lto-cgraph.c (output_profile_summary, [...]): Use gcov streaming; stream hot bb thres...
[gcc.git] / gcc / cfgcleanup.c
1 /* Control flow optimization code for GNU compiler.
2 Copyright (C) 1987-2013 Free Software Foundation, Inc.
3
4 This file is part of GCC.
5
6 GCC is free software; you can redistribute it and/or modify it under
7 the terms of the GNU General Public License as published by the Free
8 Software Foundation; either version 3, or (at your option) any later
9 version.
10
11 GCC is distributed in the hope that it will be useful, but WITHOUT ANY
12 WARRANTY; without even the implied warranty of MERCHANTABILITY or
13 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
14 for more details.
15
16 You should have received a copy of the GNU General Public License
17 along with GCC; see the file COPYING3. If not see
18 <http://www.gnu.org/licenses/>. */
19
20 /* This file contains optimizer of the control flow. The main entry point is
21 cleanup_cfg. Following optimizations are performed:
22
23 - Unreachable blocks removal
24 - Edge forwarding (edge to the forwarder block is forwarded to its
25 successor. Simplification of the branch instruction is performed by
26 underlying infrastructure so branch can be converted to simplejump or
27 eliminated).
28 - Cross jumping (tail merging)
29 - Conditional jump-around-simplejump simplification
30 - Basic block merging. */
31
32 #include "config.h"
33 #include "system.h"
34 #include "coretypes.h"
35 #include "tm.h"
36 #include "rtl.h"
37 #include "hard-reg-set.h"
38 #include "regs.h"
39 #include "insn-config.h"
40 #include "flags.h"
41 #include "recog.h"
42 #include "diagnostic-core.h"
43 #include "cselib.h"
44 #include "params.h"
45 #include "tm_p.h"
46 #include "target.h"
47 #include "function.h" /* For inline functions in emit-rtl.h they need crtl. */
48 #include "emit-rtl.h"
49 #include "tree-pass.h"
50 #include "cfgloop.h"
51 #include "expr.h"
52 #include "df.h"
53 #include "dce.h"
54 #include "dbgcnt.h"
55
56 #define FORWARDER_BLOCK_P(BB) ((BB)->flags & BB_FORWARDER_BLOCK)
57
58 /* Set to true when we are running first pass of try_optimize_cfg loop. */
59 static bool first_pass;
60
61 /* Set to true if crossjumps occurred in the latest run of try_optimize_cfg. */
62 static bool crossjumps_occured;
63
64 /* Set to true if we couldn't run an optimization due to stale liveness
65 information; we should run df_analyze to enable more opportunities. */
66 static bool block_was_dirty;
67
68 static bool try_crossjump_to_edge (int, edge, edge, enum replace_direction);
69 static bool try_crossjump_bb (int, basic_block);
70 static bool outgoing_edges_match (int, basic_block, basic_block);
71 static enum replace_direction old_insns_match_p (int, rtx, rtx);
72
73 static void merge_blocks_move_predecessor_nojumps (basic_block, basic_block);
74 static void merge_blocks_move_successor_nojumps (basic_block, basic_block);
75 static bool try_optimize_cfg (int);
76 static bool try_simplify_condjump (basic_block);
77 static bool try_forward_edges (int, basic_block);
78 static edge thread_jump (edge, basic_block);
79 static bool mark_effect (rtx, bitmap);
80 static void notice_new_block (basic_block);
81 static void update_forwarder_flag (basic_block);
82 static int mentions_nonequal_regs (rtx *, void *);
83 static void merge_memattrs (rtx, rtx);
84 \f
85 /* Set flags for newly created block. */
86
87 static void
88 notice_new_block (basic_block bb)
89 {
90 if (!bb)
91 return;
92
93 if (forwarder_block_p (bb))
94 bb->flags |= BB_FORWARDER_BLOCK;
95 }
96
97 /* Recompute forwarder flag after block has been modified. */
98
99 static void
100 update_forwarder_flag (basic_block bb)
101 {
102 if (forwarder_block_p (bb))
103 bb->flags |= BB_FORWARDER_BLOCK;
104 else
105 bb->flags &= ~BB_FORWARDER_BLOCK;
106 }
107 \f
108 /* Simplify a conditional jump around an unconditional jump.
109 Return true if something changed. */
110
111 static bool
112 try_simplify_condjump (basic_block cbranch_block)
113 {
114 basic_block jump_block, jump_dest_block, cbranch_dest_block;
115 edge cbranch_jump_edge, cbranch_fallthru_edge;
116 rtx cbranch_insn;
117
118 /* Verify that there are exactly two successors. */
119 if (EDGE_COUNT (cbranch_block->succs) != 2)
120 return false;
121
122 /* Verify that we've got a normal conditional branch at the end
123 of the block. */
124 cbranch_insn = BB_END (cbranch_block);
125 if (!any_condjump_p (cbranch_insn))
126 return false;
127
128 cbranch_fallthru_edge = FALLTHRU_EDGE (cbranch_block);
129 cbranch_jump_edge = BRANCH_EDGE (cbranch_block);
130
131 /* The next block must not have multiple predecessors, must not
132 be the last block in the function, and must contain just the
133 unconditional jump. */
134 jump_block = cbranch_fallthru_edge->dest;
135 if (!single_pred_p (jump_block)
136 || jump_block->next_bb == EXIT_BLOCK_PTR
137 || !FORWARDER_BLOCK_P (jump_block))
138 return false;
139 jump_dest_block = single_succ (jump_block);
140
141 /* If we are partitioning hot/cold basic blocks, we don't want to
142 mess up unconditional or indirect jumps that cross between hot
143 and cold sections.
144
145 Basic block partitioning may result in some jumps that appear to
146 be optimizable (or blocks that appear to be mergeable), but which really
147 must be left untouched (they are required to make it safely across
148 partition boundaries). See the comments at the top of
149 bb-reorder.c:partition_hot_cold_basic_blocks for complete details. */
150
151 if (BB_PARTITION (jump_block) != BB_PARTITION (jump_dest_block)
152 || (cbranch_jump_edge->flags & EDGE_CROSSING))
153 return false;
154
155 /* The conditional branch must target the block after the
156 unconditional branch. */
157 cbranch_dest_block = cbranch_jump_edge->dest;
158
159 if (cbranch_dest_block == EXIT_BLOCK_PTR
160 || !can_fallthru (jump_block, cbranch_dest_block))
161 return false;
162
163 /* Invert the conditional branch. */
164 if (!invert_jump (cbranch_insn, block_label (jump_dest_block), 0))
165 return false;
166
167 if (dump_file)
168 fprintf (dump_file, "Simplifying condjump %i around jump %i\n",
169 INSN_UID (cbranch_insn), INSN_UID (BB_END (jump_block)));
170
171 /* Success. Update the CFG to match. Note that after this point
172 the edge variable names appear backwards; the redirection is done
173 this way to preserve edge profile data. */
174 cbranch_jump_edge = redirect_edge_succ_nodup (cbranch_jump_edge,
175 cbranch_dest_block);
176 cbranch_fallthru_edge = redirect_edge_succ_nodup (cbranch_fallthru_edge,
177 jump_dest_block);
178 cbranch_jump_edge->flags |= EDGE_FALLTHRU;
179 cbranch_fallthru_edge->flags &= ~EDGE_FALLTHRU;
180 update_br_prob_note (cbranch_block);
181
182 /* Delete the block with the unconditional jump, and clean up the mess. */
183 delete_basic_block (jump_block);
184 tidy_fallthru_edge (cbranch_jump_edge);
185 update_forwarder_flag (cbranch_block);
186
187 return true;
188 }
189 \f
190 /* Attempt to prove that operation is NOOP using CSElib or mark the effect
191 on register. Used by jump threading. */
192
193 static bool
194 mark_effect (rtx exp, regset nonequal)
195 {
196 int regno;
197 rtx dest;
198 switch (GET_CODE (exp))
199 {
200 /* In case we do clobber the register, mark it as equal, as we know the
201 value is dead so it don't have to match. */
202 case CLOBBER:
203 if (REG_P (XEXP (exp, 0)))
204 {
205 dest = XEXP (exp, 0);
206 regno = REGNO (dest);
207 if (HARD_REGISTER_NUM_P (regno))
208 bitmap_clear_range (nonequal, regno,
209 hard_regno_nregs[regno][GET_MODE (dest)]);
210 else
211 bitmap_clear_bit (nonequal, regno);
212 }
213 return false;
214
215 case SET:
216 if (rtx_equal_for_cselib_p (SET_DEST (exp), SET_SRC (exp)))
217 return false;
218 dest = SET_DEST (exp);
219 if (dest == pc_rtx)
220 return false;
221 if (!REG_P (dest))
222 return true;
223 regno = REGNO (dest);
224 if (HARD_REGISTER_NUM_P (regno))
225 bitmap_set_range (nonequal, regno,
226 hard_regno_nregs[regno][GET_MODE (dest)]);
227 else
228 bitmap_set_bit (nonequal, regno);
229 return false;
230
231 default:
232 return false;
233 }
234 }
235
236 /* Return nonzero if X is a register set in regset DATA.
237 Called via for_each_rtx. */
238 static int
239 mentions_nonequal_regs (rtx *x, void *data)
240 {
241 regset nonequal = (regset) data;
242 if (REG_P (*x))
243 {
244 int regno;
245
246 regno = REGNO (*x);
247 if (REGNO_REG_SET_P (nonequal, regno))
248 return 1;
249 if (regno < FIRST_PSEUDO_REGISTER)
250 {
251 int n = hard_regno_nregs[regno][GET_MODE (*x)];
252 while (--n > 0)
253 if (REGNO_REG_SET_P (nonequal, regno + n))
254 return 1;
255 }
256 }
257 return 0;
258 }
259 /* Attempt to prove that the basic block B will have no side effects and
260 always continues in the same edge if reached via E. Return the edge
261 if exist, NULL otherwise. */
262
263 static edge
264 thread_jump (edge e, basic_block b)
265 {
266 rtx set1, set2, cond1, cond2, insn;
267 enum rtx_code code1, code2, reversed_code2;
268 bool reverse1 = false;
269 unsigned i;
270 regset nonequal;
271 bool failed = false;
272 reg_set_iterator rsi;
273
274 if (b->flags & BB_NONTHREADABLE_BLOCK)
275 return NULL;
276
277 /* At the moment, we do handle only conditional jumps, but later we may
278 want to extend this code to tablejumps and others. */
279 if (EDGE_COUNT (e->src->succs) != 2)
280 return NULL;
281 if (EDGE_COUNT (b->succs) != 2)
282 {
283 b->flags |= BB_NONTHREADABLE_BLOCK;
284 return NULL;
285 }
286
287 /* Second branch must end with onlyjump, as we will eliminate the jump. */
288 if (!any_condjump_p (BB_END (e->src)))
289 return NULL;
290
291 if (!any_condjump_p (BB_END (b)) || !onlyjump_p (BB_END (b)))
292 {
293 b->flags |= BB_NONTHREADABLE_BLOCK;
294 return NULL;
295 }
296
297 set1 = pc_set (BB_END (e->src));
298 set2 = pc_set (BB_END (b));
299 if (((e->flags & EDGE_FALLTHRU) != 0)
300 != (XEXP (SET_SRC (set1), 1) == pc_rtx))
301 reverse1 = true;
302
303 cond1 = XEXP (SET_SRC (set1), 0);
304 cond2 = XEXP (SET_SRC (set2), 0);
305 if (reverse1)
306 code1 = reversed_comparison_code (cond1, BB_END (e->src));
307 else
308 code1 = GET_CODE (cond1);
309
310 code2 = GET_CODE (cond2);
311 reversed_code2 = reversed_comparison_code (cond2, BB_END (b));
312
313 if (!comparison_dominates_p (code1, code2)
314 && !comparison_dominates_p (code1, reversed_code2))
315 return NULL;
316
317 /* Ensure that the comparison operators are equivalent.
318 ??? This is far too pessimistic. We should allow swapped operands,
319 different CCmodes, or for example comparisons for interval, that
320 dominate even when operands are not equivalent. */
321 if (!rtx_equal_p (XEXP (cond1, 0), XEXP (cond2, 0))
322 || !rtx_equal_p (XEXP (cond1, 1), XEXP (cond2, 1)))
323 return NULL;
324
325 /* Short circuit cases where block B contains some side effects, as we can't
326 safely bypass it. */
327 for (insn = NEXT_INSN (BB_HEAD (b)); insn != NEXT_INSN (BB_END (b));
328 insn = NEXT_INSN (insn))
329 if (INSN_P (insn) && side_effects_p (PATTERN (insn)))
330 {
331 b->flags |= BB_NONTHREADABLE_BLOCK;
332 return NULL;
333 }
334
335 cselib_init (0);
336
337 /* First process all values computed in the source basic block. */
338 for (insn = NEXT_INSN (BB_HEAD (e->src));
339 insn != NEXT_INSN (BB_END (e->src));
340 insn = NEXT_INSN (insn))
341 if (INSN_P (insn))
342 cselib_process_insn (insn);
343
344 nonequal = BITMAP_ALLOC (NULL);
345 CLEAR_REG_SET (nonequal);
346
347 /* Now assume that we've continued by the edge E to B and continue
348 processing as if it were same basic block.
349 Our goal is to prove that whole block is an NOOP. */
350
351 for (insn = NEXT_INSN (BB_HEAD (b));
352 insn != NEXT_INSN (BB_END (b)) && !failed;
353 insn = NEXT_INSN (insn))
354 {
355 if (INSN_P (insn))
356 {
357 rtx pat = PATTERN (insn);
358
359 if (GET_CODE (pat) == PARALLEL)
360 {
361 for (i = 0; i < (unsigned)XVECLEN (pat, 0); i++)
362 failed |= mark_effect (XVECEXP (pat, 0, i), nonequal);
363 }
364 else
365 failed |= mark_effect (pat, nonequal);
366 }
367
368 cselib_process_insn (insn);
369 }
370
371 /* Later we should clear nonequal of dead registers. So far we don't
372 have life information in cfg_cleanup. */
373 if (failed)
374 {
375 b->flags |= BB_NONTHREADABLE_BLOCK;
376 goto failed_exit;
377 }
378
379 /* cond2 must not mention any register that is not equal to the
380 former block. */
381 if (for_each_rtx (&cond2, mentions_nonequal_regs, nonequal))
382 goto failed_exit;
383
384 EXECUTE_IF_SET_IN_REG_SET (nonequal, 0, i, rsi)
385 goto failed_exit;
386
387 BITMAP_FREE (nonequal);
388 cselib_finish ();
389 if ((comparison_dominates_p (code1, code2) != 0)
390 != (XEXP (SET_SRC (set2), 1) == pc_rtx))
391 return BRANCH_EDGE (b);
392 else
393 return FALLTHRU_EDGE (b);
394
395 failed_exit:
396 BITMAP_FREE (nonequal);
397 cselib_finish ();
398 return NULL;
399 }
400 \f
401 /* Attempt to forward edges leaving basic block B.
