]> git.ipfire.org Git - thirdparty/gcc.git/blob - gcc/sel-sched-ir.c
Merge in trunk.
[thirdparty/gcc.git] / gcc / sel-sched-ir.c
1 /* Instruction scheduling pass. Selective scheduler and pipeliner.
2 Copyright (C) 2006-2013 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 #include "config.h"
21 #include "system.h"
22 #include "coretypes.h"
23 #include "tm.h"
24 #include "diagnostic-core.h"
25 #include "rtl.h"
26 #include "tm_p.h"
27 #include "hard-reg-set.h"
28 #include "regs.h"
29 #include "function.h"
30 #include "flags.h"
31 #include "insn-config.h"
32 #include "insn-attr.h"
33 #include "except.h"
34 #include "recog.h"
35 #include "params.h"
36 #include "target.h"
37 #include "sched-int.h"
38 #include "ggc.h"
39 #include "tree.h"
40 #include "vec.h"
41 #include "langhooks.h"
42 #include "rtlhooks-def.h"
43 #include "emit-rtl.h" /* FIXME: Can go away once crtl is moved to rtl.h. */
44
45 #ifdef INSN_SCHEDULING
46 #include "sel-sched-ir.h"
47 /* We don't have to use it except for sel_print_insn. */
48 #include "sel-sched-dump.h"
49
50 /* A vector holding bb info for whole scheduling pass. */
51 vec<sel_global_bb_info_def>
52 sel_global_bb_info = vNULL;
53
54 /* A vector holding bb info. */
55 vec<sel_region_bb_info_def>
56 sel_region_bb_info = vNULL;
57
58 /* A pool for allocating all lists. */
59 alloc_pool sched_lists_pool;
60
61 /* This contains information about successors for compute_av_set. */
62 struct succs_info current_succs;
63
64 /* Data structure to describe interaction with the generic scheduler utils. */
65 static struct common_sched_info_def sel_common_sched_info;
66
67 /* The loop nest being pipelined. */
68 struct loop *current_loop_nest;
69
70 /* LOOP_NESTS is a vector containing the corresponding loop nest for
71 each region. */
72 static vec<loop_p> loop_nests = vNULL;
73
74 /* Saves blocks already in loop regions, indexed by bb->index. */
75 static sbitmap bbs_in_loop_rgns = NULL;
76
77 /* CFG hooks that are saved before changing create_basic_block hook. */
78 static struct cfg_hooks orig_cfg_hooks;
79 \f
80
81 /* Array containing reverse topological index of function basic blocks,
82 indexed by BB->INDEX. */
83 static int *rev_top_order_index = NULL;
84
85 /* Length of the above array. */
86 static int rev_top_order_index_len = -1;
87
88 /* A regset pool structure. */
89 static struct
90 {
91 /* The stack to which regsets are returned. */
92 regset *v;
93
94 /* Its pointer. */
95 int n;
96
97 /* Its size. */
98 int s;
99
100 /* In VV we save all generated regsets so that, when destructing the
101 pool, we can compare it with V and check that every regset was returned
102 back to pool. */
103 regset *vv;
104
105 /* The pointer of VV stack. */
106 int nn;
107
108 /* Its size. */
109 int ss;
110
111 /* The difference between allocated and returned regsets. */
112 int diff;
113 } regset_pool = { NULL, 0, 0, NULL, 0, 0, 0 };
114
115 /* This represents the nop pool. */
116 static struct
117 {
118 /* The vector which holds previously emitted nops. */
119 insn_t *v;
120
121 /* Its pointer. */
122 int n;
123
124 /* Its size. */
125 int s;
126 } nop_pool = { NULL, 0, 0 };
127
128 /* The pool for basic block notes. */
129 static rtx_vec_t bb_note_pool;
130
131 /* A NOP pattern used to emit placeholder insns. */
132 rtx nop_pattern = NULL_RTX;
133 /* A special instruction that resides in EXIT_BLOCK.
134 EXIT_INSN is successor of the insns that lead to EXIT_BLOCK. */
135 rtx exit_insn = NULL_RTX;
136
137 /* TRUE if while scheduling current region, which is loop, its preheader
138 was removed. */
139 bool preheader_removed = false;
140 \f
141
142 /* Forward static declarations. */
143 static void fence_clear (fence_t);
144
145 static void deps_init_id (idata_t, insn_t, bool);
146 static void init_id_from_df (idata_t, insn_t, bool);
147 static expr_t set_insn_init (expr_t, vinsn_t, int);
148
149 static void cfg_preds (basic_block, insn_t **, int *);
150 static void prepare_insn_expr (insn_t, int);
151 static void free_history_vect (vec<expr_history_def> &);
152
153 static void move_bb_info (basic_block, basic_block);
154 static void remove_empty_bb (basic_block, bool);
155 static void sel_merge_blocks (basic_block, basic_block);
156 static void sel_remove_loop_preheader (void);
157 static bool bb_has_removable_jump_to_p (basic_block, basic_block);
158
159 static bool insn_is_the_only_one_in_bb_p (insn_t);
160 static void create_initial_data_sets (basic_block);
161
162 static void free_av_set (basic_block);
163 static void invalidate_av_set (basic_block);
164 static void extend_insn_data (void);
165 static void sel_init_new_insn (insn_t, int);
166 static void finish_insns (void);
167 \f
168 /* Various list functions. */
169
170 /* Copy an instruction list L. */
171 ilist_t
172 ilist_copy (ilist_t l)
173 {
174 ilist_t head = NULL, *tailp = &head;
175
176 while (l)
177 {
178 ilist_add (tailp, ILIST_INSN (l));
179 tailp = &ILIST_NEXT (*tailp);
180 l = ILIST_NEXT (l);
181 }
182
183 return head;
184 }
185
186 /* Invert an instruction list L. */
187 ilist_t
188 ilist_invert (ilist_t l)
189 {
190 ilist_t res = NULL;
191
192 while (l)
193 {
194 ilist_add (&res, ILIST_INSN (l));
195 l = ILIST_NEXT (l);
196 }
197
198 return res;
199 }
200
201 /* Add a new boundary to the LP list with parameters TO, PTR, and DC. */
202 void
203 blist_add (blist_t *lp, insn_t to, ilist_t ptr, deps_t dc)
204 {
205 bnd_t bnd;
206
207 _list_add (lp);
208 bnd = BLIST_BND (*lp);
209
210 BND_TO (bnd) = to;
211 BND_PTR (bnd) = ptr;
212 BND_AV (bnd) = NULL;
213 BND_AV1 (bnd) = NULL;
214 BND_DC (bnd) = dc;
215 }
216
217 /* Remove the list note pointed to by LP. */
218 void
219 blist_remove (blist_t *lp)
220 {
221 bnd_t b = BLIST_BND (*lp);
222
223 av_set_clear (&BND_AV (b));
224 av_set_clear (&BND_AV1 (b));
225 ilist_clear (&BND_PTR (b));
226
227 _list_remove (lp);
228 }
229
230 /* Init a fence tail L. */
231 void
232 flist_tail_init (flist_tail_t l)
233 {
234 FLIST_TAIL_HEAD (l) = NULL;
235 FLIST_TAIL_TAILP (l) = &FLIST_TAIL_HEAD (l);
236 }
237
238 /* Try to find fence corresponding to INSN in L. */
239 fence_t
240 flist_lookup (flist_t l, insn_t insn)
241 {
242 while (l)
243 {
244 if (FENCE_INSN (FLIST_FENCE (l)) == insn)
245 return FLIST_FENCE (l);
246
247 l = FLIST_NEXT (l);
248 }
249
250 return NULL;
251 }
252
253 /* Init the fields of F before running fill_insns. */
254 static void
255 init_fence_for_scheduling (fence_t f)
256 {
257 FENCE_BNDS (f) = NULL;
258 FENCE_PROCESSED_P (f) = false;
259 FENCE_SCHEDULED_P (f) = false;
260 }
261
262 /* Add new fence consisting of INSN and STATE to the list pointed to by LP. */
263 static void
264 flist_add (flist_t *lp, insn_t insn, state_t state, deps_t dc, void *tc,
265 insn_t last_scheduled_insn, vec<rtx, va_gc> *executing_insns,
266 int *ready_ticks, int ready_ticks_size, insn_t sched_next,
267 int cycle, int cycle_issued_insns, int issue_more,
268 bool starts_cycle_p, bool after_stall_p)
269 {
270 fence_t f;
271
272 _list_add (lp);
273 f = FLIST_FENCE (*lp);
274
275 FENCE_INSN (f) = insn;
276
277 gcc_assert (state != NULL);
278 FENCE_STATE (f) = state;
279
280 FENCE_CYCLE (f) = cycle;
281 FENCE_ISSUED_INSNS (f) = cycle_issued_insns;
282 FENCE_STARTS_CYCLE_P (f) = starts_cycle_p;
283 FENCE_AFTER_STALL_P (f) = after_stall_p;
284
285 gcc_assert (dc != NULL);
286 FENCE_DC (f) = dc;
287
288 gcc_assert (tc != NULL || targetm.sched.alloc_sched_context == NULL);
289 FENCE_TC (f) = tc;
290
291 FENCE_LAST_SCHEDULED_INSN (f) = last_scheduled_insn;
292 FENCE_ISSUE_MORE (f) = issue_more;
293 FENCE_EXECUTING_INSNS (f) = executing_insns;
294 FENCE_READY_TICKS (f) = ready_ticks;
295 FENCE_READY_TICKS_SIZE (f) = ready_ticks_size;
296 FENCE_SCHED_NEXT (f) = sched_next;
297
298 init_fence_for_scheduling (f);
299 }
300
301 /* Remove the head node of the list pointed to by LP. */
302 static void
303 flist_remove (flist_t *lp)
304 {
305 if (FENCE_INSN (FLIST_FENCE (*lp)))
306 fence_clear (FLIST_FENCE (*lp));
307 _list_remove (lp);
308 }
309
310 /* Clear the fence list pointed to by LP. */
311 void
312 flist_clear (flist_t *lp)
313 {
314 while (*lp)
315 flist_remove (lp);
316 }
317
318 /* Add ORIGINAL_INSN the def list DL honoring CROSSES_CALL. */
319 void
320 def_list_add (def_list_t *dl, insn_t original_insn, bool crosses_call)
321 {
322 def_t d;
323
324 _list_add (dl);
325 d = DEF_LIST_DEF (*dl);
326
327 d->orig_insn = original_insn;
328 d->crosses_call = crosses_call;
329 }
330 \f
331
332 /* Functions to work with target contexts. */
333
334 /* Bulk target context. It is convenient for debugging purposes to ensure
335 that there are no uninitialized (null) target contexts. */
336 static tc_t bulk_tc = (tc_t) 1;
337
338 /* Target hooks wrappers. In the future we can provide some default
339 implementations for them. */
340
341 /* Allocate a store for the target context. */
342 static tc_t
343 alloc_target_context (void)
344 {
345 return (targetm.sched.alloc_sched_context
346 ? targetm.sched.alloc_sched_context () : bulk_tc);
347 }
348
349 /* Init target context TC.
350 If CLEAN_P is true, then make TC as it is beginning of the scheduler.
351 Overwise, copy current backend context to TC. */
352 static void
353 init_target_context (tc_t tc, bool clean_p)
354 {
355 if (targetm.sched.init_sched_context)
356 targetm.sched.init_sched_context (tc, clean_p);
357 }
358
359 /* Allocate and initialize a target context. Meaning of CLEAN_P is the same as
360 int init_target_context (). */
361 tc_t
362 create_target_context (bool clean_p)
363 {
364 tc_t tc = alloc_target_context ();
365
366 init_target_context (tc, clean_p);
367 return tc;
368 }
369
370 /* Copy TC to the current backend context. */
371 void
372 set_target_context (tc_t tc)
373 {
374 if (targetm.sched.set_sched_context)
375 targetm.sched.set_sched_context (tc);
376 }
377
378 /* TC is about to be destroyed. Free any internal data. */
379 static void
380 clear_target_context (tc_t tc)
381 {
382 if (targetm.sched.clear_sched_context)
383 targetm.sched.clear_sched_context (tc);
384 }
385
386 /* Clear and free it. */
387 static void
388 delete_target_context (tc_t tc)
389 {
390 clear_target_context (tc);
391
392 if (targetm.sched.free_sched_context)
393 targetm.sched.free_sched_context (tc);
394 }
395
396 /* Make a copy of FROM in TO.
397 NB: May be this should be a hook. */
398 static void
399 copy_target_context (tc_t to, tc_t from)
400 {
401 tc_t tmp = create_target_context (false);
402
403 set_target_context (from);
404 init_target_context (to, false);
405
406 set_target_context (tmp);
407 delete_target_context (tmp);
408 }
409
410 /* Create a copy of TC. */
411 static tc_t
412 create_copy_of_target_context (tc_t tc)
413 {
414 tc_t copy = alloc_target_context ();
415
416 copy_target_context (copy, tc);
417
418 return copy;
419 }
420
421 /* Clear TC and initialize it according to CLEAN_P. The meaning of CLEAN_P
422 is the same as in init_target_context (). */
423 void
424 reset_target_context (tc_t tc, bool clean_p)
425 {
426 clear_target_context (tc);
427 init_target_context (tc, clean_p);
428 }
429 \f
430 /* Functions to work with dependence contexts.
431 Dc (aka deps context, aka deps_t, aka struct deps_desc *) is short for dependence
432 context. It accumulates information about processed insns to decide if
433 current insn is dependent on the processed ones. */
434
435 /* Make a copy of FROM in TO. */
436 static void
437 copy_deps_context (deps_t to, deps_t from)
438 {
439 init_deps (to, false);
440 deps_join (to, from);
441 }
442
443 /* Allocate store for dep context. */
444 static deps_t
445 alloc_deps_context (void)
446 {
447 return XNEW (struct deps_desc);
448 }
449
450 /* Allocate and initialize dep context. */
451 static deps_t
452 create_deps_context (void)
453 {
454 deps_t dc = alloc_deps_context ();
455
456 init_deps (dc, false);
457 return dc;
458 }
459
460 /* Create a copy of FROM. */
461 static deps_t
462 create_copy_of_deps_context (deps_t from)
463 {
464 deps_t to = alloc_deps_context ();
465
466 copy_deps_context (to, from);
467 return to;
468 }
469
470 /* Clean up internal data of DC. */
471 static void
472 clear_deps_context (deps_t dc)
473 {
474 free_deps (dc);
475 }
476
477 /* Clear and free DC. */
478 static void
479 delete_deps_context (deps_t dc)
480 {
481 clear_deps_context (dc);
482 free (dc);
483 }
484
485 /* Clear and init DC. */
486 static void
487 reset_deps_context (deps_t dc)
488 {
489 clear_deps_context (dc);
490 init_deps (dc, false);
491 }
492
493 /* This structure describes the dependence analysis hooks for advancing
494 dependence context. */
495 static struct sched_deps_info_def advance_deps_context_sched_deps_info =
496 {
497 NULL,
498
499 NULL, /* start_insn */
500 NULL, /* finish_insn */
501 NULL, /* start_lhs */
502 NULL, /* finish_lhs */
503 NULL, /* start_rhs */
504 NULL, /* finish_rhs */
505 haifa_note_reg_set,
506 haifa_note_reg_clobber,
507 haifa_note_reg_use,
508 NULL, /* note_mem_dep */
509 NULL, /* note_dep */
510
511 0, 0, 0
512 };
513
514 /* Process INSN and add its impact on DC. */
515 void
516 advance_deps_context (deps_t dc, insn_t insn)
517 {
518 sched_deps_info = &advance_deps_context_sched_deps_info;
519 deps_analyze_insn (dc, insn);
520 }
521 \f
522
523 /* Functions to work with DFA states. */
524
525 /* Allocate store for a DFA state. */
526 static state_t
527 state_alloc (void)
528 {
529 return xmalloc (dfa_state_size);
530 }
531
532 /* Allocate and initialize DFA state. */
533 static state_t
534 state_create (void)
535 {
536 state_t state = state_alloc ();
537
538 state_reset (state);
539 advance_state (state);
540 return state;
541 }
542
543 /* Free DFA state. */
544 static void
545 state_free (state_t state)
546 {
547 free (state);
548 }
549
550 /* Make a copy of FROM in TO. */
551 static void
552 state_copy (state_t to, state_t from)
553 {
554 memcpy (to, from, dfa_state_size);
555 }
556
557 /* Create a copy of FROM. */
558 static state_t
559 state_create_copy (state_t from)
560 {
561 state_t to = state_alloc ();
562
563 state_copy (to, from);
564 return to;
565 }
566 \f
567
568 /* Functions to work with fences. */
569
570 /* Clear the fence. */
571 static void
572 fence_clear (fence_t f)
573 {
574 state_t s = FENCE_STATE (f);
575 deps_t dc = FENCE_DC (f);
576 void *tc = FENCE_TC (f);
577
578 ilist_clear (&FENCE_BNDS (f));
579
580 gcc_assert ((s != NULL && dc != NULL && tc != NULL)
581 || (s == NULL && dc == NULL && tc == NULL));
582
583 free (s);
584
585 if (dc != NULL)
586 delete_deps_context (dc);
587
588 if (tc != NULL)
589 delete_target_context (tc);
590 vec_free (FENCE_EXECUTING_INSNS (f));
591 free (FENCE_READY_TICKS (f));
592 FENCE_READY_TICKS (f) = NULL;
593 }
594
595 /* Init a list of fences with successors of OLD_FENCE. */
596 void
597 init_fences (insn_t old_fence)
598 {
599 insn_t succ;
600 succ_iterator si;
601 bool first = true;
602 int ready_ticks_size = get_max_uid () + 1;
603
604 FOR_EACH_SUCC_1 (succ, si, old_fence,
605 SUCCS_NORMAL | SUCCS_SKIP_TO_LOOP_EXITS)
606 {
607
608 if (first)
609 first = false;
610 else
611 gcc_assert (flag_sel_sched_pipelining_outer_loops);
612
613 flist_add (&fences, succ,
614 state_create (),
615 create_deps_context () /* dc */,
616 create_target_context (true) /* tc */,
617 NULL_RTX /* last_scheduled_insn */,
618 NULL, /* executing_insns */
619 XCNEWVEC (int, ready_ticks_size), /* ready_ticks */
620 ready_ticks_size,
621 NULL_RTX /* sched_next */,
622 1 /* cycle */, 0 /* cycle_issued_insns */,
623 issue_rate, /* issue_more */
624 1 /* starts_cycle_p */, 0 /* after_stall_p */);
625 }
626 }
627
628 /* Merges two fences (filling fields of fence F with resulting values) by
629 following rules: 1) state, target context and last scheduled insn are
630 propagated from fallthrough edge if it is available;
631 2) deps context and cycle is propagated from more probable edge;
632 3) all other fields are set to corresponding constant values.
633
634 INSN, STATE, DC, TC, LAST_SCHEDULED_INSN, EXECUTING_INSNS,
635 READY_TICKS, READY_TICKS_SIZE, SCHED_NEXT, CYCLE, ISSUE_MORE
636 and AFTER_STALL_P are the corresponding fields of the second fence. */
637 static void
638 merge_fences (fence_t f, insn_t insn,
639 state_t state, deps_t dc, void *tc,
640 rtx last_scheduled_insn, vec<rtx, va_gc> *executing_insns,
641 int *ready_ticks, int ready_ticks_size,
642 rtx sched_next, int cycle, int issue_more, bool after_stall_p)
643 {
644 insn_t last_scheduled_insn_old = FENCE_LAST_SCHEDULED_INSN (f);
645
646 gcc_assert (sel_bb_head_p (FENCE_INSN (f))
647 && !sched_next && !FENCE_SCHED_NEXT (f));
648
649 /* Check if we can decide which path fences came.
