* gcc.dg/vect/slp-perm-1.c (main): Make sure loops aren't vectorized.
[official-gcc.git] / gcc / cfgcleanup.c
blob9ded1e6215b744214d4d66bc337e37341a5bb513
1 /* Control flow optimization code for GNU compiler.
2 Copyright (C) 1987, 1988, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
3 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2010
4 Free Software Foundation, Inc.
6 This file is part of GCC.
8 GCC is free software; you can redistribute it and/or modify it under
9 the terms of the GNU General Public License as published by the Free
10 Software Foundation; either version 3, or (at your option) any later
11 version.
13 GCC is distributed in the hope that it will be useful, but WITHOUT ANY
14 WARRANTY; without even the implied warranty of MERCHANTABILITY or
15 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
16 for more details.
18 You should have received a copy of the GNU General Public License
19 along with GCC; see the file COPYING3. If not see
20 <http://www.gnu.org/licenses/>. */
22 /* This file contains optimizer of the control flow. The main entry point is
23 cleanup_cfg. Following optimizations are performed:
25 - Unreachable blocks removal
26 - Edge forwarding (edge to the forwarder block is forwarded to its
27 successor. Simplification of the branch instruction is performed by
28 underlying infrastructure so branch can be converted to simplejump or
29 eliminated).
30 - Cross jumping (tail merging)
31 - Conditional jump-around-simplejump simplification
32 - Basic block merging. */
34 #include "config.h"
35 #include "system.h"
36 #include "coretypes.h"
37 #include "tm.h"
38 #include "rtl.h"
39 #include "hard-reg-set.h"
40 #include "regs.h"
41 #include "timevar.h"
42 #include "output.h"
43 #include "insn-config.h"
44 #include "flags.h"
45 #include "recog.h"
46 #include "diagnostic-core.h"
47 #include "toplev.h"
48 #include "cselib.h"
49 #include "params.h"
50 #include "tm_p.h"
51 #include "target.h"
52 #include "cfglayout.h"
53 #include "emit-rtl.h"
54 #include "tree-pass.h"
55 #include "cfgloop.h"
56 #include "expr.h"
57 #include "df.h"
58 #include "dce.h"
59 #include "dbgcnt.h"
61 #define FORWARDER_BLOCK_P(BB) ((BB)->flags & BB_FORWARDER_BLOCK)
63 /* Set to true when we are running first pass of try_optimize_cfg loop. */
64 static bool first_pass;
66 /* Set to true if crossjumps occured in the latest run of try_optimize_cfg. */
67 static bool crossjumps_occured;
69 static bool try_crossjump_to_edge (int, edge, edge);
70 static bool try_crossjump_bb (int, basic_block);
71 static bool outgoing_edges_match (int, basic_block, basic_block);
72 static bool old_insns_match_p (int, rtx, rtx);
74 static void merge_blocks_move_predecessor_nojumps (basic_block, basic_block);
75 static void merge_blocks_move_successor_nojumps (basic_block, basic_block);
76 static bool try_optimize_cfg (int);
77 static bool try_simplify_condjump (basic_block);
78 static bool try_forward_edges (int, basic_block);
79 static edge thread_jump (edge, basic_block);
80 static bool mark_effect (rtx, bitmap);
81 static void notice_new_block (basic_block);
82 static void update_forwarder_flag (basic_block);
83 static int mentions_nonequal_regs (rtx *, void *);
84 static void merge_memattrs (rtx, rtx);
86 /* Set flags for newly created block. */
88 static void
89 notice_new_block (basic_block bb)
91 if (!bb)
92 return;
94 if (forwarder_block_p (bb))
95 bb->flags |= BB_FORWARDER_BLOCK;
98 /* Recompute forwarder flag after block has been modified. */
100 static void
101 update_forwarder_flag (basic_block bb)
103 if (forwarder_block_p (bb))
104 bb->flags |= BB_FORWARDER_BLOCK;
105 else
106 bb->flags &= ~BB_FORWARDER_BLOCK;
109 /* Simplify a conditional jump around an unconditional jump.
110 Return true if something changed. */
112 static bool
113 try_simplify_condjump (basic_block cbranch_block)
115 basic_block jump_block, jump_dest_block, cbranch_dest_block;
116 edge cbranch_jump_edge, cbranch_fallthru_edge;
117 rtx cbranch_insn;
119 /* Verify that there are exactly two successors. */
120 if (EDGE_COUNT (cbranch_block->succs) != 2)
121 return false;
123 /* Verify that we've got a normal conditional branch at the end
124 of the block. */
125 cbranch_insn = BB_END (cbranch_block);
126 if (!any_condjump_p (cbranch_insn))
127 return false;
129 cbranch_fallthru_edge = FALLTHRU_EDGE (cbranch_block);
130 cbranch_jump_edge = BRANCH_EDGE (cbranch_block);
132 /* The next block must not have multiple predecessors, must not
133 be the last block in the function, and must contain just the
134 unconditional jump. */
135 jump_block = cbranch_fallthru_edge->dest;
136 if (!single_pred_p (jump_block)
137 || jump_block->next_bb == EXIT_BLOCK_PTR
138 || !FORWARDER_BLOCK_P (jump_block))
139 return false;
140 jump_dest_block = single_succ (jump_block);
142 /* If we are partitioning hot/cold basic blocks, we don't want to
143 mess up unconditional or indirect jumps that cross between hot
144 and cold sections.
146 Basic block partitioning may result in some jumps that appear to
147 be optimizable (or blocks that appear to be mergeable), but which really
148 must be left untouched (they are required to make it safely across
149 partition boundaries). See the comments at the top of
150 bb-reorder.c:partition_hot_cold_basic_blocks for complete details. */
152 if (BB_PARTITION (jump_block) != BB_PARTITION (jump_dest_block)
153 || (cbranch_jump_edge->flags & EDGE_CROSSING))
154 return false;
156 /* The conditional branch must target the block after the
157 unconditional branch. */
158 cbranch_dest_block = cbranch_jump_edge->dest;
160 if (cbranch_dest_block == EXIT_BLOCK_PTR
161 || !can_fallthru (jump_block, cbranch_dest_block))
162 return false;
164 /* Invert the conditional branch. */
165 if (!invert_jump (cbranch_insn, block_label (jump_dest_block), 0))
166 return false;
168 if (dump_file)
169 fprintf (dump_file, "Simplifying condjump %i around jump %i\n",
170 INSN_UID (cbranch_insn), INSN_UID (BB_END (jump_block)));
172 /* Success. Update the CFG to match. Note that after this point
173 the edge variable names appear backwards; the redirection is done
174 this way to preserve edge profile data. */
175 cbranch_jump_edge = redirect_edge_succ_nodup (cbranch_jump_edge,
176 cbranch_dest_block);
177 cbranch_fallthru_edge = redirect_edge_succ_nodup (cbranch_fallthru_edge,
178 jump_dest_block);
179 cbranch_jump_edge->flags |= EDGE_FALLTHRU;
180 cbranch_fallthru_edge->flags &= ~EDGE_FALLTHRU;
181 update_br_prob_note (cbranch_block);
183 /* Delete the block with the unconditional jump, and clean up the mess. */
184 delete_basic_block (jump_block);
185 tidy_fallthru_edge (cbranch_jump_edge);
186 update_forwarder_flag (cbranch_block);
188 return true;
191 /* Attempt to prove that operation is NOOP using CSElib or mark the effect
192 on register. Used by jump threading. */
194 static bool
195 mark_effect (rtx exp, regset nonequal)
197 int regno;
198 rtx dest;
199 switch (GET_CODE (exp))
201 /* In case we do clobber the register, mark it as equal, as we know the
202 value is dead so it don't have to match. */
203 case CLOBBER:
204 if (REG_P (XEXP (exp, 0)))
206 dest = XEXP (exp, 0);
207 regno = REGNO (dest);
208 CLEAR_REGNO_REG_SET (nonequal, regno);
209 if (regno < FIRST_PSEUDO_REGISTER)
211 int n = hard_regno_nregs[regno][GET_MODE (dest)];
212 while (--n > 0)
213 CLEAR_REGNO_REG_SET (nonequal, regno + n);
216 return false;
218 case SET:
219 if (rtx_equal_for_cselib_p (SET_DEST (exp), SET_SRC (exp)))
220 return false;
221 dest = SET_DEST (exp);
222 if (dest == pc_rtx)
223 return false;
224 if (!REG_P (dest))
225 return true;
226 regno = REGNO (dest);
227 SET_REGNO_REG_SET (nonequal, regno);
228 if (regno < FIRST_PSEUDO_REGISTER)
230 int n = hard_regno_nregs[regno][GET_MODE (dest)];
231 while (--n > 0)
232 SET_REGNO_REG_SET (nonequal, regno + n);
234 return false;
236 default:
237 return false;
241 /* Return nonzero if X is a register set in regset DATA.
242 Called via for_each_rtx. */
243 static int
244 mentions_nonequal_regs (rtx *x, void *data)
246 regset nonequal = (regset) data;
247 if (REG_P (*x))
249 int regno;
251 regno = REGNO (*x);
252 if (REGNO_REG_SET_P (nonequal, regno))
253 return 1;
254 if (regno < FIRST_PSEUDO_REGISTER)
256 int n = hard_regno_nregs[regno][GET_MODE (*x)];
257 while (--n > 0)
258 if (REGNO_REG_SET_P (nonequal, regno + n))
259 return 1;
262 return 0;
264 /* Attempt to prove that the basic block B will have no side effects and
265 always continues in the same edge if reached via E. Return the edge
266 if exist, NULL otherwise. */
268 static edge
269 thread_jump (edge e, basic_block b)
271 rtx set1, set2, cond1, cond2, insn;
272 enum rtx_code code1, code2, reversed_code2;
273 bool reverse1 = false;
274 unsigned i;
275 regset nonequal;
276 bool failed = false;
277 reg_set_iterator rsi;
279 if (b->flags & BB_NONTHREADABLE_BLOCK)
280 return NULL;
282 /* At the moment, we do handle only conditional jumps, but later we may
283 want to extend this code to tablejumps and others. */
284 if (EDGE_COUNT (e->src->succs) != 2)
285 return NULL;
286 if (EDGE_COUNT (b->succs) != 2)
288 b->flags |= BB_NONTHREADABLE_BLOCK;
289 return NULL;
292 /* Second branch must end with onlyjump, as we will eliminate the jump. */
293 if (!any_condjump_p (BB_END (e->src)))
294 return NULL;
296 if (!any_condjump_p (BB_END (b)) || !onlyjump_p (BB_END (b)))
298 b->flags |= BB_NONTHREADABLE_BLOCK;
299 return NULL;
302 set1 = pc_set (BB_END (e->src));
303 set2 = pc_set (BB_END (b));
304 if (((e->flags & EDGE_FALLTHRU) != 0)
305 != (XEXP (SET_SRC (set1), 1) == pc_rtx))
306 reverse1 = true;
308 cond1 = XEXP (SET_SRC (set1), 0);
309 cond2 = XEXP (SET_SRC (set2), 0);
310 if (reverse1)
311 code1 = reversed_comparison_code (cond1, BB_END (e->src));
312 else
313 code1 = GET_CODE (cond1);
315 code2 = GET_CODE (cond2);
316 reversed_code2 = reversed_comparison_code (cond2, BB_END (b));
318 if (!comparison_dominates_p (code1, code2)
319 && !comparison_dominates_p (code1, reversed_code2))
320 return NULL;
322 /* Ensure that the comparison operators are equivalent.