402 Return true if successful. */
403
404 static bool
405 try_forward_edges (int mode, basic_block b)
406 {
407 bool changed = false;
408 edge_iterator ei;
409 edge e, *threaded_edges = NULL;
410
411 /* If we are partitioning hot/cold basic blocks, we don't want to
412 mess up unconditional or indirect jumps that cross between hot
413 and cold sections.
414
415 Basic block partitioning may result in some jumps that appear to
416 be optimizable (or blocks that appear to be mergeable), but which really
417 must be left untouched (they are required to make it safely across
418 partition boundaries). See the comments at the top of
419 bb-reorder.c:partition_hot_cold_basic_blocks for complete details. */
420
421 if (find_reg_note (BB_END (b), REG_CROSSING_JUMP, NULL_RTX))
422 return false;
423
424 for (ei = ei_start (b->succs); (e = ei_safe_edge (ei)); )
425 {
426 basic_block target, first;
427 int counter, goto_locus;
428 bool threaded = false;
429 int nthreaded_edges = 0;
430 bool may_thread = first_pass || (b->flags & BB_MODIFIED) != 0;
431
432 /* Skip complex edges because we don't know how to update them.
433
434 Still handle fallthru edges, as we can succeed to forward fallthru
435 edge to the same place as the branch edge of conditional branch
436 and turn conditional branch to an unconditional branch. */
437 if (e->flags & EDGE_COMPLEX)
438 {
439 ei_next (&ei);
440 continue;
441 }
442
443 target = first = e->dest;
444 counter = NUM_FIXED_BLOCKS;
445 goto_locus = e->goto_locus;
446
447 /* If we are partitioning hot/cold basic_blocks, we don't want to mess
448 up jumps that cross between hot/cold sections.
449
450 Basic block partitioning may result in some jumps that appear
451 to be optimizable (or blocks that appear to be mergeable), but which
452 really must be left untouched (they are required to make it safely
453 across partition boundaries). See the comments at the top of
454 bb-reorder.c:partition_hot_cold_basic_blocks for complete
455 details. */
456
457 if (first != EXIT_BLOCK_PTR
458 && find_reg_note (BB_END (first), REG_CROSSING_JUMP, NULL_RTX))
459 return false;
460
461 while (counter < n_basic_blocks)
462 {
463 basic_block new_target = NULL;
464 bool new_target_threaded = false;
465 may_thread |= (target->flags & BB_MODIFIED) != 0;
466
467 if (FORWARDER_BLOCK_P (target)
468 && !(single_succ_edge (target)->flags & EDGE_CROSSING)
469 && single_succ (target) != EXIT_BLOCK_PTR)
470 {
471 /* Bypass trivial infinite loops. */
472 new_target = single_succ (target);
473 if (target == new_target)
474 counter = n_basic_blocks;
475 else if (!optimize)
476 {
477 /* When not optimizing, ensure that edges or forwarder
478 blocks with different locus are not optimized out. */
479 int new_locus = single_succ_edge (target)->goto_locus;
480 int locus = goto_locus;
481
482 if (new_locus != UNKNOWN_LOCATION
483 && locus != UNKNOWN_LOCATION
484 && new_locus != locus)
485 new_target = NULL;
486 else
487 {
488 rtx last;
489
490 if (new_locus != UNKNOWN_LOCATION)
491 locus = new_locus;
492
493 last = BB_END (target);
494 if (DEBUG_INSN_P (last))
495 last = prev_nondebug_insn (last);
496
497 new_locus = last && INSN_P (last)
498 ? INSN_LOCATION (last) : 0;
499
500 if (new_locus != UNKNOWN_LOCATION
501 && locus != UNKNOWN_LOCATION
502 && new_locus != locus)
503 new_target = NULL;
504 else
505 {
506 if (new_locus != UNKNOWN_LOCATION)
507 locus = new_locus;
508
509 goto_locus = locus;
510 }
511 }
512 }
513 }
514
515 /* Allow to thread only over one edge at time to simplify updating
516 of probabilities. */
517 else if ((mode & CLEANUP_THREADING) && may_thread)
518 {
519 edge t = thread_jump (e, target);
520 if (t)
521 {
522 if (!threaded_edges)
523 threaded_edges = XNEWVEC (edge, n_basic_blocks);
524 else
525 {
526 int i;
527
528 /* Detect an infinite loop across blocks not
529 including the start block. */
530 for (i = 0; i < nthreaded_edges; ++i)
531 if (threaded_edges[i] == t)
532 break;
533 if (i < nthreaded_edges)
534 {
535 counter = n_basic_blocks;
536 break;
537 }
538 }
539
540 /* Detect an infinite loop across the start block. */
541 if (t->dest == b)
542 break;
543
544 gcc_assert (nthreaded_edges < n_basic_blocks - NUM_FIXED_BLOCKS);
545 threaded_edges[nthreaded_edges++] = t;
546
547 new_target = t->dest;
548 new_target_threaded = true;
549 }
550 }
551
552 if (!new_target)
553 break;
554
555 counter++;
556 target = new_target;
557 threaded |= new_target_threaded;
558 }
559
560 if (counter >= n_basic_blocks)
561 {
562 if (dump_file)
563 fprintf (dump_file, "Infinite loop in BB %i.\n",
564 target->index);
565 }
566 else if (target == first)
567 ; /* We didn't do anything. */
568 else
569 {
570 /* Save the values now, as the edge may get removed. */
571 gcov_type edge_count = e->count;
572 int edge_probability = e->probability;
573 int edge_frequency;
574 int n = 0;
575
576 e->goto_locus = goto_locus;
577
578 /* Don't force if target is exit block. */
579 if (threaded && target != EXIT_BLOCK_PTR)
580 {
581 notice_new_block (redirect_edge_and_branch_force (e, target));
582 if (dump_file)
583 fprintf (dump_file, "Conditionals threaded.\n");
584 }
585 else if (!redirect_edge_and_branch (e, target))
586 {
587 if (dump_file)
588 fprintf (dump_file,
589 "Forwarding edge %i->%i to %i failed.\n",
590 b->index, e->dest->index, target->index);
591 ei_next (&ei);
592 continue;
593 }
594
595 /* We successfully forwarded the edge. Now update profile
596 data: for each edge we traversed in the chain, remove
597 the original edge's execution count. */
598 edge_frequency = ((edge_probability * b->frequency
599 + REG_BR_PROB_BASE / 2)
600 / REG_BR_PROB_BASE);
601
602 do
603 {
604 edge t;
605
606 if (!single_succ_p (first))
607 {
608 gcc_assert (n < nthreaded_edges);
609 t = threaded_edges [n++];
610 gcc_assert (t->src == first);
611 update_bb_profile_for_threading (first, edge_frequency,
612 edge_count, t);
613 update_br_prob_note (first);
614 }
615 else
616 {
617 first->count -= edge_count;
618 if (first->count < 0)
619 first->count = 0;
620 first->frequency -= edge_frequency;
621 if (first->frequency < 0)
622 first->frequency = 0;
623 /* It is possible that as the result of
624 threading we've removed edge as it is
625 threaded to the fallthru edge. Avoid
626 getting out of sync. */
627 if (n < nthreaded_edges
628 && first == threaded_edges [n]->src)
629 n++;
630 t = single_succ_edge (first);
631 }
632
633 t->count -= edge_count;
634 if (t->count < 0)
635 t->count = 0;
636 first = t->dest;
637 }
638 while (first != target);
639
640 changed = true;
641 continue;
642 }
643 ei_next (&ei);
644 }
645
646 free (threaded_edges);
647 return changed;
648 }
649 \f
650
651 /* Blocks A and B are to be merged into a single block. A has no incoming
652 fallthru edge, so it can be moved before B without adding or modifying
653 any jumps (aside from the jump from A to B). */
654
655 static void
656 merge_blocks_move_predecessor_nojumps (basic_block a, basic_block b)
657 {
658 rtx barrier;
659
660 /* If we are partitioning hot/cold basic blocks, we don't want to
661 mess up unconditional or indirect jumps that cross between hot
662 and cold sections.
663
664 Basic block partitioning may result in some jumps that appear to
665 be optimizable (or blocks that appear to be mergeable), but which really
666 must be left untouched (they are required to make it safely across
667 partition boundaries). See the comments at the top of
668 bb-reorder.c:partition_hot_cold_basic_blocks for complete details. */
669
670 if (BB_PARTITION (a) != BB_PARTITION (b))
671 return;
672
673 barrier = next_nonnote_insn (BB_END (a));
674 gcc_assert (BARRIER_P (barrier));
675 delete_insn (barrier);
676
677 /* Scramble the insn chain. */
678 if (BB_END (a) != PREV_INSN (BB_HEAD (b)))
679 reorder_insns_nobb (BB_HEAD (a), BB_END (a), PREV_INSN (BB_HEAD (b)));
680 df_set_bb_dirty (a);
681
682 if (dump_file)
683 fprintf (dump_file, "Moved block %d before %d and merged.\n",
684 a->index, b->index);
685
686 /* Swap the records for the two blocks around. */
687
688 unlink_block (a);
689 link_block (a, b->prev_bb);
690
691 /* Now blocks A and B are contiguous. Merge them. */
692 merge_blocks (a, b);
693 }
694
695 /* Blocks A and B are to be merged into a single block. B has no outgoing
696 fallthru edge, so it can be moved after A without adding or modifying
697 any jumps (aside from the jump from A to B). */
698
699 static void
700 merge_blocks_move_successor_nojumps (basic_block a, basic_block b)
701 {
702 rtx barrier, real_b_end;
703 rtx label, table;
704
705 /* If we are partitioning hot/cold basic blocks, we don't want to
706 mess up unconditional or indirect jumps that cross between hot
707 and cold sections.
708
709 Basic block partitioning may result in some jumps that appear to
710 be optimizable (or blocks that appear to be mergeable), but which really
711 must be left untouched (they are required to make it safely across
712 partition boundaries). See the comments at the top of
713 bb-reorder.c:partition_hot_cold_basic_blocks for complete details. */
714
715 if (BB_PARTITION (a) != BB_PARTITION (b))
716 return;
717
718 real_b_end = BB_END (b);
719
720 /* If there is a jump table following block B temporarily add the jump table
721 to block B so that it will also be moved to the correct location. */
722 if (tablejump_p (BB_END (b), &label, &table)
723 && prev_active_insn (label) == BB_END (b))
724 {
725 BB_END (b) = table;
726 }
727
728 /* There had better have been a barrier there. Delete it. */
729 barrier = NEXT_INSN (BB_END (b));
730 if (barrier && BARRIER_P (barrier))
731 delete_insn (barrier);
732
733
734 /* Scramble the insn chain. */
735 reorder_insns_nobb (BB_HEAD (b), BB_END (b), BB_END (a));
736
737 /* Restore the real end of b. */
738 BB_END (b) = real_b_end;
739
740 if (dump_file)
741 fprintf (dump_file, "Moved block %d after %d and merged.\n",
742 b->index, a->index);
743
744 /* Now blocks A and B are contiguous. Merge them. */
745 merge_blocks (a, b);
746 }
747
748 /* Attempt to merge basic blocks that are potentially non-adjacent.
749 Return NULL iff the attempt failed, otherwise return basic block
750 where cleanup_cfg should continue. Because the merging commonly
751 moves basic block away or introduces another optimization
752 possibility, return basic block just before B so cleanup_cfg don't
753 need to iterate.
754
755 It may be good idea to return basic block before C in the case
756 C has been moved after B and originally appeared earlier in the
757 insn sequence, but we have no information available about the
758 relative ordering of these two. Hopefully it is not too common. */
759
760 static basic_block
761 merge_blocks_move (edge e, basic_block b, basic_block c, int mode)
762 {
763 basic_block next;
764
765 /* If we are partitioning hot/cold basic blocks, we don't want to
766 mess up unconditional or indirect jumps that cross between hot
767 and cold sections.
768
769 Basic block partitioning may result in some jumps that appear to
770 be optimizable (or blocks that appear to be mergeable), but which really
771 must be left untouched (they are required to make it safely across
772 partition boundaries). See the comments at the top of
773 bb-reorder.c:partition_hot_cold_basic_blocks for complete details. */
774
775 if (BB_PARTITION (b) != BB_PARTITION (c))
776 return NULL;
777
778 /* If B has a fallthru edge to C, no need to move anything. */
779 if (e->flags & EDGE_FALLTHRU)
780 {
781 int b_index = b->index, c_index = c->index;
782
783 /* Protect the loop latches. */
784 if (current_loops && c->loop_father->latch == c)
785 return NULL;
786
787 merge_blocks (b, c);
788 update_forwarder_flag (b);
789
790 if (dump_file)
791 fprintf (dump_file, "Merged %d and %d without moving.\n",
792 b_index, c_index);
793
794 return b->prev_bb == ENTRY_BLOCK_PTR ? b : b->prev_bb;
795 }
796
797 /* Otherwise we will need to move code around. Do that only if expensive
798 transformations are allowed. */
799 else if (mode & CLEANUP_EXPENSIVE)
800 {
801 edge tmp_edge, b_fallthru_edge;
802 bool c_has_outgoing_fallthru;
803 bool b_has_incoming_fallthru;
804
805 /* Avoid overactive code motion, as the forwarder blocks should be
806 eliminated by edge redirection instead. One exception might have
807 been if B is a forwarder block and C has no fallthru edge, but
808 that should be cleaned up by bb-reorder instead. */
809 if (FORWARDER_BLOCK_P (b) || FORWARDER_BLOCK_P (c))
810 return NULL;
811
812 /* We must make sure to not munge nesting of lexical blocks,
813 and loop notes. This is done by squeezing out all the notes
814 and leaving them there to lie. Not ideal, but functional. */
815
816 tmp_edge = find_fallthru_edge (c->succs);
817 c_has_outgoing_fallthru = (tmp_edge != NULL);
818
819 tmp_edge = find_fallthru_edge (b->preds);
820 b_has_incoming_fallthru = (tmp_edge != NULL);
821 b_fallthru_edge = tmp_edge;
822 next = b->prev_bb;
823 if (next == c)
824 next = next->prev_bb;
825
826 /* Otherwise, we're going to try to move C after B. If C does
827 not have an outgoing fallthru, then it can be moved
828 immediately after B without introducing or modifying jumps. */
829 if (! c_has_outgoing_fallthru)
830 {
831 merge_blocks_move_successor_nojumps (b, c);
832 return next == ENTRY_BLOCK_PTR ? next->next_bb : next;
833 }
834
835 /* If B does not have an incoming fallthru, then it can be moved
836 immediately before C without introducing or modifying jumps.