650 If we can't (or don't want to) - reset all. */
651 if (last_scheduled_insn == NULL
652 || last_scheduled_insn_old == NULL
653 /* This is a case when INSN is reachable on several paths from
654 one insn (this can happen when pipelining of outer loops is on and
655 there are two edges: one going around of inner loop and the other -
656 right through it; in such case just reset everything). */
657 || last_scheduled_insn == last_scheduled_insn_old)
658 {
659 state_reset (FENCE_STATE (f));
660 state_free (state);
661
662 reset_deps_context (FENCE_DC (f));
663 delete_deps_context (dc);
664
665 reset_target_context (FENCE_TC (f), true);
666 delete_target_context (tc);
667
668 if (cycle > FENCE_CYCLE (f))
669 FENCE_CYCLE (f) = cycle;
670
671 FENCE_LAST_SCHEDULED_INSN (f) = NULL;
672 FENCE_ISSUE_MORE (f) = issue_rate;
673 vec_free (executing_insns);
674 free (ready_ticks);
675 if (FENCE_EXECUTING_INSNS (f))
676 FENCE_EXECUTING_INSNS (f)->block_remove (0,
677 FENCE_EXECUTING_INSNS (f)->length ());
678 if (FENCE_READY_TICKS (f))
679 memset (FENCE_READY_TICKS (f), 0, FENCE_READY_TICKS_SIZE (f));
680 }
681 else
682 {
683 edge edge_old = NULL, edge_new = NULL;
684 edge candidate;
685 succ_iterator si;
686 insn_t succ;
687
688 /* Find fallthrough edge. */
689 gcc_assert (BLOCK_FOR_INSN (insn)->prev_bb);
690 candidate = find_fallthru_edge_from (BLOCK_FOR_INSN (insn)->prev_bb);
691
692 if (!candidate
693 || (candidate->src != BLOCK_FOR_INSN (last_scheduled_insn)
694 && candidate->src != BLOCK_FOR_INSN (last_scheduled_insn_old)))
695 {
696 /* No fallthrough edge leading to basic block of INSN. */
697 state_reset (FENCE_STATE (f));
698 state_free (state);
699
700 reset_target_context (FENCE_TC (f), true);
701 delete_target_context (tc);
702
703 FENCE_LAST_SCHEDULED_INSN (f) = NULL;
704 FENCE_ISSUE_MORE (f) = issue_rate;
705 }
706 else
707 if (candidate->src == BLOCK_FOR_INSN (last_scheduled_insn))
708 {
709 /* Would be weird if same insn is successor of several fallthrough
710 edges. */
711 gcc_assert (BLOCK_FOR_INSN (insn)->prev_bb
712 != BLOCK_FOR_INSN (last_scheduled_insn_old));
713
714 state_free (FENCE_STATE (f));
715 FENCE_STATE (f) = state;
716
717 delete_target_context (FENCE_TC (f));
718 FENCE_TC (f) = tc;
719
720 FENCE_LAST_SCHEDULED_INSN (f) = last_scheduled_insn;
721 FENCE_ISSUE_MORE (f) = issue_more;
722 }
723 else
724 {
725 /* Leave STATE, TC and LAST_SCHEDULED_INSN fields untouched. */
726 state_free (state);
727 delete_target_context (tc);
728
729 gcc_assert (BLOCK_FOR_INSN (insn)->prev_bb
730 != BLOCK_FOR_INSN (last_scheduled_insn));
731 }
732
733 /* Find edge of first predecessor (last_scheduled_insn_old->insn). */
734 FOR_EACH_SUCC_1 (succ, si, last_scheduled_insn_old,
735 SUCCS_NORMAL | SUCCS_SKIP_TO_LOOP_EXITS)
736 {
737 if (succ == insn)
738 {
739 /* No same successor allowed from several edges. */
740 gcc_assert (!edge_old);
741 edge_old = si.e1;
742 }
743 }
744 /* Find edge of second predecessor (last_scheduled_insn->insn). */
745 FOR_EACH_SUCC_1 (succ, si, last_scheduled_insn,
746 SUCCS_NORMAL | SUCCS_SKIP_TO_LOOP_EXITS)
747 {
748 if (succ == insn)
749 {
750 /* No same successor allowed from several edges. */
751 gcc_assert (!edge_new);
752 edge_new = si.e1;
753 }
754 }
755
756 /* Check if we can choose most probable predecessor. */
757 if (edge_old == NULL || edge_new == NULL)
758 {
759 reset_deps_context (FENCE_DC (f));
760 delete_deps_context (dc);
761 vec_free (executing_insns);
762 free (ready_ticks);
763
764 FENCE_CYCLE (f) = MAX (FENCE_CYCLE (f), cycle);
765 if (FENCE_EXECUTING_INSNS (f))
766 FENCE_EXECUTING_INSNS (f)->block_remove (0,
767 FENCE_EXECUTING_INSNS (f)->length ());
768 if (FENCE_READY_TICKS (f))
769 memset (FENCE_READY_TICKS (f), 0, FENCE_READY_TICKS_SIZE (f));
770 }
771 else
772 if (edge_new->probability > edge_old->probability)
773 {
774 delete_deps_context (FENCE_DC (f));
775 FENCE_DC (f) = dc;
776 vec_free (FENCE_EXECUTING_INSNS (f));
777 FENCE_EXECUTING_INSNS (f) = executing_insns;
778 free (FENCE_READY_TICKS (f));
779 FENCE_READY_TICKS (f) = ready_ticks;
780 FENCE_READY_TICKS_SIZE (f) = ready_ticks_size;
781 FENCE_CYCLE (f) = cycle;
782 }
783 else
784 {
785 /* Leave DC and CYCLE untouched. */
786 delete_deps_context (dc);
787 vec_free (executing_insns);
788 free (ready_ticks);
789 }
790 }
791
792 /* Fill remaining invariant fields. */
793 if (after_stall_p)
794 FENCE_AFTER_STALL_P (f) = 1;
795
796 FENCE_ISSUED_INSNS (f) = 0;
797 FENCE_STARTS_CYCLE_P (f) = 1;
798 FENCE_SCHED_NEXT (f) = NULL;
799 }
800
801 /* Add a new fence to NEW_FENCES list, initializing it from all
802 other parameters. */
803 static void
804 add_to_fences (flist_tail_t new_fences, insn_t insn,
805 state_t state, deps_t dc, void *tc, rtx last_scheduled_insn,
806 vec<rtx, va_gc> *executing_insns, int *ready_ticks,
807 int ready_ticks_size, rtx sched_next, int cycle,
808 int cycle_issued_insns, int issue_rate,
809 bool starts_cycle_p, bool after_stall_p)
810 {
811 fence_t f = flist_lookup (FLIST_TAIL_HEAD (new_fences), insn);
812
813 if (! f)
814 {
815 flist_add (FLIST_TAIL_TAILP (new_fences), insn, state, dc, tc,
816 last_scheduled_insn, executing_insns, ready_ticks,
817 ready_ticks_size, sched_next, cycle, cycle_issued_insns,
818 issue_rate, starts_cycle_p, after_stall_p);
819
820 FLIST_TAIL_TAILP (new_fences)
821 = &FLIST_NEXT (*FLIST_TAIL_TAILP (new_fences));
822 }
823 else
824 {
825 merge_fences (f, insn, state, dc, tc, last_scheduled_insn,
826 executing_insns, ready_ticks, ready_ticks_size,
827 sched_next, cycle, issue_rate, after_stall_p);
828 }
829 }
830
831 /* Move the first fence in the OLD_FENCES list to NEW_FENCES. */
832 void
833 move_fence_to_fences (flist_t old_fences, flist_tail_t new_fences)
834 {
835 fence_t f, old;
836 flist_t *tailp = FLIST_TAIL_TAILP (new_fences);
837
838 old = FLIST_FENCE (old_fences);
839 f = flist_lookup (FLIST_TAIL_HEAD (new_fences),
840 FENCE_INSN (FLIST_FENCE (old_fences)));
841 if (f)
842 {
843 merge_fences (f, old->insn, old->state, old->dc, old->tc,
844 old->last_scheduled_insn, old->executing_insns,
845 old->ready_ticks, old->ready_ticks_size,
846 old->sched_next, old->cycle, old->issue_more,
847 old->after_stall_p);
848 }
849 else
850 {
851 _list_add (tailp);
852 FLIST_TAIL_TAILP (new_fences) = &FLIST_NEXT (*tailp);
853 *FLIST_FENCE (*tailp) = *old;
854 init_fence_for_scheduling (FLIST_FENCE (*tailp));
855 }
856 FENCE_INSN (old) = NULL;
857 }
858
859 /* Add a new fence to NEW_FENCES list and initialize most of its data
860 as a clean one. */
861 void
862 add_clean_fence_to_fences (flist_tail_t new_fences, insn_t succ, fence_t fence)
863 {
864 int ready_ticks_size = get_max_uid () + 1;
865
866 add_to_fences (new_fences,
867 succ, state_create (), create_deps_context (),
868 create_target_context (true),
869 NULL_RTX, NULL,
870 XCNEWVEC (int, ready_ticks_size), ready_ticks_size,
871 NULL_RTX, FENCE_CYCLE (fence) + 1,
872 0, issue_rate, 1, FENCE_AFTER_STALL_P (fence));
873 }
874
875 /* Add a new fence to NEW_FENCES list and initialize all of its data
876 from FENCE and SUCC. */
877 void
878 add_dirty_fence_to_fences (flist_tail_t new_fences, insn_t succ, fence_t fence)
879 {
880 int * new_ready_ticks
881 = XNEWVEC (int, FENCE_READY_TICKS_SIZE (fence));
882
883 memcpy (new_ready_ticks, FENCE_READY_TICKS (fence),
884 FENCE_READY_TICKS_SIZE (fence) * sizeof (int));
885 add_to_fences (new_fences,
886 succ, state_create_copy (FENCE_STATE (fence)),
887 create_copy_of_deps_context (FENCE_DC (fence)),
888 create_copy_of_target_context (FENCE_TC (fence)),
889 FENCE_LAST_SCHEDULED_INSN (fence),
890 vec_safe_copy (FENCE_EXECUTING_INSNS (fence)),
891 new_ready_ticks,
892 FENCE_READY_TICKS_SIZE (fence),
893 FENCE_SCHED_NEXT (fence),
894 FENCE_CYCLE (fence),
895 FENCE_ISSUED_INSNS (fence),
896 FENCE_ISSUE_MORE (fence),
897 FENCE_STARTS_CYCLE_P (fence),
898 FENCE_AFTER_STALL_P (fence));
899 }
900 \f
901
902 /* Functions to work with regset and nop pools. */
903
904 /* Returns the new regset from pool. It might have some of the bits set
905 from the previous usage. */
906 regset
907 get_regset_from_pool (void)
908 {
909 regset rs;
910
911 if (regset_pool.n != 0)
912 rs = regset_pool.v[--regset_pool.n];
913 else
914 /* We need to create the regset. */
915 {
916 rs = ALLOC_REG_SET (&reg_obstack);
917
918 if (regset_pool.nn == regset_pool.ss)
919 regset_pool.vv = XRESIZEVEC (regset, regset_pool.vv,
920 (regset_pool.ss = 2 * regset_pool.ss + 1));
921 regset_pool.vv[regset_pool.nn++] = rs;
922 }
923
924 regset_pool.diff++;
925
926 return rs;
927 }
928
929 /* Same as above, but returns the empty regset. */
930 regset
931 get_clear_regset_from_pool (void)
932 {
933 regset rs = get_regset_from_pool ();
934
935 CLEAR_REG_SET (rs);
936 return rs;
937 }
938
939 /* Return regset RS to the pool for future use. */
940 void
941 return_regset_to_pool (regset rs)
942 {
943 gcc_assert (rs);
944 regset_pool.diff--;
945
946 if (regset_pool.n == regset_pool.s)
947 regset_pool.v = XRESIZEVEC (regset, regset_pool.v,
948 (regset_pool.s = 2 * regset_pool.s + 1));
949 regset_pool.v[regset_pool.n++] = rs;
950 }
951
952 #ifdef ENABLE_CHECKING
953 /* This is used as a qsort callback for sorting regset pool stacks.
954 X and XX are addresses of two regsets. They are never equal. */
955 static int
956 cmp_v_in_regset_pool (const void *x, const void *xx)
957 {
958 uintptr_t r1 = (uintptr_t) *((const regset *) x);
959 uintptr_t r2 = (uintptr_t) *((const regset *) xx);
960 if (r1 > r2)
961 return 1;
962 else if (r1 < r2)
963 return -1;
964 gcc_unreachable ();
965 }
966 #endif
967
968 /* Free the regset pool possibly checking for memory leaks. */
969 void
970 free_regset_pool (void)
971 {
972 #ifdef ENABLE_CHECKING
973 {
974 regset *v = regset_pool.v;
975 int i = 0;
976 int n = regset_pool.n;
977
978 regset *vv = regset_pool.vv;
979 int ii = 0;
980 int nn = regset_pool.nn;
981
982 int diff = 0;
983
984 gcc_assert (n <= nn);
985
986 /* Sort both vectors so it will be possible to compare them. */
987 qsort (v, n, sizeof (*v), cmp_v_in_regset_pool);
988 qsort (vv, nn, sizeof (*vv), cmp_v_in_regset_pool);
989
990 while (ii < nn)
991 {
992 if (v[i] == vv[ii])
993 i++;
994 else
995 /* VV[II] was lost. */
996 diff++;
997
998 ii++;
999 }
1000
1001 gcc_assert (diff == regset_pool.diff);
1002 }
1003 #endif
1004
1005 /* If not true - we have a memory leak. */
1006 gcc_assert (regset_pool.diff == 0);
1007
1008 while (regset_pool.n)
1009 {
1010 --regset_pool.n;
1011 FREE_REG_SET (regset_pool.v[regset_pool.n]);
1012 }
1013
1014 free (regset_pool.v);
1015 regset_pool.v = NULL;
1016 regset_pool.s = 0;
1017
1018 free (regset_pool.vv);
1019 regset_pool.vv = NULL;
1020 regset_pool.nn = 0;
1021 regset_pool.ss = 0;
1022
1023 regset_pool.diff = 0;
1024 }
1025 \f
1026
1027 /* Functions to work with nop pools. NOP insns are used as temporary
1028 placeholders of the insns being scheduled to allow correct update of
1029 the data sets. When update is finished, NOPs are deleted. */
1030
1031 /* A vinsn that is used to represent a nop. This vinsn is shared among all
1032 nops sel-sched generates. */
1033 static vinsn_t nop_vinsn = NULL;
1034
1035 /* Emit a nop before INSN, taking it from pool. */
1036 insn_t
1037 get_nop_from_pool (insn_t insn)
1038 {
1039 insn_t nop;
1040 bool old_p = nop_pool.n != 0;
1041 int flags;
1042
1043 if (old_p)
1044 nop = nop_pool.v[--nop_pool.n];
1045 else
1046 nop = nop_pattern;
1047
1048 nop = emit_insn_before (nop, insn);
1049
1050 if (old_p)
1051 flags = INSN_INIT_TODO_SSID;
1052 else
1053 flags = INSN_INIT_TODO_LUID | INSN_INIT_TODO_SSID;
1054
1055 set_insn_init (INSN_EXPR (insn), nop_vinsn, INSN_SEQNO (insn));
1056 sel_init_new_insn (nop, flags);
1057
1058 return nop;
1059 }
1060
1061 /* Remove NOP from the instruction stream and return it to the pool. */
1062 void
1063 return_nop_to_pool (insn_t nop, bool full_tidying)
1064 {
1065 gcc_assert (INSN_IN_STREAM_P (nop));
1066 sel_remove_insn (nop, false, full_tidying);
1067
1068 /* We'll recycle this nop. */
1069 INSN_DELETED_P (nop) = 0;
1070
1071 if (nop_pool.n == nop_pool.s)
1072 nop_pool.v = XRESIZEVEC (rtx, nop_pool.v,
1073 (nop_pool.s = 2 * nop_pool.s + 1));
1074 nop_pool.v[nop_pool.n++] = nop;
1075 }
1076
1077 /* Free the nop pool. */
1078 void
1079 free_nop_pool (void)
1080 {
1081 nop_pool.n = 0;
1082 nop_pool.s = 0;
1083 free (nop_pool.v);
1084 nop_pool.v = NULL;
1085 }
1086 \f
1087
1088 /* Skip unspec to support ia64 speculation. Called from rtx_equal_p_cb.
1089 The callback is given two rtxes XX and YY and writes the new rtxes
1090 to NX and NY in case some needs to be skipped. */
1091 static int
1092 skip_unspecs_callback (const_rtx *xx, const_rtx *yy, rtx *nx, rtx* ny)
1093 {
1094 const_rtx x = *xx;
1095 const_rtx y = *yy;
1096
1097 if (GET_CODE (x) == UNSPEC
1098 && (targetm.sched.skip_rtx_p == NULL
1099 || targetm.sched.skip_rtx_p (x)))
1100 {
1101 *nx = XVECEXP (x, 0, 0);
1102 *ny = CONST_CAST_RTX (y);
1103 return 1;
1104 }
1105
1106 if (GET_CODE (y) == UNSPEC
1107 && (targetm.sched.skip_rtx_p == NULL
1108 || targetm.sched.skip_rtx_p (y)))
1109 {
1110 *nx = CONST_CAST_RTX (x);
1111 *ny = XVECEXP (y, 0, 0);
1112 return 1;
1113 }
1114
1115 return 0;
1116 }
1117
1118 /* Callback, called from hash_rtx_cb. Helps to hash UNSPEC rtx X in a correct way
1119 to support ia64 speculation. When changes are needed, new rtx X and new mode
1120 NMODE are written, and the callback returns true. */
1121 static int
1122 hash_with_unspec_callback (const_rtx x, enum machine_mode mode ATTRIBUTE_UNUSED,
1123 rtx *nx, enum machine_mode* nmode)
1124 {
1125 if (GET_CODE (x) == UNSPEC
1126 && targetm.sched.skip_rtx_p
1127 && targetm.sched.skip_rtx_p (x))
1128 {
1129 *nx = XVECEXP (x, 0 ,0);
1130 *nmode = VOIDmode;
1131 return 1;
1132 }
1133
1134 return 0;
1135 }
1136
1137 /* Returns LHS and RHS are ok to be scheduled separately. */
1138 static bool
1139 lhs_and_rhs_separable_p (rtx lhs, rtx rhs)
1140 {
1141 if (lhs == NULL || rhs == NULL)
1142 return false;
1143
1144 /* Do not schedule constants as rhs: no point to use reg, if const
1145 can be used. Moreover, scheduling const as rhs may lead to mode
1146 mismatch cause consts don't have modes but they could be merged
1147 from branches where the same const used in different modes. */
1148 if (CONSTANT_P (rhs))
1149 return false;
1150
1151 /* ??? Do not rename predicate registers to avoid ICEs in bundling. */
1152 if (COMPARISON_P (rhs))
1153 return false;
1154
1155 /* Do not allow single REG to be an rhs. */
1156 if (REG_P (rhs))
1157 return false;
1158
1159 /* See comment at find_used_regs_1 (*1) for explanation of this
1160 restriction. */
1161 /* FIXME: remove this later. */
1162 if (MEM_P (lhs))
1163 return false;
1164
1165 /* This will filter all tricky things like ZERO_EXTRACT etc.
1166 For now we don't handle it. */
1167 if (!REG_P (lhs) && !MEM_P (lhs))
1168 return false;
1169
1170 return true;
1171 }
1172
1173 /* Initialize vinsn VI for INSN. Only for use from vinsn_create (). When
1174 FORCE_UNIQUE_P is true, the resulting vinsn will not be clonable. This is
1175 used e.g. for insns from recovery blocks. */
1176 static void
1177 vinsn_init (vinsn_t vi, insn_t insn, bool force_unique_p)
1178 {
1179 hash_rtx_callback_function hrcf;
1180 int insn_class;
1181
1182 VINSN_INSN_RTX (vi) = insn;
1183 VINSN_COUNT (vi) = 0;
1184 vi->cost = -1;
1185
1186 if (INSN_NOP_P (insn))
1187 return;
1188
1189 if (DF_INSN_UID_SAFE_GET (INSN_UID (insn)) != NULL)
1190 init_id_from_df (VINSN_ID (vi), insn, force_unique_p);
1191 else
1192 deps_init_id (VINSN_ID (vi), insn, force_unique_p);
1193
1194 /* Hash vinsn depending on whether it is separable or not. */
1195 hrcf = targetm.sched.skip_rtx_p ? hash_with_unspec_callback : NULL;
1196 if (VINSN_SEPARABLE_P (vi))
1197 {
1198 rtx rhs = VINSN_RHS (vi);
1199
1200 VINSN_HASH (vi) = hash_rtx_cb (rhs, GET_MODE (rhs),
1201 NULL, NULL, false, hrcf);
1202 VINSN_HASH_RTX (vi) = hash_rtx_cb (VINSN_PATTERN (vi),
1203 VOIDmode, NULL, NULL,
1204 false, hrcf);
1205 }
1206 else
1207 {
1208 VINSN_HASH (vi) = hash_rtx_cb (VINSN_PATTERN (vi), VOIDmode,
1209 NULL, NULL, false, hrcf);
1210 VINSN_HASH_RTX (vi) = VINSN_HASH (vi);
1211 }
1212
1213 insn_class = haifa_classify_insn (insn);
1214 if (insn_class >= 2
1215 && (!targetm.sched.get_insn_spec_ds
1216 || ((targetm.sched.get_insn_spec_ds (insn) & BEGIN_CONTROL)
1217 == 0)))
1218 VINSN_MAY_TRAP_P (vi) = true;
1219 else
1220 VINSN_MAY_TRAP_P (vi) = false;
1221 }
1222
1223 /* Indicate that VI has become the part of an rtx object. */
1224 void
1225 vinsn_attach (vinsn_t vi)
1226 {
1227 /* Assert that VI is not pending for deletion. */
1228 gcc_assert (VINSN_INSN_RTX (vi));
1229
1230 VINSN_COUNT (vi)++;
1231 }
1232
1233 /* Create and init VI from the INSN. Use UNIQUE_P for determining the correct
1234 VINSN_TYPE (VI). */
1235 static vinsn_t
1236 vinsn_create (insn_t insn, bool force_unique_p)
1237 {
1238 vinsn_t vi = XCNEW (struct vinsn_def);
1239
1240 vinsn_init (vi, insn, force_unique_p);
1241 return vi;
1242 }
1243
1244 /* Return a copy of VI. When REATTACH_P is true, detach VI and attach
1245 the copy. */
1246 vinsn_t
1247 vinsn_copy (vinsn_t vi, bool reattach_p)
1248 {
1249 rtx copy;
1250 bool unique = VINSN_UNIQUE_P (vi);
1251 vinsn_t new_vi;
1252
1253 copy = create_copy_of_insn_rtx (VINSN_INSN_RTX (vi));
1254 new_vi = create_vinsn_from_insn_rtx (copy, unique);
1255 if (reattach_p)
1256 {
1257 vinsn_detach (vi);
1258 vinsn_attach (new_vi);
1259 }
1260
1261 return new_vi;
1262 }
1263
1264 /* Delete the VI vinsn and free its data. */
1265 static void
1266 vinsn_delete (vinsn_t vi)
1267 {
1268 gcc_assert (VINSN_COUNT (vi) == 0);
1269
1270 if (!INSN_NOP_P (VINSN_INSN_RTX (vi)))
1271 {
1272 return_regset_to_pool (VINSN_REG_SETS (vi));
1273 return_regset_to_pool (VINSN_REG_USES (vi));
1274 return_regset_to_pool (VINSN_REG_CLOBBERS (vi));
1275 }
1276
1277 free (vi);
1278 }
1279
1280 /* Indicate that VI is no longer a part of some rtx object.