323 ??? This is far too pessimistic. We should allow swapped operands,
324 different CCmodes, or for example comparisons for interval, that
325 dominate even when operands are not equivalent. */
326 if (!rtx_equal_p (XEXP (cond1, 0), XEXP (cond2, 0))
327 || !rtx_equal_p (XEXP (cond1, 1), XEXP (cond2, 1)))
328 return NULL;
330 /* Short circuit cases where block B contains some side effects, as we can't
331 safely bypass it. */
332 for (insn = NEXT_INSN (BB_HEAD (b)); insn != NEXT_INSN (BB_END (b));
333 insn = NEXT_INSN (insn))
334 if (INSN_P (insn) && side_effects_p (PATTERN (insn)))
336 b->flags |= BB_NONTHREADABLE_BLOCK;
337 return NULL;
340 cselib_init (0);
342 /* First process all values computed in the source basic block. */
343 for (insn = NEXT_INSN (BB_HEAD (e->src));
344 insn != NEXT_INSN (BB_END (e->src));
345 insn = NEXT_INSN (insn))
346 if (INSN_P (insn))
347 cselib_process_insn (insn);
349 nonequal = BITMAP_ALLOC (NULL);
350 CLEAR_REG_SET (nonequal);
352 /* Now assume that we've continued by the edge E to B and continue
353 processing as if it were same basic block.
354 Our goal is to prove that whole block is an NOOP. */
356 for (insn = NEXT_INSN (BB_HEAD (b));
357 insn != NEXT_INSN (BB_END (b)) && !failed;
358 insn = NEXT_INSN (insn))
360 if (INSN_P (insn))
362 rtx pat = PATTERN (insn);
364 if (GET_CODE (pat) == PARALLEL)
366 for (i = 0; i < (unsigned)XVECLEN (pat, 0); i++)
367 failed |= mark_effect (XVECEXP (pat, 0, i), nonequal);
369 else
370 failed |= mark_effect (pat, nonequal);
373 cselib_process_insn (insn);
376 /* Later we should clear nonequal of dead registers. So far we don't
377 have life information in cfg_cleanup. */
378 if (failed)
380 b->flags |= BB_NONTHREADABLE_BLOCK;
381 goto failed_exit;
384 /* cond2 must not mention any register that is not equal to the
385 former block. */
386 if (for_each_rtx (&cond2, mentions_nonequal_regs, nonequal))
387 goto failed_exit;
389 EXECUTE_IF_SET_IN_REG_SET (nonequal, 0, i, rsi)
390 goto failed_exit;
392 BITMAP_FREE (nonequal);
393 cselib_finish ();
394 if ((comparison_dominates_p (code1, code2) != 0)
395 != (XEXP (SET_SRC (set2), 1) == pc_rtx))
396 return BRANCH_EDGE (b);
397 else
398 return FALLTHRU_EDGE (b);
400 failed_exit:
401 BITMAP_FREE (nonequal);
402 cselib_finish ();
403 return NULL;
406 /* Attempt to forward edges leaving basic block B.
407 Return true if successful. */
409 static bool
410 try_forward_edges (int mode, basic_block b)
412 bool changed = false;
413 edge_iterator ei;
414 edge e, *threaded_edges = NULL;
416 /* If we are partitioning hot/cold basic blocks, we don't want to
417 mess up unconditional or indirect jumps that cross between hot
418 and cold sections.
420 Basic block partitioning may result in some jumps that appear to
421 be optimizable (or blocks that appear to be mergeable), but which really
422 must be left untouched (they are required to make it safely across
423 partition boundaries). See the comments at the top of
424 bb-reorder.c:partition_hot_cold_basic_blocks for complete details. */
426 if (find_reg_note (BB_END (b), REG_CROSSING_JUMP, NULL_RTX))
427 return false;
429 for (ei = ei_start (b->succs); (e = ei_safe_edge (ei)); )
431 basic_block target, first;
432 int counter, goto_locus;
433 bool threaded = false;
434 int nthreaded_edges = 0;
435 bool may_thread = first_pass | df_get_bb_dirty (b);
437 /* Skip complex edges because we don't know how to update them.
439 Still handle fallthru edges, as we can succeed to forward fallthru
440 edge to the same place as the branch edge of conditional branch
441 and turn conditional branch to an unconditional branch. */
442 if (e->flags & EDGE_COMPLEX)
444 ei_next (&ei);
445 continue;
448 target = first = e->dest;
449 counter = NUM_FIXED_BLOCKS;
450 goto_locus = e->goto_locus;
452 /* If we are partitioning hot/cold basic_blocks, we don't want to mess
453 up jumps that cross between hot/cold sections.
455 Basic block partitioning may result in some jumps that appear
456 to be optimizable (or blocks that appear to be mergeable), but which
457 really must be left untouched (they are required to make it safely
458 across partition boundaries). See the comments at the top of
459 bb-reorder.c:partition_hot_cold_basic_blocks for complete
460 details. */
462 if (first != EXIT_BLOCK_PTR
463 && find_reg_note (BB_END (first), REG_CROSSING_JUMP, NULL_RTX))
464 return false;
466 while (counter < n_basic_blocks)
468 basic_block new_target = NULL;
469 bool new_target_threaded = false;
470 may_thread |= df_get_bb_dirty (target);
472 if (FORWARDER_BLOCK_P (target)
473 && !(single_succ_edge (target)->flags & EDGE_CROSSING)
474 && single_succ (target) != EXIT_BLOCK_PTR)
476 /* Bypass trivial infinite loops. */
477 new_target = single_succ (target);
478 if (target == new_target)
479 counter = n_basic_blocks;
480 else if (!optimize)
482 /* When not optimizing, ensure that edges or forwarder
483 blocks with different locus are not optimized out. */
484 int locus = single_succ_edge (target)->goto_locus;
486 if (locus && goto_locus && !locator_eq (locus, goto_locus))
487 counter = n_basic_blocks;
488 else if (locus)
489 goto_locus = locus;
491 if (INSN_P (BB_END (target)))
493 locus = INSN_LOCATOR (BB_END (target));
495 if (locus && goto_locus
496 && !locator_eq (locus, goto_locus))
497 counter = n_basic_blocks;
498 else if (locus)
499 goto_locus = locus;
504 /* Allow to thread only over one edge at time to simplify updating
505 of probabilities. */
506 else if ((mode & CLEANUP_THREADING) && may_thread)
508 edge t = thread_jump (e, target);
509 if (t)
511 if (!threaded_edges)
512 threaded_edges = XNEWVEC (edge, n_basic_blocks);
513 else
515 int i;
517 /* Detect an infinite loop across blocks not
518 including the start block. */
519 for (i = 0; i < nthreaded_edges; ++i)
520 if (threaded_edges[i] == t)
521 break;
522 if (i < nthreaded_edges)
524 counter = n_basic_blocks;
525 break;
529 /* Detect an infinite loop across the start block. */
530 if (t->dest == b)
531 break;
533 gcc_assert (nthreaded_edges < n_basic_blocks - NUM_FIXED_BLOCKS);
534 threaded_edges[nthreaded_edges++] = t;
536 new_target = t->dest;
537 new_target_threaded = true;
541 if (!new_target)
542 break;
544 counter++;
545 target = new_target;
546 threaded |= new_target_threaded;
549 if (counter >= n_basic_blocks)
551 if (dump_file)
552 fprintf (dump_file, "Infinite loop in BB %i.\n",
553 target->index);
555 else if (target == first)
556 ; /* We didn't do anything. */
557 else
559 /* Save the values now, as the edge may get removed. */
560 gcov_type edge_count = e->count;
561 int edge_probability = e->probability;
562 int edge_frequency;
563 int n = 0;
565 e->goto_locus = goto_locus;
567 /* Don't force if target is exit block. */
568 if (threaded && target != EXIT_BLOCK_PTR)
570 notice_new_block (redirect_edge_and_branch_force (e, target));
571 if (dump_file)
572 fprintf (dump_file, "Conditionals threaded.\n");
574 else if (!redirect_edge_and_branch (e, target))
576 if (dump_file)
577 fprintf (dump_file,
578 "Forwarding edge %i->%i to %i failed.\n",
579 b->index, e->dest->index, target->index);
580 ei_next (&ei);
581 continue;
584 /* We successfully forwarded the edge. Now update profile
585 data: for each edge we traversed in the chain, remove
586 the original edge's execution count. */
587 edge_frequency = ((edge_probability * b->frequency
588 + REG_BR_PROB_BASE / 2)
589 / REG_BR_PROB_BASE);
591 if (!FORWARDER_BLOCK_P (b) && forwarder_block_p (b))
592 b->flags |= BB_FORWARDER_BLOCK;
596 edge t;
598 if (!single_succ_p (first))
600 gcc_assert (n < nthreaded_edges);
601 t = threaded_edges [n++];
602 gcc_assert (t->src == first);
603 update_bb_profile_for_threading (first, edge_frequency,
604 edge_count, t);
605 update_br_prob_note (first);
607 else
609 first->count -= edge_count;
610 if (first->count < 0)
611 first->count = 0;
612 first->frequency -= edge_frequency;
613 if (first->frequency < 0)
614 first->frequency = 0;
615 /* It is possible that as the result of
616 threading we've removed edge as it is
617 threaded to the fallthru edge. Avoid
618 getting out of sync. */
619 if (n < nthreaded_edges
620 && first == threaded_edges [n]->src)
621 n++;
622 t = single_succ_edge (first);
625 t->count -= edge_count;
626 if (t->count < 0)
627 t->count = 0;
628 first = t->dest;
630 while (first != target);
632 changed = true;
633 continue;
635 ei_next (&ei);
638 if (threaded_edges)
639 free (threaded_edges);
640 return changed;
644 /* Blocks A and B are to be merged into a single block. A has no incoming
645 fallthru edge, so it can be moved before B without adding or modifying
646 any jumps (aside from the jump from A to B). */
648 static void
649 merge_blocks_move_predecessor_nojumps (basic_block a, basic_block b)
651 rtx barrier;
653 /* If we are partitioning hot/cold basic blocks, we don't want to
654 mess up unconditional or indirect jumps that cross between hot
655 and cold sections.