837 C cannot be the first block, so we do not have to worry about
838 accessing a non-existent block. */
839
840 if (b_has_incoming_fallthru)
841 {
842 basic_block bb;
843
844 if (b_fallthru_edge->src == ENTRY_BLOCK_PTR)
845 return NULL;
846 bb = force_nonfallthru (b_fallthru_edge);
847 if (bb)
848 notice_new_block (bb);
849 }
850
851 merge_blocks_move_predecessor_nojumps (b, c);
852 return next == ENTRY_BLOCK_PTR ? next->next_bb : next;
853 }
854
855 return NULL;
856 }
857 \f
858
859 /* Removes the memory attributes of MEM expression
860 if they are not equal. */
861
862 void
863 merge_memattrs (rtx x, rtx y)
864 {
865 int i;
866 int j;
867 enum rtx_code code;
868 const char *fmt;
869
870 if (x == y)
871 return;
872 if (x == 0 || y == 0)
873 return;
874
875 code = GET_CODE (x);
876
877 if (code != GET_CODE (y))
878 return;
879
880 if (GET_MODE (x) != GET_MODE (y))
881 return;
882
883 if (code == MEM && MEM_ATTRS (x) != MEM_ATTRS (y))
884 {
885 if (! MEM_ATTRS (x))
886 MEM_ATTRS (y) = 0;
887 else if (! MEM_ATTRS (y))
888 MEM_ATTRS (x) = 0;
889 else
890 {
891 HOST_WIDE_INT mem_size;
892
893 if (MEM_ALIAS_SET (x) != MEM_ALIAS_SET (y))
894 {
895 set_mem_alias_set (x, 0);
896 set_mem_alias_set (y, 0);
897 }
898
899 if (! mem_expr_equal_p (MEM_EXPR (x), MEM_EXPR (y)))
900 {
901 set_mem_expr (x, 0);
902 set_mem_expr (y, 0);
903 clear_mem_offset (x);
904 clear_mem_offset (y);
905 }
906 else if (MEM_OFFSET_KNOWN_P (x) != MEM_OFFSET_KNOWN_P (y)
907 || (MEM_OFFSET_KNOWN_P (x)
908 && MEM_OFFSET (x) != MEM_OFFSET (y)))
909 {
910 clear_mem_offset (x);
911 clear_mem_offset (y);
912 }
913
914 if (MEM_SIZE_KNOWN_P (x) && MEM_SIZE_KNOWN_P (y))
915 {
916 mem_size = MAX (MEM_SIZE (x), MEM_SIZE (y));
917 set_mem_size (x, mem_size);
918 set_mem_size (y, mem_size);
919 }
920 else
921 {
922 clear_mem_size (x);
923 clear_mem_size (y);
924 }
925
926 set_mem_align (x, MIN (MEM_ALIGN (x), MEM_ALIGN (y)));
927 set_mem_align (y, MEM_ALIGN (x));
928 }
929 }
930
931 fmt = GET_RTX_FORMAT (code);
932 for (i = GET_RTX_LENGTH (code) - 1; i >= 0; i--)
933 {
934 switch (fmt[i])
935 {
936 case 'E':
937 /* Two vectors must have the same length. */
938 if (XVECLEN (x, i) != XVECLEN (y, i))
939 return;
940
941 for (j = 0; j < XVECLEN (x, i); j++)
942 merge_memattrs (XVECEXP (x, i, j), XVECEXP (y, i, j));
943
944 break;
945
946 case 'e':
947 merge_memattrs (XEXP (x, i), XEXP (y, i));
948 }
949 }
950 return;
951 }
952
953
954 /* Checks if patterns P1 and P2 are equivalent, apart from the possibly
955 different single sets S1 and S2. */
956
957 static bool
958 equal_different_set_p (rtx p1, rtx s1, rtx p2, rtx s2)
959 {
960 int i;
961 rtx e1, e2;
962
963 if (p1 == s1 && p2 == s2)
964 return true;
965
966 if (GET_CODE (p1) != PARALLEL || GET_CODE (p2) != PARALLEL)
967 return false;
968
969 if (XVECLEN (p1, 0) != XVECLEN (p2, 0))
970 return false;
971
972 for (i = 0; i < XVECLEN (p1, 0); i++)
973 {
974 e1 = XVECEXP (p1, 0, i);
975 e2 = XVECEXP (p2, 0, i);
976 if (e1 == s1 && e2 == s2)
977 continue;
978 if (reload_completed
979 ? rtx_renumbered_equal_p (e1, e2) : rtx_equal_p (e1, e2))
980 continue;
981
982 return false;
983 }
984
985 return true;
986 }
987
988 /* Examine register notes on I1 and I2 and return:
989 - dir_forward if I1 can be replaced by I2, or
990 - dir_backward if I2 can be replaced by I1, or
991 - dir_both if both are the case. */
992
993 static enum replace_direction
994 can_replace_by (rtx i1, rtx i2)
995 {
996 rtx s1, s2, d1, d2, src1, src2, note1, note2;
997 bool c1, c2;
998
999 /* Check for 2 sets. */
1000 s1 = single_set (i1);
1001 s2 = single_set (i2);
1002 if (s1 == NULL_RTX || s2 == NULL_RTX)
1003 return dir_none;
1004
1005 /* Check that the 2 sets set the same dest. */
1006 d1 = SET_DEST (s1);
1007 d2 = SET_DEST (s2);
1008 if (!(reload_completed
1009 ? rtx_renumbered_equal_p (d1, d2) : rtx_equal_p (d1, d2)))
1010 return dir_none;
1011
1012 /* Find identical req_equiv or reg_equal note, which implies that the 2 sets
1013 set dest to the same value. */
1014 note1 = find_reg_equal_equiv_note (i1);
1015 note2 = find_reg_equal_equiv_note (i2);
1016 if (!note1 || !note2 || !rtx_equal_p (XEXP (note1, 0), XEXP (note2, 0))
1017 || !CONST_INT_P (XEXP (note1, 0)))
1018 return dir_none;
1019
1020 if (!equal_different_set_p (PATTERN (i1), s1, PATTERN (i2), s2))
1021 return dir_none;
1022
1023 /* Although the 2 sets set dest to the same value, we cannot replace
1024 (set (dest) (const_int))
1025 by
1026 (set (dest) (reg))
1027 because we don't know if the reg is live and has the same value at the
1028 location of replacement. */
1029 src1 = SET_SRC (s1);
1030 src2 = SET_SRC (s2);
1031 c1 = CONST_INT_P (src1);
1032 c2 = CONST_INT_P (src2);
1033 if (c1 && c2)
1034 return dir_both;
1035 else if (c2)
1036 return dir_forward;
1037 else if (c1)
1038 return dir_backward;
1039
1040 return dir_none;
1041 }
1042
1043 /* Merges directions A and B. */
1044
1045 static enum replace_direction
1046 merge_dir (enum replace_direction a, enum replace_direction b)
1047 {
1048 /* Implements the following table:
1049 |bo fw bw no
1050 ---+-----------
1051 bo |bo fw bw no
1052 fw |-- fw no no
1053 bw |-- -- bw no
1054 no |-- -- -- no. */
1055
1056 if (a == b)
1057 return a;
1058
1059 if (a == dir_both)
1060 return b;
1061 if (b == dir_both)
1062 return a;
1063
1064 return dir_none;
1065 }
1066
1067 /* Examine I1 and I2 and return:
1068 - dir_forward if I1 can be replaced by I2, or
1069 - dir_backward if I2 can be replaced by I1, or
1070 - dir_both if both are the case. */
1071
1072 static enum replace_direction
1073 old_insns_match_p (int mode ATTRIBUTE_UNUSED, rtx i1, rtx i2)
1074 {
1075 rtx p1, p2;
1076
1077 /* Verify that I1 and I2 are equivalent. */
1078 if (GET_CODE (i1) != GET_CODE (i2))
1079 return dir_none;
1080
1081 /* __builtin_unreachable() may lead to empty blocks (ending with
1082 NOTE_INSN_BASIC_BLOCK). They may be crossjumped. */
1083 if (NOTE_INSN_BASIC_BLOCK_P (i1) && NOTE_INSN_BASIC_BLOCK_P (i2))
1084 return dir_both;
1085
1086 /* ??? Do not allow cross-jumping between different stack levels. */
1087 p1 = find_reg_note (i1, REG_ARGS_SIZE, NULL);
1088 p2 = find_reg_note (i2, REG_ARGS_SIZE, NULL);
1089 if (p1 && p2)
1090 {
1091 p1 = XEXP (p1, 0);
1092 p2 = XEXP (p2, 0);
1093 if (!rtx_equal_p (p1, p2))
1094 return dir_none;
1095
1096 /* ??? Worse, this adjustment had better be constant lest we
1097 have differing incoming stack levels. */
1098 if (!frame_pointer_needed
1099 && find_args_size_adjust (i1) == HOST_WIDE_INT_MIN)
1100 return dir_none;
1101 }
1102 else if (p1 || p2)
1103 return dir_none;
1104
1105 p1 = PATTERN (i1);
1106 p2 = PATTERN (i2);
1107
1108 if (GET_CODE (p1) != GET_CODE (p2))
1109 return dir_none;
1110
1111 /* If this is a CALL_INSN, compare register usage information.
1112 If we don't check this on stack register machines, the two
1113 CALL_INSNs might be merged leaving reg-stack.c with mismatching
1114 numbers of stack registers in the same basic block.
1115 If we don't check this on machines with delay slots, a delay slot may
1116 be filled that clobbers a parameter expected by the subroutine.
1117
1118 ??? We take the simple route for now and assume that if they're
1119 equal, they were constructed identically.
1120
1121 Also check for identical exception regions. */
1122
1123 if (CALL_P (i1))
1124 {
1125 /* Ensure the same EH region. */
1126 rtx n1 = find_reg_note (i1, REG_EH_REGION, 0);
1127 rtx n2 = find_reg_note (i2, REG_EH_REGION, 0);
1128
1129 if (!n1 && n2)
1130 return dir_none;
1131
1132 if (n1 && (!n2 || XEXP (n1, 0) != XEXP (n2, 0)))
1133 return dir_none;
1134
1135 if (!rtx_equal_p (CALL_INSN_FUNCTION_USAGE (i1),
1136 CALL_INSN_FUNCTION_USAGE (i2))
1137 || SIBLING_CALL_P (i1) != SIBLING_CALL_P (i2))
1138 return dir_none;
1139
1140 /* For address sanitizer, never crossjump __asan_report_* builtins,
1141 otherwise errors might be reported on incorrect lines. */
1142 if (flag_asan)
1143 {
1144 rtx call = get_call_rtx_from (i1);
1145 if (call && GET_CODE (XEXP (XEXP (call, 0), 0)) == SYMBOL_REF)
1146 {
1147 rtx symbol = XEXP (XEXP (call, 0), 0);
1148 if (SYMBOL_REF_DECL (symbol)
1149 && TREE_CODE (SYMBOL_REF_DECL (symbol)) == FUNCTION_DECL)
1150 {
1151 if ((DECL_BUILT_IN_CLASS (SYMBOL_REF_DECL (symbol))
1152 == BUILT_IN_NORMAL)
1153 && DECL_FUNCTION_CODE (SYMBOL_REF_DECL (symbol))
1154 >= BUILT_IN_ASAN_REPORT_LOAD1
1155 && DECL_FUNCTION_CODE (SYMBOL_REF_DECL (symbol))
1156 <= BUILT_IN_ASAN_REPORT_STORE16)
1157 return dir_none;
1158 }
1159 }
1160 }
1161 }
1162
1163 #ifdef STACK_REGS
1164 /* If cross_jump_death_matters is not 0, the insn's mode
1165 indicates whether or not the insn contains any stack-like
1166 regs. */
1167
1168 if ((mode & CLEANUP_POST_REGSTACK) && stack_regs_mentioned (i1))
1169 {
1170 /* If register stack conversion has already been done, then
1171 death notes must also be compared before it is certain that
1172 the two instruction streams match. */
1173
1174 rtx note;
1175 HARD_REG_SET i1_regset, i2_regset;
1176
1177 CLEAR_HARD_REG_SET (i1_regset);
1178 CLEAR_HARD_REG_SET (i2_regset);
1179
1180 for (note = REG_NOTES (i1); note; note = XEXP (note, 1))
1181 if (REG_NOTE_KIND (note) == REG_DEAD && STACK_REG_P (XEXP (note, 0)))
1182 SET_HARD_REG_BIT (i1_regset, REGNO (XEXP (note, 0)));
1183
1184 for (note = REG_NOTES (i2); note; note = XEXP (note, 1))
1185 if (REG_NOTE_KIND (note) == REG_DEAD && STACK_REG_P (XEXP (note, 0)))
1186 SET_HARD_REG_BIT (i2_regset, REGNO (XEXP (note, 0)));
1187
1188 if (!hard_reg_set_equal_p (i1_regset, i2_regset))
1189 return dir_none;
1190 }
1191 #endif
1192
1193 if (reload_completed
1194 ? rtx_renumbered_equal_p (p1, p2) : rtx_equal_p (p1, p2))
1195 return dir_both;
1196
1197 return can_replace_by (i1, i2);
1198 }
1199 \f
1200 /* When comparing insns I1 and I2 in flow_find_cross_jump or
1201 flow_find_head_matching_sequence, ensure the notes match. */
1202
1203 static void
1204 merge_notes (rtx i1, rtx i2)
1205 {
1206 /* If the merged insns have different REG_EQUAL notes, then
1207 remove them. */
1208 rtx equiv1 = find_reg_equal_equiv_note (i1);
1209 rtx equiv2 = find_reg_equal_equiv_note (i2);
1210
1211 if (equiv1 && !equiv2)
1212 remove_note (i1, equiv1);
1213 else if (!equiv1 && equiv2)
1214 remove_note (i2, equiv2);
1215 else if (equiv1 && equiv2
1216 && !rtx_equal_p (XEXP (equiv1, 0), XEXP (equiv2, 0)))
1217 {
1218 remove_note (i1, equiv1);
1219 remove_note (i2, equiv2);
1220 }
1221 }
1222
1223 /* Walks from I1 in BB1 backward till the next non-debug insn, and returns the
1224 resulting insn in I1, and the corresponding bb in BB1. At the head of a
1225 bb, if there is a predecessor bb that reaches this bb via fallthru, and
1226 FOLLOW_FALLTHRU, walks further in the predecessor bb and registers this in
1227 DID_FALLTHRU. Otherwise, stops at the head of the bb. */
1228
1229 static void
1230 walk_to_nondebug_insn (rtx *i1, basic_block *bb1, bool follow_fallthru,
1231 bool *did_fallthru)
1232 {
1233 edge fallthru;
1234
1235 *did_fallthru = false;
1236
1237 /* Ignore notes. */
1238 while (!NONDEBUG_INSN_P (*i1))
1239 {
1240 if (*i1 != BB_HEAD (*bb1))
1241 {
1242 *i1 = PREV_INSN (*i1);
1243 continue;
1244 }
1245
1246 if (!follow_fallthru)
1247 return;
1248
1249 fallthru = find_fallthru_edge ((*bb1)->preds);
1250 if (!fallthru || fallthru->src == ENTRY_BLOCK_PTR_FOR_FUNCTION (cfun)
1251 || !single_succ_p (fallthru->src))
1252 return;
1253
1254 *bb1 = fallthru->src;
1255 *i1 = BB_END (*bb1);
1256 *did_fallthru = true;
1257 }
1258 }
1259
1260 /* Look through the insns at the end of BB1 and BB2 and find the longest
1261 sequence that are either equivalent, or allow forward or backward
1262 replacement. Store the first insns for that sequence in *F1 and *F2 and
1263 return the sequence length.