1281 Remove VI if it is no longer needed. */
1282 void
1283 vinsn_detach (vinsn_t vi)
1284 {
1285 gcc_assert (VINSN_COUNT (vi) > 0);
1286
1287 if (--VINSN_COUNT (vi) == 0)
1288 vinsn_delete (vi);
1289 }
1290
1291 /* Returns TRUE if VI is a branch. */
1292 bool
1293 vinsn_cond_branch_p (vinsn_t vi)
1294 {
1295 insn_t insn;
1296
1297 if (!VINSN_UNIQUE_P (vi))
1298 return false;
1299
1300 insn = VINSN_INSN_RTX (vi);
1301 if (BB_END (BLOCK_FOR_INSN (insn)) != insn)
1302 return false;
1303
1304 return control_flow_insn_p (insn);
1305 }
1306
1307 /* Return latency of INSN. */
1308 static int
1309 sel_insn_rtx_cost (rtx insn)
1310 {
1311 int cost;
1312
1313 /* A USE insn, or something else we don't need to
1314 understand. We can't pass these directly to
1315 result_ready_cost or insn_default_latency because it will
1316 trigger a fatal error for unrecognizable insns. */
1317 if (recog_memoized (insn) < 0)
1318 cost = 0;
1319 else
1320 {
1321 cost = insn_default_latency (insn);
1322
1323 if (cost < 0)
1324 cost = 0;
1325 }
1326
1327 return cost;
1328 }
1329
1330 /* Return the cost of the VI.
1331 !!! FIXME: Unify with haifa-sched.c: insn_cost (). */
1332 int
1333 sel_vinsn_cost (vinsn_t vi)
1334 {
1335 int cost = vi->cost;
1336
1337 if (cost < 0)
1338 {
1339 cost = sel_insn_rtx_cost (VINSN_INSN_RTX (vi));
1340 vi->cost = cost;
1341 }
1342
1343 return cost;
1344 }
1345 \f
1346
1347 /* Functions for insn emitting. */
1348
1349 /* Emit new insn after AFTER based on PATTERN and initialize its data from
1350 EXPR and SEQNO. */
1351 insn_t
1352 sel_gen_insn_from_rtx_after (rtx pattern, expr_t expr, int seqno, insn_t after)
1353 {
1354 insn_t new_insn;
1355
1356 gcc_assert (EXPR_TARGET_AVAILABLE (expr) == true);
1357
1358 new_insn = emit_insn_after (pattern, after);
1359 set_insn_init (expr, NULL, seqno);
1360 sel_init_new_insn (new_insn, INSN_INIT_TODO_LUID | INSN_INIT_TODO_SSID);
1361
1362 return new_insn;
1363 }
1364
1365 /* Force newly generated vinsns to be unique. */
1366 static bool init_insn_force_unique_p = false;
1367
1368 /* Emit new speculation recovery insn after AFTER based on PATTERN and
1369 initialize its data from EXPR and SEQNO. */
1370 insn_t
1371 sel_gen_recovery_insn_from_rtx_after (rtx pattern, expr_t expr, int seqno,
1372 insn_t after)
1373 {
1374 insn_t insn;
1375
1376 gcc_assert (!init_insn_force_unique_p);
1377
1378 init_insn_force_unique_p = true;
1379 insn = sel_gen_insn_from_rtx_after (pattern, expr, seqno, after);
1380 CANT_MOVE (insn) = 1;
1381 init_insn_force_unique_p = false;
1382
1383 return insn;
1384 }
1385
1386 /* Emit new insn after AFTER based on EXPR and SEQNO. If VINSN is not NULL,
1387 take it as a new vinsn instead of EXPR's vinsn.
1388 We simplify insns later, after scheduling region in
1389 simplify_changed_insns. */
1390 insn_t
1391 sel_gen_insn_from_expr_after (expr_t expr, vinsn_t vinsn, int seqno,
1392 insn_t after)
1393 {
1394 expr_t emit_expr;
1395 insn_t insn;
1396 int flags;
1397
1398 emit_expr = set_insn_init (expr, vinsn ? vinsn : EXPR_VINSN (expr),
1399 seqno);
1400 insn = EXPR_INSN_RTX (emit_expr);
1401 add_insn_after (insn, after, BLOCK_FOR_INSN (insn));
1402
1403 flags = INSN_INIT_TODO_SSID;
1404 if (INSN_LUID (insn) == 0)
1405 flags |= INSN_INIT_TODO_LUID;
1406 sel_init_new_insn (insn, flags);
1407
1408 return insn;
1409 }
1410
1411 /* Move insn from EXPR after AFTER. */
1412 insn_t
1413 sel_move_insn (expr_t expr, int seqno, insn_t after)
1414 {
1415 insn_t insn = EXPR_INSN_RTX (expr);
1416 basic_block bb = BLOCK_FOR_INSN (after);
1417 insn_t next = NEXT_INSN (after);
1418
1419 /* Assert that in move_op we disconnected this insn properly. */
1420 gcc_assert (EXPR_VINSN (INSN_EXPR (insn)) != NULL);
1421 PREV_INSN (insn) = after;
1422 NEXT_INSN (insn) = next;
1423
1424 NEXT_INSN (after) = insn;
1425 PREV_INSN (next) = insn;
1426
1427 /* Update links from insn to bb and vice versa. */
1428 df_insn_change_bb (insn, bb);
1429 if (BB_END (bb) == after)
1430 BB_END (bb) = insn;
1431
1432 prepare_insn_expr (insn, seqno);
1433 return insn;
1434 }
1435
1436 \f
1437 /* Functions to work with right-hand sides. */
1438
1439 /* Search for a hash value determined by UID/NEW_VINSN in a sorted vector
1440 VECT and return true when found. Use NEW_VINSN for comparison only when
1441 COMPARE_VINSNS is true. Write to INDP the index on which
1442 the search has stopped, such that inserting the new element at INDP will
1443 retain VECT's sort order. */
1444 static bool
1445 find_in_history_vect_1 (vec<expr_history_def> vect,
1446 unsigned uid, vinsn_t new_vinsn,
1447 bool compare_vinsns, int *indp)
1448 {
1449 expr_history_def *arr;
1450 int i, j, len = vect.length ();
1451
1452 if (len == 0)
1453 {
1454 *indp = 0;
1455 return false;
1456 }
1457
1458 arr = vect.address ();
1459 i = 0, j = len - 1;
1460
1461 while (i <= j)
1462 {
1463 unsigned auid = arr[i].uid;
1464 vinsn_t avinsn = arr[i].new_expr_vinsn;
1465
1466 if (auid == uid
1467 /* When undoing transformation on a bookkeeping copy, the new vinsn
1468 may not be exactly equal to the one that is saved in the vector.
1469 This is because the insn whose copy we're checking was possibly
1470 substituted itself. */
1471 && (! compare_vinsns
1472 || vinsn_equal_p (avinsn, new_vinsn)))
1473 {
1474 *indp = i;
1475 return true;
1476 }
1477 else if (auid > uid)
1478 break;
1479 i++;
1480 }
1481
1482 *indp = i;
1483 return false;
1484 }
1485
1486 /* Search for a uid of INSN and NEW_VINSN in a sorted vector VECT. Return
1487 the position found or -1, if no such value is in vector.
1488 Search also for UIDs of insn's originators, if ORIGINATORS_P is true. */
1489 int
1490 find_in_history_vect (vec<expr_history_def> vect, rtx insn,
1491 vinsn_t new_vinsn, bool originators_p)
1492 {
1493 int ind;
1494
1495 if (find_in_history_vect_1 (vect, INSN_UID (insn), new_vinsn,
1496 false, &ind))
1497 return ind;
1498
1499 if (INSN_ORIGINATORS (insn) && originators_p)
1500 {
1501 unsigned uid;
1502 bitmap_iterator bi;
1503
1504 EXECUTE_IF_SET_IN_BITMAP (INSN_ORIGINATORS (insn), 0, uid, bi)
1505 if (find_in_history_vect_1 (vect, uid, new_vinsn, false, &ind))
1506 return ind;
1507 }
1508
1509 return -1;
1510 }
1511
1512 /* Insert new element in a sorted history vector pointed to by PVECT,
1513 if it is not there already. The element is searched using
1514 UID/NEW_EXPR_VINSN pair. TYPE, OLD_EXPR_VINSN and SPEC_DS save
1515 the history of a transformation. */
1516 void
1517 insert_in_history_vect (vec<expr_history_def> *pvect,
1518 unsigned uid, enum local_trans_type type,
1519 vinsn_t old_expr_vinsn, vinsn_t new_expr_vinsn,
1520 ds_t spec_ds)
1521 {
1522 vec<expr_history_def> vect = *pvect;
1523 expr_history_def temp;
1524 bool res;
1525 int ind;
1526
1527 res = find_in_history_vect_1 (vect, uid, new_expr_vinsn, true, &ind);
1528
1529 if (res)
1530 {
1531 expr_history_def *phist = &vect[ind];
1532
1533 /* It is possible that speculation types of expressions that were
1534 propagated through different paths will be different here. In this
1535 case, merge the status to get the correct check later. */
1536 if (phist->spec_ds != spec_ds)
1537 phist->spec_ds = ds_max_merge (phist->spec_ds, spec_ds);
1538 return;
1539 }
1540
1541 temp.uid = uid;
1542 temp.old_expr_vinsn = old_expr_vinsn;
1543 temp.new_expr_vinsn = new_expr_vinsn;
1544 temp.spec_ds = spec_ds;
1545 temp.type = type;
1546
1547 vinsn_attach (old_expr_vinsn);
1548 vinsn_attach (new_expr_vinsn);
1549 vect.safe_insert (ind, temp);
1550 *pvect = vect;
1551 }
1552
1553 /* Free history vector PVECT. */
1554 static void
1555 free_history_vect (vec<expr_history_def> &pvect)
1556 {
1557 unsigned i;
1558 expr_history_def *phist;
1559
1560 if (! pvect.exists ())
1561 return;
1562
1563 for (i = 0; pvect.iterate (i, &phist); i++)
1564 {
1565 vinsn_detach (phist->old_expr_vinsn);
1566 vinsn_detach (phist->new_expr_vinsn);
1567 }
1568
1569 pvect.release ();
1570 }
1571
1572 /* Merge vector FROM to PVECT. */
1573 static void
1574 merge_history_vect (vec<expr_history_def> *pvect,
1575 vec<expr_history_def> from)
1576 {
1577 expr_history_def *phist;
1578 int i;
1579
1580 /* We keep this vector sorted. */
1581 for (i = 0; from.iterate (i, &phist); i++)
1582 insert_in_history_vect (pvect, phist->uid, phist->type,
1583 phist->old_expr_vinsn, phist->new_expr_vinsn,
1584 phist->spec_ds);
1585 }
1586
1587 /* Compare two vinsns as rhses if possible and as vinsns otherwise. */
1588 bool
1589 vinsn_equal_p (vinsn_t x, vinsn_t y)
1590 {
1591 rtx_equal_p_callback_function repcf;
1592
1593 if (x == y)
1594 return true;
1595
1596 if (VINSN_TYPE (x) != VINSN_TYPE (y))
1597 return false;
1598
1599 if (VINSN_HASH (x) != VINSN_HASH (y))
1600 return false;
1601
1602 repcf = targetm.sched.skip_rtx_p ? skip_unspecs_callback : NULL;
1603 if (VINSN_SEPARABLE_P (x))
1604 {
1605 /* Compare RHSes of VINSNs. */
1606 gcc_assert (VINSN_RHS (x));
1607 gcc_assert (VINSN_RHS (y));
1608
1609 return rtx_equal_p_cb (VINSN_RHS (x), VINSN_RHS (y), repcf);
1610 }
1611
1612 return rtx_equal_p_cb (VINSN_PATTERN (x), VINSN_PATTERN (y), repcf);
1613 }
1614 \f
1615
1616 /* Functions for working with expressions. */
1617
1618 /* Initialize EXPR. */
1619 static void
1620 init_expr (expr_t expr, vinsn_t vi, int spec, int use, int priority,
1621 int sched_times, int orig_bb_index, ds_t spec_done_ds,
1622 ds_t spec_to_check_ds, int orig_sched_cycle,
1623 vec<expr_history_def> history,
1624 signed char target_available,
1625 bool was_substituted, bool was_renamed, bool needs_spec_check_p,
1626 bool cant_move)
1627 {
1628 vinsn_attach (vi);
1629
1630 EXPR_VINSN (expr) = vi;
1631 EXPR_SPEC (expr) = spec;
1632 EXPR_USEFULNESS (expr) = use;
1633 EXPR_PRIORITY (expr) = priority;
1634 EXPR_PRIORITY_ADJ (expr) = 0;
1635 EXPR_SCHED_TIMES (expr) = sched_times;
1636 EXPR_ORIG_BB_INDEX (expr) = orig_bb_index;
1637 EXPR_ORIG_SCHED_CYCLE (expr) = orig_sched_cycle;
1638 EXPR_SPEC_DONE_DS (expr) = spec_done_ds;
1639 EXPR_SPEC_TO_CHECK_DS (expr) = spec_to_check_ds;
1640
1641 if (history.exists ())
1642 EXPR_HISTORY_OF_CHANGES (expr) = history;
1643 else
1644 EXPR_HISTORY_OF_CHANGES (expr).create (0);
1645
1646 EXPR_TARGET_AVAILABLE (expr) = target_available;
1647 EXPR_WAS_SUBSTITUTED (expr) = was_substituted;
1648 EXPR_WAS_RENAMED (expr) = was_renamed;
1649 EXPR_NEEDS_SPEC_CHECK_P (expr) = needs_spec_check_p;
1650 EXPR_CANT_MOVE (expr) = cant_move;
1651 }
1652
1653 /* Make a copy of the expr FROM into the expr TO. */
1654 void
1655 copy_expr (expr_t to, expr_t from)
1656 {
1657 vec<expr_history_def> temp = vNULL;
1658
1659 if (EXPR_HISTORY_OF_CHANGES (from).exists ())
1660 {
1661 unsigned i;
1662 expr_history_def *phist;
1663
1664 temp = EXPR_HISTORY_OF_CHANGES (from).copy ();
1665 for (i = 0;
1666 temp.iterate (i, &phist);
1667 i++)
1668 {
1669 vinsn_attach (phist->old_expr_vinsn);
1670 vinsn_attach (phist->new_expr_vinsn);
1671 }
1672 }
1673
1674 init_expr (to, EXPR_VINSN (from), EXPR_SPEC (from),
1675 EXPR_USEFULNESS (from), EXPR_PRIORITY (from),
1676 EXPR_SCHED_TIMES (from), EXPR_ORIG_BB_INDEX (from),
1677 EXPR_SPEC_DONE_DS (from), EXPR_SPEC_TO_CHECK_DS (from),
1678 EXPR_ORIG_SCHED_CYCLE (from), temp,
1679 EXPR_TARGET_AVAILABLE (from), EXPR_WAS_SUBSTITUTED (from),
1680 EXPR_WAS_RENAMED (from), EXPR_NEEDS_SPEC_CHECK_P (from),
1681 EXPR_CANT_MOVE (from));
1682 }
1683
1684 /* Same, but the final expr will not ever be in av sets, so don't copy
1685 "uninteresting" data such as bitmap cache. */
1686 void
1687 copy_expr_onside (expr_t to, expr_t from)
1688 {
1689 init_expr (to, EXPR_VINSN (from), EXPR_SPEC (from), EXPR_USEFULNESS (from),
1690 EXPR_PRIORITY (from), EXPR_SCHED_TIMES (from), 0,
1691 EXPR_SPEC_DONE_DS (from), EXPR_SPEC_TO_CHECK_DS (from), 0,
1692 vNULL,
1693 EXPR_TARGET_AVAILABLE (from), EXPR_WAS_SUBSTITUTED (from),
1694 EXPR_WAS_RENAMED (from), EXPR_NEEDS_SPEC_CHECK_P (from),
1695 EXPR_CANT_MOVE (from));
1696 }
1697
1698 /* Prepare the expr of INSN for scheduling. Used when moving insn and when
1699 initializing new insns. */
1700 static void
1701 prepare_insn_expr (insn_t insn, int seqno)
1702 {
1703 expr_t expr = INSN_EXPR (insn);
1704 ds_t ds;
1705
1706 INSN_SEQNO (insn) = seqno;
1707 EXPR_ORIG_BB_INDEX (expr) = BLOCK_NUM (insn);
1708 EXPR_SPEC (expr) = 0;
1709 EXPR_ORIG_SCHED_CYCLE (expr) = 0;
1710 EXPR_WAS_SUBSTITUTED (expr) = 0;
1711 EXPR_WAS_RENAMED (expr) = 0;
1712 EXPR_TARGET_AVAILABLE (expr) = 1;
1713 INSN_LIVE_VALID_P (insn) = false;
1714
1715 /* ??? If this expression is speculative, make its dependence
1716 as weak as possible. We can filter this expression later
1717 in process_spec_exprs, because we do not distinguish
1718 between the status we got during compute_av_set and the
1719 existing status. To be fixed. */
1720 ds = EXPR_SPEC_DONE_DS (expr);
1721 if (ds)
1722 EXPR_SPEC_DONE_DS (expr) = ds_get_max_dep_weak (ds);
1723
1724 free_history_vect (EXPR_HISTORY_OF_CHANGES (expr));
1725 }
1726
1727 /* Update target_available bits when merging exprs TO and FROM. SPLIT_POINT
1728 is non-null when expressions are merged from different successors at
1729 a split point. */
1730 static void
1731 update_target_availability (expr_t to, expr_t from, insn_t split_point)
1732 {
1733 if (EXPR_TARGET_AVAILABLE (to) < 0
1734 || EXPR_TARGET_AVAILABLE (from) < 0)
1735 EXPR_TARGET_AVAILABLE (to) = -1;
1736 else
1737 {
1738 /* We try to detect the case when one of the expressions
1739 can only be reached through another one. In this case,
1740 we can do better. */
1741 if (split_point == NULL)
1742 {
1743 int toind, fromind;
1744
1745 toind = EXPR_ORIG_BB_INDEX (to);
1746 fromind = EXPR_ORIG_BB_INDEX (from);
1747
1748 if (toind && toind == fromind)
1749 /* Do nothing -- everything is done in
1750 merge_with_other_exprs. */
1751 ;
1752 else
1753 EXPR_TARGET_AVAILABLE (to) = -1;
1754 }
1755 else if (EXPR_TARGET_AVAILABLE (from) == 0
1756 && EXPR_LHS (from)
1757 && REG_P (EXPR_LHS (from))
1758 && REGNO (EXPR_LHS (to)) != REGNO (EXPR_LHS (from)))
1759 EXPR_TARGET_AVAILABLE (to) = -1;
1760 else
1761 EXPR_TARGET_AVAILABLE (to) &= EXPR_TARGET_AVAILABLE (from);
1762 }
1763 }
1764
1765 /* Update speculation bits when merging exprs TO and FROM. SPLIT_POINT
1766 is non-null when expressions are merged from different successors at
1767 a split point. */
1768 static void
1769 update_speculative_bits (expr_t to, expr_t from, insn_t split_point)
1770 {
1771 ds_t old_to_ds, old_from_ds;
1772
1773 old_to_ds = EXPR_SPEC_DONE_DS (to);
1774 old_from_ds = EXPR_SPEC_DONE_DS (from);
1775
1776 EXPR_SPEC_DONE_DS (to) = ds_max_merge (old_to_ds, old_from_ds);
1777 EXPR_SPEC_TO_CHECK_DS (to) |= EXPR_SPEC_TO_CHECK_DS (from);
1778 EXPR_NEEDS_SPEC_CHECK_P (to) |= EXPR_NEEDS_SPEC_CHECK_P (from);
1779
1780 /* When merging e.g. control & data speculative exprs, or a control
1781 speculative with a control&data speculative one, we really have
1782 to change vinsn too. Also, when speculative status is changed,
1783 we also need to record this as a transformation in expr's history. */
1784 if ((old_to_ds & SPECULATIVE) || (old_from_ds & SPECULATIVE))
1785 {
1786 old_to_ds = ds_get_speculation_types (old_to_ds);
1787 old_from_ds = ds_get_speculation_types (old_from_ds);
1788
1789 if (old_to_ds != old_from_ds)
1790 {
1791 ds_t record_ds;
1792
1793 /* When both expressions are speculative, we need to change
1794 the vinsn first. */
1795 if ((old_to_ds & SPECULATIVE) && (old_from_ds & SPECULATIVE))
1796 {
1797 int res;
1798
1799 res = speculate_expr (to, EXPR_SPEC_DONE_DS (to));
1800 gcc_assert (res >= 0);
1801 }
1802
1803 if (split_point != NULL)
1804 {
1805 /* Record the change with proper status. */
1806 record_ds = EXPR_SPEC_DONE_DS (to) & SPECULATIVE;
1807 record_ds &= ~(old_to_ds & SPECULATIVE);
1808 record_ds &= ~(old_from_ds & SPECULATIVE);
1809
1810 insert_in_history_vect (&EXPR_HISTORY_OF_CHANGES (to),
1811 INSN_UID (split_point), TRANS_SPECULATION,
1812 EXPR_VINSN (from), EXPR_VINSN (to),
1813 record_ds);
1814 }
1815 }
1816 }
1817 }
1818
1819
1820 /* Merge bits of FROM expr to TO expr. When SPLIT_POINT is not NULL,
1821 this is done along different paths. */
1822 void
1823 merge_expr_data (expr_t to, expr_t from, insn_t split_point)
1824 {
1825 /* Choose the maximum of the specs of merged exprs. This is required
1826 for correctness of bookkeeping. */
1827 if (EXPR_SPEC (to) < EXPR_SPEC (from))
1828 EXPR_SPEC (to) = EXPR_SPEC (from);
1829
1830 if (split_point)
1831 EXPR_USEFULNESS (to) += EXPR_USEFULNESS (from);
1832 else
1833 EXPR_USEFULNESS (to) = MAX (EXPR_USEFULNESS (to),
1834 EXPR_USEFULNESS (from));
1835
1836 if (EXPR_PRIORITY (to) < EXPR_PRIORITY (from))
1837 EXPR_PRIORITY (to) = EXPR_PRIORITY (from);
1838
1839 if (EXPR_SCHED_TIMES (to) > EXPR_SCHED_TIMES (from))
1840 EXPR_SCHED_TIMES (to) = EXPR_SCHED_TIMES (from);
1841
1842 if (EXPR_ORIG_BB_INDEX (to) != EXPR_ORIG_BB_INDEX (from))
1843 EXPR_ORIG_BB_INDEX (to) = 0;
1844
1845 EXPR_ORIG_SCHED_CYCLE (to) = MIN (EXPR_ORIG_SCHED_CYCLE (to),
1846 EXPR_ORIG_SCHED_CYCLE (from));
1847
1848 EXPR_WAS_SUBSTITUTED (to) |= EXPR_WAS_SUBSTITUTED (from);
1849 EXPR_WAS_RENAMED (to) |= EXPR_WAS_RENAMED (from);
1850 EXPR_CANT_MOVE (to) |= EXPR_CANT_MOVE (from);
1851
1852 merge_history_vect (&EXPR_HISTORY_OF_CHANGES (to),
1853 EXPR_HISTORY_OF_CHANGES (from));
1854 update_target_availability (to, from, split_point);
1855 update_speculative_bits (to, from, split_point);
1856 }
1857
1858 /* Merge bits of FROM expr to TO expr. Vinsns in the exprs should be equal
1859 in terms of vinsn_equal_p. SPLIT_POINT is non-null when expressions
1860 are merged from different successors at a split point. */
1861 void
1862 merge_expr (expr_t to, expr_t from, insn_t split_point)
1863 {
1864 vinsn_t to_vi = EXPR_VINSN (to);
1865 vinsn_t from_vi = EXPR_VINSN (from);
1866
1867 gcc_assert (vinsn_equal_p (to_vi, from_vi));
1868
1869 /* Make sure that speculative pattern is propagated into exprs that
1870 have non-speculative one. This will provide us with consistent
1871 speculative bits and speculative patterns inside expr. */
1872 if ((EXPR_SPEC_DONE_DS (from) != 0
1873 && EXPR_SPEC_DONE_DS (to) == 0)
1874 /* Do likewise for volatile insns, so that we always retain
1875 the may_trap_p bit on the resulting expression. */
1876 || (VINSN_MAY_TRAP_P (EXPR_VINSN (from))
1877 && !VINSN_MAY_TRAP_P (EXPR_VINSN (to))))
1878 change_vinsn_in_expr (to, EXPR_VINSN (from));
1879
1880 merge_expr_data (to, from, split_point);
1881 gcc_assert (EXPR_USEFULNESS (to) <= REG_BR_PROB_BASE);
1882 }
1883
1884 /* Clear the information of this EXPR. */
1885 void
1886 clear_expr (expr_t expr)
1887 {
1888
1889 vinsn_detach (EXPR_VINSN (expr));
1890 EXPR_VINSN (expr) = NULL;
1891
1892 free_history_vect (EXPR_HISTORY_OF_CHANGES (expr));
1893 }
1894
1895 /* For a given LV_SET, mark EXPR having unavailable target register. */
1896 static void
1897 set_unavailable_target_for_expr (expr_t expr, regset lv_set)
1898 {
1899 if (EXPR_SEPARABLE_P (expr))
1900 {
1901 if (REG_P (EXPR_LHS (expr))
1902 && register_unavailable_p (lv_set, EXPR_LHS (expr)))
1903 {
1904 /* If it's an insn like r1 = use (r1, ...), and it exists in
1905 different forms in each of the av_sets being merged, we can't say
1906 whether original destination register is available or not.