657 Basic block partitioning may result in some jumps that appear to
658 be optimizable (or blocks that appear to be mergeable), but which really
659 must be left untouched (they are required to make it safely across
660 partition boundaries). See the comments at the top of
661 bb-reorder.c:partition_hot_cold_basic_blocks for complete details. */
663 if (BB_PARTITION (a) != BB_PARTITION (b))
664 return;
666 barrier = next_nonnote_insn (BB_END (a));
667 gcc_assert (BARRIER_P (barrier));
668 delete_insn (barrier);
670 /* Scramble the insn chain. */
671 if (BB_END (a) != PREV_INSN (BB_HEAD (b)))
672 reorder_insns_nobb (BB_HEAD (a), BB_END (a), PREV_INSN (BB_HEAD (b)));
673 df_set_bb_dirty (a);
675 if (dump_file)
676 fprintf (dump_file, "Moved block %d before %d and merged.\n",
677 a->index, b->index);
679 /* Swap the records for the two blocks around. */
681 unlink_block (a);
682 link_block (a, b->prev_bb);
684 /* Now blocks A and B are contiguous. Merge them. */
685 merge_blocks (a, b);
688 /* Blocks A and B are to be merged into a single block. B has no outgoing
689 fallthru edge, so it can be moved after A without adding or modifying
690 any jumps (aside from the jump from A to B). */
692 static void
693 merge_blocks_move_successor_nojumps (basic_block a, basic_block b)
695 rtx barrier, real_b_end;
696 rtx label, table;
698 /* If we are partitioning hot/cold basic blocks, we don't want to
699 mess up unconditional or indirect jumps that cross between hot
700 and cold sections.
702 Basic block partitioning may result in some jumps that appear to
703 be optimizable (or blocks that appear to be mergeable), but which really
704 must be left untouched (they are required to make it safely across
705 partition boundaries). See the comments at the top of
706 bb-reorder.c:partition_hot_cold_basic_blocks for complete details. */
708 if (BB_PARTITION (a) != BB_PARTITION (b))
709 return;
711 real_b_end = BB_END (b);
713 /* If there is a jump table following block B temporarily add the jump table
714 to block B so that it will also be moved to the correct location. */
715 if (tablejump_p (BB_END (b), &label, &table)
716 && prev_active_insn (label) == BB_END (b))
718 BB_END (b) = table;
721 /* There had better have been a barrier there. Delete it. */
722 barrier = NEXT_INSN (BB_END (b));
723 if (barrier && BARRIER_P (barrier))
724 delete_insn (barrier);
727 /* Scramble the insn chain. */
728 reorder_insns_nobb (BB_HEAD (b), BB_END (b), BB_END (a));
730 /* Restore the real end of b. */
731 BB_END (b) = real_b_end;
733 if (dump_file)
734 fprintf (dump_file, "Moved block %d after %d and merged.\n",
735 b->index, a->index);
737 /* Now blocks A and B are contiguous. Merge them. */
738 merge_blocks (a, b);
741 /* Attempt to merge basic blocks that are potentially non-adjacent.
742 Return NULL iff the attempt failed, otherwise return basic block
743 where cleanup_cfg should continue. Because the merging commonly
744 moves basic block away or introduces another optimization
745 possibility, return basic block just before B so cleanup_cfg don't
746 need to iterate.
748 It may be good idea to return basic block before C in the case
749 C has been moved after B and originally appeared earlier in the
750 insn sequence, but we have no information available about the
751 relative ordering of these two. Hopefully it is not too common. */
753 static basic_block
754 merge_blocks_move (edge e, basic_block b, basic_block c, int mode)
756 basic_block next;
758 /* If we are partitioning hot/cold basic blocks, we don't want to
759 mess up unconditional or indirect jumps that cross between hot
760 and cold sections.
762 Basic block partitioning may result in some jumps that appear to
763 be optimizable (or blocks that appear to be mergeable), but which really
764 must be left untouched (they are required to make it safely across
765 partition boundaries). See the comments at the top of
766 bb-reorder.c:partition_hot_cold_basic_blocks for complete details. */
768 if (BB_PARTITION (b) != BB_PARTITION (c))
769 return NULL;
771 /* If B has a fallthru edge to C, no need to move anything. */
772 if (e->flags & EDGE_FALLTHRU)
774 int b_index = b->index, c_index = c->index;
775 merge_blocks (b, c);
776 update_forwarder_flag (b);
778 if (dump_file)
779 fprintf (dump_file, "Merged %d and %d without moving.\n",
780 b_index, c_index);
782 return b->prev_bb == ENTRY_BLOCK_PTR ? b : b->prev_bb;
785 /* Otherwise we will need to move code around. Do that only if expensive
786 transformations are allowed. */
787 else if (mode & CLEANUP_EXPENSIVE)
789 edge tmp_edge, b_fallthru_edge;
790 bool c_has_outgoing_fallthru;
791 bool b_has_incoming_fallthru;
792 edge_iterator ei;
794 /* Avoid overactive code motion, as the forwarder blocks should be
795 eliminated by edge redirection instead. One exception might have
796 been if B is a forwarder block and C has no fallthru edge, but
797 that should be cleaned up by bb-reorder instead. */
798 if (FORWARDER_BLOCK_P (b) || FORWARDER_BLOCK_P (c))
799 return NULL;
801 /* We must make sure to not munge nesting of lexical blocks,
802 and loop notes. This is done by squeezing out all the notes
803 and leaving them there to lie. Not ideal, but functional. */
805 FOR_EACH_EDGE (tmp_edge, ei, c->succs)
806 if (tmp_edge->flags & EDGE_FALLTHRU)
807 break;
809 c_has_outgoing_fallthru = (tmp_edge != NULL);
811 FOR_EACH_EDGE (tmp_edge, ei, b->preds)
812 if (tmp_edge->flags & EDGE_FALLTHRU)
813 break;
815 b_has_incoming_fallthru = (tmp_edge != NULL);
816 b_fallthru_edge = tmp_edge;
817 next = b->prev_bb;
818 if (next == c)
819 next = next->prev_bb;
821 /* Otherwise, we're going to try to move C after B. If C does
822 not have an outgoing fallthru, then it can be moved
823 immediately after B without introducing or modifying jumps. */
824 if (! c_has_outgoing_fallthru)
826 merge_blocks_move_successor_nojumps (b, c);
827 return next == ENTRY_BLOCK_PTR ? next->next_bb : next;
830 /* If B does not have an incoming fallthru, then it can be moved
831 immediately before C without introducing or modifying jumps.
832 C cannot be the first block, so we do not have to worry about
833 accessing a non-existent block. */
835 if (b_has_incoming_fallthru)
837 basic_block bb;
839 if (b_fallthru_edge->src == ENTRY_BLOCK_PTR)
840 return NULL;
841 bb = force_nonfallthru (b_fallthru_edge);
842 if (bb)
843 notice_new_block (bb);
846 merge_blocks_move_predecessor_nojumps (b, c);
847 return next == ENTRY_BLOCK_PTR ? next->next_bb : next;
850 return NULL;
854 /* Removes the memory attributes of MEM expression
855 if they are not equal. */
857 void
858 merge_memattrs (rtx x, rtx y)
860 int i;
861 int j;
862 enum rtx_code code;
863 const char *fmt;
865 if (x == y)
866 return;
867 if (x == 0 || y == 0)
868 return;
870 code = GET_CODE (x);
872 if (code != GET_CODE (y))
873 return;
875 if (GET_MODE (x) != GET_MODE (y))
876 return;
878 if (code == MEM && MEM_ATTRS (x) != MEM_ATTRS (y))
880 if (! MEM_ATTRS (x))
881 MEM_ATTRS (y) = 0;
882 else if (! MEM_ATTRS (y))
883 MEM_ATTRS (x) = 0;
884 else
886 rtx mem_size;
888 if (MEM_ALIAS_SET (x) != MEM_ALIAS_SET (y))
890 set_mem_alias_set (x, 0);
891 set_mem_alias_set (y, 0);
894 if (! mem_expr_equal_p (MEM_EXPR (x), MEM_EXPR (y)))
896 set_mem_expr (x, 0);
897 set_mem_expr (y, 0);
898 set_mem_offset (x, 0);
899 set_mem_offset (y, 0);
901 else if (MEM_OFFSET (x) != MEM_OFFSET (y))
903 set_mem_offset (x, 0);
904 set_mem_offset (y, 0);
907 if (!MEM_SIZE (x))
908 mem_size = NULL_RTX;
909 else if (!MEM_SIZE (y))
910 mem_size = NULL_RTX;
911 else
912 mem_size = GEN_INT (MAX (INTVAL (MEM_SIZE (x)),
913 INTVAL (MEM_SIZE (y))));
914 set_mem_size (x, mem_size);
915 set_mem_size (y, mem_size);
917 set_mem_align (x, MIN (MEM_ALIGN (x), MEM_ALIGN (y)));
918 set_mem_align (y, MEM_ALIGN (x));
922 fmt = GET_RTX_FORMAT (code);
923 for (i = GET_RTX_LENGTH (code) - 1; i >= 0; i--)
925 switch (fmt[i])
927 case 'E':
928 /* Two vectors must have the same length. */
929 if (XVECLEN (x, i) != XVECLEN (y, i))
930 return;
932 for (j = 0; j < XVECLEN (x, i); j++)
933 merge_memattrs (XVECEXP (x, i, j), XVECEXP (y, i, j));
935 break;
937 case 'e':
938 merge_memattrs (XEXP (x, i), XEXP (y, i));
941 return;
945 /* Return true if I1 and I2 are equivalent and thus can be crossjumped. */
947 static bool
948 old_insns_match_p (int mode ATTRIBUTE_UNUSED, rtx i1, rtx i2)
950 rtx p1, p2;
952 /* Verify that I1 and I2 are equivalent. */
953 if (GET_CODE (i1) != GET_CODE (i2))
954 return false;
956 /* __builtin_unreachable() may lead to empty blocks (ending with
957 NOTE_INSN_BASIC_BLOCK). They may be crossjumped. */
958 if (NOTE_INSN_BASIC_BLOCK_P (i1) && NOTE_INSN_BASIC_BLOCK_P (i2))
959 return true;
961 p1 = PATTERN (i1);
962 p2 = PATTERN (i2);
964 if (GET_CODE (p1) != GET_CODE (p2))
965 return false;
967 /* If this is a CALL_INSN, compare register usage information.