1264
1265 DIR_P indicates the allowed replacement direction on function entry, and
1266 the actual replacement direction on function exit. If NULL, only equivalent
1267 sequences are allowed.
1268
1269 To simplify callers of this function, if the blocks match exactly,
1270 store the head of the blocks in *F1 and *F2. */
1271
1272 int
1273 flow_find_cross_jump (basic_block bb1, basic_block bb2, rtx *f1, rtx *f2,
1274 enum replace_direction *dir_p)
1275 {
1276 rtx i1, i2, last1, last2, afterlast1, afterlast2;
1277 int ninsns = 0;
1278 rtx p1;
1279 enum replace_direction dir, last_dir, afterlast_dir;
1280 bool follow_fallthru, did_fallthru;
1281
1282 if (dir_p)
1283 dir = *dir_p;
1284 else
1285 dir = dir_both;
1286 afterlast_dir = dir;
1287 last_dir = afterlast_dir;
1288
1289 /* Skip simple jumps at the end of the blocks. Complex jumps still
1290 need to be compared for equivalence, which we'll do below. */
1291
1292 i1 = BB_END (bb1);
1293 last1 = afterlast1 = last2 = afterlast2 = NULL_RTX;
1294 if (onlyjump_p (i1)
1295 || (returnjump_p (i1) && !side_effects_p (PATTERN (i1))))
1296 {
1297 last1 = i1;
1298 i1 = PREV_INSN (i1);
1299 }
1300
1301 i2 = BB_END (bb2);
1302 if (onlyjump_p (i2)
1303 || (returnjump_p (i2) && !side_effects_p (PATTERN (i2))))
1304 {
1305 last2 = i2;
1306 /* Count everything except for unconditional jump as insn. */
1307 if (!simplejump_p (i2) && !returnjump_p (i2) && last1)
1308 ninsns++;
1309 i2 = PREV_INSN (i2);
1310 }
1311
1312 while (true)
1313 {
1314 /* In the following example, we can replace all jumps to C by jumps to A.
1315
1316 This removes 4 duplicate insns.
1317 [bb A] insn1 [bb C] insn1
1318 insn2 insn2
1319 [bb B] insn3 insn3
1320 insn4 insn4
1321 jump_insn jump_insn
1322
1323 We could also replace all jumps to A by jumps to C, but that leaves B
1324 alive, and removes only 2 duplicate insns. In a subsequent crossjump
1325 step, all jumps to B would be replaced with jumps to the middle of C,
1326 achieving the same result with more effort.
1327 So we allow only the first possibility, which means that we don't allow
1328 fallthru in the block that's being replaced. */
1329
1330 follow_fallthru = dir_p && dir != dir_forward;
1331 walk_to_nondebug_insn (&i1, &bb1, follow_fallthru, &did_fallthru);
1332 if (did_fallthru)
1333 dir = dir_backward;
1334
1335 follow_fallthru = dir_p && dir != dir_backward;
1336 walk_to_nondebug_insn (&i2, &bb2, follow_fallthru, &did_fallthru);
1337 if (did_fallthru)
1338 dir = dir_forward;
1339
1340 if (i1 == BB_HEAD (bb1) || i2 == BB_HEAD (bb2))
1341 break;
1342
1343 dir = merge_dir (dir, old_insns_match_p (0, i1, i2));
1344 if (dir == dir_none || (!dir_p && dir != dir_both))
1345 break;
1346
1347 merge_memattrs (i1, i2);
1348
1349 /* Don't begin a cross-jump with a NOTE insn. */
1350 if (INSN_P (i1))
1351 {
1352 merge_notes (i1, i2);
1353
1354 afterlast1 = last1, afterlast2 = last2;
1355 last1 = i1, last2 = i2;
1356 afterlast_dir = last_dir;
1357 last_dir = dir;
1358 p1 = PATTERN (i1);
1359 if (!(GET_CODE (p1) == USE || GET_CODE (p1) == CLOBBER))
1360 ninsns++;
1361 }
1362
1363 i1 = PREV_INSN (i1);
1364 i2 = PREV_INSN (i2);
1365 }
1366
1367 #ifdef HAVE_cc0
1368 /* Don't allow the insn after a compare to be shared by
1369 cross-jumping unless the compare is also shared. */
1370 if (ninsns && reg_mentioned_p (cc0_rtx, last1) && ! sets_cc0_p (last1))
1371 last1 = afterlast1, last2 = afterlast2, last_dir = afterlast_dir, ninsns--;
1372 #endif
1373
1374 /* Include preceding notes and labels in the cross-jump. One,
1375 this may bring us to the head of the blocks as requested above.
1376 Two, it keeps line number notes as matched as may be. */
1377 if (ninsns)
1378 {
1379 bb1 = BLOCK_FOR_INSN (last1);
1380 while (last1 != BB_HEAD (bb1) && !NONDEBUG_INSN_P (PREV_INSN (last1)))
1381 last1 = PREV_INSN (last1);
1382
1383 if (last1 != BB_HEAD (bb1) && LABEL_P (PREV_INSN (last1)))
1384 last1 = PREV_INSN (last1);
1385
1386 bb2 = BLOCK_FOR_INSN (last2);
1387 while (last2 != BB_HEAD (bb2) && !NONDEBUG_INSN_P (PREV_INSN (last2)))
1388 last2 = PREV_INSN (last2);
1389
1390 if (last2 != BB_HEAD (bb2) && LABEL_P (PREV_INSN (last2)))
1391 last2 = PREV_INSN (last2);
1392
1393 *f1 = last1;
1394 *f2 = last2;
1395 }
1396
1397 if (dir_p)
1398 *dir_p = last_dir;
1399 return ninsns;
1400 }
1401
1402 /* Like flow_find_cross_jump, except start looking for a matching sequence from
1403 the head of the two blocks. Do not include jumps at the end.
1404 If STOP_AFTER is nonzero, stop after finding that many matching
1405 instructions. */
1406
1407 int
1408 flow_find_head_matching_sequence (basic_block bb1, basic_block bb2, rtx *f1,
1409 rtx *f2, int stop_after)
1410 {
1411 rtx i1, i2, last1, last2, beforelast1, beforelast2;
1412 int ninsns = 0;
1413 edge e;
1414 edge_iterator ei;
1415 int nehedges1 = 0, nehedges2 = 0;
1416
1417 FOR_EACH_EDGE (e, ei, bb1->succs)
1418 if (e->flags & EDGE_EH)
1419 nehedges1++;
1420 FOR_EACH_EDGE (e, ei, bb2->succs)
1421 if (e->flags & EDGE_EH)
1422 nehedges2++;
1423
1424 i1 = BB_HEAD (bb1);
1425 i2 = BB_HEAD (bb2);
1426 last1 = beforelast1 = last2 = beforelast2 = NULL_RTX;
1427
1428 while (true)
1429 {
1430 /* Ignore notes, except NOTE_INSN_EPILOGUE_BEG. */
1431 while (!NONDEBUG_INSN_P (i1) && i1 != BB_END (bb1))
1432 {
1433 if (NOTE_P (i1) && NOTE_KIND (i1) == NOTE_INSN_EPILOGUE_BEG)
1434 break;
1435 i1 = NEXT_INSN (i1);
1436 }
1437
1438 while (!NONDEBUG_INSN_P (i2) && i2 != BB_END (bb2))
1439 {
1440 if (NOTE_P (i2) && NOTE_KIND (i2) == NOTE_INSN_EPILOGUE_BEG)
1441 break;
1442 i2 = NEXT_INSN (i2);
1443 }
1444
1445 if ((i1 == BB_END (bb1) && !NONDEBUG_INSN_P (i1))
1446 || (i2 == BB_END (bb2) && !NONDEBUG_INSN_P (i2)))
1447 break;
1448
1449 if (NOTE_P (i1) || NOTE_P (i2)
1450 || JUMP_P (i1) || JUMP_P (i2))
1451 break;
1452
1453 /* A sanity check to make sure we're not merging insns with different
1454 effects on EH. If only one of them ends a basic block, it shouldn't
1455 have an EH edge; if both end a basic block, there should be the same
1456 number of EH edges. */
1457 if ((i1 == BB_END (bb1) && i2 != BB_END (bb2)
1458 && nehedges1 > 0)
1459 || (i2 == BB_END (bb2) && i1 != BB_END (bb1)
1460 && nehedges2 > 0)
1461 || (i1 == BB_END (bb1) && i2 == BB_END (bb2)
1462 && nehedges1 != nehedges2))
1463 break;
1464
1465 if (old_insns_match_p (0, i1, i2) != dir_both)
1466 break;
1467
1468 merge_memattrs (i1, i2);
1469
1470 /* Don't begin a cross-jump with a NOTE insn. */
1471 if (INSN_P (i1))
1472 {
1473 merge_notes (i1, i2);
1474
1475 beforelast1 = last1, beforelast2 = last2;
1476 last1 = i1, last2 = i2;
1477 ninsns++;
1478 }
1479
1480 if (i1 == BB_END (bb1) || i2 == BB_END (bb2)
1481 || (stop_after > 0 && ninsns == stop_after))
1482 break;
1483
1484 i1 = NEXT_INSN (i1);
1485 i2 = NEXT_INSN (i2);
1486 }
1487
1488 #ifdef HAVE_cc0
1489 /* Don't allow a compare to be shared by cross-jumping unless the insn
1490 after the compare is also shared. */
1491 if (ninsns && reg_mentioned_p (cc0_rtx, last1) && sets_cc0_p (last1))
1492 last1 = beforelast1, last2 = beforelast2, ninsns--;
1493 #endif
1494
1495 if (ninsns)
1496 {
1497 *f1 = last1;
1498 *f2 = last2;
1499 }
1500
1501 return ninsns;
1502 }
1503
1504 /* Return true iff outgoing edges of BB1 and BB2 match, together with
1505 the branch instruction. This means that if we commonize the control
1506 flow before end of the basic block, the semantic remains unchanged.