1907 However, this still works if destination register is not used
1908 in the original expression: if the branch at which LV_SET we're
1909 looking here is not actually 'other branch' in sense that same
1910 expression is available through it (but it can't be determined
1911 at computation stage because of transformations on one of the
1912 branches), it still won't affect the availability.
1913 Liveness of a register somewhere on a code motion path means
1914 it's either read somewhere on a codemotion path, live on
1915 'other' branch, live at the point immediately following
1916 the original operation, or is read by the original operation.
1917 The latter case is filtered out in the condition below.
1918 It still doesn't cover the case when register is defined and used
1919 somewhere within the code motion path, and in this case we could
1920 miss a unifying code motion along both branches using a renamed
1921 register, but it won't affect a code correctness since upon
1922 an actual code motion a bookkeeping code would be generated. */
1923 if (register_unavailable_p (VINSN_REG_USES (EXPR_VINSN (expr)),
1924 EXPR_LHS (expr)))
1925 EXPR_TARGET_AVAILABLE (expr) = -1;
1926 else
1927 EXPR_TARGET_AVAILABLE (expr) = false;
1928 }
1929 }
1930 else
1931 {
1932 unsigned regno;
1933 reg_set_iterator rsi;
1934
1935 EXECUTE_IF_SET_IN_REG_SET (VINSN_REG_SETS (EXPR_VINSN (expr)),
1936 0, regno, rsi)
1937 if (bitmap_bit_p (lv_set, regno))
1938 {
1939 EXPR_TARGET_AVAILABLE (expr) = false;
1940 break;
1941 }
1942
1943 EXECUTE_IF_SET_IN_REG_SET (VINSN_REG_CLOBBERS (EXPR_VINSN (expr)),
1944 0, regno, rsi)
1945 if (bitmap_bit_p (lv_set, regno))
1946 {
1947 EXPR_TARGET_AVAILABLE (expr) = false;
1948 break;
1949 }
1950 }
1951 }
1952
1953 /* Try to make EXPR speculative. Return 1 when EXPR's pattern
1954 or dependence status have changed, 2 when also the target register
1955 became unavailable, 0 if nothing had to be changed. */
1956 int
1957 speculate_expr (expr_t expr, ds_t ds)
1958 {
1959 int res;
1960 rtx orig_insn_rtx;
1961 rtx spec_pat;
1962 ds_t target_ds, current_ds;
1963
1964 /* Obtain the status we need to put on EXPR. */
1965 target_ds = (ds & SPECULATIVE);
1966 current_ds = EXPR_SPEC_DONE_DS (expr);
1967 ds = ds_full_merge (current_ds, target_ds, NULL_RTX, NULL_RTX);
1968
1969 orig_insn_rtx = EXPR_INSN_RTX (expr);
1970
1971 res = sched_speculate_insn (orig_insn_rtx, ds, &spec_pat);
1972
1973 switch (res)
1974 {
1975 case 0:
1976 EXPR_SPEC_DONE_DS (expr) = ds;
1977 return current_ds != ds ? 1 : 0;
1978
1979 case 1:
1980 {
1981 rtx spec_insn_rtx = create_insn_rtx_from_pattern (spec_pat, NULL_RTX);
1982 vinsn_t spec_vinsn = create_vinsn_from_insn_rtx (spec_insn_rtx, false);
1983
1984 change_vinsn_in_expr (expr, spec_vinsn);
1985 EXPR_SPEC_DONE_DS (expr) = ds;
1986 EXPR_NEEDS_SPEC_CHECK_P (expr) = true;
1987
1988 /* Do not allow clobbering the address register of speculative
1989 insns. */
1990 if (register_unavailable_p (VINSN_REG_USES (EXPR_VINSN (expr)),
1991 expr_dest_reg (expr)))
1992 {
1993 EXPR_TARGET_AVAILABLE (expr) = false;
1994 return 2;
1995 }
1996
1997 return 1;
1998 }
1999
2000 case -1:
2001 return -1;
2002
2003 default:
2004 gcc_unreachable ();
2005 return -1;
2006 }
2007 }
2008
2009 /* Return a destination register, if any, of EXPR. */
2010 rtx
2011 expr_dest_reg (expr_t expr)
2012 {
2013 rtx dest = VINSN_LHS (EXPR_VINSN (expr));
2014
2015 if (dest != NULL_RTX && REG_P (dest))
2016 return dest;
2017
2018 return NULL_RTX;
2019 }
2020
2021 /* Returns the REGNO of the R's destination. */
2022 unsigned
2023 expr_dest_regno (expr_t expr)
2024 {
2025 rtx dest = expr_dest_reg (expr);
2026
2027 gcc_assert (dest != NULL_RTX);
2028 return REGNO (dest);
2029 }
2030
2031 /* For a given LV_SET, mark all expressions in JOIN_SET, but not present in
2032 AV_SET having unavailable target register. */
2033 void
2034 mark_unavailable_targets (av_set_t join_set, av_set_t av_set, regset lv_set)
2035 {
2036 expr_t expr;
2037 av_set_iterator avi;
2038
2039 FOR_EACH_EXPR (expr, avi, join_set)
2040 if (av_set_lookup (av_set, EXPR_VINSN (expr)) == NULL)
2041 set_unavailable_target_for_expr (expr, lv_set);
2042 }
2043 \f
2044
2045 /* Returns true if REG (at least partially) is present in REGS. */
2046 bool
2047 register_unavailable_p (regset regs, rtx reg)
2048 {
2049 unsigned regno, end_regno;
2050
2051 regno = REGNO (reg);
2052 if (bitmap_bit_p (regs, regno))
2053 return true;
2054
2055 end_regno = END_REGNO (reg);
2056
2057 while (++regno < end_regno)
2058 if (bitmap_bit_p (regs, regno))
2059 return true;
2060
2061 return false;
2062 }
2063
2064 /* Av set functions. */
2065
2066 /* Add a new element to av set SETP.
2067 Return the element added. */
2068 static av_set_t
2069 av_set_add_element (av_set_t *setp)
2070 {
2071 /* Insert at the beginning of the list. */
2072 _list_add (setp);
2073 return *setp;
2074 }
2075
2076 /* Add EXPR to SETP. */
2077 void
2078 av_set_add (av_set_t *setp, expr_t expr)
2079 {
2080 av_set_t elem;
2081
2082 gcc_assert (!INSN_NOP_P (EXPR_INSN_RTX (expr)));
2083 elem = av_set_add_element (setp);
2084 copy_expr (_AV_SET_EXPR (elem), expr);
2085 }
2086
2087 /* Same, but do not copy EXPR. */
2088 static void
2089 av_set_add_nocopy (av_set_t *setp, expr_t expr)
2090 {
2091 av_set_t elem;
2092
2093 elem = av_set_add_element (setp);
2094 *_AV_SET_EXPR (elem) = *expr;
2095 }
2096
2097 /* Remove expr pointed to by IP from the av_set. */
2098 void
2099 av_set_iter_remove (av_set_iterator *ip)
2100 {
2101 clear_expr (_AV_SET_EXPR (*ip->lp));
2102 _list_iter_remove (ip);
2103 }
2104
2105 /* Search for an expr in SET, such that it's equivalent to SOUGHT_VINSN in the
2106 sense of vinsn_equal_p function. Return NULL if no such expr is
2107 in SET was found. */
2108 expr_t
2109 av_set_lookup (av_set_t set, vinsn_t sought_vinsn)
2110 {
2111 expr_t expr;
2112 av_set_iterator i;
2113
2114 FOR_EACH_EXPR (expr, i, set)
2115 if (vinsn_equal_p (EXPR_VINSN (expr), sought_vinsn))
2116 return expr;
2117 return NULL;
2118 }
2119
2120 /* Same, but also remove the EXPR found. */
2121 static expr_t
2122 av_set_lookup_and_remove (av_set_t *setp, vinsn_t sought_vinsn)
2123 {
2124 expr_t expr;
2125 av_set_iterator i;
2126
2127 FOR_EACH_EXPR_1 (expr, i, setp)
2128 if (vinsn_equal_p (EXPR_VINSN (expr), sought_vinsn))
2129 {
2130 _list_iter_remove_nofree (&i);
2131 return expr;
2132 }
2133 return NULL;
2134 }
2135
2136 /* Search for an expr in SET, such that it's equivalent to EXPR in the
2137 sense of vinsn_equal_p function of their vinsns, but not EXPR itself.
2138 Returns NULL if no such expr is in SET was found. */
2139 static expr_t
2140 av_set_lookup_other_equiv_expr (av_set_t set, expr_t expr)
2141 {
2142 expr_t cur_expr;
2143 av_set_iterator i;
2144
2145 FOR_EACH_EXPR (cur_expr, i, set)
2146 {
2147 if (cur_expr == expr)
2148 continue;
2149 if (vinsn_equal_p (EXPR_VINSN (cur_expr), EXPR_VINSN (expr)))
2150 return cur_expr;
2151 }
2152
2153 return NULL;
2154 }
2155
2156 /* If other expression is already in AVP, remove one of them. */
2157 expr_t
2158 merge_with_other_exprs (av_set_t *avp, av_set_iterator *ip, expr_t expr)
2159 {
2160 expr_t expr2;
2161
2162 expr2 = av_set_lookup_other_equiv_expr (*avp, expr);
2163 if (expr2 != NULL)
2164 {
2165 /* Reset target availability on merge, since taking it only from one
2166 of the exprs would be controversial for different code. */
2167 EXPR_TARGET_AVAILABLE (expr2) = -1;
2168 EXPR_USEFULNESS (expr2) = 0;
2169
2170 merge_expr (expr2, expr, NULL);
2171
2172 /* Fix usefulness as it should be now REG_BR_PROB_BASE. */
2173 EXPR_USEFULNESS (expr2) = REG_BR_PROB_BASE;
2174
2175 av_set_iter_remove (ip);
2176 return expr2;
2177 }
2178
2179 return expr;
2180 }
2181
2182 /* Return true if there is an expr that correlates to VI in SET. */
2183 bool
2184 av_set_is_in_p (av_set_t set, vinsn_t vi)
2185 {
2186 return av_set_lookup (set, vi) != NULL;
2187 }
2188
2189 /* Return a copy of SET. */
2190 av_set_t
2191 av_set_copy (av_set_t set)
2192 {
2193 expr_t expr;
2194 av_set_iterator i;
2195 av_set_t res = NULL;
2196
2197 FOR_EACH_EXPR (expr, i, set)
2198 av_set_add (&res, expr);
2199
2200 return res;
2201 }
2202
2203 /* Join two av sets that do not have common elements by attaching second set
2204 (pointed to by FROMP) to the end of first set (TO_TAILP must point to
2205 _AV_SET_NEXT of first set's last element). */
2206 static void
2207 join_distinct_sets (av_set_t *to_tailp, av_set_t *fromp)
2208 {
2209 gcc_assert (*to_tailp == NULL);
2210 *to_tailp = *fromp;
2211 *fromp = NULL;
2212 }
2213
2214 /* Makes set pointed to by TO to be the union of TO and FROM. Clear av_set
2215 pointed to by FROMP afterwards. */
2216 void
2217 av_set_union_and_clear (av_set_t *top, av_set_t *fromp, insn_t insn)
2218 {
2219 expr_t expr1;
2220 av_set_iterator i;
2221
2222 /* Delete from TOP all exprs, that present in FROMP. */
2223 FOR_EACH_EXPR_1 (expr1, i, top)
2224 {
2225 expr_t expr2 = av_set_lookup (*fromp, EXPR_VINSN (expr1));
2226
2227 if (expr2)
2228 {
2229 merge_expr (expr2, expr1, insn);
2230 av_set_iter_remove (&i);
2231 }
2232 }
2233
2234 join_distinct_sets (i.lp, fromp);
2235 }
2236
2237 /* Same as above, but also update availability of target register in
2238 TOP judging by TO_LV_SET and FROM_LV_SET. */
2239 void
2240 av_set_union_and_live (av_set_t *top, av_set_t *fromp, regset to_lv_set,
2241 regset from_lv_set, insn_t insn)
2242 {
2243 expr_t expr1;
2244 av_set_iterator i;
2245 av_set_t *to_tailp, in_both_set = NULL;
2246
2247 /* Delete from TOP all expres, that present in FROMP. */
2248 FOR_EACH_EXPR_1 (expr1, i, top)
2249 {
2250 expr_t expr2 = av_set_lookup_and_remove (fromp, EXPR_VINSN (expr1));
2251
2252 if (expr2)
2253 {
2254 /* It may be that the expressions have different destination
2255 registers, in which case we need to check liveness here. */
2256 if (EXPR_SEPARABLE_P (expr1))
2257 {
2258 int regno1 = (REG_P (EXPR_LHS (expr1))
2259 ? (int) expr_dest_regno (expr1) : -1);
2260 int regno2 = (REG_P (EXPR_LHS (expr2))
2261 ? (int) expr_dest_regno (expr2) : -1);
2262
2263 /* ??? We don't have a way to check restrictions for
2264 *other* register on the current path, we did it only
2265 for the current target register. Give up. */
2266 if (regno1 != regno2)
2267 EXPR_TARGET_AVAILABLE (expr2) = -1;
2268 }
2269 else if (EXPR_INSN_RTX (expr1) != EXPR_INSN_RTX (expr2))
2270 EXPR_TARGET_AVAILABLE (expr2) = -1;
2271
2272 merge_expr (expr2, expr1, insn);
2273 av_set_add_nocopy (&in_both_set, expr2);
2274 av_set_iter_remove (&i);
2275 }
2276 else
2277 /* EXPR1 is present in TOP, but not in FROMP. Check it on
2278 FROM_LV_SET. */
2279 set_unavailable_target_for_expr (expr1, from_lv_set);
2280 }
2281 to_tailp = i.lp;
2282
2283 /* These expressions are not present in TOP. Check liveness
2284 restrictions on TO_LV_SET. */
2285 FOR_EACH_EXPR (expr1, i, *fromp)
2286 set_unavailable_target_for_expr (expr1, to_lv_set);
2287
2288 join_distinct_sets (i.lp, &in_both_set);
2289 join_distinct_sets (to_tailp, fromp);
2290 }
2291
2292 /* Clear av_set pointed to by SETP. */
2293 void
2294 av_set_clear (av_set_t *setp)
2295 {
2296 expr_t expr;
2297 av_set_iterator i;
2298
2299 FOR_EACH_EXPR_1 (expr, i, setp)
2300 av_set_iter_remove (&i);
2301
2302 gcc_assert (*setp == NULL);
2303 }
2304
2305 /* Leave only one non-speculative element in the SETP. */
2306 void
2307 av_set_leave_one_nonspec (av_set_t *setp)
2308 {
2309 expr_t expr;
2310 av_set_iterator i;
2311 bool has_one_nonspec = false;
2312
2313 /* Keep all speculative exprs, and leave one non-speculative
2314 (the first one). */
2315 FOR_EACH_EXPR_1 (expr, i, setp)
2316 {
2317 if (!EXPR_SPEC_DONE_DS (expr))
2318 {
2319 if (has_one_nonspec)
2320 av_set_iter_remove (&i);
2321 else
2322 has_one_nonspec = true;
2323 }
2324 }
2325 }
2326
2327 /* Return the N'th element of the SET. */
2328 expr_t
2329 av_set_element (av_set_t set, int n)
2330 {
2331 expr_t expr;
2332 av_set_iterator i;
2333
2334 FOR_EACH_EXPR (expr, i, set)
2335 if (n-- == 0)
2336 return expr;
2337
2338 gcc_unreachable ();
2339 return NULL;
2340 }
2341
2342 /* Deletes all expressions from AVP that are conditional branches (IFs). */
2343 void
2344 av_set_substract_cond_branches (av_set_t *avp)
2345 {
2346 av_set_iterator i;
2347 expr_t expr;
2348
2349 FOR_EACH_EXPR_1 (expr, i, avp)
2350 if (vinsn_cond_branch_p (EXPR_VINSN (expr)))
2351 av_set_iter_remove (&i);
2352 }
2353
2354 /* Multiplies usefulness attribute of each member of av-set *AVP by
2355 value PROB / ALL_PROB. */
2356 void
2357 av_set_split_usefulness (av_set_t av, int prob, int all_prob)
2358 {
2359 av_set_iterator i;
2360 expr_t expr;
2361
2362 FOR_EACH_EXPR (expr, i, av)
2363 EXPR_USEFULNESS (expr) = (all_prob
2364 ? (EXPR_USEFULNESS (expr) * prob) / all_prob
2365 : 0);
2366 }
2367
2368 /* Leave in AVP only those expressions, which are present in AV,
2369 and return it, merging history expressions. */
2370 void
2371 av_set_code_motion_filter (av_set_t *avp, av_set_t av)
2372 {
2373 av_set_iterator i;
2374 expr_t expr, expr2;
2375
2376 FOR_EACH_EXPR_1 (expr, i, avp)
2377 if ((expr2 = av_set_lookup (av, EXPR_VINSN (expr))) == NULL)
2378 av_set_iter_remove (&i);
2379 else
2380 /* When updating av sets in bookkeeping blocks, we can add more insns
2381 there which will be transformed but the upper av sets will not
2382 reflect those transformations. We then fail to undo those
2383 when searching for such insns. So merge the history saved
2384 in the av set of the block we are processing. */
2385 merge_history_vect (&EXPR_HISTORY_OF_CHANGES (expr),
2386 EXPR_HISTORY_OF_CHANGES (expr2));
2387 }
2388
2389 \f
2390
2391 /* Dependence hooks to initialize insn data. */
2392
2393 /* This is used in hooks callable from dependence analysis when initializing
2394 instruction's data. */
2395 static struct
2396 {
2397 /* Where the dependence was found (lhs/rhs). */
2398 deps_where_t where;
2399
2400 /* The actual data object to initialize. */
2401 idata_t id;
2402
2403 /* True when the insn should not be made clonable. */
2404 bool force_unique_p;
2405
2406 /* True when insn should be treated as of type USE, i.e. never renamed. */
2407 bool force_use_p;
2408 } deps_init_id_data;
2409
2410
2411 /* Setup ID for INSN. FORCE_UNIQUE_P is true when INSN should not be
2412 clonable. */
2413 static void
2414 setup_id_for_insn (idata_t id, insn_t insn, bool force_unique_p)
2415 {
2416 int type;
2417
2418 /* Determine whether INSN could be cloned and return appropriate vinsn type.