968 If we don't check this on stack register machines, the two
969 CALL_INSNs might be merged leaving reg-stack.c with mismatching
970 numbers of stack registers in the same basic block.
971 If we don't check this on machines with delay slots, a delay slot may
972 be filled that clobbers a parameter expected by the subroutine.
974 ??? We take the simple route for now and assume that if they're
975 equal, they were constructed identically.
977 Also check for identical exception regions. */
979 if (CALL_P (i1))
981 /* Ensure the same EH region. */
982 rtx n1 = find_reg_note (i1, REG_EH_REGION, 0);
983 rtx n2 = find_reg_note (i2, REG_EH_REGION, 0);
985 if (!n1 && n2)
986 return false;
988 if (n1 && (!n2 || XEXP (n1, 0) != XEXP (n2, 0)))
989 return false;
991 if (!rtx_equal_p (CALL_INSN_FUNCTION_USAGE (i1),
992 CALL_INSN_FUNCTION_USAGE (i2))
993 || SIBLING_CALL_P (i1) != SIBLING_CALL_P (i2))
994 return false;
997 #ifdef STACK_REGS
998 /* If cross_jump_death_matters is not 0, the insn's mode
999 indicates whether or not the insn contains any stack-like
1000 regs. */
1002 if ((mode & CLEANUP_POST_REGSTACK) && stack_regs_mentioned (i1))
1004 /* If register stack conversion has already been done, then
1005 death notes must also be compared before it is certain that
1006 the two instruction streams match. */
1008 rtx note;
1009 HARD_REG_SET i1_regset, i2_regset;
1011 CLEAR_HARD_REG_SET (i1_regset);
1012 CLEAR_HARD_REG_SET (i2_regset);
1014 for (note = REG_NOTES (i1); note; note = XEXP (note, 1))
1015 if (REG_NOTE_KIND (note) == REG_DEAD && STACK_REG_P (XEXP (note, 0)))
1016 SET_HARD_REG_BIT (i1_regset, REGNO (XEXP (note, 0)));
1018 for (note = REG_NOTES (i2); note; note = XEXP (note, 1))
1019 if (REG_NOTE_KIND (note) == REG_DEAD && STACK_REG_P (XEXP (note, 0)))
1020 SET_HARD_REG_BIT (i2_regset, REGNO (XEXP (note, 0)));
1022 if (!hard_reg_set_equal_p (i1_regset, i2_regset))
1023 return false;
1025 #endif
1027 if (reload_completed
1028 ? rtx_renumbered_equal_p (p1, p2) : rtx_equal_p (p1, p2))
1029 return true;
1031 return false;
1034 /* When comparing insns I1 and I2 in flow_find_cross_jump or
1035 flow_find_head_matching_sequence, ensure the notes match. */
1037 static void
1038 merge_notes (rtx i1, rtx i2)
1040 /* If the merged insns have different REG_EQUAL notes, then
1041 remove them. */
1042 rtx equiv1 = find_reg_equal_equiv_note (i1);
1043 rtx equiv2 = find_reg_equal_equiv_note (i2);
1045 if (equiv1 && !equiv2)
1046 remove_note (i1, equiv1);
1047 else if (!equiv1 && equiv2)
1048 remove_note (i2, equiv2);
1049 else if (equiv1 && equiv2
1050 && !rtx_equal_p (XEXP (equiv1, 0), XEXP (equiv2, 0)))
1052 remove_note (i1, equiv1);
1053 remove_note (i2, equiv2);
1057 /* Look through the insns at the end of BB1 and BB2 and find the longest
1058 sequence that are equivalent. Store the first insns for that sequence
1059 in *F1 and *F2 and return the sequence length.
1061 To simplify callers of this function, if the blocks match exactly,
1062 store the head of the blocks in *F1 and *F2. */
1065 flow_find_cross_jump (basic_block bb1, basic_block bb2, rtx *f1, rtx *f2)
1067 rtx i1, i2, last1, last2, afterlast1, afterlast2;
1068 int ninsns = 0;
1070 /* Skip simple jumps at the end of the blocks. Complex jumps still
1071 need to be compared for equivalence, which we'll do below. */
1073 i1 = BB_END (bb1);
1074 last1 = afterlast1 = last2 = afterlast2 = NULL_RTX;
1075 if (onlyjump_p (i1)
1076 || (returnjump_p (i1) && !side_effects_p (PATTERN (i1))))
1078 last1 = i1;
1079 i1 = PREV_INSN (i1);
1082 i2 = BB_END (bb2);
1083 if (onlyjump_p (i2)
1084 || (returnjump_p (i2) && !side_effects_p (PATTERN (i2))))
1086 last2 = i2;
1087 /* Count everything except for unconditional jump as insn. */
1088 if (!simplejump_p (i2) && !returnjump_p (i2) && last1)
1089 ninsns++;
1090 i2 = PREV_INSN (i2);
1093 while (true)
1095 /* Ignore notes. */
1096 while (!NONDEBUG_INSN_P (i1) && i1 != BB_HEAD (bb1))
1097 i1 = PREV_INSN (i1);
1099 while (!NONDEBUG_INSN_P (i2) && i2 != BB_HEAD (bb2))
1100 i2 = PREV_INSN (i2);
1102 if (i1 == BB_HEAD (bb1) || i2 == BB_HEAD (bb2))
1103 break;
1105 if (!old_insns_match_p (0, i1, i2))
1106 break;
1108 merge_memattrs (i1, i2);
1110 /* Don't begin a cross-jump with a NOTE insn. */
1111 if (INSN_P (i1))
1113 merge_notes (i1, i2);
1115 afterlast1 = last1, afterlast2 = last2;
1116 last1 = i1, last2 = i2;
1117 ninsns++;
1120 i1 = PREV_INSN (i1);
1121 i2 = PREV_INSN (i2);
1124 #ifdef HAVE_cc0
1125 /* Don't allow the insn after a compare to be shared by
1126 cross-jumping unless the compare is also shared. */
1127 if (ninsns && reg_mentioned_p (cc0_rtx, last1) && ! sets_cc0_p (last1))
1128 last1 = afterlast1, last2 = afterlast2, ninsns--;
1129 #endif
1131 /* Include preceding notes and labels in the cross-jump. One,
1132 this may bring us to the head of the blocks as requested above.
1133 Two, it keeps line number notes as matched as may be. */
1134 if (ninsns)
1136 while (last1 != BB_HEAD (bb1) && !NONDEBUG_INSN_P (PREV_INSN (last1)))
1137 last1 = PREV_INSN (last1);
1139 if (last1 != BB_HEAD (bb1) && LABEL_P (PREV_INSN (last1)))
1140 last1 = PREV_INSN (last1);
1142 while (last2 != BB_HEAD (bb2) && !NONDEBUG_INSN_P (PREV_INSN (last2)))
1143 last2 = PREV_INSN (last2);
1145 if (last2 != BB_HEAD (bb2) && LABEL_P (PREV_INSN (last2)))
1146 last2 = PREV_INSN (last2);
1148 *f1 = last1;
1149 *f2 = last2;
1152 return ninsns;
1155 /* Like flow_find_cross_jump, except start looking for a matching sequence from
1156 the head of the two blocks. Do not include jumps at the end.
1157 If STOP_AFTER is nonzero, stop after finding that many matching
1158 instructions. */
1161 flow_find_head_matching_sequence (basic_block bb1, basic_block bb2, rtx *f1,
1162 rtx *f2, int stop_after)
1164 rtx i1, i2, last1, last2, beforelast1, beforelast2;
1165 int ninsns = 0;
1166 edge e;
1167 edge_iterator ei;
1168 int nehedges1 = 0, nehedges2 = 0;
1170 FOR_EACH_EDGE (e, ei, bb1->succs)
1171 if (e->flags & EDGE_EH)
1172 nehedges1++;
1173 FOR_EACH_EDGE (e, ei, bb2->succs)
1174 if (e->flags & EDGE_EH)
1175 nehedges2++;
1177 i1 = BB_HEAD (bb1);
1178 i2 = BB_HEAD (bb2);
1179 last1 = beforelast1 = last2 = beforelast2 = NULL_RTX;
1181 while (true)
1184 /* Ignore notes. */
1185 while (!NONDEBUG_INSN_P (i1) && i1 != BB_END (bb1))
1186 i1 = NEXT_INSN (i1);
1188 while (!NONDEBUG_INSN_P (i2) && i2 != BB_END (bb2))
1189 i2 = NEXT_INSN (i2);
1191 if (NOTE_P (i1) || NOTE_P (i2)
1192 || JUMP_P (i1) || JUMP_P (i2))
1193 break;
1195 /* A sanity check to make sure we're not merging insns with different
1196 effects on EH. If only one of them ends a basic block, it shouldn't
1197 have an EH edge; if both end a basic block, there should be the same
1198 number of EH edges. */
1199 if ((i1 == BB_END (bb1) && i2 != BB_END (bb2)
1200 && nehedges1 > 0)
1201 || (i2 == BB_END (bb2) && i1 != BB_END (bb1)
1202 && nehedges2 > 0)
1203 || (i1 == BB_END (bb1) && i2 == BB_END (bb2)
1204 && nehedges1 != nehedges2))
1205 break;
1207 if (!old_insns_match_p (0, i1, i2))
1208 break;
1210 merge_memattrs (i1, i2);
1212 /* Don't begin a cross-jump with a NOTE insn. */
1213 if (INSN_P (i1))
1215 merge_notes (i1, i2);
1217 beforelast1 = last1, beforelast2 = last2;
1218 last1 = i1, last2 = i2;
1219 ninsns++;
1222 if (i1 == BB_END (bb1) || i2 == BB_END (bb2)
1223 || (stop_after > 0 && ninsns == stop_after))
1224 break;
1226 i1 = NEXT_INSN (i1);
1227 i2 = NEXT_INSN (i2);
1230 #ifdef HAVE_cc0
1231 /* Don't allow a compare to be shared by cross-jumping unless the insn
1232 after the compare is also shared. */
1233 if (ninsns && reg_mentioned_p (cc0_rtx, last1) && sets_cc0_p (last1))
1234 last1 = beforelast1, last2 = beforelast2, ninsns--;
1235 #endif
1237 if (ninsns)
1239 *f1 = last1;
1240 *f2 = last2;
1243 return ninsns;
1246 /* Return true iff outgoing edges of BB1 and BB2 match, together with
1247 the branch instruction. This means that if we commonize the control
1248 flow before end of the basic block, the semantic remains unchanged.