1507
1508 We may assume that there exists one edge with a common destination. */
1509
1510 static bool
1511 outgoing_edges_match (int mode, basic_block bb1, basic_block bb2)
1512 {
1513 int nehedges1 = 0, nehedges2 = 0;
1514 edge fallthru1 = 0, fallthru2 = 0;
1515 edge e1, e2;
1516 edge_iterator ei;
1517
1518 /* If we performed shrink-wrapping, edges to the EXIT_BLOCK_PTR can
1519 only be distinguished for JUMP_INSNs. The two paths may differ in
1520 whether they went through the prologue. Sibcalls are fine, we know
1521 that we either didn't need or inserted an epilogue before them. */
1522 if (crtl->shrink_wrapped
1523 && single_succ_p (bb1) && single_succ (bb1) == EXIT_BLOCK_PTR
1524 && !JUMP_P (BB_END (bb1))
1525 && !(CALL_P (BB_END (bb1)) && SIBLING_CALL_P (BB_END (bb1))))
1526 return false;
1527
1528 /* If BB1 has only one successor, we may be looking at either an
1529 unconditional jump, or a fake edge to exit. */
1530 if (single_succ_p (bb1)
1531 && (single_succ_edge (bb1)->flags & (EDGE_COMPLEX | EDGE_FAKE)) == 0
1532 && (!JUMP_P (BB_END (bb1)) || simplejump_p (BB_END (bb1))))
1533 return (single_succ_p (bb2)
1534 && (single_succ_edge (bb2)->flags
1535 & (EDGE_COMPLEX | EDGE_FAKE)) == 0
1536 && (!JUMP_P (BB_END (bb2)) || simplejump_p (BB_END (bb2))));
1537
1538 /* Match conditional jumps - this may get tricky when fallthru and branch
1539 edges are crossed. */
1540 if (EDGE_COUNT (bb1->succs) == 2
1541 && any_condjump_p (BB_END (bb1))
1542 && onlyjump_p (BB_END (bb1)))
1543 {
1544 edge b1, f1, b2, f2;
1545 bool reverse, match;
1546 rtx set1, set2, cond1, cond2;
1547 enum rtx_code code1, code2;
1548
1549 if (EDGE_COUNT (bb2->succs) != 2
1550 || !any_condjump_p (BB_END (bb2))
1551 || !onlyjump_p (BB_END (bb2)))
1552 return false;
1553
1554 b1 = BRANCH_EDGE (bb1);
1555 b2 = BRANCH_EDGE (bb2);
1556 f1 = FALLTHRU_EDGE (bb1);
1557 f2 = FALLTHRU_EDGE (bb2);
1558
1559 /* Get around possible forwarders on fallthru edges. Other cases
1560 should be optimized out already. */
1561 if (FORWARDER_BLOCK_P (f1->dest))
1562 f1 = single_succ_edge (f1->dest);
1563
1564 if (FORWARDER_BLOCK_P (f2->dest))
1565 f2 = single_succ_edge (f2->dest);
1566
1567 /* To simplify use of this function, return false if there are
1568 unneeded forwarder blocks. These will get eliminated later
1569 during cleanup_cfg. */
1570 if (FORWARDER_BLOCK_P (f1->dest)
1571 || FORWARDER_BLOCK_P (f2->dest)
1572 || FORWARDER_BLOCK_P (b1->dest)
1573 || FORWARDER_BLOCK_P (b2->dest))
1574 return false;
1575
1576 if (f1->dest == f2->dest && b1->dest == b2->dest)
1577 reverse = false;
1578 else if (f1->dest == b2->dest && b1->dest == f2->dest)
1579 reverse = true;
1580 else
1581 return false;
1582
1583 set1 = pc_set (BB_END (bb1));
1584 set2 = pc_set (BB_END (bb2));
1585 if ((XEXP (SET_SRC (set1), 1) == pc_rtx)
1586 != (XEXP (SET_SRC (set2), 1) == pc_rtx))
1587 reverse = !reverse;
1588
1589 cond1 = XEXP (SET_SRC (set1), 0);
1590 cond2 = XEXP (SET_SRC (set2), 0);
1591 code1 = GET_CODE (cond1);
1592 if (reverse)
1593 code2 = reversed_comparison_code (cond2, BB_END (bb2));
1594 else
1595 code2 = GET_CODE (cond2);
1596
1597 if (code2 == UNKNOWN)
1598 return false;
1599
1600 /* Verify codes and operands match. */
1601 match = ((code1 == code2
1602 && rtx_renumbered_equal_p (XEXP (cond1, 0), XEXP (cond2, 0))
1603 && rtx_renumbered_equal_p (XEXP (cond1, 1), XEXP (cond2, 1)))
1604 || (code1 == swap_condition (code2)
1605 && rtx_renumbered_equal_p (XEXP (cond1, 1),
1606 XEXP (cond2, 0))
1607 && rtx_renumbered_equal_p (XEXP (cond1, 0),
1608 XEXP (cond2, 1))));
1609
1610 /* If we return true, we will join the blocks. Which means that
1611 we will only have one branch prediction bit to work with. Thus
1612 we require the existing branches to have probabilities that are
1613 roughly similar. */
1614 if (match
1615 && optimize_bb_for_speed_p (bb1)
1616 && optimize_bb_for_speed_p (bb2))
1617 {
1618 int prob2;
1619
1620 if (b1->dest == b2->dest)
1621 prob2 = b2->probability;
1622 else
1623 /* Do not use f2 probability as f2 may be forwarded. */
1624 prob2 = REG_BR_PROB_BASE - b2->probability;
1625
1626 /* Fail if the difference in probabilities is greater than 50%.
1627 This rules out two well-predicted branches with opposite
1628 outcomes. */
1629 if (abs (b1->probability - prob2) > REG_BR_PROB_BASE / 2)
1630 {
1631 if (dump_file)
1632 fprintf (dump_file,
1633 "Outcomes of branch in bb %i and %i differ too much (%i %i)\n",
1634 bb1->index, bb2->index, b1->probability, prob2);
1635
1636 return false;
1637 }
1638 }
1639
1640 if (dump_file && match)
1641 fprintf (dump_file, "Conditionals in bb %i and %i match.\n",
1642 bb1->index, bb2->index);
1643
1644 return match;
1645 }
1646
1647 /* Generic case - we are seeing a computed jump, table jump or trapping
1648 instruction. */
1649
1650 /* Check whether there are tablejumps in the end of BB1 and BB2.
1651 Return true if they are identical. */
1652 {
1653 rtx label1, label2;
1654 rtx table1, table2;
1655
1656 if (tablejump_p (BB_END (bb1), &label1, &table1)
1657 && tablejump_p (BB_END (bb2), &label2, &table2)
1658 && GET_CODE (PATTERN (table1)) == GET_CODE (PATTERN (table2)))
1659 {
1660 /* The labels should never be the same rtx. If they really are same
1661 the jump tables are same too. So disable crossjumping of blocks BB1
1662 and BB2 because when deleting the common insns in the end of BB1
1663 by delete_basic_block () the jump table would be deleted too. */
1664 /* If LABEL2 is referenced in BB1->END do not do anything
1665 because we would loose information when replacing
1666 LABEL1 by LABEL2 and then LABEL2 by LABEL1 in BB1->END. */
1667 if (label1 != label2 && !rtx_referenced_p (label2, BB_END (bb1)))
1668 {
1669 /* Set IDENTICAL to true when the tables are identical. */
1670 bool identical = false;
1671 rtx p1, p2;
1672
1673 p1 = PATTERN (table1);
1674 p2 = PATTERN (table2);
1675 if (GET_CODE (p1) == ADDR_VEC && rtx_equal_p (p1, p2))
1676 {
1677 identical = true;
1678 }
1679 else if (GET_CODE (p1) == ADDR_DIFF_VEC
1680 && (XVECLEN (p1, 1) == XVECLEN (p2, 1))
1681 && rtx_equal_p (XEXP (p1, 2), XEXP (p2, 2))
1682 && rtx_equal_p (XEXP (p1, 3), XEXP (p2, 3)))
1683 {
1684 int i;
1685
1686 identical = true;
1687 for (i = XVECLEN (p1, 1) - 1; i >= 0 && identical; i--)
1688 if (!rtx_equal_p (XVECEXP (p1, 1, i), XVECEXP (p2, 1, i)))
1689 identical = false;
1690 }
1691
1692 if (identical)
1693 {
1694 replace_label_data rr;
1695 bool match;
1696
1697 /* Temporarily replace references to LABEL1 with LABEL2
1698 in BB1->END so that we could compare the instructions. */
1699 rr.r1 = label1;
1700 rr.r2 = label2;
1701 rr.update_label_nuses = false;
1702 for_each_rtx (&BB_END (bb1), replace_label, &rr);
1703
1704 match = (old_insns_match_p (mode, BB_END (bb1), BB_END (bb2))
1705 == dir_both);
1706 if (dump_file && match)
1707 fprintf (dump_file,
1708 "Tablejumps in bb %i and %i match.\n",
1709 bb1->index, bb2->index);
1710
1711 /* Set the original label in BB1->END because when deleting
1712 a block whose end is a tablejump, the tablejump referenced
1713 from the instruction is deleted too. */
1714 rr.r1 = label2;
1715 rr.r2 = label1;
1716 for_each_rtx (&BB_END (bb1), replace_label, &rr);
1717
1718 return match;
1719 }
1720 }
1721 return false;
1722 }
1723 }
1724
1725 rtx last1 = BB_END (bb1);
1726 rtx last2 = BB_END (bb2);
1727 if (DEBUG_INSN_P (last1))
1728 last1 = prev_nondebug_insn (last1);
1729 if (DEBUG_INSN_P (last2))
1730 last2 = prev_nondebug_insn (last2);
1731 /* First ensure that the instructions match. There may be many outgoing
1732 edges so this test is generally cheaper. */
1733 if (old_insns_match_p (mode, last1, last2) != dir_both)
1734 return false;
1735
1736 /* Search the outgoing edges, ensure that the counts do match, find possible
1737 fallthru and exception handling edges since these needs more
1738 validation. */
1739 if (EDGE_COUNT (bb1->succs) != EDGE_COUNT (bb2->succs))
1740 return false;
1741
1742 bool nonfakeedges = false;
1743 FOR_EACH_EDGE (e1, ei, bb1->succs)
1744 {
1745 e2 = EDGE_SUCC (bb2, ei.index);
1746
1747 if ((e1->flags & EDGE_FAKE) == 0)
1748 nonfakeedges = true;
1749
1750 if (e1->flags & EDGE_EH)
1751 nehedges1++;
1752
1753 if (e2->flags & EDGE_EH)
1754 nehedges2++;
1755
1756 if (e1->flags & EDGE_FALLTHRU)
1757 fallthru1 = e1;
1758 if (e2->flags & EDGE_FALLTHRU)
1759 fallthru2 = e2;
1760 }
1761
1762 /* If number of edges of various types does not match, fail. */
1763 if (nehedges1 != nehedges2
1764 || (fallthru1 != 0) != (fallthru2 != 0))
1765 return false;
1766
1767 /* If !ACCUMULATE_OUTGOING_ARGS, bb1 (and bb2) have no successors
1768 and the last real insn doesn't have REG_ARGS_SIZE note, don't
1769 attempt to optimize, as the two basic blocks might have different
1770 REG_ARGS_SIZE depths. For noreturn calls and unconditional
1771 traps there should be REG_ARG_SIZE notes, they could be missing
1772 for __builtin_unreachable () uses though. */
1773 if (!nonfakeedges
1774 && !ACCUMULATE_OUTGOING_ARGS
1775 && (!INSN_P (last1)
1776 || !find_reg_note (last1, REG_ARGS_SIZE, NULL)))
1777 return false;
1778
1779 /* fallthru edges must be forwarded to the same destination. */
1780 if (fallthru1)
1781 {
1782 basic_block d1 = (forwarder_block_p (fallthru1->dest)
1783 ? single_succ (fallthru1->dest): fallthru1->dest);
1784 basic_block d2 = (forwarder_block_p (fallthru2->dest)
1785 ? single_succ (fallthru2->dest): fallthru2->dest);
1786
1787 if (d1 != d2)
1788 return false;
1789 }
1790
1791 /* Ensure the same EH region. */
1792 {
1793 rtx n1 = find_reg_note (BB_END (bb1), REG_EH_REGION, 0);
1794 rtx n2 = find_reg_note (BB_END (bb2), REG_EH_REGION, 0);
1795
1796 if (!n1 && n2)
1797 return false;
1798
1799 if (n1 && (!n2 || XEXP (n1, 0) != XEXP (n2, 0)))
1800 return false;
1801 }
1802
1803 /* The same checks as in try_crossjump_to_edge. It is required for RTL
1804 version of sequence abstraction. */
1805 FOR_EACH_EDGE (e1, ei, bb2->succs)
1806 {
1807 edge e2;
1808 edge_iterator ei;
1809 basic_block d1 = e1->dest;
1810
1811 if (FORWARDER_BLOCK_P (d1))
1812 d1 = EDGE_SUCC (d1, 0)->dest;
1813
1814 FOR_EACH_EDGE (e2, ei, bb1->succs)
1815 {
1816 basic_block d2 = e2->dest;
1817 if (FORWARDER_BLOCK_P (d2))
1818 d2 = EDGE_SUCC (d2, 0)->dest;
1819 if (d1 == d2)
1820 break;
1821 }
1822
1823 if (!e2)
1824 return false;
1825 }
1826
1827 return true;
1828 }
1829
1830 /* Returns true if BB basic block has a preserve label. */
1831
1832 static bool
1833 block_has_preserve_label (basic_block bb)
1834 {
1835 return (bb
1836 && block_label (bb)
1837 && LABEL_PRESERVE_P (block_label (bb)));
1838 }
1839
1840 /* E1 and E2 are edges with the same destination block. Search their
1841 predecessors for common code. If found, redirect control flow from
1842 (maybe the middle of) E1->SRC to (maybe the middle of) E2->SRC (dir_forward),
1843 or the other way around (dir_backward). DIR specifies the allowed
1844 replacement direction. */
1845
1846 static bool
1847 try_crossjump_to_edge (int mode, edge e1, edge e2,
1848 enum replace_direction dir)
1849 {
1850 int nmatch;
1851 basic_block src1 = e1->src, src2 = e2->src;
1852 basic_block redirect_to, redirect_from, to_remove;
1853 basic_block osrc1, osrc2, redirect_edges_to, tmp;
1854 rtx newpos1, newpos2;
1855 edge s;
1856 edge_iterator ei;
1857
1858 newpos1 = newpos2 = NULL_RTX;
1859
1860 /* If we have partitioned hot/cold basic blocks, it is a bad idea
1861 to try this optimization.
1862
1863 Basic block partitioning may result in some jumps that appear to
1864 be optimizable (or blocks that appear to be mergeable), but which really
1865 must be left untouched (they are required to make it safely across
1866 partition boundaries). See the comments at the top of
1867 bb-reorder.c:partition_hot_cold_basic_blocks for complete details. */
1868
1869 if (flag_reorder_blocks_and_partition && reload_completed)
1870 return false;
1871
1872 /* Search backward through forwarder blocks. We don't need to worry
1873 about multiple entry or chained forwarders, as they will be optimized
1874 away. We do this to look past the unconditional jump following a
1875 conditional jump that is required due to the current CFG shape. */
1876 if (single_pred_p (src1)
1877 && FORWARDER_BLOCK_P (src1))
1878 e1 = single_pred_edge (src1), src1 = e1->src;
1879
1880 if (single_pred_p (src2)
1881 && FORWARDER_BLOCK_P (src2))
1882 e2 = single_pred_edge (src2), src2 = e2->src;
1883
1884 /* Nothing to do if we reach ENTRY, or a common source block. */
1885 if (src1 == ENTRY_BLOCK_PTR || src2 == ENTRY_BLOCK_PTR)
1886 return false;
1887 if (src1 == src2)
1888 return false;
1889
1890 /* Seeing more than 1 forwarder blocks would confuse us later... */
1891 if (FORWARDER_BLOCK_P (e1->dest)
1892 && FORWARDER_BLOCK_P (single_succ (e1->dest)))
1893 return false;
1894
1895 if (FORWARDER_BLOCK_P (e2->dest)
1896 && FORWARDER_BLOCK_P (single_succ (e2->dest)))
1897 return false;
1898
1899 /* Likewise with dead code (possibly newly created by the other optimizations
1900 of cfg_cleanup). */
1901 if (EDGE_COUNT (src1->preds) == 0 || EDGE_COUNT (src2->preds) == 0)
1902 return false;
1903
1904 /* Look for the common insn sequence, part the first ... */
1905 if (!outgoing_edges_match (mode, src1, src2))
1906 return false;
1907
1908 /* ... and part the second. */
1909 nmatch = flow_find_cross_jump (src1, src2, &newpos1, &newpos2, &dir);
1910
1911 osrc1 = src1;
1912 osrc2 = src2;
1913 if (newpos1 != NULL_RTX)
1914 src1 = BLOCK_FOR_INSN (newpos1);
1915 if (newpos2 != NULL_RTX)
1916 src2 = BLOCK_FOR_INSN (newpos2);
1917
1918 if (dir == dir_backward)
1919 {
1920 #define SWAP(T, X, Y) do { T tmp = (X); (X) = (Y); (Y) = tmp; } while (0)
1921 SWAP (basic_block, osrc1, osrc2);
1922 SWAP (basic_block, src1, src2);
1923 SWAP (edge, e1, e2);
1924 SWAP (rtx, newpos1, newpos2);
1925 #undef SWAP
1926 }
1927
1928 /* Don't proceed with the crossjump unless we found a sufficient number
1929 of matching instructions or the 'from' block was totally matched
1930 (such that its predecessors will hopefully be redirected and the
1931 block removed). */
1932 if ((nmatch < PARAM_VALUE (PARAM_MIN_CROSSJUMP_INSNS))
1933 && (newpos1 != BB_HEAD (src1)))
1934 return false;
1935
1936 /* Avoid deleting preserve label when redirecting ABNORMAL edges. */
1937 if (block_has_preserve_label (e1->dest)
1938 && (e1->flags & EDGE_ABNORMAL))
1939 return false;
1940
1941 /* Here we know that the insns in the end of SRC1 which are common with SRC2
1942 will be deleted.