2419 That clonable insns which can be separated into lhs and rhs have type SET.
2420 Other clonable insns have type USE. */
2421 type = GET_CODE (insn);
2422
2423 /* Only regular insns could be cloned. */
2424 if (type == INSN && !force_unique_p)
2425 type = SET;
2426 else if (type == JUMP_INSN && simplejump_p (insn))
2427 type = PC;
2428 else if (type == DEBUG_INSN)
2429 type = !force_unique_p ? USE : INSN;
2430
2431 IDATA_TYPE (id) = type;
2432 IDATA_REG_SETS (id) = get_clear_regset_from_pool ();
2433 IDATA_REG_USES (id) = get_clear_regset_from_pool ();
2434 IDATA_REG_CLOBBERS (id) = get_clear_regset_from_pool ();
2435 }
2436
2437 /* Start initializing insn data. */
2438 static void
2439 deps_init_id_start_insn (insn_t insn)
2440 {
2441 gcc_assert (deps_init_id_data.where == DEPS_IN_NOWHERE);
2442
2443 setup_id_for_insn (deps_init_id_data.id, insn,
2444 deps_init_id_data.force_unique_p);
2445 deps_init_id_data.where = DEPS_IN_INSN;
2446 }
2447
2448 /* Start initializing lhs data. */
2449 static void
2450 deps_init_id_start_lhs (rtx lhs)
2451 {
2452 gcc_assert (deps_init_id_data.where == DEPS_IN_INSN);
2453 gcc_assert (IDATA_LHS (deps_init_id_data.id) == NULL);
2454
2455 if (IDATA_TYPE (deps_init_id_data.id) == SET)
2456 {
2457 IDATA_LHS (deps_init_id_data.id) = lhs;
2458 deps_init_id_data.where = DEPS_IN_LHS;
2459 }
2460 }
2461
2462 /* Finish initializing lhs data. */
2463 static void
2464 deps_init_id_finish_lhs (void)
2465 {
2466 deps_init_id_data.where = DEPS_IN_INSN;
2467 }
2468
2469 /* Note a set of REGNO. */
2470 static void
2471 deps_init_id_note_reg_set (int regno)
2472 {
2473 haifa_note_reg_set (regno);
2474
2475 if (deps_init_id_data.where == DEPS_IN_RHS)
2476 deps_init_id_data.force_use_p = true;
2477
2478 if (IDATA_TYPE (deps_init_id_data.id) != PC)
2479 SET_REGNO_REG_SET (IDATA_REG_SETS (deps_init_id_data.id), regno);
2480
2481 #ifdef STACK_REGS
2482 /* Make instructions that set stack registers to be ineligible for
2483 renaming to avoid issues with find_used_regs. */
2484 if (IN_RANGE (regno, FIRST_STACK_REG, LAST_STACK_REG))
2485 deps_init_id_data.force_use_p = true;
2486 #endif
2487 }
2488
2489 /* Note a clobber of REGNO. */
2490 static void
2491 deps_init_id_note_reg_clobber (int regno)
2492 {
2493 haifa_note_reg_clobber (regno);
2494
2495 if (deps_init_id_data.where == DEPS_IN_RHS)
2496 deps_init_id_data.force_use_p = true;
2497
2498 if (IDATA_TYPE (deps_init_id_data.id) != PC)
2499 SET_REGNO_REG_SET (IDATA_REG_CLOBBERS (deps_init_id_data.id), regno);
2500 }
2501
2502 /* Note a use of REGNO. */
2503 static void
2504 deps_init_id_note_reg_use (int regno)
2505 {
2506 haifa_note_reg_use (regno);
2507
2508 if (IDATA_TYPE (deps_init_id_data.id) != PC)
2509 SET_REGNO_REG_SET (IDATA_REG_USES (deps_init_id_data.id), regno);
2510 }
2511
2512 /* Start initializing rhs data. */
2513 static void
2514 deps_init_id_start_rhs (rtx rhs)
2515 {
2516 gcc_assert (deps_init_id_data.where == DEPS_IN_INSN);
2517
2518 /* And there was no sel_deps_reset_to_insn (). */
2519 if (IDATA_LHS (deps_init_id_data.id) != NULL)
2520 {
2521 IDATA_RHS (deps_init_id_data.id) = rhs;
2522 deps_init_id_data.where = DEPS_IN_RHS;
2523 }
2524 }
2525
2526 /* Finish initializing rhs data. */
2527 static void
2528 deps_init_id_finish_rhs (void)
2529 {
2530 gcc_assert (deps_init_id_data.where == DEPS_IN_RHS
2531 || deps_init_id_data.where == DEPS_IN_INSN);
2532 deps_init_id_data.where = DEPS_IN_INSN;
2533 }
2534
2535 /* Finish initializing insn data. */
2536 static void
2537 deps_init_id_finish_insn (void)
2538 {
2539 gcc_assert (deps_init_id_data.where == DEPS_IN_INSN);
2540
2541 if (IDATA_TYPE (deps_init_id_data.id) == SET)
2542 {
2543 rtx lhs = IDATA_LHS (deps_init_id_data.id);
2544 rtx rhs = IDATA_RHS (deps_init_id_data.id);
2545
2546 if (lhs == NULL || rhs == NULL || !lhs_and_rhs_separable_p (lhs, rhs)
2547 || deps_init_id_data.force_use_p)
2548 {
2549 /* This should be a USE, as we don't want to schedule its RHS
2550 separately. However, we still want to have them recorded
2551 for the purposes of substitution. That's why we don't
2552 simply call downgrade_to_use () here. */
2553 gcc_assert (IDATA_TYPE (deps_init_id_data.id) == SET);
2554 gcc_assert (!lhs == !rhs);
2555
2556 IDATA_TYPE (deps_init_id_data.id) = USE;
2557 }
2558 }
2559
2560 deps_init_id_data.where = DEPS_IN_NOWHERE;
2561 }
2562
2563 /* This is dependence info used for initializing insn's data. */
2564 static struct sched_deps_info_def deps_init_id_sched_deps_info;
2565
2566 /* This initializes most of the static part of the above structure. */
2567 static const struct sched_deps_info_def const_deps_init_id_sched_deps_info =
2568 {
2569 NULL,
2570
2571 deps_init_id_start_insn,
2572 deps_init_id_finish_insn,
2573 deps_init_id_start_lhs,
2574 deps_init_id_finish_lhs,
2575 deps_init_id_start_rhs,
2576 deps_init_id_finish_rhs,
2577 deps_init_id_note_reg_set,
2578 deps_init_id_note_reg_clobber,
2579 deps_init_id_note_reg_use,
2580 NULL, /* note_mem_dep */
2581 NULL, /* note_dep */
2582
2583 0, /* use_cselib */
2584 0, /* use_deps_list */
2585 0 /* generate_spec_deps */
2586 };
2587
2588 /* Initialize INSN's lhs and rhs in ID. When FORCE_UNIQUE_P is true,
2589 we don't actually need information about lhs and rhs. */
2590 static void
2591 setup_id_lhs_rhs (idata_t id, insn_t insn, bool force_unique_p)
2592 {
2593 rtx pat = PATTERN (insn);
2594
2595 if (NONJUMP_INSN_P (insn)
2596 && GET_CODE (pat) == SET
2597 && !force_unique_p)
2598 {
2599 IDATA_RHS (id) = SET_SRC (pat);
2600 IDATA_LHS (id) = SET_DEST (pat);
2601 }
2602 else
2603 IDATA_LHS (id) = IDATA_RHS (id) = NULL;
2604 }
2605
2606 /* Possibly downgrade INSN to USE. */
2607 static void
2608 maybe_downgrade_id_to_use (idata_t id, insn_t insn)
2609 {
2610 bool must_be_use = false;
2611 unsigned uid = INSN_UID (insn);
2612 df_ref *rec;
2613 rtx lhs = IDATA_LHS (id);
2614 rtx rhs = IDATA_RHS (id);
2615
2616 /* We downgrade only SETs. */
2617 if (IDATA_TYPE (id) != SET)
2618 return;
2619
2620 if (!lhs || !lhs_and_rhs_separable_p (lhs, rhs))
2621 {
2622 IDATA_TYPE (id) = USE;
2623 return;
2624 }
2625
2626 for (rec = DF_INSN_UID_DEFS (uid); *rec; rec++)
2627 {
2628 df_ref def = *rec;
2629
2630 if (DF_REF_INSN (def)
2631 && DF_REF_FLAGS_IS_SET (def, DF_REF_PRE_POST_MODIFY)
2632 && loc_mentioned_in_p (DF_REF_LOC (def), IDATA_RHS (id)))
2633 {
2634 must_be_use = true;
2635 break;
2636 }
2637
2638 #ifdef STACK_REGS
2639 /* Make instructions that set stack registers to be ineligible for
2640 renaming to avoid issues with find_used_regs. */
2641 if (IN_RANGE (DF_REF_REGNO (def), FIRST_STACK_REG, LAST_STACK_REG))
2642 {
2643 must_be_use = true;
2644 break;
2645 }
2646 #endif
2647 }
2648
2649 if (must_be_use)
2650 IDATA_TYPE (id) = USE;
2651 }
2652
2653 /* Setup register sets describing INSN in ID. */
2654 static void
2655 setup_id_reg_sets (idata_t id, insn_t insn)
2656 {
2657 unsigned uid = INSN_UID (insn);
2658 df_ref *rec;
2659 regset tmp = get_clear_regset_from_pool ();
2660
2661 for (rec = DF_INSN_UID_DEFS (uid); *rec; rec++)
2662 {
2663 df_ref def = *rec;
2664 unsigned int regno = DF_REF_REGNO (def);
2665
2666 /* Post modifies are treated like clobbers by sched-deps.c. */
2667 if (DF_REF_FLAGS_IS_SET (def, (DF_REF_MUST_CLOBBER
2668 | DF_REF_PRE_POST_MODIFY)))
2669 SET_REGNO_REG_SET (IDATA_REG_CLOBBERS (id), regno);
2670 else if (! DF_REF_FLAGS_IS_SET (def, DF_REF_MAY_CLOBBER))
2671 {
2672 SET_REGNO_REG_SET (IDATA_REG_SETS (id), regno);
2673
2674 #ifdef STACK_REGS
2675 /* For stack registers, treat writes to them as writes
2676 to the first one to be consistent with sched-deps.c. */
2677 if (IN_RANGE (regno, FIRST_STACK_REG, LAST_STACK_REG))
2678 SET_REGNO_REG_SET (IDATA_REG_SETS (id), FIRST_STACK_REG);
2679 #endif
2680 }
2681 /* Mark special refs that generate read/write def pair. */
2682 if (DF_REF_FLAGS_IS_SET (def, DF_REF_CONDITIONAL)
2683 || regno == STACK_POINTER_REGNUM)
2684 bitmap_set_bit (tmp, regno);
2685 }
2686
2687 for (rec = DF_INSN_UID_USES (uid); *rec; rec++)
2688 {
2689 df_ref use = *rec;
2690 unsigned int regno = DF_REF_REGNO (use);
2691
2692 /* When these refs are met for the first time, skip them, as
2693 these uses are just counterparts of some defs. */
2694 if (bitmap_bit_p (tmp, regno))
2695 bitmap_clear_bit (tmp, regno);
2696 else if (! DF_REF_FLAGS_IS_SET (use, DF_REF_CALL_STACK_USAGE))
2697 {
2698 SET_REGNO_REG_SET (IDATA_REG_USES (id), regno);
2699
2700 #ifdef STACK_REGS
2701 /* For stack registers, treat reads from them as reads from
2702 the first one to be consistent with sched-deps.c. */
2703 if (IN_RANGE (regno, FIRST_STACK_REG, LAST_STACK_REG))
2704 SET_REGNO_REG_SET (IDATA_REG_USES (id), FIRST_STACK_REG);
2705 #endif
2706 }
2707 }
2708
2709 return_regset_to_pool (tmp);
2710 }
2711
2712 /* Initialize instruction data for INSN in ID using DF's data. */
2713 static void
2714 init_id_from_df (idata_t id, insn_t insn, bool force_unique_p)
2715 {
2716 gcc_assert (DF_INSN_UID_SAFE_GET (INSN_UID (insn)) != NULL);
2717
2718 setup_id_for_insn (id, insn, force_unique_p);
2719 setup_id_lhs_rhs (id, insn, force_unique_p);
2720
2721 if (INSN_NOP_P (insn))
2722 return;
2723
2724 maybe_downgrade_id_to_use (id, insn);
2725 setup_id_reg_sets (id, insn);
2726 }
2727
2728 /* Initialize instruction data for INSN in ID. */
2729 static void
2730 deps_init_id (idata_t id, insn_t insn, bool force_unique_p)
2731 {
2732 struct deps_desc _dc, *dc = &_dc;
2733
2734 deps_init_id_data.where = DEPS_IN_NOWHERE;
2735 deps_init_id_data.id = id;
2736 deps_init_id_data.force_unique_p = force_unique_p;
2737 deps_init_id_data.force_use_p = false;
2738
2739 init_deps (dc, false);
2740
2741 memcpy (&deps_init_id_sched_deps_info,
2742 &const_deps_init_id_sched_deps_info,
2743 sizeof (deps_init_id_sched_deps_info));
2744
2745 if (spec_info != NULL)
2746 deps_init_id_sched_deps_info.generate_spec_deps = 1;
2747
2748 sched_deps_info = &deps_init_id_sched_deps_info;
2749
2750 deps_analyze_insn (dc, insn);
2751
2752 free_deps (dc);
2753
2754 deps_init_id_data.id = NULL;
2755 }
2756
2757 \f
2758 struct sched_scan_info_def
2759 {
2760 /* This hook notifies scheduler frontend to extend its internal per basic
2761 block data structures. This hook should be called once before a series of
2762 calls to bb_init (). */
2763 void (*extend_bb) (void);
2764
2765 /* This hook makes scheduler frontend to initialize its internal data
2766 structures for the passed basic block. */
2767 void (*init_bb) (basic_block);
2768
2769 /* This hook notifies scheduler frontend to extend its internal per insn data
2770 structures. This hook should be called once before a series of calls to
2771 insn_init (). */
2772 void (*extend_insn) (void);
2773
2774 /* This hook makes scheduler frontend to initialize its internal data
2775 structures for the passed insn. */
2776 void (*init_insn) (rtx);
2777 };
2778
2779 /* A driver function to add a set of basic blocks (BBS) to the
2780 scheduling region. */
2781 static void
2782 sched_scan (const struct sched_scan_info_def *ssi, bb_vec_t bbs)
2783 {
2784 unsigned i;
2785 basic_block bb;
2786
2787 if (ssi->extend_bb)
2788 ssi->extend_bb ();
2789
2790 if (ssi->init_bb)
2791 FOR_EACH_VEC_ELT (bbs, i, bb)
2792 ssi->init_bb (bb);
2793
2794 if (ssi->extend_insn)
2795 ssi->extend_insn ();
2796
2797 if (ssi->init_insn)
2798 FOR_EACH_VEC_ELT (bbs, i, bb)
2799 {
2800 rtx insn;
2801
2802 FOR_BB_INSNS (bb, insn)
2803 ssi->init_insn (insn);
2804 }
2805 }
2806
2807 /* Implement hooks for collecting fundamental insn properties like if insn is
2808 an ASM or is within a SCHED_GROUP. */
2809
2810 /* True when a "one-time init" data for INSN was already inited. */
2811 static bool
2812 first_time_insn_init (insn_t insn)
2813 {
2814 return INSN_LIVE (insn) == NULL;
2815 }
2816
2817 /* Hash an entry in a transformed_insns hashtable. */
2818 static hashval_t
2819 hash_transformed_insns (const void *p)
2820 {
2821 return VINSN_HASH_RTX (((const struct transformed_insns *) p)->vinsn_old);
2822 }
2823
2824 /* Compare the entries in a transformed_insns hashtable. */
2825 static int
2826 eq_transformed_insns (const void *p, const void *q)
2827 {
2828 rtx i1 = VINSN_INSN_RTX (((const struct transformed_insns *) p)->vinsn_old);
2829 rtx i2 = VINSN_INSN_RTX (((const struct transformed_insns *) q)->vinsn_old);
2830
2831 if (INSN_UID (i1) == INSN_UID (i2))
2832 return 1;
2833 return rtx_equal_p (PATTERN (i1), PATTERN (i2));
2834 }
2835
2836 /* Free an entry in a transformed_insns hashtable. */
2837 static void
2838 free_transformed_insns (void *p)
2839 {
2840 struct transformed_insns *pti = (struct transformed_insns *) p;
2841
2842 vinsn_detach (pti->vinsn_old);
2843 vinsn_detach (pti->vinsn_new);
2844 free (pti);
2845 }
2846
2847 /* Init the s_i_d data for INSN which should be inited just once, when
2848 we first see the insn. */
2849 static void
2850 init_first_time_insn_data (insn_t insn)
2851 {
2852 /* This should not be set if this is the first time we init data for
2853 insn. */
2854 gcc_assert (first_time_insn_init (insn));
2855
2856 /* These are needed for nops too. */
2857 INSN_LIVE (insn) = get_regset_from_pool ();
2858 INSN_LIVE_VALID_P (insn) = false;
2859
2860 if (!INSN_NOP_P (insn))
2861 {
2862 INSN_ANALYZED_DEPS (insn) = BITMAP_ALLOC (NULL);
2863 INSN_FOUND_DEPS (insn) = BITMAP_ALLOC (NULL);
2864 INSN_TRANSFORMED_INSNS (insn)
2865 = htab_create (16, hash_transformed_insns,
2866 eq_transformed_insns, free_transformed_insns);
2867 init_deps (&INSN_DEPS_CONTEXT (insn), true);
2868 }
2869 }
2870
2871 /* Free almost all above data for INSN that is scheduled already.