1250 We may assume that there exists one edge with a common destination. */
1252 static bool
1253 outgoing_edges_match (int mode, basic_block bb1, basic_block bb2)
1255 int nehedges1 = 0, nehedges2 = 0;
1256 edge fallthru1 = 0, fallthru2 = 0;
1257 edge e1, e2;
1258 edge_iterator ei;
1260 /* If BB1 has only one successor, we may be looking at either an
1261 unconditional jump, or a fake edge to exit. */
1262 if (single_succ_p (bb1)
1263 && (single_succ_edge (bb1)->flags & (EDGE_COMPLEX | EDGE_FAKE)) == 0
1264 && (!JUMP_P (BB_END (bb1)) || simplejump_p (BB_END (bb1))))
1265 return (single_succ_p (bb2)
1266 && (single_succ_edge (bb2)->flags
1267 & (EDGE_COMPLEX | EDGE_FAKE)) == 0
1268 && (!JUMP_P (BB_END (bb2)) || simplejump_p (BB_END (bb2))));
1270 /* Match conditional jumps - this may get tricky when fallthru and branch
1271 edges are crossed. */
1272 if (EDGE_COUNT (bb1->succs) == 2
1273 && any_condjump_p (BB_END (bb1))
1274 && onlyjump_p (BB_END (bb1)))
1276 edge b1, f1, b2, f2;
1277 bool reverse, match;
1278 rtx set1, set2, cond1, cond2;
1279 enum rtx_code code1, code2;
1281 if (EDGE_COUNT (bb2->succs) != 2
1282 || !any_condjump_p (BB_END (bb2))
1283 || !onlyjump_p (BB_END (bb2)))
1284 return false;
1286 b1 = BRANCH_EDGE (bb1);
1287 b2 = BRANCH_EDGE (bb2);
1288 f1 = FALLTHRU_EDGE (bb1);
1289 f2 = FALLTHRU_EDGE (bb2);
1291 /* Get around possible forwarders on fallthru edges. Other cases
1292 should be optimized out already. */
1293 if (FORWARDER_BLOCK_P (f1->dest))
1294 f1 = single_succ_edge (f1->dest);
1296 if (FORWARDER_BLOCK_P (f2->dest))
1297 f2 = single_succ_edge (f2->dest);
1299 /* To simplify use of this function, return false if there are
1300 unneeded forwarder blocks. These will get eliminated later
1301 during cleanup_cfg. */
1302 if (FORWARDER_BLOCK_P (f1->dest)
1303 || FORWARDER_BLOCK_P (f2->dest)
1304 || FORWARDER_BLOCK_P (b1->dest)
1305 || FORWARDER_BLOCK_P (b2->dest))
1306 return false;
1308 if (f1->dest == f2->dest && b1->dest == b2->dest)
1309 reverse = false;
1310 else if (f1->dest == b2->dest && b1->dest == f2->dest)
1311 reverse = true;
1312 else
1313 return false;
1315 set1 = pc_set (BB_END (bb1));
1316 set2 = pc_set (BB_END (bb2));
1317 if ((XEXP (SET_SRC (set1), 1) == pc_rtx)
1318 != (XEXP (SET_SRC (set2), 1) == pc_rtx))
1319 reverse = !reverse;
1321 cond1 = XEXP (SET_SRC (set1), 0);
1322 cond2 = XEXP (SET_SRC (set2), 0);
1323 code1 = GET_CODE (cond1);
1324 if (reverse)
1325 code2 = reversed_comparison_code (cond2, BB_END (bb2));
1326 else
1327 code2 = GET_CODE (cond2);
1329 if (code2 == UNKNOWN)
1330 return false;
1332 /* Verify codes and operands match. */
1333 match = ((code1 == code2
1334 && rtx_renumbered_equal_p (XEXP (cond1, 0), XEXP (cond2, 0))
1335 && rtx_renumbered_equal_p (XEXP (cond1, 1), XEXP (cond2, 1)))
1336 || (code1 == swap_condition (code2)
1337 && rtx_renumbered_equal_p (XEXP (cond1, 1),
1338 XEXP (cond2, 0))
1339 && rtx_renumbered_equal_p (XEXP (cond1, 0),
1340 XEXP (cond2, 1))));
1342 /* If we return true, we will join the blocks. Which means that
1343 we will only have one branch prediction bit to work with. Thus
1344 we require the existing branches to have probabilities that are
1345 roughly similar. */
1346 if (match
1347 && optimize_bb_for_speed_p (bb1)
1348 && optimize_bb_for_speed_p (bb2))
1350 int prob2;
1352 if (b1->dest == b2->dest)
1353 prob2 = b2->probability;
1354 else
1355 /* Do not use f2 probability as f2 may be forwarded. */
1356 prob2 = REG_BR_PROB_BASE - b2->probability;
1358 /* Fail if the difference in probabilities is greater than 50%.
1359 This rules out two well-predicted branches with opposite
1360 outcomes. */
1361 if (abs (b1->probability - prob2) > REG_BR_PROB_BASE / 2)
1363 if (dump_file)
1364 fprintf (dump_file,
1365 "Outcomes of branch in bb %i and %i differ too much (%i %i)\n",
1366 bb1->index, bb2->index, b1->probability, prob2);
1368 return false;
1372 if (dump_file && match)
1373 fprintf (dump_file, "Conditionals in bb %i and %i match.\n",
1374 bb1->index, bb2->index);
1376 return match;
1379 /* Generic case - we are seeing a computed jump, table jump or trapping
1380 instruction. */
1382 /* Check whether there are tablejumps in the end of BB1 and BB2.
1383 Return true if they are identical. */
1385 rtx label1, label2;
1386 rtx table1, table2;
1388 if (tablejump_p (BB_END (bb1), &label1, &table1)
1389 && tablejump_p (BB_END (bb2), &label2, &table2)
1390 && GET_CODE (PATTERN (table1)) == GET_CODE (PATTERN (table2)))
1392 /* The labels should never be the same rtx. If they really are same
1393 the jump tables are same too. So disable crossjumping of blocks BB1
1394 and BB2 because when deleting the common insns in the end of BB1
1395 by delete_basic_block () the jump table would be deleted too. */
1396 /* If LABEL2 is referenced in BB1->END do not do anything
1397 because we would loose information when replacing
1398 LABEL1 by LABEL2 and then LABEL2 by LABEL1 in BB1->END. */
1399 if (label1 != label2 && !rtx_referenced_p (label2, BB_END (bb1)))
1401 /* Set IDENTICAL to true when the tables are identical. */
1402 bool identical = false;
1403 rtx p1, p2;
1405 p1 = PATTERN (table1);
1406 p2 = PATTERN (table2);
1407 if (GET_CODE (p1) == ADDR_VEC && rtx_equal_p (p1, p2))
1409 identical = true;
1411 else if (GET_CODE (p1) == ADDR_DIFF_VEC
1412 && (XVECLEN (p1, 1) == XVECLEN (p2, 1))
1413 && rtx_equal_p (XEXP (p1, 2), XEXP (p2, 2))
1414 && rtx_equal_p (XEXP (p1, 3), XEXP (p2, 3)))
1416 int i;
1418 identical = true;
1419 for (i = XVECLEN (p1, 1) - 1; i >= 0 && identical; i--)
1420 if (!rtx_equal_p (XVECEXP (p1, 1, i), XVECEXP (p2, 1, i)))
1421 identical = false;
1424 if (identical)
1426 replace_label_data rr;
1427 bool match;
1429 /* Temporarily replace references to LABEL1 with LABEL2
1430 in BB1->END so that we could compare the instructions. */
1431 rr.r1 = label1;
1432 rr.r2 = label2;
1433 rr.update_label_nuses = false;
1434 for_each_rtx (&BB_END (bb1), replace_label, &rr);
1436 match = old_insns_match_p (mode, BB_END (bb1), BB_END (bb2));
1437 if (dump_file && match)
1438 fprintf (dump_file,
1439 "Tablejumps in bb %i and %i match.\n",
1440 bb1->index, bb2->index);
1442 /* Set the original label in BB1->END because when deleting
1443 a block whose end is a tablejump, the tablejump referenced
1444 from the instruction is deleted too. */
1445 rr.r1 = label2;
1446 rr.r2 = label1;
1447 for_each_rtx (&BB_END (bb1), replace_label, &rr);
1449 return match;
1452 return false;
1456 /* First ensure that the instructions match. There may be many outgoing
1457 edges so this test is generally cheaper. */
1458 if (!old_insns_match_p (mode, BB_END (bb1), BB_END (bb2)))
1459 return false;
1461 /* Search the outgoing edges, ensure that the counts do match, find possible
1462 fallthru and exception handling edges since these needs more
1463 validation. */
1464 if (EDGE_COUNT (bb1->succs) != EDGE_COUNT (bb2->succs))
1465 return false;
1467 FOR_EACH_EDGE (e1, ei, bb1->succs)
1469 e2 = EDGE_SUCC (bb2, ei.index);
1471 if (e1->flags & EDGE_EH)
1472 nehedges1++;
1474 if (e2->flags & EDGE_EH)
1475 nehedges2++;
1477 if (e1->flags & EDGE_FALLTHRU)
1478 fallthru1 = e1;
1479 if (e2->flags & EDGE_FALLTHRU)
1480 fallthru2 = e2;
1483 /* If number of edges of various types does not match, fail. */
1484 if (nehedges1 != nehedges2
1485 || (fallthru1 != 0) != (fallthru2 != 0))
1486 return false;
1488 /* fallthru edges must be forwarded to the same destination. */
1489 if (fallthru1)
1491 basic_block d1 = (forwarder_block_p (fallthru1->dest)
1492 ? single_succ (fallthru1->dest): fallthru1->dest);
1493 basic_block d2 = (forwarder_block_p (fallthru2->dest)
1494 ? single_succ (fallthru2->dest): fallthru2->dest);
1496 if (d1 != d2)
1497 return false;
1500 /* Ensure the same EH region. */
1502 rtx n1 = find_reg_note (BB_END (bb1), REG_EH_REGION, 0);
1503 rtx n2 = find_reg_note (BB_END (bb2), REG_EH_REGION, 0);
1505 if (!n1 && n2)
1506 return false;
1508 if (n1 && (!n2 || XEXP (n1, 0) != XEXP (n2, 0)))
1509 return false;
1512 /* The same checks as in try_crossjump_to_edge. It is required for RTL
1513 version of sequence abstraction. */
1514 FOR_EACH_EDGE (e1, ei, bb2->succs)
1516 edge e2;
1517 edge_iterator ei;
1518 basic_block d1 = e1->dest;
1520 if (FORWARDER_BLOCK_P (d1))
1521 d1 = EDGE_SUCC (d1, 0)->dest;
1523 FOR_EACH_EDGE (e2, ei, bb1->succs)
1525 basic_block d2 = e2->dest;
1526 if (FORWARDER_BLOCK_P (d2))
1527 d2 = EDGE_SUCC (d2, 0)->dest;
1528 if (d1 == d2)
1529 break;
1532 if (!e2)
1533 return false;
1536 return true;
1539 /* Returns true if BB basic block has a preserve label. */
1541 static bool
1542 block_has_preserve_label (basic_block bb)
1544 return (bb
1545 && block_label (bb)
1546 && LABEL_PRESERVE_P (block_label (bb)));
1549 /* E1 and E2 are edges with the same destination block. Search their
1550 predecessors for common code. If found, redirect control flow from
1551 (maybe the middle of) E1->SRC to (maybe the middle of) E2->SRC. */
1553 static bool
1554 try_crossjump_to_edge (int mode, edge e1, edge e2)
1556 int nmatch;
1557 basic_block src1 = e1->src, src2 = e2->src;
1558 basic_block redirect_to, redirect_from, to_remove;
1559 rtx newpos1, newpos2;
1560 edge s;
1561 edge_iterator ei;
1563 newpos1 = newpos2 = NULL_RTX;
1565 /* If we have partitioned hot/cold basic blocks, it is a bad idea
1566 to try this optimization.