1943 If we have tablejumps in the end of SRC1 and SRC2
1944 they have been already compared for equivalence in outgoing_edges_match ()
1945 so replace the references to TABLE1 by references to TABLE2. */
1946 {
1947 rtx label1, label2;
1948 rtx table1, table2;
1949
1950 if (tablejump_p (BB_END (osrc1), &label1, &table1)
1951 && tablejump_p (BB_END (osrc2), &label2, &table2)
1952 && label1 != label2)
1953 {
1954 replace_label_data rr;
1955 rtx insn;
1956
1957 /* Replace references to LABEL1 with LABEL2. */
1958 rr.r1 = label1;
1959 rr.r2 = label2;
1960 rr.update_label_nuses = true;
1961 for (insn = get_insns (); insn; insn = NEXT_INSN (insn))
1962 {
1963 /* Do not replace the label in SRC1->END because when deleting
1964 a block whose end is a tablejump, the tablejump referenced
1965 from the instruction is deleted too. */
1966 if (insn != BB_END (osrc1))
1967 for_each_rtx (&insn, replace_label, &rr);
1968 }
1969 }
1970 }
1971
1972 /* Avoid splitting if possible. We must always split when SRC2 has
1973 EH predecessor edges, or we may end up with basic blocks with both
1974 normal and EH predecessor edges. */
1975 if (newpos2 == BB_HEAD (src2)
1976 && !(EDGE_PRED (src2, 0)->flags & EDGE_EH))
1977 redirect_to = src2;
1978 else
1979 {
1980 if (newpos2 == BB_HEAD (src2))
1981 {
1982 /* Skip possible basic block header. */
1983 if (LABEL_P (newpos2))
1984 newpos2 = NEXT_INSN (newpos2);
1985 while (DEBUG_INSN_P (newpos2))
1986 newpos2 = NEXT_INSN (newpos2);
1987 if (NOTE_P (newpos2))
1988 newpos2 = NEXT_INSN (newpos2);
1989 while (DEBUG_INSN_P (newpos2))
1990 newpos2 = NEXT_INSN (newpos2);
1991 }
1992
1993 if (dump_file)
1994 fprintf (dump_file, "Splitting bb %i before %i insns\n",
1995 src2->index, nmatch);
1996 redirect_to = split_block (src2, PREV_INSN (newpos2))->dest;
1997 }
1998
1999 if (dump_file)
2000 fprintf (dump_file,
2001 "Cross jumping from bb %i to bb %i; %i common insns\n",
2002 src1->index, src2->index, nmatch);
2003
2004 /* We may have some registers visible through the block. */
2005 df_set_bb_dirty (redirect_to);
2006
2007 if (osrc2 == src2)
2008 redirect_edges_to = redirect_to;
2009 else
2010 redirect_edges_to = osrc2;
2011
2012 /* Recompute the frequencies and counts of outgoing edges. */
2013 FOR_EACH_EDGE (s, ei, redirect_edges_to->succs)
2014 {
2015 edge s2;
2016 edge_iterator ei;
2017 basic_block d = s->dest;
2018
2019 if (FORWARDER_BLOCK_P (d))
2020 d = single_succ (d);
2021
2022 FOR_EACH_EDGE (s2, ei, src1->succs)
2023 {
2024 basic_block d2 = s2->dest;
2025 if (FORWARDER_BLOCK_P (d2))
2026 d2 = single_succ (d2);
2027 if (d == d2)
2028 break;
2029 }
2030
2031 s->count += s2->count;
2032
2033 /* Take care to update possible forwarder blocks. We verified
2034 that there is no more than one in the chain, so we can't run
2035 into infinite loop. */
2036 if (FORWARDER_BLOCK_P (s->dest))
2037 {
2038 single_succ_edge (s->dest)->count += s2->count;
2039 s->dest->count += s2->count;
2040 s->dest->frequency += EDGE_FREQUENCY (s);
2041 }
2042
2043 if (FORWARDER_BLOCK_P (s2->dest))
2044 {
2045 single_succ_edge (s2->dest)->count -= s2->count;
2046 if (single_succ_edge (s2->dest)->count < 0)
2047 single_succ_edge (s2->dest)->count = 0;
2048 s2->dest->count -= s2->count;
2049 s2->dest->frequency -= EDGE_FREQUENCY (s);
2050 if (s2->dest->frequency < 0)
2051 s2->dest->frequency = 0;
2052 if (s2->dest->count < 0)
2053 s2->dest->count = 0;
2054 }
2055
2056 if (!redirect_edges_to->frequency && !src1->frequency)
2057 s->probability = (s->probability + s2->probability) / 2;
2058 else
2059 s->probability
2060 = ((s->probability * redirect_edges_to->frequency +
2061 s2->probability * src1->frequency)
2062 / (redirect_edges_to->frequency + src1->frequency));
2063 }
2064
2065 /* Adjust count and frequency for the block. An earlier jump
2066 threading pass may have left the profile in an inconsistent
2067 state (see update_bb_profile_for_threading) so we must be
2068 prepared for overflows. */
2069 tmp = redirect_to;
2070 do
2071 {
2072 tmp->count += src1->count;
2073 tmp->frequency += src1->frequency;
2074 if (tmp->frequency > BB_FREQ_MAX)
2075 tmp->frequency = BB_FREQ_MAX;
2076 if (tmp == redirect_edges_to)
2077 break;
2078 tmp = find_fallthru_edge (tmp->succs)->dest;
2079 }
2080 while (true);
2081 update_br_prob_note (redirect_edges_to);
2082
2083 /* Edit SRC1 to go to REDIRECT_TO at NEWPOS1. */
2084
2085 /* Skip possible basic block header. */
2086 if (LABEL_P (newpos1))
2087 newpos1 = NEXT_INSN (newpos1);
2088
2089 while (DEBUG_INSN_P (newpos1))
2090 newpos1 = NEXT_INSN (newpos1);
2091
2092 if (NOTE_INSN_BASIC_BLOCK_P (newpos1))
2093 newpos1 = NEXT_INSN (newpos1);
2094
2095 while (DEBUG_INSN_P (newpos1))
2096 newpos1 = NEXT_INSN (newpos1);
2097
2098 redirect_from = split_block (src1, PREV_INSN (newpos1))->src;
2099 to_remove = single_succ (redirect_from);
2100
2101 redirect_edge_and_branch_force (single_succ_edge (redirect_from), redirect_to);
2102 delete_basic_block (to_remove);
2103
2104 update_forwarder_flag (redirect_from);
2105 if (redirect_to != src2)
2106 update_forwarder_flag (src2);
2107
2108 return true;
2109 }
2110
2111 /* Search the predecessors of BB for common insn sequences. When found,
2112 share code between them by redirecting control flow. Return true if
2113 any changes made. */
2114
2115 static bool
2116 try_crossjump_bb (int mode, basic_block bb)
2117 {
2118 edge e, e2, fallthru;
2119 bool changed;
2120 unsigned max, ix, ix2;
2121
2122 /* Nothing to do if there is not at least two incoming edges. */
2123 if (EDGE_COUNT (bb->preds) < 2)
2124 return false;
2125
2126 /* Don't crossjump if this block ends in a computed jump,
2127 unless we are optimizing for size. */
2128 if (optimize_bb_for_size_p (bb)
2129 && bb != EXIT_BLOCK_PTR
2130 && computed_jump_p (BB_END (bb)))
2131 return false;
2132
2133 /* If we are partitioning hot/cold basic blocks, we don't want to
2134 mess up unconditional or indirect jumps that cross between hot
2135 and cold sections.
2136
2137 Basic block partitioning may result in some jumps that appear to
2138 be optimizable (or blocks that appear to be mergeable), but which really
2139 must be left untouched (they are required to make it safely across
2140 partition boundaries). See the comments at the top of
2141 bb-reorder.c:partition_hot_cold_basic_blocks for complete details. */
2142
2143 if (BB_PARTITION (EDGE_PRED (bb, 0)->src) !=
2144 BB_PARTITION (EDGE_PRED (bb, 1)->src)
2145 || (EDGE_PRED (bb, 0)->flags & EDGE_CROSSING))
2146 return false;
2147
2148 /* It is always cheapest to redirect a block that ends in a branch to
2149 a block that falls through into BB, as that adds no branches to the
2150 program. We'll try that combination first. */
2151 fallthru = NULL;
2152 max = PARAM_VALUE (PARAM_MAX_CROSSJUMP_EDGES);
2153
2154 if (EDGE_COUNT (bb->preds) > max)
2155 return false;
2156
2157 fallthru = find_fallthru_edge (bb->preds);
2158
2159 changed = false;
2160 for (ix = 0; ix < EDGE_COUNT (bb->preds);)
2161 {
2162 e = EDGE_PRED (bb, ix);
2163 ix++;
2164
2165 /* As noted above, first try with the fallthru predecessor (or, a
2166 fallthru predecessor if we are in cfglayout mode). */
2167 if (fallthru)
2168 {
2169 /* Don't combine the fallthru edge into anything else.
2170 If there is a match, we'll do it the other way around. */
2171 if (e == fallthru)
2172 continue;
2173 /* If nothing changed since the last attempt, there is nothing
2174 we can do. */
2175 if (!first_pass
2176 && !((e->src->flags & BB_MODIFIED)
2177 || (fallthru->src->flags & BB_MODIFIED)))
2178 continue;
2179
2180 if (try_crossjump_to_edge (mode, e, fallthru, dir_forward))
2181 {
2182 changed = true;
2183 ix = 0;
2184 continue;
2185 }
2186 }
2187
2188 /* Non-obvious work limiting check: Recognize that we're going
2189 to call try_crossjump_bb on every basic block. So if we have
2190 two blocks with lots of outgoing edges (a switch) and they
2191 share lots of common destinations, then we would do the
2192 cross-jump check once for each common destination.
2193
2194 Now, if the blocks actually are cross-jump candidates, then
2195 all of their destinations will be shared. Which means that
2196 we only need check them for cross-jump candidacy once. We
2197 can eliminate redundant checks of crossjump(A,B) by arbitrarily
2198 choosing to do the check from the block for which the edge
2199 in question is the first successor of A. */
2200 if (EDGE_SUCC (e->src, 0) != e)
2201 continue;
2202
2203 for (ix2 = 0; ix2 < EDGE_COUNT (bb->preds); ix2++)
2204 {
2205 e2 = EDGE_PRED (bb, ix2);
2206
2207 if (e2 == e)
2208 continue;
2209
2210 /* We've already checked the fallthru edge above. */
2211 if (e2 == fallthru)
2212 continue;
2213
2214 /* The "first successor" check above only prevents multiple
2215 checks of crossjump(A,B). In order to prevent redundant
2216 checks of crossjump(B,A), require that A be the block
2217 with the lowest index. */
2218 if (e->src->index > e2->src->index)
2219 continue;
2220
2221 /* If nothing changed since the last attempt, there is nothing
2222 we can do. */
2223 if (!first_pass
2224 && !((e->src->flags & BB_MODIFIED)
2225 || (e2->src->flags & BB_MODIFIED)))
2226 continue;
2227
2228 /* Both e and e2 are not fallthru edges, so we can crossjump in either
2229 direction. */
2230 if (try_crossjump_to_edge (mode, e, e2, dir_both))
2231 {
2232 changed = true;
2233 ix = 0;
2234 break;
2235 }
2236 }
2237 }
2238
2239 if (changed)
2240 crossjumps_occured = true;
2241
2242 return changed;
2243 }
2244
2245 /* Search the successors of BB for common insn sequences. When found,
2246 share code between them by moving it across the basic block
2247 boundary. Return true if any changes made. */
2248
2249 static bool
2250 try_head_merge_bb (basic_block bb)
2251 {
2252 basic_block final_dest_bb = NULL;
2253 int max_match = INT_MAX;
2254 edge e0;
2255 rtx *headptr, *currptr, *nextptr;
2256 bool changed, moveall;
2257 unsigned ix;
2258 rtx e0_last_head, cond, move_before;
2259 unsigned nedges = EDGE_COUNT (bb->succs);
2260 rtx jump = BB_END (bb);
2261 regset live, live_union;
2262
2263 /* Nothing to do if there is not at least two outgoing edges. */
2264 if (nedges < 2)
2265 return false;
2266
2267 /* Don't crossjump if this block ends in a computed jump,
2268 unless we are optimizing for size. */
2269 if (optimize_bb_for_size_p (bb)
2270 && bb != EXIT_BLOCK_PTR
2271 && computed_jump_p (BB_END (bb)))
2272 return false;
2273
2274 cond = get_condition (jump, &move_before, true, false);
2275 if (cond == NULL_RTX)
2276 {
2277 #ifdef HAVE_cc0
2278 if (reg_mentioned_p (cc0_rtx, jump))
2279 move_before = prev_nonnote_nondebug_insn (jump);
2280 else
2281 #endif
2282 move_before = jump;
2283 }
2284
2285 for (ix = 0; ix < nedges; ix++)
2286 if (EDGE_SUCC (bb, ix)->dest == EXIT_BLOCK_PTR)
2287 return false;
2288
2289 for (ix = 0; ix < nedges; ix++)
2290 {
2291 edge e = EDGE_SUCC (bb, ix);
2292 basic_block other_bb = e->dest;
2293
2294 if (df_get_bb_dirty (other_bb))
2295 {
2296 block_was_dirty = true;
2297 return false;
2298 }
2299
2300 if (e->flags & EDGE_ABNORMAL)
2301 return false;
2302
2303 /* Normally, all destination blocks must only be reachable from this
2304 block, i.e. they must have one incoming edge.