2872 Used for extra-large basic blocks. */
2873 void
2874 free_data_for_scheduled_insn (insn_t insn)
2875 {
2876 gcc_assert (! first_time_insn_init (insn));
2877
2878 if (! INSN_ANALYZED_DEPS (insn))
2879 return;
2880
2881 BITMAP_FREE (INSN_ANALYZED_DEPS (insn));
2882 BITMAP_FREE (INSN_FOUND_DEPS (insn));
2883 htab_delete (INSN_TRANSFORMED_INSNS (insn));
2884
2885 /* This is allocated only for bookkeeping insns. */
2886 if (INSN_ORIGINATORS (insn))
2887 BITMAP_FREE (INSN_ORIGINATORS (insn));
2888 free_deps (&INSN_DEPS_CONTEXT (insn));
2889
2890 INSN_ANALYZED_DEPS (insn) = NULL;
2891
2892 /* Clear the readonly flag so we would ICE when trying to recalculate
2893 the deps context (as we believe that it should not happen). */
2894 (&INSN_DEPS_CONTEXT (insn))->readonly = 0;
2895 }
2896
2897 /* Free the same data as above for INSN. */
2898 static void
2899 free_first_time_insn_data (insn_t insn)
2900 {
2901 gcc_assert (! first_time_insn_init (insn));
2902
2903 free_data_for_scheduled_insn (insn);
2904 return_regset_to_pool (INSN_LIVE (insn));
2905 INSN_LIVE (insn) = NULL;
2906 INSN_LIVE_VALID_P (insn) = false;
2907 }
2908
2909 /* Initialize region-scope data structures for basic blocks. */
2910 static void
2911 init_global_and_expr_for_bb (basic_block bb)
2912 {
2913 if (sel_bb_empty_p (bb))
2914 return;
2915
2916 invalidate_av_set (bb);
2917 }
2918
2919 /* Data for global dependency analysis (to initialize CANT_MOVE and
2920 SCHED_GROUP_P). */
2921 static struct
2922 {
2923 /* Previous insn. */
2924 insn_t prev_insn;
2925 } init_global_data;
2926
2927 /* Determine if INSN is in the sched_group, is an asm or should not be
2928 cloned. After that initialize its expr. */
2929 static void
2930 init_global_and_expr_for_insn (insn_t insn)
2931 {
2932 if (LABEL_P (insn))
2933 return;
2934
2935 if (NOTE_INSN_BASIC_BLOCK_P (insn))
2936 {
2937 init_global_data.prev_insn = NULL_RTX;
2938 return;
2939 }
2940
2941 gcc_assert (INSN_P (insn));
2942
2943 if (SCHED_GROUP_P (insn))
2944 /* Setup a sched_group. */
2945 {
2946 insn_t prev_insn = init_global_data.prev_insn;
2947
2948 if (prev_insn)
2949 INSN_SCHED_NEXT (prev_insn) = insn;
2950
2951 init_global_data.prev_insn = insn;
2952 }
2953 else
2954 init_global_data.prev_insn = NULL_RTX;
2955
2956 if (GET_CODE (PATTERN (insn)) == ASM_INPUT
2957 || asm_noperands (PATTERN (insn)) >= 0)
2958 /* Mark INSN as an asm. */
2959 INSN_ASM_P (insn) = true;
2960
2961 {
2962 bool force_unique_p;
2963 ds_t spec_done_ds;
2964
2965 /* Certain instructions cannot be cloned, and frame related insns and
2966 the insn adjacent to NOTE_INSN_EPILOGUE_BEG cannot be moved out of
2967 their block. */
2968 if (prologue_epilogue_contains (insn))
2969 {
2970 if (RTX_FRAME_RELATED_P (insn))
2971 CANT_MOVE (insn) = 1;
2972 else
2973 {
2974 rtx note;
2975 for (note = REG_NOTES (insn); note; note = XEXP (note, 1))
2976 if (REG_NOTE_KIND (note) == REG_SAVE_NOTE
2977 && ((enum insn_note) INTVAL (XEXP (note, 0))
2978 == NOTE_INSN_EPILOGUE_BEG))
2979 {
2980 CANT_MOVE (insn) = 1;
2981 break;
2982 }
2983 }
2984 force_unique_p = true;
2985 }
2986 else
2987 if (CANT_MOVE (insn)
2988 || INSN_ASM_P (insn)
2989 || SCHED_GROUP_P (insn)
2990 || CALL_P (insn)
2991 /* Exception handling insns are always unique. */
2992 || (cfun->can_throw_non_call_exceptions && can_throw_internal (insn))
2993 /* TRAP_IF though have an INSN code is control_flow_insn_p (). */
2994 || control_flow_insn_p (insn)
2995 || volatile_insn_p (PATTERN (insn))
2996 || (targetm.cannot_copy_insn_p
2997 && targetm.cannot_copy_insn_p (insn)))
2998 force_unique_p = true;
2999 else
3000 force_unique_p = false;
3001
3002 if (targetm.sched.get_insn_spec_ds)
3003 {
3004 spec_done_ds = targetm.sched.get_insn_spec_ds (insn);
3005 spec_done_ds = ds_get_max_dep_weak (spec_done_ds);
3006 }
3007 else
3008 spec_done_ds = 0;
3009
3010 /* Initialize INSN's expr. */
3011 init_expr (INSN_EXPR (insn), vinsn_create (insn, force_unique_p), 0,
3012 REG_BR_PROB_BASE, INSN_PRIORITY (insn), 0, BLOCK_NUM (insn),
3013 spec_done_ds, 0, 0, vNULL, true,
3014 false, false, false, CANT_MOVE (insn));
3015 }
3016
3017 init_first_time_insn_data (insn);
3018 }
3019
3020 /* Scan the region and initialize instruction data for basic blocks BBS. */
3021 void
3022 sel_init_global_and_expr (bb_vec_t bbs)
3023 {
3024 /* ??? It would be nice to implement push / pop scheme for sched_infos. */
3025 const struct sched_scan_info_def ssi =
3026 {
3027 NULL, /* extend_bb */
3028 init_global_and_expr_for_bb, /* init_bb */
3029 extend_insn_data, /* extend_insn */
3030 init_global_and_expr_for_insn /* init_insn */
3031 };
3032
3033 sched_scan (&ssi, bbs);
3034 }
3035
3036 /* Finalize region-scope data structures for basic blocks. */
3037 static void
3038 finish_global_and_expr_for_bb (basic_block bb)
3039 {
3040 av_set_clear (&BB_AV_SET (bb));
3041 BB_AV_LEVEL (bb) = 0;
3042 }
3043
3044 /* Finalize INSN's data. */
3045 static void
3046 finish_global_and_expr_insn (insn_t insn)
3047 {
3048 if (LABEL_P (insn) || NOTE_INSN_BASIC_BLOCK_P (insn))
3049 return;
3050
3051 gcc_assert (INSN_P (insn));
3052
3053 if (INSN_LUID (insn) > 0)
3054 {
3055 free_first_time_insn_data (insn);
3056 INSN_WS_LEVEL (insn) = 0;
3057 CANT_MOVE (insn) = 0;
3058
3059 /* We can no longer assert this, as vinsns of this insn could be
3060 easily live in other insn's caches. This should be changed to
3061 a counter-like approach among all vinsns. */
3062 gcc_assert (true || VINSN_COUNT (INSN_VINSN (insn)) == 1);
3063 clear_expr (INSN_EXPR (insn));
3064 }
3065 }
3066
3067 /* Finalize per instruction data for the whole region. */
3068 void
3069 sel_finish_global_and_expr (void)
3070 {
3071 {
3072 bb_vec_t bbs;
3073 int i;
3074
3075 bbs.create (current_nr_blocks);
3076
3077 for (i = 0; i < current_nr_blocks; i++)
3078 bbs.quick_push (BASIC_BLOCK_FOR_FN (cfun, BB_TO_BLOCK (i)));
3079
3080 /* Clear AV_SETs and INSN_EXPRs. */
3081 {
3082 const struct sched_scan_info_def ssi =
3083 {
3084 NULL, /* extend_bb */
3085 finish_global_and_expr_for_bb, /* init_bb */
3086 NULL, /* extend_insn */
3087 finish_global_and_expr_insn /* init_insn */
3088 };
3089
3090 sched_scan (&ssi, bbs);
3091 }
3092
3093 bbs.release ();
3094 }
3095
3096 finish_insns ();
3097 }
3098 \f
3099
3100 /* In the below hooks, we merely calculate whether or not a dependence
3101 exists, and in what part of insn. However, we will need more data
3102 when we'll start caching dependence requests. */
3103
3104 /* Container to hold information for dependency analysis. */
3105 static struct
3106 {
3107 deps_t dc;
3108
3109 /* A variable to track which part of rtx we are scanning in
3110 sched-deps.c: sched_analyze_insn (). */
3111 deps_where_t where;
3112
3113 /* Current producer. */
3114 insn_t pro;
3115
3116 /* Current consumer. */
3117 vinsn_t con;
3118
3119 /* Is SEL_DEPS_HAS_DEP_P[DEPS_IN_X] is true, then X has a dependence.
3120 X is from { INSN, LHS, RHS }. */
3121 ds_t has_dep_p[DEPS_IN_NOWHERE];
3122 } has_dependence_data;
3123
3124 /* Start analyzing dependencies of INSN. */
3125 static void
3126 has_dependence_start_insn (insn_t insn ATTRIBUTE_UNUSED)
3127 {
3128 gcc_assert (has_dependence_data.where == DEPS_IN_NOWHERE);
3129
3130 has_dependence_data.where = DEPS_IN_INSN;
3131 }
3132
3133 /* Finish analyzing dependencies of an insn. */
3134 static void
3135 has_dependence_finish_insn (void)
3136 {
3137 gcc_assert (has_dependence_data.where == DEPS_IN_INSN);
3138
3139 has_dependence_data.where = DEPS_IN_NOWHERE;
3140 }
3141
3142 /* Start analyzing dependencies of LHS. */
3143 static void
3144 has_dependence_start_lhs (rtx lhs ATTRIBUTE_UNUSED)
3145 {
3146 gcc_assert (has_dependence_data.where == DEPS_IN_INSN);
3147
3148 if (VINSN_LHS (has_dependence_data.con) != NULL)
3149 has_dependence_data.where = DEPS_IN_LHS;
3150 }
3151
3152 /* Finish analyzing dependencies of an lhs. */
3153 static void
3154 has_dependence_finish_lhs (void)
3155 {
3156 has_dependence_data.where = DEPS_IN_INSN;
3157 }
3158
3159 /* Start analyzing dependencies of RHS. */
3160 static void
3161 has_dependence_start_rhs (rtx rhs ATTRIBUTE_UNUSED)
3162 {
3163 gcc_assert (has_dependence_data.where == DEPS_IN_INSN);
3164
3165 if (VINSN_RHS (has_dependence_data.con) != NULL)
3166 has_dependence_data.where = DEPS_IN_RHS;
3167 }
3168
3169 /* Start analyzing dependencies of an rhs. */
3170 static void
3171 has_dependence_finish_rhs (void)
3172 {
3173 gcc_assert (has_dependence_data.where == DEPS_IN_RHS
3174 || has_dependence_data.where == DEPS_IN_INSN);
3175
3176 has_dependence_data.where = DEPS_IN_INSN;
3177 }
3178
3179 /* Note a set of REGNO. */
3180 static void
3181 has_dependence_note_reg_set (int regno)
3182 {
3183 struct deps_reg *reg_last = &has_dependence_data.dc->reg_last[regno];
3184
3185 if (!sched_insns_conditions_mutex_p (has_dependence_data.pro,
3186 VINSN_INSN_RTX
3187 (has_dependence_data.con)))
3188 {
3189 ds_t *dsp = &has_dependence_data.has_dep_p[has_dependence_data.where];
3190
3191 if (reg_last->sets != NULL
3192 || reg_last->clobbers != NULL)
3193 *dsp = (*dsp & ~SPECULATIVE) | DEP_OUTPUT;
3194
3195 if (reg_last->uses || reg_last->implicit_sets)
3196 *dsp = (*dsp & ~SPECULATIVE) | DEP_ANTI;
3197 }
3198 }
3199
3200 /* Note a clobber of REGNO. */
3201 static void
3202 has_dependence_note_reg_clobber (int regno)
3203 {
3204 struct deps_reg *reg_last = &has_dependence_data.dc->reg_last[regno];
3205
3206 if (!sched_insns_conditions_mutex_p (has_dependence_data.pro,
3207 VINSN_INSN_RTX
3208 (has_dependence_data.con)))
3209 {
3210 ds_t *dsp = &has_dependence_data.has_dep_p[has_dependence_data.where];
3211
3212 if (reg_last->sets)
3213 *dsp = (*dsp & ~SPECULATIVE) | DEP_OUTPUT;
3214
3215 if (reg_last->uses || reg_last->implicit_sets)
3216 *dsp = (*dsp & ~SPECULATIVE) | DEP_ANTI;
3217 }
3218 }
3219
3220 /* Note a use of REGNO. */
3221 static void
3222 has_dependence_note_reg_use (int regno)
3223 {
3224 struct deps_reg *reg_last = &has_dependence_data.dc->reg_last[regno];
3225
3226 if (!sched_insns_conditions_mutex_p (has_dependence_data.pro,
3227 VINSN_INSN_RTX
3228 (has_dependence_data.con)))
3229 {
3230 ds_t *dsp = &has_dependence_data.has_dep_p[has_dependence_data.where];
3231
3232 if (reg_last->sets)
3233 *dsp = (*dsp & ~SPECULATIVE) | DEP_TRUE;
3234
3235 if (reg_last->clobbers || reg_last->implicit_sets)
3236 *dsp = (*dsp & ~SPECULATIVE) | DEP_ANTI;
3237
3238 /* Merge BE_IN_SPEC bits into *DSP when the dependency producer
3239 is actually a check insn. We need to do this for any register
3240 read-read dependency with the check unless we track properly
3241 all registers written by BE_IN_SPEC-speculated insns, as
3242 we don't have explicit dependence lists. See PR 53975. */
3243 if (reg_last->uses)
3244 {
3245 ds_t pro_spec_checked_ds;
3246
3247 pro_spec_checked_ds = INSN_SPEC_CHECKED_DS (has_dependence_data.pro);
3248 pro_spec_checked_ds = ds_get_max_dep_weak (pro_spec_checked_ds);
3249
3250 if (pro_spec_checked_ds != 0)
3251 *dsp = ds_full_merge (*dsp, pro_spec_checked_ds,
3252 NULL_RTX, NULL_RTX);
3253 }
3254 }
3255 }
3256
3257 /* Note a memory dependence. */
3258 static void
3259 has_dependence_note_mem_dep (rtx mem ATTRIBUTE_UNUSED,
3260 rtx pending_mem ATTRIBUTE_UNUSED,
3261 insn_t pending_insn ATTRIBUTE_UNUSED,
3262 ds_t ds ATTRIBUTE_UNUSED)
3263 {
3264 if (!sched_insns_conditions_mutex_p (has_dependence_data.pro,
3265 VINSN_INSN_RTX (has_dependence_data.con)))
3266 {
3267 ds_t *dsp = &has_dependence_data.has_dep_p[has_dependence_data.where];
3268
3269 *dsp = ds_full_merge (ds, *dsp, pending_mem, mem);
3270 }
3271 }
3272
3273 /* Note a dependence. */
3274 static void
3275 has_dependence_note_dep (insn_t pro ATTRIBUTE_UNUSED,
3276 ds_t ds ATTRIBUTE_UNUSED)
3277 {
3278 if (!sched_insns_conditions_mutex_p (has_dependence_data.pro,
3279 VINSN_INSN_RTX (has_dependence_data.con)))
3280 {
3281 ds_t *dsp = &has_dependence_data.has_dep_p[has_dependence_data.where];
3282
3283 *dsp = ds_full_merge (ds, *dsp, NULL_RTX, NULL_RTX);
3284 }
3285 }
3286
3287 /* Mark the insn as having a hard dependence that prevents speculation. */
3288 void
3289 sel_mark_hard_insn (rtx insn)
3290 {
3291 int i;
3292
3293 /* Only work when we're in has_dependence_p mode.
3294 ??? This is a hack, this should actually be a hook. */
3295 if (!has_dependence_data.dc || !has_dependence_data.pro)
3296 return;
3297
3298 gcc_assert (insn == VINSN_INSN_RTX (has_dependence_data.con));
3299 gcc_assert (has_dependence_data.where == DEPS_IN_INSN);
3300
3301 for (i = 0; i < DEPS_IN_NOWHERE; i++)
3302 has_dependence_data.has_dep_p[i] &= ~SPECULATIVE;
3303 }
3304
3305 /* This structure holds the hooks for the dependency analysis used when
3306 actually processing dependencies in the scheduler. */
3307 static struct sched_deps_info_def has_dependence_sched_deps_info;
3308
3309 /* This initializes most of the fields of the above structure. */
3310 static const struct sched_deps_info_def const_has_dependence_sched_deps_info =
3311 {
3312 NULL,
3313
3314 has_dependence_start_insn,
3315 has_dependence_finish_insn,
3316 has_dependence_start_lhs,
3317 has_dependence_finish_lhs,
3318 has_dependence_start_rhs,
3319 has_dependence_finish_rhs,
3320 has_dependence_note_reg_set,
3321 has_dependence_note_reg_clobber,
3322 has_dependence_note_reg_use,
3323 has_dependence_note_mem_dep,
3324 has_dependence_note_dep,
3325
3326 0, /* use_cselib */
3327 0, /* use_deps_list */
3328 0 /* generate_spec_deps */
3329 };
3330
3331 /* Initialize has_dependence_sched_deps_info with extra spec field. */
3332 static void
3333 setup_has_dependence_sched_deps_info (void)
3334 {
3335 memcpy (&has_dependence_sched_deps_info,
3336 &const_has_dependence_sched_deps_info,
3337 sizeof (has_dependence_sched_deps_info));
3338
3339 if (spec_info != NULL)
3340 has_dependence_sched_deps_info.generate_spec_deps = 1;
3341
3342 sched_deps_info = &has_dependence_sched_deps_info;
3343 }
3344
3345 /* Remove all dependences found and recorded in has_dependence_data array. */
3346 void
3347 sel_clear_has_dependence (void)
3348 {
3349 int i;
3350
3351 for (i = 0; i < DEPS_IN_NOWHERE; i++)
3352 has_dependence_data.has_dep_p[i] = 0;
3353 }
3354
3355 /* Return nonzero if EXPR has is dependent upon PRED. Return the pointer
3356 to the dependence information array in HAS_DEP_PP. */
3357 ds_t
3358 has_dependence_p (expr_t expr, insn_t pred, ds_t **has_dep_pp)
3359 {
3360 int i;
3361 ds_t ds;
3362 struct deps_desc *dc;
3363
3364 if (INSN_SIMPLEJUMP_P (pred))
3365 /* Unconditional jump is just a transfer of control flow.
3366 Ignore it. */
3367 return false;
3368
3369 dc = &INSN_DEPS_CONTEXT (pred);
3370
3371 /* We init this field lazily. */
3372 if (dc->reg_last == NULL)
3373 init_deps_reg_last (dc);
3374
3375 if (!dc->readonly)
3376 {
3377 has_dependence_data.pro = NULL;
3378 /* Initialize empty dep context with information about PRED. */
3379 advance_deps_context (dc, pred);
3380 dc->readonly = 1;
3381 }
3382
3383 has_dependence_data.where = DEPS_IN_NOWHERE;
3384 has_dependence_data.pro = pred;
3385 has_dependence_data.con = EXPR_VINSN (expr);
3386 has_dependence_data.dc = dc;
3387
3388 sel_clear_has_dependence ();
3389
3390 /* Now catch all dependencies that would be generated between PRED and
3391 INSN. */
3392 setup_has_dependence_sched_deps_info ();
3393 deps_analyze_insn (dc, EXPR_INSN_RTX (expr));
3394 has_dependence_data.dc = NULL;
3395
3396 /* When a barrier was found, set DEPS_IN_INSN bits. */
3397 if (dc->last_reg_pending_barrier == TRUE_BARRIER)
3398 has_dependence_data.has_dep_p[DEPS_IN_INSN] = DEP_TRUE;
3399 else if (dc->last_reg_pending_barrier == MOVE_BARRIER)
3400 has_dependence_data.has_dep_p[DEPS_IN_INSN] = DEP_ANTI;
3401
3402 /* Do not allow stores to memory to move through checks. Currently
3403 we don't move this to sched-deps.c as the check doesn't have
3404 obvious places to which this dependence can be attached.