1568 Basic block partitioning may result in some jumps that appear to
1569 be optimizable (or blocks that appear to be mergeable), but which really
1570 must be left untouched (they are required to make it safely across
1571 partition boundaries). See the comments at the top of
1572 bb-reorder.c:partition_hot_cold_basic_blocks for complete details. */
1574 if (flag_reorder_blocks_and_partition && reload_completed)
1575 return false;
1577 /* Search backward through forwarder blocks. We don't need to worry
1578 about multiple entry or chained forwarders, as they will be optimized
1579 away. We do this to look past the unconditional jump following a
1580 conditional jump that is required due to the current CFG shape. */
1581 if (single_pred_p (src1)
1582 && FORWARDER_BLOCK_P (src1))
1583 e1 = single_pred_edge (src1), src1 = e1->src;
1585 if (single_pred_p (src2)
1586 && FORWARDER_BLOCK_P (src2))
1587 e2 = single_pred_edge (src2), src2 = e2->src;
1589 /* Nothing to do if we reach ENTRY, or a common source block. */
1590 if (src1 == ENTRY_BLOCK_PTR || src2 == ENTRY_BLOCK_PTR)
1591 return false;
1592 if (src1 == src2)
1593 return false;
1595 /* Seeing more than 1 forwarder blocks would confuse us later... */
1596 if (FORWARDER_BLOCK_P (e1->dest)
1597 && FORWARDER_BLOCK_P (single_succ (e1->dest)))
1598 return false;
1600 if (FORWARDER_BLOCK_P (e2->dest)
1601 && FORWARDER_BLOCK_P (single_succ (e2->dest)))
1602 return false;
1604 /* Likewise with dead code (possibly newly created by the other optimizations
1605 of cfg_cleanup). */
1606 if (EDGE_COUNT (src1->preds) == 0 || EDGE_COUNT (src2->preds) == 0)
1607 return false;
1609 /* Look for the common insn sequence, part the first ... */
1610 if (!outgoing_edges_match (mode, src1, src2))
1611 return false;
1613 /* ... and part the second. */
1614 nmatch = flow_find_cross_jump (src1, src2, &newpos1, &newpos2);
1616 /* Don't proceed with the crossjump unless we found a sufficient number
1617 of matching instructions or the 'from' block was totally matched
1618 (such that its predecessors will hopefully be redirected and the
1619 block removed). */
1620 if ((nmatch < PARAM_VALUE (PARAM_MIN_CROSSJUMP_INSNS))
1621 && (newpos1 != BB_HEAD (src1)))
1622 return false;
1624 /* Avoid deleting preserve label when redirecting ABNORMAL edges. */
1625 if (block_has_preserve_label (e1->dest)
1626 && (e1->flags & EDGE_ABNORMAL))
1627 return false;
1629 /* Here we know that the insns in the end of SRC1 which are common with SRC2
1630 will be deleted.
1631 If we have tablejumps in the end of SRC1 and SRC2
1632 they have been already compared for equivalence in outgoing_edges_match ()
1633 so replace the references to TABLE1 by references to TABLE2. */
1635 rtx label1, label2;
1636 rtx table1, table2;
1638 if (tablejump_p (BB_END (src1), &label1, &table1)
1639 && tablejump_p (BB_END (src2), &label2, &table2)
1640 && label1 != label2)
1642 replace_label_data rr;
1643 rtx insn;
1645 /* Replace references to LABEL1 with LABEL2. */
1646 rr.r1 = label1;
1647 rr.r2 = label2;
1648 rr.update_label_nuses = true;
1649 for (insn = get_insns (); insn; insn = NEXT_INSN (insn))
1651 /* Do not replace the label in SRC1->END because when deleting
1652 a block whose end is a tablejump, the tablejump referenced
1653 from the instruction is deleted too. */
1654 if (insn != BB_END (src1))
1655 for_each_rtx (&insn, replace_label, &rr);
1660 /* Avoid splitting if possible. We must always split when SRC2 has
1661 EH predecessor edges, or we may end up with basic blocks with both
1662 normal and EH predecessor edges. */
1663 if (newpos2 == BB_HEAD (src2)
1664 && !(EDGE_PRED (src2, 0)->flags & EDGE_EH))
1665 redirect_to = src2;
1666 else
1668 if (newpos2 == BB_HEAD (src2))
1670 /* Skip possible basic block header. */
1671 if (LABEL_P (newpos2))
1672 newpos2 = NEXT_INSN (newpos2);
1673 while (DEBUG_INSN_P (newpos2))
1674 newpos2 = NEXT_INSN (newpos2);
1675 if (NOTE_P (newpos2))
1676 newpos2 = NEXT_INSN (newpos2);
1677 while (DEBUG_INSN_P (newpos2))
1678 newpos2 = NEXT_INSN (newpos2);
1681 if (dump_file)
1682 fprintf (dump_file, "Splitting bb %i before %i insns\n",
1683 src2->index, nmatch);
1684 redirect_to = split_block (src2, PREV_INSN (newpos2))->dest;
1687 if (dump_file)
1688 fprintf (dump_file,
1689 "Cross jumping from bb %i to bb %i; %i common insns\n",
1690 src1->index, src2->index, nmatch);
1692 /* We may have some registers visible through the block. */
1693 df_set_bb_dirty (redirect_to);
1695 /* Recompute the frequencies and counts of outgoing edges. */
1696 FOR_EACH_EDGE (s, ei, redirect_to->succs)
1698 edge s2;
1699 edge_iterator ei;
1700 basic_block d = s->dest;
1702 if (FORWARDER_BLOCK_P (d))
1703 d = single_succ (d);
1705 FOR_EACH_EDGE (s2, ei, src1->succs)
1707 basic_block d2 = s2->dest;
1708 if (FORWARDER_BLOCK_P (d2))
1709 d2 = single_succ (d2);
1710 if (d == d2)
1711 break;
1714 s->count += s2->count;
1716 /* Take care to update possible forwarder blocks. We verified
1717 that there is no more than one in the chain, so we can't run
1718 into infinite loop. */
1719 if (FORWARDER_BLOCK_P (s->dest))
1721 single_succ_edge (s->dest)->count += s2->count;
1722 s->dest->count += s2->count;
1723 s->dest->frequency += EDGE_FREQUENCY (s);
1726 if (FORWARDER_BLOCK_P (s2->dest))
1728 single_succ_edge (s2->dest)->count -= s2->count;
1729 if (single_succ_edge (s2->dest)->count < 0)
1730 single_succ_edge (s2->dest)->count = 0;
1731 s2->dest->count -= s2->count;
1732 s2->dest->frequency -= EDGE_FREQUENCY (s);
1733 if (s2->dest->frequency < 0)
1734 s2->dest->frequency = 0;
1735 if (s2->dest->count < 0)
1736 s2->dest->count = 0;
1739 if (!redirect_to->frequency && !src1->frequency)
1740 s->probability = (s->probability + s2->probability) / 2;
1741 else
1742 s->probability
1743 = ((s->probability * redirect_to->frequency +
1744 s2->probability * src1->frequency)
1745 / (redirect_to->frequency + src1->frequency));
1748 /* Adjust count and frequency for the block. An earlier jump
1749 threading pass may have left the profile in an inconsistent
1750 state (see update_bb_profile_for_threading) so we must be
1751 prepared for overflows. */
1752 redirect_to->count += src1->count;
1753 redirect_to->frequency += src1->frequency;
1754 if (redirect_to->frequency > BB_FREQ_MAX)
1755 redirect_to->frequency = BB_FREQ_MAX;
1756 update_br_prob_note (redirect_to);
1758 /* Edit SRC1 to go to REDIRECT_TO at NEWPOS1. */
1760 /* Skip possible basic block header. */
1761 if (LABEL_P (newpos1))
1762 newpos1 = NEXT_INSN (newpos1);
1764 while (DEBUG_INSN_P (newpos1))
1765 newpos1 = NEXT_INSN (newpos1);
1767 if (NOTE_INSN_BASIC_BLOCK_P (newpos1))
1768 newpos1 = NEXT_INSN (newpos1);
1770 while (DEBUG_INSN_P (newpos1))
1771 newpos1 = NEXT_INSN (newpos1);
1773 redirect_from = split_block (src1, PREV_INSN (newpos1))->src;
1774 to_remove = single_succ (redirect_from);
1776 redirect_edge_and_branch_force (single_succ_edge (redirect_from), redirect_to);
1777 delete_basic_block (to_remove);
1779 update_forwarder_flag (redirect_from);
1780 if (redirect_to != src2)
1781 update_forwarder_flag (src2);
1783 return true;
1786 /* Search the predecessors of BB for common insn sequences. When found,
1787 share code between them by redirecting control flow. Return true if
1788 any changes made. */
1790 static bool
1791 try_crossjump_bb (int mode, basic_block bb)
1793 edge e, e2, fallthru;
1794 bool changed;
1795 unsigned max, ix, ix2;
1796 basic_block ev, ev2;
1797 edge_iterator ei;
1799 /* Nothing to do if there is not at least two incoming edges. */
1800 if (EDGE_COUNT (bb->preds) < 2)
1801 return false;
1803 /* Don't crossjump if this block ends in a computed jump,
1804 unless we are optimizing for size. */
1805 if (optimize_bb_for_size_p (bb)
1806 && bb != EXIT_BLOCK_PTR
1807 && computed_jump_p (BB_END (bb)))
1808 return false;
1810 /* If we are partitioning hot/cold basic blocks, we don't want to
1811 mess up unconditional or indirect jumps that cross between hot
1812 and cold sections.