2305
2306 There is one special case we can handle, that of multiple consecutive
2307 jumps where the first jumps to one of the targets of the second jump.
2308 This happens frequently in switch statements for default labels.
2309 The structure is as follows:
2310 FINAL_DEST_BB
2311 ....
2312 if (cond) jump A;
2313 fall through
2314 BB
2315 jump with targets A, B, C, D...
2316 A
2317 has two incoming edges, from FINAL_DEST_BB and BB
2318
2319 In this case, we can try to move the insns through BB and into
2320 FINAL_DEST_BB. */
2321 if (EDGE_COUNT (other_bb->preds) != 1)
2322 {
2323 edge incoming_edge, incoming_bb_other_edge;
2324 edge_iterator ei;
2325
2326 if (final_dest_bb != NULL
2327 || EDGE_COUNT (other_bb->preds) != 2)
2328 return false;
2329
2330 /* We must be able to move the insns across the whole block. */
2331 move_before = BB_HEAD (bb);
2332 while (!NONDEBUG_INSN_P (move_before))
2333 move_before = NEXT_INSN (move_before);
2334
2335 if (EDGE_COUNT (bb->preds) != 1)
2336 return false;
2337 incoming_edge = EDGE_PRED (bb, 0);
2338 final_dest_bb = incoming_edge->src;
2339 if (EDGE_COUNT (final_dest_bb->succs) != 2)
2340 return false;
2341 FOR_EACH_EDGE (incoming_bb_other_edge, ei, final_dest_bb->succs)
2342 if (incoming_bb_other_edge != incoming_edge)
2343 break;
2344 if (incoming_bb_other_edge->dest != other_bb)
2345 return false;
2346 }
2347 }
2348
2349 e0 = EDGE_SUCC (bb, 0);
2350 e0_last_head = NULL_RTX;
2351 changed = false;
2352
2353 for (ix = 1; ix < nedges; ix++)
2354 {
2355 edge e = EDGE_SUCC (bb, ix);
2356 rtx e0_last, e_last;
2357 int nmatch;
2358
2359 nmatch = flow_find_head_matching_sequence (e0->dest, e->dest,
2360 &e0_last, &e_last, 0);
2361 if (nmatch == 0)
2362 return false;
2363
2364 if (nmatch < max_match)
2365 {
2366 max_match = nmatch;
2367 e0_last_head = e0_last;
2368 }
2369 }
2370
2371 /* If we matched an entire block, we probably have to avoid moving the
2372 last insn. */
2373 if (max_match > 0
2374 && e0_last_head == BB_END (e0->dest)
2375 && (find_reg_note (e0_last_head, REG_EH_REGION, 0)
2376 || control_flow_insn_p (e0_last_head)))
2377 {
2378 max_match--;
2379 if (max_match == 0)
2380 return false;
2381 do
2382 e0_last_head = prev_real_insn (e0_last_head);
2383 while (DEBUG_INSN_P (e0_last_head));
2384 }
2385
2386 if (max_match == 0)
2387 return false;
2388
2389 /* We must find a union of the live registers at each of the end points. */
2390 live = BITMAP_ALLOC (NULL);
2391 live_union = BITMAP_ALLOC (NULL);
2392
2393 currptr = XNEWVEC (rtx, nedges);
2394 headptr = XNEWVEC (rtx, nedges);
2395 nextptr = XNEWVEC (rtx, nedges);
2396
2397 for (ix = 0; ix < nedges; ix++)
2398 {
2399 int j;
2400 basic_block merge_bb = EDGE_SUCC (bb, ix)->dest;
2401 rtx head = BB_HEAD (merge_bb);
2402
2403 while (!NONDEBUG_INSN_P (head))
2404 head = NEXT_INSN (head);
2405 headptr[ix] = head;
2406 currptr[ix] = head;
2407
2408 /* Compute the end point and live information */
2409 for (j = 1; j < max_match; j++)
2410 do
2411 head = NEXT_INSN (head);
2412 while (!NONDEBUG_INSN_P (head));
2413 simulate_backwards_to_point (merge_bb, live, head);
2414 IOR_REG_SET (live_union, live);
2415 }
2416
2417 /* If we're moving across two blocks, verify the validity of the
2418 first move, then adjust the target and let the loop below deal
2419 with the final move. */
2420 if (final_dest_bb != NULL)
2421 {
2422 rtx move_upto;
2423
2424 moveall = can_move_insns_across (currptr[0], e0_last_head, move_before,
2425 jump, e0->dest, live_union,
2426 NULL, &move_upto);
2427 if (!moveall)
2428 {
2429 if (move_upto == NULL_RTX)
2430 goto out;
2431
2432 while (e0_last_head != move_upto)
2433 {
2434 df_simulate_one_insn_backwards (e0->dest, e0_last_head,
2435 live_union);
2436 e0_last_head = PREV_INSN (e0_last_head);
2437 }
2438 }
2439 if (e0_last_head == NULL_RTX)
2440 goto out;
2441
2442 jump = BB_END (final_dest_bb);
2443 cond = get_condition (jump, &move_before, true, false);
2444 if (cond == NULL_RTX)
2445 {
2446 #ifdef HAVE_cc0
2447 if (reg_mentioned_p (cc0_rtx, jump))
2448 move_before = prev_nonnote_nondebug_insn (jump);
2449 else
2450 #endif
2451 move_before = jump;
2452 }
2453 }
2454
2455 do
2456 {
2457 rtx move_upto;
2458 moveall = can_move_insns_across (currptr[0], e0_last_head,
2459 move_before, jump, e0->dest, live_union,
2460 NULL, &move_upto);
2461 if (!moveall && move_upto == NULL_RTX)
2462 {
2463 if (jump == move_before)
2464 break;
2465
2466 /* Try again, using a different insertion point. */
2467 move_before = jump;
2468
2469 #ifdef HAVE_cc0
2470 /* Don't try moving before a cc0 user, as that may invalidate
2471 the cc0. */
2472 if (reg_mentioned_p (cc0_rtx, jump))
2473 break;
2474 #endif
2475
2476 continue;
2477 }
2478
2479 if (final_dest_bb && !moveall)
2480 /* We haven't checked whether a partial move would be OK for the first
2481 move, so we have to fail this case. */
2482 break;
2483
2484 changed = true;
2485 for (;;)
2486 {
2487 if (currptr[0] == move_upto)
2488 break;
2489 for (ix = 0; ix < nedges; ix++)
2490 {
2491 rtx curr = currptr[ix];
2492 do
2493 curr = NEXT_INSN (curr);
2494 while (!NONDEBUG_INSN_P (curr));
2495 currptr[ix] = curr;
2496 }
2497 }
2498
2499 /* If we can't currently move all of the identical insns, remember
2500 each insn after the range that we'll merge. */
2501 if (!moveall)
2502 for (ix = 0; ix < nedges; ix++)
2503 {
2504 rtx curr = currptr[ix];
2505 do
2506 curr = NEXT_INSN (curr);
2507 while (!NONDEBUG_INSN_P (curr));
2508 nextptr[ix] = curr;
2509 }
2510
2511 reorder_insns (headptr[0], currptr[0], PREV_INSN (move_before));
2512 df_set_bb_dirty (EDGE_SUCC (bb, 0)->dest);
2513 if (final_dest_bb != NULL)
2514 df_set_bb_dirty (final_dest_bb);
2515 df_set_bb_dirty (bb);
2516 for (ix = 1; ix < nedges; ix++)
2517 {
2518 df_set_bb_dirty (EDGE_SUCC (bb, ix)->dest);
2519 delete_insn_chain (headptr[ix], currptr[ix], false);
2520 }
2521 if (!moveall)
2522 {
2523 if (jump == move_before)
2524 break;
2525
2526 /* For the unmerged insns, try a different insertion point. */
2527 move_before = jump;
2528
2529 #ifdef HAVE_cc0
2530 /* Don't try moving before a cc0 user, as that may invalidate
2531 the cc0. */
2532 if (reg_mentioned_p (cc0_rtx, jump))
2533 break;
2534 #endif
2535
2536 for (ix = 0; ix < nedges; ix++)
2537 currptr[ix] = headptr[ix] = nextptr[ix];
2538 }
2539 }
2540 while (!moveall);
2541
2542 out:
2543 free (currptr);
2544 free (headptr);
2545 free (nextptr);
2546
2547 crossjumps_occured |= changed;
2548
2549 return changed;
2550 }
2551
2552 /* Return true if BB contains just bb note, or bb note followed
2553 by only DEBUG_INSNs. */
2554
2555 static bool
2556 trivially_empty_bb_p (basic_block bb)
2557 {
2558 rtx insn = BB_END (bb);
2559
2560 while (1)
2561 {
2562 if (insn == BB_HEAD (bb))
2563 return true;
2564 if (!DEBUG_INSN_P (insn))
2565 return false;
2566 insn = PREV_INSN (insn);
2567 }
2568 }
2569
2570 /* Do simple CFG optimizations - basic block merging, simplifying of jump
2571 instructions etc. Return nonzero if changes were made. */
2572
2573 static bool
2574 try_optimize_cfg (int mode)
2575 {
2576 bool changed_overall = false;
2577 bool changed;
2578 int iterations = 0;
2579 basic_block bb, b, next;
2580
2581 if (mode & (CLEANUP_CROSSJUMP | CLEANUP_THREADING))
2582 clear_bb_flags ();
2583
2584 crossjumps_occured = false;
2585
2586 FOR_EACH_BB (bb)
2587 update_forwarder_flag (bb);
2588
2589 if (! targetm.cannot_modify_jumps_p ())
2590 {
2591 first_pass = true;
2592 /* Attempt to merge blocks as made possible by edge removal. If
2593 a block has only one successor, and the successor has only
2594 one predecessor, they may be combined. */
2595 do
2596 {
2597 block_was_dirty = false;
2598 changed = false;
2599 iterations++;
2600
2601 if (dump_file)
2602 fprintf (dump_file,
2603 "\n\ntry_optimize_cfg iteration %i\n\n",
2604 iterations);
2605
2606 for (b = ENTRY_BLOCK_PTR->next_bb; b != EXIT_BLOCK_PTR;)
2607 {
2608 basic_block c;
2609 edge s;
2610 bool changed_here = false;
2611
2612 /* Delete trivially dead basic blocks. This is either
2613 blocks with no predecessors, or empty blocks with no
2614 successors. However if the empty block with no
2615 successors is the successor of the ENTRY_BLOCK, it is
2616 kept. This ensures that the ENTRY_BLOCK will have a
2617 successor which is a precondition for many RTL
2618 passes. Empty blocks may result from expanding
2619 __builtin_unreachable (). */
2620 if (EDGE_COUNT (b->preds) == 0
2621 || (EDGE_COUNT (b->succs) == 0
2622 && trivially_empty_bb_p (b)
2623 && single_succ_edge (ENTRY_BLOCK_PTR)->dest != b))
2624 {
2625 c = b->prev_bb;
2626 if (EDGE_COUNT (b->preds) > 0)
2627 {
2628 edge e;
2629 edge_iterator ei;
2630
2631 if (current_ir_type () == IR_RTL_CFGLAYOUT)
2632 {
2633 if (BB_FOOTER (b)
2634 && BARRIER_P (BB_FOOTER (b)))
2635 FOR_EACH_EDGE (e, ei, b->preds)
2636 if ((e->flags & EDGE_FALLTHRU)
2637 && BB_FOOTER (e->src) == NULL)
2638 {
2639 if (BB_FOOTER (b))
2640 {
2641 BB_FOOTER (e->src) = BB_FOOTER (b);
2642 BB_FOOTER (b) = NULL;
2643 }
2644 else
2645 {
2646 start_sequence ();
2647 BB_FOOTER (e->src) = emit_barrier ();
2648 end_sequence ();
2649 }
2650 }
2651 }
2652 else
2653 {
2654 rtx last = get_last_bb_insn (b);
2655 if (last && BARRIER_P (last))
2656 FOR_EACH_EDGE (e, ei, b->preds)
2657 if ((e->flags & EDGE_FALLTHRU))
2658 emit_barrier_after (BB_END (e->src));
2659 }
2660 }
2661 delete_basic_block (b);
2662 changed = true;
2663 /* Avoid trying to remove ENTRY_BLOCK_PTR. */
2664 b = (c == ENTRY_BLOCK_PTR ? c->next_bb : c);
2665 continue;
2666 }
2667
2668 /* Remove code labels no longer used. */
2669 if (single_pred_p (b)
2670 && (single_pred_edge (b)->flags & EDGE_FALLTHRU)
2671 && !(single_pred_edge (b)->flags & EDGE_COMPLEX)
2672 && LABEL_P (BB_HEAD (b))
2673 /* If the previous block ends with a branch to this
2674 block, we can't delete the label. Normally this
2675 is a condjump that is yet to be simplified, but
2676 if CASE_DROPS_THRU, this can be a tablejump with
2677 some element going to the same place as the
2678 default (fallthru). */
2679 && (single_pred (b) == ENTRY_BLOCK_PTR
2680 || !JUMP_P (BB_END (single_pred (b)))
2681 || ! label_is_jump_target_p (BB_HEAD (b),
2682 BB_END (single_pred (b)))))
2683 {
2684 delete_insn (BB_HEAD (b));
2685 if (dump_file)
2686 fprintf (dump_file, "Deleted label in block %i.\n",
2687 b->index);
2688 }
2689
2690 /* If we fall through an empty block, we can remove it. */
2691 if (!(mode & (CLEANUP_CFGLAYOUT | CLEANUP_NO_INSN_DEL))
2692 && single_pred_p (b)
2693 && (single_pred_edge (b)->flags & EDGE_FALLTHRU)
2694 && !LABEL_P (BB_HEAD (b))
2695 && FORWARDER_BLOCK_P (b)
2696 /* Note that forwarder_block_p true ensures that
2697 there is a successor for this block. */
2698 && (single_succ_edge (b)->flags & EDGE_FALLTHRU)
2699 && n_basic_blocks > NUM_FIXED_BLOCKS + 1)
2700 {
2701 if (dump_file)
2702 fprintf (dump_file,
2703 "Deleting fallthru block %i.\n",
2704 b->index);