3405 FIMXE: this should go to a hook. */
3406 if (EXPR_LHS (expr)
3407 && MEM_P (EXPR_LHS (expr))
3408 && sel_insn_is_speculation_check (pred))
3409 has_dependence_data.has_dep_p[DEPS_IN_INSN] = DEP_ANTI;
3410
3411 *has_dep_pp = has_dependence_data.has_dep_p;
3412 ds = 0;
3413 for (i = 0; i < DEPS_IN_NOWHERE; i++)
3414 ds = ds_full_merge (ds, has_dependence_data.has_dep_p[i],
3415 NULL_RTX, NULL_RTX);
3416
3417 return ds;
3418 }
3419 \f
3420
3421 /* Dependence hooks implementation that checks dependence latency constraints
3422 on the insns being scheduled. The entry point for these routines is
3423 tick_check_p predicate. */
3424
3425 static struct
3426 {
3427 /* An expr we are currently checking. */
3428 expr_t expr;
3429
3430 /* A minimal cycle for its scheduling. */
3431 int cycle;
3432
3433 /* Whether we have seen a true dependence while checking. */
3434 bool seen_true_dep_p;
3435 } tick_check_data;
3436
3437 /* Update minimal scheduling cycle for tick_check_insn given that it depends
3438 on PRO with status DS and weight DW. */
3439 static void
3440 tick_check_dep_with_dw (insn_t pro_insn, ds_t ds, dw_t dw)
3441 {
3442 expr_t con_expr = tick_check_data.expr;
3443 insn_t con_insn = EXPR_INSN_RTX (con_expr);
3444
3445 if (con_insn != pro_insn)
3446 {
3447 enum reg_note dt;
3448 int tick;
3449
3450 if (/* PROducer was removed from above due to pipelining. */
3451 !INSN_IN_STREAM_P (pro_insn)
3452 /* Or PROducer was originally on the next iteration regarding the
3453 CONsumer. */
3454 || (INSN_SCHED_TIMES (pro_insn)
3455 - EXPR_SCHED_TIMES (con_expr)) > 1)
3456 /* Don't count this dependence. */
3457 return;
3458
3459 dt = ds_to_dt (ds);
3460 if (dt == REG_DEP_TRUE)
3461 tick_check_data.seen_true_dep_p = true;
3462
3463 gcc_assert (INSN_SCHED_CYCLE (pro_insn) > 0);
3464
3465 {
3466 dep_def _dep, *dep = &_dep;
3467
3468 init_dep (dep, pro_insn, con_insn, dt);
3469
3470 tick = INSN_SCHED_CYCLE (pro_insn) + dep_cost_1 (dep, dw);
3471 }
3472
3473 /* When there are several kinds of dependencies between pro and con,
3474 only REG_DEP_TRUE should be taken into account. */
3475 if (tick > tick_check_data.cycle
3476 && (dt == REG_DEP_TRUE || !tick_check_data.seen_true_dep_p))
3477 tick_check_data.cycle = tick;
3478 }
3479 }
3480
3481 /* An implementation of note_dep hook. */
3482 static void
3483 tick_check_note_dep (insn_t pro, ds_t ds)
3484 {
3485 tick_check_dep_with_dw (pro, ds, 0);
3486 }
3487
3488 /* An implementation of note_mem_dep hook. */
3489 static void
3490 tick_check_note_mem_dep (rtx mem1, rtx mem2, insn_t pro, ds_t ds)
3491 {
3492 dw_t dw;
3493
3494 dw = (ds_to_dt (ds) == REG_DEP_TRUE
3495 ? estimate_dep_weak (mem1, mem2)
3496 : 0);
3497
3498 tick_check_dep_with_dw (pro, ds, dw);
3499 }
3500
3501 /* This structure contains hooks for dependence analysis used when determining
3502 whether an insn is ready for scheduling. */
3503 static struct sched_deps_info_def tick_check_sched_deps_info =
3504 {
3505 NULL,
3506
3507 NULL,
3508 NULL,
3509 NULL,
3510 NULL,
3511 NULL,
3512 NULL,
3513 haifa_note_reg_set,
3514 haifa_note_reg_clobber,
3515 haifa_note_reg_use,
3516 tick_check_note_mem_dep,
3517 tick_check_note_dep,
3518
3519 0, 0, 0
3520 };
3521
3522 /* Estimate number of cycles from the current cycle of FENCE until EXPR can be
3523 scheduled. Return 0 if all data from producers in DC is ready. */
3524 int
3525 tick_check_p (expr_t expr, deps_t dc, fence_t fence)
3526 {
3527 int cycles_left;
3528 /* Initialize variables. */
3529 tick_check_data.expr = expr;
3530 tick_check_data.cycle = 0;
3531 tick_check_data.seen_true_dep_p = false;
3532 sched_deps_info = &tick_check_sched_deps_info;
3533
3534 gcc_assert (!dc->readonly);
3535 dc->readonly = 1;
3536 deps_analyze_insn (dc, EXPR_INSN_RTX (expr));
3537 dc->readonly = 0;
3538
3539 cycles_left = tick_check_data.cycle - FENCE_CYCLE (fence);
3540
3541 return cycles_left >= 0 ? cycles_left : 0;
3542 }
3543 \f
3544
3545 /* Functions to work with insns. */
3546
3547 /* Returns true if LHS of INSN is the same as DEST of an insn
3548 being moved. */
3549 bool
3550 lhs_of_insn_equals_to_dest_p (insn_t insn, rtx dest)
3551 {
3552 rtx lhs = INSN_LHS (insn);
3553
3554 if (lhs == NULL || dest == NULL)
3555 return false;
3556
3557 return rtx_equal_p (lhs, dest);
3558 }
3559
3560 /* Return s_i_d entry of INSN. Callable from debugger. */
3561 sel_insn_data_def
3562 insn_sid (insn_t insn)
3563 {
3564 return *SID (insn);
3565 }
3566
3567 /* True when INSN is a speculative check. We can tell this by looking
3568 at the data structures of the selective scheduler, not by examining
3569 the pattern. */
3570 bool
3571 sel_insn_is_speculation_check (rtx insn)
3572 {
3573 return s_i_d.exists () && !! INSN_SPEC_CHECKED_DS (insn);
3574 }
3575
3576 /* Extracts machine mode MODE and destination location DST_LOC
3577 for given INSN. */
3578 void
3579 get_dest_and_mode (rtx insn, rtx *dst_loc, enum machine_mode *mode)
3580 {
3581 rtx pat = PATTERN (insn);
3582
3583 gcc_assert (dst_loc);
3584 gcc_assert (GET_CODE (pat) == SET);
3585
3586 *dst_loc = SET_DEST (pat);
3587
3588 gcc_assert (*dst_loc);
3589 gcc_assert (MEM_P (*dst_loc) || REG_P (*dst_loc));
3590
3591 if (mode)
3592 *mode = GET_MODE (*dst_loc);
3593 }
3594
3595 /* Returns true when moving through JUMP will result in bookkeeping
3596 creation. */
3597 bool
3598 bookkeeping_can_be_created_if_moved_through_p (insn_t jump)
3599 {
3600 insn_t succ;
3601 succ_iterator si;
3602
3603 FOR_EACH_SUCC (succ, si, jump)
3604 if (sel_num_cfg_preds_gt_1 (succ))
3605 return true;
3606
3607 return false;
3608 }
3609
3610 /* Return 'true' if INSN is the only one in its basic block. */
3611 static bool
3612 insn_is_the_only_one_in_bb_p (insn_t insn)
3613 {
3614 return sel_bb_head_p (insn) && sel_bb_end_p (insn);
3615 }
3616
3617 #ifdef ENABLE_CHECKING
3618 /* Check that the region we're scheduling still has at most one
3619 backedge. */
3620 static void
3621 verify_backedges (void)
3622 {
3623 if (pipelining_p)
3624 {
3625 int i, n = 0;
3626 edge e;
3627 edge_iterator ei;
3628
3629 for (i = 0; i < current_nr_blocks; i++)
3630 FOR_EACH_EDGE (e, ei, BASIC_BLOCK_FOR_FN (cfun, BB_TO_BLOCK (i))->succs)
3631 if (in_current_region_p (e->dest)
3632 && BLOCK_TO_BB (e->dest->index) < i)
3633 n++;
3634
3635 gcc_assert (n <= 1);
3636 }
3637 }
3638 #endif
3639 \f
3640
3641 /* Functions to work with control flow. */
3642
3643 /* Recompute BLOCK_TO_BB and BB_FOR_BLOCK for current region so that blocks
3644 are sorted in topological order (it might have been invalidated by
3645 redirecting an edge). */
3646 static void
3647 sel_recompute_toporder (void)
3648 {
3649 int i, n, rgn;
3650 int *postorder, n_blocks;
3651
3652 postorder = XALLOCAVEC (int, n_basic_blocks_for_fn (cfun));
3653 n_blocks = post_order_compute (postorder, false, false);
3654
3655 rgn = CONTAINING_RGN (BB_TO_BLOCK (0));
3656 for (n = 0, i = n_blocks - 1; i >= 0; i--)
3657 if (CONTAINING_RGN (postorder[i]) == rgn)
3658 {
3659 BLOCK_TO_BB (postorder[i]) = n;
3660 BB_TO_BLOCK (n) = postorder[i];
3661 n++;
3662 }
3663
3664 /* Assert that we updated info for all blocks. We may miss some blocks if
3665 this function is called when redirecting an edge made a block
3666 unreachable, but that block is not deleted yet. */
3667 gcc_assert (n == RGN_NR_BLOCKS (rgn));
3668 }
3669
3670 /* Tidy the possibly empty block BB. */
3671 static bool
3672 maybe_tidy_empty_bb (basic_block bb)
3673 {
3674 basic_block succ_bb, pred_bb, note_bb;
3675 vec<basic_block> dom_bbs;
3676 edge e;
3677 edge_iterator ei;
3678 bool rescan_p;
3679
3680 /* Keep empty bb only if this block immediately precedes EXIT and
3681 has incoming non-fallthrough edge, or it has no predecessors or
3682 successors. Otherwise remove it. */
3683 if (!sel_bb_empty_p (bb)
3684 || (single_succ_p (bb)
3685 && single_succ (bb) == EXIT_BLOCK_PTR_FOR_FN (cfun)
3686 && (!single_pred_p (bb)
3687 || !(single_pred_edge (bb)->flags & EDGE_FALLTHRU)))
3688 || EDGE_COUNT (bb->preds) == 0
3689 || EDGE_COUNT (bb->succs) == 0)
3690 return false;
3691
3692 /* Do not attempt to redirect complex edges. */
3693 FOR_EACH_EDGE (e, ei, bb->preds)
3694 if (e->flags & EDGE_COMPLEX)
3695 return false;
3696 else if (e->flags & EDGE_FALLTHRU)
3697 {
3698 rtx note;
3699 /* If prev bb ends with asm goto, see if any of the
3700 ASM_OPERANDS_LABELs don't point to the fallthru
3701 label. Do not attempt to redirect it in that case. */
3702 if (JUMP_P (BB_END (e->src))
3703 && (note = extract_asm_operands (PATTERN (BB_END (e->src)))))
3704 {
3705 int i, n = ASM_OPERANDS_LABEL_LENGTH (note);
3706
3707 for (i = 0; i < n; ++i)
3708 if (XEXP (ASM_OPERANDS_LABEL (note, i), 0) == BB_HEAD (bb))
3709 return false;
3710 }
3711 }
3712
3713 free_data_sets (bb);
3714
3715 /* Do not delete BB if it has more than one successor.
3716 That can occur when we moving a jump. */
3717 if (!single_succ_p (bb))
3718 {
3719 gcc_assert (can_merge_blocks_p (bb->prev_bb, bb));
3720 sel_merge_blocks (bb->prev_bb, bb);
3721 return true;
3722 }
3723
3724 succ_bb = single_succ (bb);
3725 rescan_p = true;
3726 pred_bb = NULL;
3727 dom_bbs.create (0);
3728
3729 /* Save a pred/succ from the current region to attach the notes to. */
3730 note_bb = NULL;
3731 FOR_EACH_EDGE (e, ei, bb->preds)
3732 if (in_current_region_p (e->src))
3733 {
3734 note_bb = e->src;
3735 break;
3736 }
3737 if (note_bb == NULL)
3738 note_bb = succ_bb;
3739
3740 /* Redirect all non-fallthru edges to the next bb. */
3741 while (rescan_p)
3742 {
3743 rescan_p = false;
3744
3745 FOR_EACH_EDGE (e, ei, bb->preds)
3746 {
3747 pred_bb = e->src;
3748
3749 if (!(e->flags & EDGE_FALLTHRU))
3750 {
3751 /* We can not invalidate computed topological order by moving
3752 the edge destination block (E->SUCC) along a fallthru edge.
3753
3754 We will update dominators here only when we'll get
3755 an unreachable block when redirecting, otherwise
3756 sel_redirect_edge_and_branch will take care of it. */
3757 if (e->dest != bb
3758 && single_pred_p (e->dest))
3759 dom_bbs.safe_push (e->dest);
3760 sel_redirect_edge_and_branch (e, succ_bb);
3761 rescan_p = true;
3762 break;
3763 }
3764 /* If the edge is fallthru, but PRED_BB ends in a conditional jump
3765 to BB (so there is no non-fallthru edge from PRED_BB to BB), we
3766 still have to adjust it. */
3767 else if (single_succ_p (pred_bb) && any_condjump_p (BB_END (pred_bb)))
3768 {
3769 /* If possible, try to remove the unneeded conditional jump. */
3770 if (INSN_SCHED_TIMES (BB_END (pred_bb)) == 0
3771 && !IN_CURRENT_FENCE_P (BB_END (pred_bb)))
3772 {
3773 if (!sel_remove_insn (BB_END (pred_bb), false, false))
3774 tidy_fallthru_edge (e);
3775 }
3776 else
3777 sel_redirect_edge_and_branch (e, succ_bb);
3778 rescan_p = true;
3779 break;
3780 }
3781 }
3782 }
3783
3784 if (can_merge_blocks_p (bb->prev_bb, bb))
3785 sel_merge_blocks (bb->prev_bb, bb);
3786 else
3787 {
3788 /* This is a block without fallthru predecessor. Just delete it. */
3789 gcc_assert (note_bb);
3790 move_bb_info (note_bb, bb);
3791 remove_empty_bb (bb, true);
3792 }
3793
3794 if (!dom_bbs.is_empty ())
3795 {
3796 dom_bbs.safe_push (succ_bb);
3797 iterate_fix_dominators (CDI_DOMINATORS, dom_bbs, false);
3798 dom_bbs.release ();
3799 }
3800
3801 return true;
3802 }
3803
3804 /* Tidy the control flow after we have removed original insn from
3805 XBB. Return true if we have removed some blocks. When FULL_TIDYING
3806 is true, also try to optimize control flow on non-empty blocks. */
3807 bool
3808 tidy_control_flow (basic_block xbb, bool full_tidying)
3809 {
3810 bool changed = true;
3811 insn_t first, last;
3812
3813 /* First check whether XBB is empty. */
3814 changed = maybe_tidy_empty_bb (xbb);
3815 if (changed || !full_tidying)
3816 return changed;
3817
3818 /* Check if there is a unnecessary jump after insn left. */
3819 if (bb_has_removable_jump_to_p (xbb, xbb->next_bb)
3820 && INSN_SCHED_TIMES (BB_END (xbb)) == 0
3821 && !IN_CURRENT_FENCE_P (BB_END (xbb)))
3822 {
3823 if (sel_remove_insn (BB_END (xbb), false, false))
3824 return true;
3825 tidy_fallthru_edge (EDGE_SUCC (xbb, 0));
3826 }
3827
3828 first = sel_bb_head (xbb);
3829 last = sel_bb_end (xbb);
3830 if (MAY_HAVE_DEBUG_INSNS)
3831 {
3832 if (first != last && DEBUG_INSN_P (first))
3833 do
3834 first = NEXT_INSN (first);
3835 while (first != last && (DEBUG_INSN_P (first) || NOTE_P (first)));
3836
3837 if (first != last && DEBUG_INSN_P (last))
3838 do
3839 last = PREV_INSN (last);
3840 while (first != last && (DEBUG_INSN_P (last) || NOTE_P (last)));
3841 }
3842 /* Check if there is an unnecessary jump in previous basic block leading
3843 to next basic block left after removing INSN from stream.
3844 If it is so, remove that jump and redirect edge to current
3845 basic block (where there was INSN before deletion). This way
3846 when NOP will be deleted several instructions later with its
3847 basic block we will not get a jump to next instruction, which
3848 can be harmful. */
3849 if (first == last
3850 && !sel_bb_empty_p (xbb)
3851 && INSN_NOP_P (last)
3852 /* Flow goes fallthru from current block to the next. */
3853 && EDGE_COUNT (xbb->succs) == 1
3854 && (EDGE_SUCC (xbb, 0)->flags & EDGE_FALLTHRU)
3855 /* When successor is an EXIT block, it may not be the next block. */
3856 && single_succ (xbb) != EXIT_BLOCK_PTR_FOR_FN (cfun)
3857 /* And unconditional jump in previous basic block leads to
3858 next basic block of XBB and this jump can be safely removed. */
3859 && in_current_region_p (xbb->prev_bb)
3860 && bb_has_removable_jump_to_p (xbb->prev_bb, xbb->next_bb)
3861 && INSN_SCHED_TIMES (BB_END (xbb->prev_bb)) == 0
3862 /* Also this jump is not at the scheduling boundary. */
3863 && !IN_CURRENT_FENCE_P (BB_END (xbb->prev_bb)))
3864 {
3865 bool recompute_toporder_p;
3866 /* Clear data structures of jump - jump itself will be removed
3867 by sel_redirect_edge_and_branch. */
3868 clear_expr (INSN_EXPR (BB_END (xbb->prev_bb)));
3869 recompute_toporder_p
3870 = sel_redirect_edge_and_branch (EDGE_SUCC (xbb->prev_bb, 0), xbb);
3871
3872 gcc_assert (EDGE_SUCC (xbb->prev_bb, 0)->flags & EDGE_FALLTHRU);
3873
3874 /* It can turn out that after removing unused jump, basic block
3875 that contained that jump, becomes empty too. In such case
3876 remove it too. */
3877 if (sel_bb_empty_p (xbb->prev_bb))
3878 changed = maybe_tidy_empty_bb (xbb->prev_bb);
3879 if (recompute_toporder_p)
3880 sel_recompute_toporder ();
3881 }
3882
3883 #ifdef ENABLE_CHECKING
3884 verify_backedges ();
3885 verify_dominators (CDI_DOMINATORS);
3886 #endif
3887
3888 return changed;
3889 }
3890
3891 /* Purge meaningless empty blocks in the middle of a region. */
3892 void
3893 purge_empty_blocks (void)
3894 {
3895 int i;
3896
3897 /* Do not attempt to delete the first basic block in the region. */
3898 for (i = 1; i < current_nr_blocks; )
3899 {
3900 basic_block b = BASIC_BLOCK_FOR_FN (cfun, BB_TO_BLOCK (i));
3901
3902 if (maybe_tidy_empty_bb (b))
3903 continue;
3904
3905 i++;
3906 }
3907 }
3908
3909 /* Rip-off INSN from the insn stream. When ONLY_DISCONNECT is true,
3910 do not delete insn's data, because it will be later re-emitted.
3911 Return true if we have removed some blocks afterwards. */
3912 bool
3913 sel_remove_insn (insn_t insn, bool only_disconnect, bool full_tidying)
3914 {
3915 basic_block bb = BLOCK_FOR_INSN (insn);
3916
3917 gcc_assert (INSN_IN_STREAM_P (insn));
3918
3919 if (DEBUG_INSN_P (insn) && BB_AV_SET_VALID_P (bb))
3920 {
3921 expr_t expr;
3922 av_set_iterator i;
3923
3924 /* When we remove a debug insn that is head of a BB, it remains
3925 in the AV_SET of the block, but it shouldn't. */
3926 FOR_EACH_EXPR_1 (expr, i, &BB_AV_SET (bb))
3927 if (EXPR_INSN_RTX (expr) == insn)
3928 {
3929 av_set_iter_remove (&i);
3930 break;
3931 }
3932 }
3933
3934 if (only_disconnect)
3935 remove_insn (insn);
3936 else
3937 {
3938 delete_insn (insn);
3939 clear_expr (INSN_EXPR (insn));
3940 }
3941
3942 /* It is necessary to NULL these fields in case we are going to re-insert
3943 INSN into the insns stream, as will usually happen in the ONLY_DISCONNECT
3944 case, but also for NOPs that we will return to the nop pool. */
3945 PREV_INSN (insn) = NULL_RTX;
3946 NEXT_INSN (insn) = NULL_RTX;
3947 set_block_for_insn (insn, NULL);
3948
3949 return tidy_control_flow (bb, full_tidying);
3950 }
3951
3952 /* Estimate number of the insns in BB. */
3953 static int
3954 sel_estimate_number_of_insns (basic_block bb)
3955 {
3956 int res = 0;
3957 insn_t insn = NEXT_INSN (BB_HEAD (bb)), next_tail = NEXT_INSN (BB_END (bb));
3958
3959 for (; insn != next_tail; insn = NEXT_INSN (insn))
3960 if (NONDEBUG_INSN_P (insn))
3961 res++;
3962
3963 return res;
3964 }
3965
3966 /* We don't need separate luids for notes or labels. */
3967 static int
3968 sel_luid_for_non_insn (rtx x)
3969 {
3970 gcc_assert (NOTE_P (x) || LABEL_P (x));
3971
3972 return -1;
3973 }
3974
3975 /* Find the proper seqno for inserting at INSN by successors.