1814 Basic block partitioning may result in some jumps that appear to
1815 be optimizable (or blocks that appear to be mergeable), but which really
1816 must be left untouched (they are required to make it safely across
1817 partition boundaries). See the comments at the top of
1818 bb-reorder.c:partition_hot_cold_basic_blocks for complete details. */
1820 if (BB_PARTITION (EDGE_PRED (bb, 0)->src) !=
1821 BB_PARTITION (EDGE_PRED (bb, 1)->src)
1822 || (EDGE_PRED (bb, 0)->flags & EDGE_CROSSING))
1823 return false;
1825 /* It is always cheapest to redirect a block that ends in a branch to
1826 a block that falls through into BB, as that adds no branches to the
1827 program. We'll try that combination first. */
1828 fallthru = NULL;
1829 max = PARAM_VALUE (PARAM_MAX_CROSSJUMP_EDGES);
1831 if (EDGE_COUNT (bb->preds) > max)
1832 return false;
1834 FOR_EACH_EDGE (e, ei, bb->preds)
1836 if (e->flags & EDGE_FALLTHRU)
1838 fallthru = e;
1839 break;
1843 changed = false;
1844 for (ix = 0, ev = bb; ix < EDGE_COUNT (ev->preds); )
1846 e = EDGE_PRED (ev, ix);
1847 ix++;
1849 /* As noted above, first try with the fallthru predecessor (or, a
1850 fallthru predecessor if we are in cfglayout mode). */
1851 if (fallthru)
1853 /* Don't combine the fallthru edge into anything else.
1854 If there is a match, we'll do it the other way around. */
1855 if (e == fallthru)
1856 continue;
1857 /* If nothing changed since the last attempt, there is nothing
1858 we can do. */
1859 if (!first_pass
1860 && (!(df_get_bb_dirty (e->src))
1861 && !(df_get_bb_dirty (fallthru->src))))
1862 continue;
1864 if (try_crossjump_to_edge (mode, e, fallthru))
1866 changed = true;
1867 ix = 0;
1868 ev = bb;
1869 continue;
1873 /* Non-obvious work limiting check: Recognize that we're going
1874 to call try_crossjump_bb on every basic block. So if we have
1875 two blocks with lots of outgoing edges (a switch) and they
1876 share lots of common destinations, then we would do the
1877 cross-jump check once for each common destination.
1879 Now, if the blocks actually are cross-jump candidates, then
1880 all of their destinations will be shared. Which means that
1881 we only need check them for cross-jump candidacy once. We
1882 can eliminate redundant checks of crossjump(A,B) by arbitrarily
1883 choosing to do the check from the block for which the edge
1884 in question is the first successor of A. */
1885 if (EDGE_SUCC (e->src, 0) != e)
1886 continue;
1888 for (ix2 = 0, ev2 = bb; ix2 < EDGE_COUNT (ev2->preds); )
1890 e2 = EDGE_PRED (ev2, ix2);
1891 ix2++;
1893 if (e2 == e)
1894 continue;
1896 /* We've already checked the fallthru edge above. */
1897 if (e2 == fallthru)
1898 continue;
1900 /* The "first successor" check above only prevents multiple
1901 checks of crossjump(A,B). In order to prevent redundant
1902 checks of crossjump(B,A), require that A be the block
1903 with the lowest index. */
1904 if (e->src->index > e2->src->index)
1905 continue;
1907 /* If nothing changed since the last attempt, there is nothing
1908 we can do. */
1909 if (!first_pass
1910 && (!(df_get_bb_dirty (e->src))
1911 && !(df_get_bb_dirty (e2->src))))
1912 continue;
1914 if (try_crossjump_to_edge (mode, e, e2))
1916 changed = true;
1917 ev2 = bb;
1918 ix = 0;
1919 break;
1924 if (changed)
1925 crossjumps_occured = true;
1927 return changed;
1930 /* Return true if BB contains just bb note, or bb note followed
1931 by only DEBUG_INSNs. */
1933 static bool
1934 trivially_empty_bb_p (basic_block bb)
1936 rtx insn = BB_END (bb);
1938 while (1)
1940 if (insn == BB_HEAD (bb))
1941 return true;
1942 if (!DEBUG_INSN_P (insn))
1943 return false;
1944 insn = PREV_INSN (insn);
1948 /* Do simple CFG optimizations - basic block merging, simplifying of jump
1949 instructions etc. Return nonzero if changes were made. */
1951 static bool
1952 try_optimize_cfg (int mode)
1954 bool changed_overall = false;
1955 bool changed;
1956 int iterations = 0;
1957 basic_block bb, b, next;
1959 if (mode & (CLEANUP_CROSSJUMP | CLEANUP_THREADING))
1960 clear_bb_flags ();
1962 crossjumps_occured = false;
1964 FOR_EACH_BB (bb)
1965 update_forwarder_flag (bb);
1967 if (! targetm.cannot_modify_jumps_p ())
1969 first_pass = true;
1970 /* Attempt to merge blocks as made possible by edge removal. If
1971 a block has only one successor, and the successor has only
1972 one predecessor, they may be combined. */
1975 changed = false;
1976 iterations++;
1978 if (dump_file)
1979 fprintf (dump_file,
1980 "\n\ntry_optimize_cfg iteration %i\n\n",
1981 iterations);
1983 for (b = ENTRY_BLOCK_PTR->next_bb; b != EXIT_BLOCK_PTR;)
1985 basic_block c;
1986 edge s;
1987 bool changed_here = false;
1989 /* Delete trivially dead basic blocks. This is either
1990 blocks with no predecessors, or empty blocks with no
1991 successors. However if the empty block with no
1992 successors is the successor of the ENTRY_BLOCK, it is
1993 kept. This ensures that the ENTRY_BLOCK will have a
1994 successor which is a precondition for many RTL
1995 passes. Empty blocks may result from expanding
1996 __builtin_unreachable (). */
1997 if (EDGE_COUNT (b->preds) == 0
1998 || (EDGE_COUNT (b->succs) == 0
1999 && trivially_empty_bb_p (b)
2000 && single_succ_edge (ENTRY_BLOCK_PTR)->dest != b))
2002 c = b->prev_bb;
2003 if (EDGE_COUNT (b->preds) > 0)
2005 edge e;
2006 edge_iterator ei;
2008 if (current_ir_type () == IR_RTL_CFGLAYOUT)
2010 if (b->il.rtl->footer
2011 && BARRIER_P (b->il.rtl->footer))
2012 FOR_EACH_EDGE (e, ei, b->preds)
2013 if ((e->flags & EDGE_FALLTHRU)
2014 && e->src->il.rtl->footer == NULL)
2016 if (b->il.rtl->footer)
2018 e->src->il.rtl->footer = b->il.rtl->footer;
2019 b->il.rtl->footer = NULL;
2021 else
2023 start_sequence ();
2024 e->src->il.rtl->footer = emit_barrier ();
2025 end_sequence ();
2029 else
2031 rtx last = get_last_bb_insn (b);
2032 if (last && BARRIER_P (last))
2033 FOR_EACH_EDGE (e, ei, b->preds)
2034 if ((e->flags & EDGE_FALLTHRU))
2035 emit_barrier_after (BB_END (e->src));
2038 delete_basic_block (b);
2039 if (!(mode & CLEANUP_CFGLAYOUT))
2040 changed = true;
2041 /* Avoid trying to remove ENTRY_BLOCK_PTR. */
2042 b = (c == ENTRY_BLOCK_PTR ? c->next_bb : c);
2043 continue;
2046 /* Remove code labels no longer used. */
2047 if (single_pred_p (b)
2048 && (single_pred_edge (b)->flags & EDGE_FALLTHRU)
2049 && !(single_pred_edge (b)->flags & EDGE_COMPLEX)
2050 && LABEL_P (BB_HEAD (b))
2051 /* If the previous block ends with a branch to this
2052 block, we can't delete the label. Normally this
2053 is a condjump that is yet to be simplified, but
2054 if CASE_DROPS_THRU, this can be a tablejump with
2055 some element going to the same place as the
2056 default (fallthru). */
2057 && (single_pred (b) == ENTRY_BLOCK_PTR
2058 || !JUMP_P (BB_END (single_pred (b)))
2059 || ! label_is_jump_target_p (BB_HEAD (b),
2060 BB_END (single_pred (b)))))
2062 rtx label = BB_HEAD (b);
2064 delete_insn_chain (label, label, false);
2065 /* If the case label is undeletable, move it after the
2066 BASIC_BLOCK note. */
2067 if (NOTE_KIND (BB_HEAD (b)) == NOTE_INSN_DELETED_LABEL)
2069 rtx bb_note = NEXT_INSN (BB_HEAD (b));
2071 reorder_insns_nobb (label, label, bb_note);
2072 BB_HEAD (b) = bb_note;
2073 if (BB_END (b) == bb_note)
2074 BB_END (b) = label;
2076 if (dump_file)
2077 fprintf (dump_file, "Deleted label in block %i.\n",
2078 b->index);
2081 /* If we fall through an empty block, we can remove it. */
2082 if (!(mode & CLEANUP_CFGLAYOUT)
2083 && single_pred_p (b)
2084 && (single_pred_edge (b)->flags & EDGE_FALLTHRU)
2085 && !LABEL_P (BB_HEAD (b))
2086 && FORWARDER_BLOCK_P (b)
2087 /* Note that forwarder_block_p true ensures that
2088 there is a successor for this block. */
2089 && (single_succ_edge (b)->flags & EDGE_FALLTHRU)
2090 && n_basic_blocks > NUM_FIXED_BLOCKS + 1)
2092 if (dump_file)
2093 fprintf (dump_file,
2094 "Deleting fallthru block %i.\n",
2095 b->index);
2097 c = b->prev_bb == ENTRY_BLOCK_PTR ? b->next_bb : b->prev_bb;
2098 redirect_edge_succ_nodup (single_pred_edge (b),
2099 single_succ (b));
2100 delete_basic_block (b);
2101 changed = true;
2102 b = c;
2103 continue;
2106 if (single_succ_p (b)
2107 && (s = single_succ_edge (b))
2108 && !(s->flags & EDGE_COMPLEX)
2109 && (c = s->dest) != EXIT_BLOCK_PTR
2110 && single_pred_p (c)
2111 && b != c)
2113 /* When not in cfg_layout mode use code aware of reordering
2114 INSN. This code possibly creates new basic blocks so it
2115 does not fit merge_blocks interface and is kept here in
2116 hope that it will become useless once more of compiler
2117 is transformed to use cfg_layout mode. */
2119 if ((mode & CLEANUP_CFGLAYOUT)
