_CLOSED_SSA))
loop_closed_ssa_invalidated = BITMAP_ALLOC (NULL);
free_numbers_of_iterations_estimates (cfun);
estimate_numbers_of_iterations ();
changed = tree_unroll_loops_completely_1 (may_increase_size,
unroll_outer, father_bbs,
current_loops->tree_root);
if (changed)
{
unsigned i;
unloop_loops (loop_closed_ssa_invalidated, &irred_invalidated);
/* We can not use TODO_update_ssa_no_phi because VOPS gets confused. */
if (loop_closed_ssa_invalidated
&& !bitmap_empty_p (loop_closed_ssa_invalidated))
rewrite_into_loop_closed_ssa (loop_closed_ssa_invalidated,
TODO_update_ssa);
else
update_ssa (TODO_update_ssa);
/* father_bbs is a bitmap of loop father header BB indices.
Translate that to what non-root loops these BBs belong to now. */
bitmap_iterator bi;
bitmap fathers = BITMAP_ALLOC (NULL);
EXECUTE_IF_SET_IN_BITMAP (father_bbs, 0, i, bi)
{
basic_block unrolled_loop_bb = BASIC_BLOCK_FOR_FN (cfun, i);
if (! unrolled_loop_bb)
continue;
if (loop_outer (unrolled_loop_bb->loop_father))
bitmap_set_bit (fathers,
unrolled_loop_bb->loop_father->num);
}
bitmap_clear (father_bbs);
/* Propagate the constants within the new basic blocks. */
EXECUTE_IF_SET_IN_BITMAP (fathers, 0, i, bi)
{
loop_p father = get_loop (cfun, i);
basic_block *body = get_loop_body_in_dom_order (father);
for (unsigned j = 0; j < father->num_nodes; j++)
propagate_constants_for_unrolling (body[j]);
free (body);
}
BITMAP_FREE (fathers);
/* This will take care of removing completely unrolled loops
from the loop structures so we can continue unrolling now
innermost loops. */
if (cleanup_tree_cfg ())
update_ssa (TODO_update_ssa_only_virtuals);
/* Clean up the information about numbers of iterations, since
complete unrolling might have invalidated it. */
scev_reset ();
if (flag_checking && loops_state_satisfies_p (LOOP_CLOSED_SSA))
verify_loop_closed_ssa (true);
}
if (loop_closed_ssa_invalidated)
BITMAP_FREE (loop_closed_ssa_invalidated);
}
while (changed
&& ++iteration <= PARAM_VALUE (PARAM_MAX_UNROLL_ITERATIONS));
BITMAP_FREE (father_bbs);
if (irred_invalidated
&& loops_state_satisfies_p (LOOPS_HAVE_MARKED_IRREDUCIBLE_REGIONS))
mark_irreducible_loops ();
return 0;
}
/* Canonical induction variable creation pass. */
namespace {
const pass_data pass_data_iv_canon =
{
GIMPLE_PASS, /* type */
"ivcanon", /* name */
OPTGROUP_LOOP, /* optinfo_flags */
TV_TREE_LOOP_IVCANON, /* tv_id */
( PROP_cfg | PROP_ssa ), /* properties_required */
0, /* properties_provided */
0, /* properties_destroyed */
0, /* todo_flags_start */
0, /* todo_flags_finish */
};
class pass_iv_canon : public gimple_opt_pass
{
public:
pass_iv_canon (gcc::context *ctxt)
: gimple_opt_pass (pass_data_iv_canon, ctxt)
{}
/* opt_pass methods: */
virtual bool gate (function *) { return flag_tree_loop_ivcanon != 0; }
virtual unsigned int execute (function *fun);
}; // class pass_iv_canon
unsigned int
pass_iv_canon::execute (function *fun)
{
if (number_of_loops (fun) <= 1)
return 0;
return canonicalize_induction_variables ();
}
} // anon namespace
gimple_opt_pass *
make_pass_iv_canon (gcc::context *ctxt)
{
return new pass_iv_canon (ctxt);
}
/* Complete unrolling of loops. */
namespace {
const pass_data pass_data_complete_unroll =
{
GIMPLE_PASS, /* type */
"cunroll", /* name */
OPTGROUP_LOOP, /* optinfo_flags */
TV_COMPLETE_UNROLL, /* tv_id */
( PROP_cfg | PROP_ssa ), /* properties_required */
0, /* properties_provided */
0, /* properties_destroyed */
0, /* todo_flags_start */
0, /* todo_flags_finish */
};
class pass_complete_unroll : public gimple_opt_pass
{
public:
pass_complete_unroll (gcc::context *ctxt)
: gimple_opt_pass (pass_data_complete_unroll, ctxt)
{}
/* opt_pass methods: */
virtual unsigned int execute (function *);
}; // class pass_complete_unroll
unsigned int
pass_complete_unroll::execute (function *fun)
{
if (number_of_loops (fun) <= 1)
return 0;
/* If we ever decide to run loop peeling more than once, we will need to
track loops already peeled in loop structures themselves to avoid
re-peeling the same loop multiple times. */
if (flag_peel_loops)
peeled_loops = BITMAP_ALLOC (NULL);
int val = tree_unroll_loops_completely (flag_unroll_loops
|| flag_peel_loops
|| optimize >= 3, true);
if (peeled_loops)
{
BITMAP_FREE (peeled_loops);
peeled_loops = NULL;
}
return val;
}
} // anon namespace
gimple_opt_pass *
make_pass_complete_unroll (gcc::context *ctxt)
{
return new pass_complete_unroll (ctxt);
}
/* Complete unrolling of inner loops. */
namespace {
const pass_data pass_data_complete_unrolli =
{
GIMPLE_PASS, /* type */
"cunrolli", /* name */
OPTGROUP_LOOP, /* optinfo_flags */
TV_COMPLETE_UNROLL, /* tv_id */
( PROP_cfg | PROP_ssa ), /* properties_required */
0, /* properties_provided */
0, /* properties_destroyed */
0, /* todo_flags_start */
0, /* todo_flags_finish */
};
class pass_complete_unrolli : public gimple_opt_pass
{
public:
pass_complete_unrolli (gcc::context *ctxt)
: gimple_opt_pass (pass_data_complete_unrolli, ctxt)
{}
/* opt_pass methods: */
virtual bool gate (function *) { return optimize >= 2; }
virtual unsigned int execute (function *);
}; // class pass_complete_unrolli
unsigned int
pass_complete_unrolli::execute (function *fun)
{
unsigned ret = 0;
loop_optimizer_init (LOOPS_NORMAL
| LOOPS_HAVE_RECORDED_EXITS);
if (number_of_loops (fun) > 1)
{
scev_initialize ();
ret = tree_unroll_loops_completely (optimize >= 3, false);
free_numbers_of_iterations_estimates (fun);
scev_finalize ();
}
loop_optimizer_finalize ();
return ret;
}
} // anon namespace
gimple_opt_pass *
make_pass_complete_unrolli (gcc::context *ctxt)
{
return new pass_complete_unrolli (ctxt);
}
|