2705
2706 c = b->prev_bb == ENTRY_BLOCK_PTR ? b->next_bb : b->prev_bb;
2707 redirect_edge_succ_nodup (single_pred_edge (b),
2708 single_succ (b));
2709 delete_basic_block (b);
2710 changed = true;
2711 b = c;
2712 continue;
2713 }
2714
2715 /* Merge B with its single successor, if any. */
2716 if (single_succ_p (b)
2717 && (s = single_succ_edge (b))
2718 && !(s->flags & EDGE_COMPLEX)
2719 && (c = s->dest) != EXIT_BLOCK_PTR
2720 && single_pred_p (c)
2721 && b != c)
2722 {
2723 /* When not in cfg_layout mode use code aware of reordering
2724 INSN. This code possibly creates new basic blocks so it
2725 does not fit merge_blocks interface and is kept here in
2726 hope that it will become useless once more of compiler
2727 is transformed to use cfg_layout mode. */
2728
2729 if ((mode & CLEANUP_CFGLAYOUT)
2730 && can_merge_blocks_p (b, c))
2731 {
2732 merge_blocks (b, c);
2733 update_forwarder_flag (b);
2734 changed_here = true;
2735 }
2736 else if (!(mode & CLEANUP_CFGLAYOUT)
2737 /* If the jump insn has side effects,
2738 we can't kill the edge. */
2739 && (!JUMP_P (BB_END (b))
2740 || (reload_completed
2741 ? simplejump_p (BB_END (b))
2742 : (onlyjump_p (BB_END (b))
2743 && !tablejump_p (BB_END (b),
2744 NULL, NULL))))
2745 && (next = merge_blocks_move (s, b, c, mode)))
2746 {
2747 b = next;
2748 changed_here = true;
2749 }
2750 }
2751
2752 /* Simplify branch over branch. */
2753 if ((mode & CLEANUP_EXPENSIVE)
2754 && !(mode & CLEANUP_CFGLAYOUT)
2755 && try_simplify_condjump (b))
2756 changed_here = true;
2757
2758 /* If B has a single outgoing edge, but uses a
2759 non-trivial jump instruction without side-effects, we
2760 can either delete the jump entirely, or replace it
2761 with a simple unconditional jump. */
2762 if (single_succ_p (b)
2763 && single_succ (b) != EXIT_BLOCK_PTR
2764 && onlyjump_p (BB_END (b))
2765 && !find_reg_note (BB_END (b), REG_CROSSING_JUMP, NULL_RTX)
2766 && try_redirect_by_replacing_jump (single_succ_edge (b),
2767 single_succ (b),
2768 (mode & CLEANUP_CFGLAYOUT) != 0))
2769 {
2770 update_forwarder_flag (b);
2771 changed_here = true;
2772 }
2773
2774 /* Simplify branch to branch. */
2775 if (try_forward_edges (mode, b))
2776 {
2777 update_forwarder_flag (b);
2778 changed_here = true;
2779 }
2780
2781 /* Look for shared code between blocks. */
2782 if ((mode & CLEANUP_CROSSJUMP)
2783 && try_crossjump_bb (mode, b))
2784 changed_here = true;
2785
2786 if ((mode & CLEANUP_CROSSJUMP)
2787 /* This can lengthen register lifetimes. Do it only after
2788 reload. */
2789 && reload_completed
2790 && try_head_merge_bb (b))
2791 changed_here = true;
2792
2793 /* Don't get confused by the index shift caused by
2794 deleting blocks. */
2795 if (!changed_here)
2796 b = b->next_bb;
2797 else
2798 changed = true;
2799 }
2800
2801 if ((mode & CLEANUP_CROSSJUMP)
2802 && try_crossjump_bb (mode, EXIT_BLOCK_PTR))
2803 changed = true;
2804
2805 if (block_was_dirty)
2806 {
2807 /* This should only be set by head-merging. */
2808 gcc_assert (mode & CLEANUP_CROSSJUMP);
2809 df_analyze ();
2810 }
2811
2812 #ifdef ENABLE_CHECKING
2813 if (changed)
2814 verify_flow_info ();
2815 #endif
2816
2817 changed_overall |= changed;
2818 first_pass = false;
2819 }
2820 while (changed);
2821 }
2822
2823 FOR_ALL_BB (b)
2824 b->flags &= ~(BB_FORWARDER_BLOCK | BB_NONTHREADABLE_BLOCK);
2825
2826 return changed_overall;
2827 }
2828 \f
2829 /* Delete all unreachable basic blocks. */
2830
2831 bool
2832 delete_unreachable_blocks (void)
2833 {
2834 bool changed = false;
2835 basic_block b, prev_bb;
2836
2837 find_unreachable_blocks ();
2838
2839 /* When we're in GIMPLE mode and there may be debug insns, we should
2840 delete blocks in reverse dominator order, so as to get a chance
2841 to substitute all released DEFs into debug stmts. If we don't
2842 have dominators information, walking blocks backward gets us a
2843 better chance of retaining most debug information than
2844 otherwise. */
2845 if (MAY_HAVE_DEBUG_INSNS && current_ir_type () == IR_GIMPLE
2846 && dom_info_available_p (CDI_DOMINATORS))
2847 {
2848 for (b = EXIT_BLOCK_PTR->prev_bb; b != ENTRY_BLOCK_PTR; b = prev_bb)
2849 {
2850 prev_bb = b->prev_bb;
2851
2852 if (!(b->flags & BB_REACHABLE))
2853 {
2854 /* Speed up the removal of blocks that don't dominate
2855 others. Walking backwards, this should be the common
2856 case. */
2857 if (!first_dom_son (CDI_DOMINATORS, b))
2858 delete_basic_block (b);
2859 else
2860 {
2861 vec<basic_block> h
2862 = get_all_dominated_blocks (CDI_DOMINATORS, b);
2863
2864 while (h.length ())
2865 {
2866 b = h.pop ();
2867
2868 prev_bb = b->prev_bb;
2869
2870 gcc_assert (!(b->flags & BB_REACHABLE));
2871
2872 delete_basic_block (b);
2873 }
2874
2875 h.release ();
2876 }
2877
2878 changed = true;
2879 }
2880 }
2881 }
2882 else
2883 {
2884 for (b = EXIT_BLOCK_PTR->prev_bb; b != ENTRY_BLOCK_PTR; b = prev_bb)
2885 {
2886 prev_bb = b->prev_bb;
2887
2888 if (!(b->flags & BB_REACHABLE))
2889 {
2890 delete_basic_block (b);
2891 changed = true;
2892 }
2893 }
2894 }
2895
2896 if (changed)
2897 tidy_fallthru_edges ();
2898 return changed;
2899 }
2900
2901 /* Delete any jump tables never referenced. We can't delete them at the
2902 time of removing tablejump insn as they are referenced by the preceding
2903 insns computing the destination, so we delay deleting and garbagecollect
2904 them once life information is computed. */
2905 void
2906 delete_dead_jumptables (void)
2907 {
2908 basic_block bb;
2909
2910 /* A dead jump table does not belong to any basic block. Scan insns
2911 between two adjacent basic blocks. */
2912 FOR_EACH_BB (bb)
2913 {
2914 rtx insn, next;
2915
2916 for (insn = NEXT_INSN (BB_END (bb));
2917 insn && !NOTE_INSN_BASIC_BLOCK_P (insn);
2918 insn = next)
2919 {
2920 next = NEXT_INSN (insn);
2921 if (LABEL_P (insn)
2922 && LABEL_NUSES (insn) == LABEL_PRESERVE_P (insn)
2923 && JUMP_TABLE_DATA_P (next))
2924 {
2925 rtx label = insn, jump = next;
2926
2927 if (dump_file)
2928 fprintf (dump_file, "Dead jumptable %i removed\n",
2929 INSN_UID (insn));
2930
2931 next = NEXT_INSN (next);
2932 delete_insn (jump);
2933 delete_insn (label);
2934 }
2935 }
2936 }
2937 }
2938
2939 \f
2940 /* Tidy the CFG by deleting unreachable code and whatnot. */
2941
2942 bool
2943 cleanup_cfg (int mode)
2944 {
2945 bool changed = false;
2946
2947 /* Set the cfglayout mode flag here. We could update all the callers
2948 but that is just inconvenient, especially given that we eventually
2949 want to have cfglayout mode as the default. */
2950 if (current_ir_type () == IR_RTL_CFGLAYOUT)
2951 mode |= CLEANUP_CFGLAYOUT;
2952
2953 timevar_push (TV_CLEANUP_CFG);
2954 if (delete_unreachable_blocks ())
2955 {
2956 changed = true;
2957 /* We've possibly created trivially dead code. Cleanup it right
2958 now to introduce more opportunities for try_optimize_cfg. */
2959 if (!(mode & (CLEANUP_NO_INSN_DEL))
2960 && !reload_completed)
2961 delete_trivially_dead_insns (get_insns (), max_reg_num ());
2962 }
2963
2964 compact_blocks ();
2965
2966 /* To tail-merge blocks ending in the same noreturn function (e.g.
2967 a call to abort) we have to insert fake edges to exit. Do this
2968 here once. The fake edges do not interfere with any other CFG
2969 cleanups. */
2970 if (mode & CLEANUP_CROSSJUMP)
2971 add_noreturn_fake_exit_edges ();
2972
2973 if (!dbg_cnt (cfg_cleanup))
2974 return changed;
2975
2976 while (try_optimize_cfg (mode))
2977 {
2978 delete_unreachable_blocks (), changed = true;
2979 if (!(mode & CLEANUP_NO_INSN_DEL))
2980 {
2981 /* Try to remove some trivially dead insns when doing an expensive
2982 cleanup. But delete_trivially_dead_insns doesn't work after
2983 reload (it only handles pseudos) and run_fast_dce is too costly
2984 to run in every iteration.
2985
2986 For effective cross jumping, we really want to run a fast DCE to
2987 clean up any dead conditions, or they get in the way of performing
2988 useful tail merges.
2989
2990 Other transformations in cleanup_cfg are not so sensitive to dead
2991 code, so delete_trivially_dead_insns or even doing nothing at all
2992 is good enough. */
2993 if ((mode & CLEANUP_EXPENSIVE) && !reload_completed
2994 && !delete_trivially_dead_insns (get_insns (), max_reg_num ()))
2995 break;
2996 if ((mode & CLEANUP_CROSSJUMP) && crossjumps_occured)
2997 run_fast_dce ();
2998 }
2999 else
3000 break;
3001 }
3002
3003 if (mode & CLEANUP_CROSSJUMP)
3004 remove_fake_exit_edges ();
3005
3006 /* Don't call delete_dead_jumptables in cfglayout mode, because
3007 that function assumes that jump tables are in the insns stream.
3008 But we also don't _have_ to delete dead jumptables in cfglayout
3009 mode because we shouldn't even be looking at things that are
3010 not in a basic block. Dead jumptables are cleaned up when
3011 going out of cfglayout mode. */
3012 if (!(mode & CLEANUP_CFGLAYOUT))
3013 delete_dead_jumptables ();
3014
3015 /* ??? We probably do this way too often. */
3016 if (current_loops
3017 && (changed
3018 || (mode & CLEANUP_CFG_CHANGED)))
3019 {
3020 timevar_push (TV_REPAIR_LOOPS);
3021 /* The above doesn't preserve dominance info if available. */
3022 gcc_assert (!dom_info_available_p (CDI_DOMINATORS));
3023 calculate_dominance_info (CDI_DOMINATORS);
3024 fix_loop_structure (NULL);
3025 free_dominance_info (CDI_DOMINATORS);
3026 timevar_pop (TV_REPAIR_LOOPS);
3027 }
3028
3029 timevar_pop (TV_CLEANUP_CFG);
3030
3031 return changed;
3032 }
3033 \f
3034 static unsigned int
3035 execute_jump (void)
3036 {
3037 delete_trivially_dead_insns (get_insns (), max_reg_num ());
3038 if (dump_file)
3039 dump_flow_info (dump_file, dump_flags);
3040 cleanup_cfg ((optimize ? CLEANUP_EXPENSIVE : 0)
3041 | (flag_thread_jumps ? CLEANUP_THREADING : 0));
3042 return 0;
3043 }
3044
3045 struct rtl_opt_pass pass_jump =
3046 {
3047 {
3048 RTL_PASS,
3049 "jump", /* name */
3050 OPTGROUP_NONE, /* optinfo_flags */
3051 NULL, /* gate */
3052 execute_jump, /* execute */
3053 NULL, /* sub */
3054 NULL, /* next */
3055 0, /* static_pass_number */
3056 TV_JUMP, /* tv_id */
3057 0, /* properties_required */
3058 0, /* properties_provided */
3059 0, /* properties_destroyed */
3060 TODO_ggc_collect, /* todo_flags_start */
3061 TODO_verify_rtl_sharing, /* todo_flags_finish */
3062 }
3063 };
3064 \f
3065 static unsigned int
3066 execute_jump2 (void)
3067 {
3068 cleanup_cfg (flag_crossjumping ? CLEANUP_CROSSJUMP : 0);
3069 return 0;
3070 }
3071
3072 struct rtl_opt_pass pass_jump2 =
3073 {
3074 {
3075 RTL_PASS,
3076 "jump2", /* name */
3077 OPTGROUP_NONE, /* optinfo_flags */
3078 NULL, /* gate */
3079 execute_jump2, /* execute */
3080 NULL, /* sub */
3081 NULL, /* next */
3082 0, /* static_pass_number */
3083 TV_JUMP, /* tv_id */
3084 0, /* properties_required */
3085 0, /* properties_provided */
3086 0, /* properties_destroyed */
3087 TODO_ggc_collect, /* todo_flags_start */
3088 TODO_verify_rtl_sharing, /* todo_flags_finish */
3089 }
3090 };