2120 && can_merge_blocks_p (b, c))
2122 merge_blocks (b, c);
2123 update_forwarder_flag (b);
2124 changed_here = true;
2126 else if (!(mode & CLEANUP_CFGLAYOUT)
2127 /* If the jump insn has side effects,
2128 we can't kill the edge. */
2129 && (!JUMP_P (BB_END (b))
2130 || (reload_completed
2131 ? simplejump_p (BB_END (b))
2132 : (onlyjump_p (BB_END (b))
2133 && !tablejump_p (BB_END (b),
2134 NULL, NULL))))
2135 && (next = merge_blocks_move (s, b, c, mode)))
2137 b = next;
2138 changed_here = true;
2142 /* Simplify branch over branch. */
2143 if ((mode & CLEANUP_EXPENSIVE)
2144 && !(mode & CLEANUP_CFGLAYOUT)
2145 && try_simplify_condjump (b))
2146 changed_here = true;
2148 /* If B has a single outgoing edge, but uses a
2149 non-trivial jump instruction without side-effects, we
2150 can either delete the jump entirely, or replace it
2151 with a simple unconditional jump. */
2152 if (single_succ_p (b)
2153 && single_succ (b) != EXIT_BLOCK_PTR
2154 && onlyjump_p (BB_END (b))
2155 && !find_reg_note (BB_END (b), REG_CROSSING_JUMP, NULL_RTX)
2156 && try_redirect_by_replacing_jump (single_succ_edge (b),
2157 single_succ (b),
2158 (mode & CLEANUP_CFGLAYOUT) != 0))
2160 update_forwarder_flag (b);
2161 changed_here = true;
2164 /* Simplify branch to branch. */
2165 if (try_forward_edges (mode, b))
2166 changed_here = true;
2168 /* Look for shared code between blocks. */
2169 if ((mode & CLEANUP_CROSSJUMP)
2170 && try_crossjump_bb (mode, b))
2171 changed_here = true;
2173 /* Don't get confused by the index shift caused by
2174 deleting blocks. */
2175 if (!changed_here)
2176 b = b->next_bb;
2177 else
2178 changed = true;
2181 if ((mode & CLEANUP_CROSSJUMP)
2182 && try_crossjump_bb (mode, EXIT_BLOCK_PTR))
2183 changed = true;
2185 #ifdef ENABLE_CHECKING
2186 if (changed)
2187 verify_flow_info ();
2188 #endif
2190 changed_overall |= changed;
2191 first_pass = false;
2193 while (changed);
2196 FOR_ALL_BB (b)
2197 b->flags &= ~(BB_FORWARDER_BLOCK | BB_NONTHREADABLE_BLOCK);
2199 return changed_overall;
2202 /* Delete all unreachable basic blocks. */
2204 bool
2205 delete_unreachable_blocks (void)
2207 bool changed = false;
2208 basic_block b, prev_bb;
2210 find_unreachable_blocks ();
2212 /* When we're in GIMPLE mode and there may be debug insns, we should
2213 delete blocks in reverse dominator order, so as to get a chance
2214 to substitute all released DEFs into debug stmts. If we don't
2215 have dominators information, walking blocks backward gets us a
2216 better chance of retaining most debug information than
2217 otherwise. */
2218 if (MAY_HAVE_DEBUG_STMTS && current_ir_type () == IR_GIMPLE
2219 && dom_info_available_p (CDI_DOMINATORS))
2221 for (b = EXIT_BLOCK_PTR->prev_bb; b != ENTRY_BLOCK_PTR; b = prev_bb)
2223 prev_bb = b->prev_bb;
2225 if (!(b->flags & BB_REACHABLE))
2227 /* Speed up the removal of blocks that don't dominate
2228 others. Walking backwards, this should be the common
2229 case. */
2230 if (!first_dom_son (CDI_DOMINATORS, b))
2231 delete_basic_block (b);
2232 else
2234 VEC (basic_block, heap) *h
2235 = get_all_dominated_blocks (CDI_DOMINATORS, b);
2237 while (VEC_length (basic_block, h))
2239 b = VEC_pop (basic_block, h);
2241 prev_bb = b->prev_bb;
2243 gcc_assert (!(b->flags & BB_REACHABLE));
2245 delete_basic_block (b);
2248 VEC_free (basic_block, heap, h);
2251 changed = true;
2255 else
2257 for (b = EXIT_BLOCK_PTR->prev_bb; b != ENTRY_BLOCK_PTR; b = prev_bb)
2259 prev_bb = b->prev_bb;
2261 if (!(b->flags & BB_REACHABLE))
2263 delete_basic_block (b);
2264 changed = true;
2269 if (changed)
2270 tidy_fallthru_edges ();
2271 return changed;
2274 /* Delete any jump tables never referenced. We can't delete them at the
2275 time of removing tablejump insn as they are referenced by the preceding
2276 insns computing the destination, so we delay deleting and garbagecollect
2277 them once life information is computed. */
2278 void
2279 delete_dead_jumptables (void)
2281 basic_block bb;
2283 /* A dead jump table does not belong to any basic block. Scan insns
2284 between two adjacent basic blocks. */
2285 FOR_EACH_BB (bb)
2287 rtx insn, next;
2289 for (insn = NEXT_INSN (BB_END (bb));
2290 insn && !NOTE_INSN_BASIC_BLOCK_P (insn);
2291 insn = next)
2293 next = NEXT_INSN (insn);
2294 if (LABEL_P (insn)
2295 && LABEL_NUSES (insn) == LABEL_PRESERVE_P (insn)
2296 && JUMP_TABLE_DATA_P (next))
2298 rtx label = insn, jump = next;
2300 if (dump_file)
2301 fprintf (dump_file, "Dead jumptable %i removed\n",
2302 INSN_UID (insn));
2304 next = NEXT_INSN (next);
2305 delete_insn (jump);
2306 delete_insn (label);
2313 /* Tidy the CFG by deleting unreachable code and whatnot. */
2315 bool
2316 cleanup_cfg (int mode)
2318 bool changed = false;
2320 /* Set the cfglayout mode flag here. We could update all the callers
2321 but that is just inconvenient, especially given that we eventually
2322 want to have cfglayout mode as the default. */
2323 if (current_ir_type () == IR_RTL_CFGLAYOUT)
2324 mode |= CLEANUP_CFGLAYOUT;
2326 timevar_push (TV_CLEANUP_CFG);
2327 if (delete_unreachable_blocks ())
2329 changed = true;
2330 /* We've possibly created trivially dead code. Cleanup it right
2331 now to introduce more opportunities for try_optimize_cfg. */
2332 if (!(mode & (CLEANUP_NO_INSN_DEL))
2333 && !reload_completed)
2334 delete_trivially_dead_insns (get_insns (), max_reg_num ());
2337 compact_blocks ();
2339 /* To tail-merge blocks ending in the same noreturn function (e.g.
2340 a call to abort) we have to insert fake edges to exit. Do this
2341 here once. The fake edges do not interfere with any other CFG
2342 cleanups. */
2343 if (mode & CLEANUP_CROSSJUMP)
2344 add_noreturn_fake_exit_edges ();
2346 if (!dbg_cnt (cfg_cleanup))
2347 return changed;
2349 while (try_optimize_cfg (mode))
2351 delete_unreachable_blocks (), changed = true;
2352 if (!(mode & CLEANUP_NO_INSN_DEL))
2354 /* Try to remove some trivially dead insns when doing an expensive
2355 cleanup. But delete_trivially_dead_insns doesn't work after
2356 reload (it only handles pseudos) and run_fast_dce is too costly
2357 to run in every iteration.
2359 For effective cross jumping, we really want to run a fast DCE to
2360 clean up any dead conditions, or they get in the way of performing
2361 useful tail merges.
2363 Other transformations in cleanup_cfg are not so sensitive to dead
2364 code, so delete_trivially_dead_insns or even doing nothing at all
2365 is good enough. */
2366 if ((mode & CLEANUP_EXPENSIVE) && !reload_completed
2367 && !delete_trivially_dead_insns (get_insns (), max_reg_num ()))
2368 break;
2369 else if ((mode & CLEANUP_CROSSJUMP)
2370 && crossjumps_occured)
2371 run_fast_dce ();
2373 else
2374 break;
2377 if (mode & CLEANUP_CROSSJUMP)
2378 remove_fake_exit_edges ();
2380 /* Don't call delete_dead_jumptables in cfglayout mode, because
2381 that function assumes that jump tables are in the insns stream.
2382 But we also don't _have_ to delete dead jumptables in cfglayout
2383 mode because we shouldn't even be looking at things that are
2384 not in a basic block. Dead jumptables are cleaned up when
2385 going out of cfglayout mode. */
2386 if (!(mode & CLEANUP_CFGLAYOUT))
2387 delete_dead_jumptables ();
2389 timevar_pop (TV_CLEANUP_CFG);
2391 return changed;
2394 static unsigned int
2395 rest_of_handle_jump (void)
2397 if (crtl->tail_call_emit)
2398 fixup_tail_calls ();
2399 return 0;
2402 struct rtl_opt_pass pass_jump =
2405 RTL_PASS,
2406 "sibling", /* name */
2407 NULL, /* gate */
2408 rest_of_handle_jump, /* execute */
2409 NULL, /* sub */
2410 NULL, /* next */
2411 0, /* static_pass_number */
2412 TV_JUMP, /* tv_id */
2413 0, /* properties_required */
2414 0, /* properties_provided */
2415 0, /* properties_destroyed */
2416 TODO_ggc_collect, /* todo_flags_start */
2417 TODO_verify_flow, /* todo_flags_finish */
2422 static unsigned int
2423 rest_of_handle_jump2 (void)
2425 delete_trivially_dead_insns (get_insns (), max_reg_num ());
2426 if (dump_file)
2427 dump_flow_info (dump_file, dump_flags);
2428 cleanup_cfg ((optimize ? CLEANUP_EXPENSIVE : 0)
2429 | (flag_thread_jumps ? CLEANUP_THREADING : 0));
2430 return 0;
2434 struct rtl_opt_pass pass_jump2 =
2437 RTL_PASS,
2438 "jump", /* name */
2439 NULL, /* gate */
2440 rest_of_handle_jump2, /* execute */
2441 NULL, /* sub */
2442 NULL, /* next */
2443 0, /* static_pass_number */
2444 TV_JUMP, /* tv_id */
2445 0, /* properties_required */
2446 0, /* properties_provided */
2447 0, /* properties_destroyed */
2448 TODO_ggc_collect, /* todo_flags_start */
2449 TODO_dump_func | TODO_verify_rtl_sharing,/* todo_flags_finish */