3976 Return -1 if no successors with positive seqno exist. */
3977 static int
3978 get_seqno_by_succs (rtx insn)
3979 {
3980 basic_block bb = BLOCK_FOR_INSN (insn);
3981 rtx tmp = insn, end = BB_END (bb);
3982 int seqno;
3983 insn_t succ = NULL;
3984 succ_iterator si;
3985
3986 while (tmp != end)
3987 {
3988 tmp = NEXT_INSN (tmp);
3989 if (INSN_P (tmp))
3990 return INSN_SEQNO (tmp);
3991 }
3992
3993 seqno = INT_MAX;
3994
3995 FOR_EACH_SUCC_1 (succ, si, end, SUCCS_NORMAL)
3996 if (INSN_SEQNO (succ) > 0)
3997 seqno = MIN (seqno, INSN_SEQNO (succ));
3998
3999 if (seqno == INT_MAX)
4000 return -1;
4001
4002 return seqno;
4003 }
4004
4005 /* Compute seqno for INSN by its preds or succs. */
4006 static int
4007 get_seqno_for_a_jump (insn_t insn)
4008 {
4009 int seqno;
4010
4011 gcc_assert (INSN_SIMPLEJUMP_P (insn));
4012
4013 if (!sel_bb_head_p (insn))
4014 seqno = INSN_SEQNO (PREV_INSN (insn));
4015 else
4016 {
4017 basic_block bb = BLOCK_FOR_INSN (insn);
4018
4019 if (single_pred_p (bb)
4020 && !in_current_region_p (single_pred (bb)))
4021 {
4022 /* We can have preds outside a region when splitting edges
4023 for pipelining of an outer loop. Use succ instead.
4024 There should be only one of them. */
4025 insn_t succ = NULL;
4026 succ_iterator si;
4027 bool first = true;
4028
4029 gcc_assert (flag_sel_sched_pipelining_outer_loops
4030 && current_loop_nest);
4031 FOR_EACH_SUCC_1 (succ, si, insn,
4032 SUCCS_NORMAL | SUCCS_SKIP_TO_LOOP_EXITS)
4033 {
4034 gcc_assert (first);
4035 first = false;
4036 }
4037
4038 gcc_assert (succ != NULL);
4039 seqno = INSN_SEQNO (succ);
4040 }
4041 else
4042 {
4043 insn_t *preds;
4044 int n;
4045
4046 cfg_preds (BLOCK_FOR_INSN (insn), &preds, &n);
4047
4048 gcc_assert (n > 0);
4049 /* For one predecessor, use simple method. */
4050 if (n == 1)
4051 seqno = INSN_SEQNO (preds[0]);
4052 else
4053 seqno = get_seqno_by_preds (insn);
4054
4055 free (preds);
4056 }
4057 }
4058
4059 /* We were unable to find a good seqno among preds. */
4060 if (seqno < 0)
4061 seqno = get_seqno_by_succs (insn);
4062
4063 gcc_assert (seqno >= 0);
4064
4065 return seqno;
4066 }
4067
4068 /* Find the proper seqno for inserting at INSN. Returns -1 if no predecessors
4069 with positive seqno exist. */
4070 int
4071 get_seqno_by_preds (rtx insn)
4072 {
4073 basic_block bb = BLOCK_FOR_INSN (insn);
4074 rtx tmp = insn, head = BB_HEAD (bb);
4075 insn_t *preds;
4076 int n, i, seqno;
4077
4078 while (tmp != head)
4079 {
4080 tmp = PREV_INSN (tmp);
4081 if (INSN_P (tmp))
4082 return INSN_SEQNO (tmp);
4083 }
4084
4085 cfg_preds (bb, &preds, &n);
4086 for (i = 0, seqno = -1; i < n; i++)
4087 seqno = MAX (seqno, INSN_SEQNO (preds[i]));
4088
4089 return seqno;
4090 }
4091
4092 \f
4093
4094 /* Extend pass-scope data structures for basic blocks. */
4095 void
4096 sel_extend_global_bb_info (void)
4097 {
4098 sel_global_bb_info.safe_grow_cleared (last_basic_block_for_fn (cfun));
4099 }
4100
4101 /* Extend region-scope data structures for basic blocks. */
4102 static void
4103 extend_region_bb_info (void)
4104 {
4105 sel_region_bb_info.safe_grow_cleared (last_basic_block_for_fn (cfun));
4106 }
4107
4108 /* Extend all data structures to fit for all basic blocks. */
4109 static void
4110 extend_bb_info (void)
4111 {
4112 sel_extend_global_bb_info ();
4113 extend_region_bb_info ();
4114 }
4115
4116 /* Finalize pass-scope data structures for basic blocks. */
4117 void
4118 sel_finish_global_bb_info (void)
4119 {
4120 sel_global_bb_info.release ();
4121 }
4122
4123 /* Finalize region-scope data structures for basic blocks. */
4124 static void
4125 finish_region_bb_info (void)
4126 {
4127 sel_region_bb_info.release ();
4128 }
4129 \f
4130
4131 /* Data for each insn in current region. */
4132 vec<sel_insn_data_def> s_i_d = vNULL;
4133
4134 /* Extend data structures for insns from current region. */
4135 static void
4136 extend_insn_data (void)
4137 {
4138 int reserve;
4139
4140 sched_extend_target ();
4141 sched_deps_init (false);
4142
4143 /* Extend data structures for insns from current region. */
4144 reserve = (sched_max_luid + 1 - s_i_d.length ());
4145 if (reserve > 0 && ! s_i_d.space (reserve))
4146 {
4147 int size;
4148
4149 if (sched_max_luid / 2 > 1024)
4150 size = sched_max_luid + 1024;
4151 else
4152 size = 3 * sched_max_luid / 2;
4153
4154
4155 s_i_d.safe_grow_cleared (size);
4156 }
4157 }
4158
4159 /* Finalize data structures for insns from current region. */
4160 static void
4161 finish_insns (void)
4162 {
4163 unsigned i;
4164
4165 /* Clear here all dependence contexts that may have left from insns that were
4166 removed during the scheduling. */
4167 for (i = 0; i < s_i_d.length (); i++)
4168 {
4169 sel_insn_data_def *sid_entry = &s_i_d[i];
4170
4171 if (sid_entry->live)
4172 return_regset_to_pool (sid_entry->live);
4173 if (sid_entry->analyzed_deps)
4174 {
4175 BITMAP_FREE (sid_entry->analyzed_deps);
4176 BITMAP_FREE (sid_entry->found_deps);
4177 htab_delete (sid_entry->transformed_insns);
4178 free_deps (&sid_entry->deps_context);
4179 }
4180 if (EXPR_VINSN (&sid_entry->expr))
4181 {
4182 clear_expr (&sid_entry->expr);
4183
4184 /* Also, clear CANT_MOVE bit here, because we really don't want it
4185 to be passed to the next region. */
4186 CANT_MOVE_BY_LUID (i) = 0;
4187 }
4188 }
4189
4190 s_i_d.release ();
4191 }
4192
4193 /* A proxy to pass initialization data to init_insn (). */
4194 static sel_insn_data_def _insn_init_ssid;
4195 static sel_insn_data_t insn_init_ssid = &_insn_init_ssid;
4196
4197 /* If true create a new vinsn. Otherwise use the one from EXPR. */
4198 static bool insn_init_create_new_vinsn_p;
4199
4200 /* Set all necessary data for initialization of the new insn[s]. */
4201 static expr_t
4202 set_insn_init (expr_t expr, vinsn_t vi, int seqno)
4203 {
4204 expr_t x = &insn_init_ssid->expr;
4205
4206 copy_expr_onside (x, expr);
4207 if (vi != NULL)
4208 {
4209 insn_init_create_new_vinsn_p = false;
4210 change_vinsn_in_expr (x, vi);
4211 }
4212 else
4213 insn_init_create_new_vinsn_p = true;
4214
4215 insn_init_ssid->seqno = seqno;
4216 return x;
4217 }
4218
4219 /* Init data for INSN. */
4220 static void
4221 init_insn_data (insn_t insn)
4222 {
4223 expr_t expr;
4224 sel_insn_data_t ssid = insn_init_ssid;
4225
4226 /* The fields mentioned below are special and hence are not being
4227 propagated to the new insns. */
4228 gcc_assert (!ssid->asm_p && ssid->sched_next == NULL
4229 && !ssid->after_stall_p && ssid->sched_cycle == 0);
4230 gcc_assert (INSN_P (insn) && INSN_LUID (insn) > 0);
4231
4232 expr = INSN_EXPR (insn);
4233 copy_expr (expr, &ssid->expr);
4234 prepare_insn_expr (insn, ssid->seqno);
4235
4236 if (insn_init_create_new_vinsn_p)
4237 change_vinsn_in_expr (expr, vinsn_create (insn, init_insn_force_unique_p));
4238
4239 if (first_time_insn_init (insn))
4240 init_first_time_insn_data (insn);
4241 }
4242
4243 /* This is used to initialize spurious jumps generated by
4244 sel_redirect_edge (). */
4245 static void
4246 init_simplejump_data (insn_t insn)
4247 {
4248 init_expr (INSN_EXPR (insn), vinsn_create (insn, false), 0,
4249 REG_BR_PROB_BASE, 0, 0, 0, 0, 0, 0,
4250 vNULL, true, false, false,
4251 false, true);
4252 INSN_SEQNO (insn) = get_seqno_for_a_jump (insn);
4253 init_first_time_insn_data (insn);
4254 }
4255
4256 /* Perform deferred initialization of insns. This is used to process
4257 a new jump that may be created by redirect_edge. */
4258 void
4259 sel_init_new_insn (insn_t insn, int flags)
4260 {
4261 /* We create data structures for bb when the first insn is emitted in it. */
4262 if (INSN_P (insn)
4263 && INSN_IN_STREAM_P (insn)
4264 && insn_is_the_only_one_in_bb_p (insn))
4265 {
4266 extend_bb_info ();
4267 create_initial_data_sets (BLOCK_FOR_INSN (insn));
4268 }
4269
4270 if (flags & INSN_INIT_TODO_LUID)
4271 {
4272 sched_extend_luids ();
4273 sched_init_insn_luid (insn);
4274 }
4275
4276 if (flags & INSN_INIT_TODO_SSID)
4277 {
4278 extend_insn_data ();
4279 init_insn_data (insn);
4280 clear_expr (&insn_init_ssid->expr);
4281 }
4282
4283 if (flags & INSN_INIT_TODO_SIMPLEJUMP)
4284 {
4285 extend_insn_data ();
4286 init_simplejump_data (insn);
4287 }
4288
4289 gcc_assert (CONTAINING_RGN (BLOCK_NUM (insn))
4290 == CONTAINING_RGN (BB_TO_BLOCK (0)));
4291 }
4292 \f
4293
4294 /* Functions to init/finish work with lv sets. */
4295
4296 /* Init BB_LV_SET of BB from DF_LR_IN set of BB. */
4297 static void
4298 init_lv_set (basic_block bb)
4299 {
4300 gcc_assert (!BB_LV_SET_VALID_P (bb));
4301
4302 BB_LV_SET (bb) = get_regset_from_pool ();
4303 COPY_REG_SET (BB_LV_SET (bb), DF_LR_IN (bb));
4304 BB_LV_SET_VALID_P (bb) = true;
4305 }
4306
4307 /* Copy liveness information to BB from FROM_BB. */
4308 static void
4309 copy_lv_set_from (basic_block bb, basic_block from_bb)
4310 {
4311 gcc_assert (!BB_LV_SET_VALID_P (bb));
4312
4313 COPY_REG_SET (BB_LV_SET (bb), BB_LV_SET (from_bb));
4314 BB_LV_SET_VALID_P (bb) = true;
4315 }
4316
4317 /* Initialize lv set of all bb headers. */
4318 void
4319 init_lv_sets (void)
4320 {
4321 basic_block bb;
4322
4323 /* Initialize of LV sets. */
4324 FOR_EACH_BB_FN (bb, cfun)
4325 init_lv_set (bb);
4326
4327 /* Don't forget EXIT_BLOCK. */
4328 init_lv_set (EXIT_BLOCK_PTR_FOR_FN (cfun));
4329 }
4330
4331 /* Release lv set of HEAD. */
4332 static void
4333 free_lv_set (basic_block bb)
4334 {
4335 gcc_assert (BB_LV_SET (bb) != NULL);
4336
4337 return_regset_to_pool (BB_LV_SET (bb));
4338 BB_LV_SET (bb) = NULL;
4339 BB_LV_SET_VALID_P (bb) = false;
4340 }
4341
4342 /* Finalize lv sets of all bb headers. */
4343 void
4344 free_lv_sets (void)
4345 {
4346 basic_block bb;
4347
4348 /* Don't forget EXIT_BLOCK. */
4349 free_lv_set (EXIT_BLOCK_PTR_FOR_FN (cfun));
4350
4351 /* Free LV sets. */
4352 FOR_EACH_BB_FN (bb, cfun)
4353 if (BB_LV_SET (bb))
4354 free_lv_set (bb);
4355 }
4356
4357 /* Mark AV_SET for BB as invalid, so this set will be updated the next time
4358 compute_av() processes BB. This function is called when creating new basic
4359 blocks, as well as for blocks (either new or existing) where new jumps are
4360 created when the control flow is being updated. */
4361 static void
4362 invalidate_av_set (basic_block bb)
4363 {
4364 BB_AV_LEVEL (bb) = -1;
4365 }
4366
4367 /* Create initial data sets for BB (they will be invalid). */
4368 static void
4369 create_initial_data_sets (basic_block bb)
4370 {
4371 if (BB_LV_SET (bb))
4372 BB_LV_SET_VALID_P (bb) = false;
4373 else
4374 BB_LV_SET (bb) = get_regset_from_pool ();
4375 invalidate_av_set (bb);
4376 }
4377
4378 /* Free av set of BB. */
4379 static void
4380 free_av_set (basic_block bb)
4381 {
4382 av_set_clear (&BB_AV_SET (bb));
4383 BB_AV_LEVEL (bb) = 0;
4384 }
4385
4386 /* Free data sets of BB. */
4387 void
4388 free_data_sets (basic_block bb)
4389 {
4390 free_lv_set (bb);
4391 free_av_set (bb);
4392 }
4393
4394 /* Exchange lv sets of TO and FROM. */
4395 static void
4396 exchange_lv_sets (basic_block to, basic_block from)
4397 {
4398 {
4399 regset to_lv_set = BB_LV_SET (to);
4400
4401 BB_LV_SET (to) = BB_LV_SET (from);
4402 BB_LV_SET (from) = to_lv_set;
4403 }
4404
4405 {
4406 bool to_lv_set_valid_p = BB_LV_SET_VALID_P (to);
4407
4408 BB_LV_SET_VALID_P (to) = BB_LV_SET_VALID_P (from);
4409 BB_LV_SET_VALID_P (from) = to_lv_set_valid_p;
4410 }
4411 }
4412
4413
4414 /* Exchange av sets of TO and FROM. */
4415 static void
4416 exchange_av_sets (basic_block to, basic_block from)
4417 {
4418 {
4419 av_set_t to_av_set = BB_AV_SET (to);
4420
4421 BB_AV_SET (to) = BB_AV_SET (from);
4422 BB_AV_SET (from) = to_av_set;
4423 }
4424
4425 {
4426 int to_av_level = BB_AV_LEVEL (to);
4427
4428 BB_AV_LEVEL (to) = BB_AV_LEVEL (from);
4429 BB_AV_LEVEL (from) = to_av_level;
4430 }
4431 }
4432
4433 /* Exchange data sets of TO and FROM. */
4434 void
4435 exchange_data_sets (basic_block to, basic_block from)
4436 {
4437 exchange_lv_sets (to, from);
4438 exchange_av_sets (to, from);
4439 }
4440
4441 /* Copy data sets of FROM to TO. */
4442 void
4443 copy_data_sets (basic_block to, basic_block from)
4444 {
4445 gcc_assert (!BB_LV_SET_VALID_P (to) && !BB_AV_SET_VALID_P (to));
4446 gcc_assert (BB_AV_SET (to) == NULL);
4447
4448 BB_AV_LEVEL (to) = BB_AV_LEVEL (from);
4449 BB_LV_SET_VALID_P (to) = BB_LV_SET_VALID_P (from);
4450
4451 if (BB_AV_SET_VALID_P (from))
4452 {
4453 BB_AV_SET (to) = av_set_copy (BB_AV_SET (from));
4454 }
4455 if (BB_LV_SET_VALID_P (from))
4456 {
4457 gcc_assert (BB_LV_SET (to) != NULL);
4458 COPY_REG_SET (BB_LV_SET (to), BB_LV_SET (from));
4459 }
4460 }
4461
4462 /* Return an av set for INSN, if any. */
4463 av_set_t
4464 get_av_set (insn_t insn)
4465 {
4466 av_set_t av_set;
4467
4468 gcc_assert (AV_SET_VALID_P (insn));
4469
4470 if (sel_bb_head_p (insn))
4471 av_set = BB_AV_SET (BLOCK_FOR_INSN (insn));
4472 else
4473 av_set = NULL;
4474
4475 return av_set;
4476 }
4477
4478 /* Implementation of AV_LEVEL () macro. Return AV_LEVEL () of INSN. */
4479 int
4480 get_av_level (insn_t insn)
4481 {
4482 int av_level;
4483
4484 gcc_assert (INSN_P (insn));
4485
4486 if (sel_bb_head_p (insn))
4487 av_level = BB_AV_LEVEL (BLOCK_FOR_INSN (insn));
4488 else
4489 av_level = INSN_WS_LEVEL (insn);
4490
4491 return av_level;
4492 }
4493
4494 \f
4495
4496 /* Variables to work with control-flow graph. */
4497
4498 /* The basic block that already has been processed by the sched_data_update (),
4499 but hasn't been in sel_add_bb () yet. */
4500 static vec<basic_block>
4501 last_added_blocks = vNULL;
4502
4503 /* A pool for allocating successor infos. */
4504 static struct
4505 {
4506 /* A stack for saving succs_info structures. */
4507 struct succs_info *stack;
4508
4509 /* Its size. */
4510 int size;
4511
4512 /* Top of the stack. */
4513 int top;
4514
4515 /* Maximal value of the top. */
4516 int max_top;
4517 } succs_info_pool;
4518
4519 /* Functions to work with control-flow graph. */
4520
4521 /* Return basic block note of BB. */
4522 insn_t
4523 sel_bb_head (basic_block bb)
4524 {
4525 insn_t head;
4526
4527 if (bb == EXIT_BLOCK_PTR_FOR_FN (cfun))
4528 {
4529 gcc_assert (exit_insn != NULL_RTX);
4530 head = exit_insn;
4531 }
4532 else
4533 {
4534 insn_t note;
4535
4536 note = bb_note (bb);
4537 head = next_nonnote_insn (note);
4538
4539 if (head && (BARRIER_P (head) || BLOCK_FOR_INSN (head) != bb))
4540 head = NULL_RTX;
4541 }
4542
4543 return head;
4544 }
4545
4546 /* Return true if INSN is a basic block header. */
4547 bool
4548 sel_bb_head_p (insn_t insn)
4549 {
4550 return sel_bb_head (BLOCK_FOR_INSN (insn)) == insn;
4551 }
4552
4553 /* Return last insn of BB. */
4554 insn_t
4555 sel_bb_end (basic_block bb)
4556 {
4557 if (sel_bb_empty_p (bb))
4558 return NULL_RTX;
4559
4560 gcc_assert (bb != EXIT_BLOCK_PTR_FOR_FN (cfun));
4561
4562 return BB_END (bb);
4563 }
4564
4565 /* Return true if INSN is the last insn in its basic block. */
4566 bool
4567 sel_bb_end_p (insn_t insn)
4568 {
4569 return insn == sel_bb_end (BLOCK_FOR_INSN (insn));
4570 }
4571
4572 /* Return true if BB consist of single NOTE_INSN_BASIC_BLOCK. */
4573 bool
4574 sel_bb_empty_p (basic_block bb)
4575 {
4576 return sel_bb_head (bb) == NULL;
4577 }
4578
4579 /* True when BB belongs to the current scheduling region. */
4580 bool
4581 in_current_region_p (basic_block bb)
4582 {
4583 if (bb->index < NUM_FIXED_BLOCKS)
4584 return false;
4585
4586 return CONTAINING_RGN (bb->index) == CONTAINING_RGN (BB_TO_BLOCK (0));
4587 }
4588
4589 /* Return the block which is a fallthru bb of a conditional jump JUMP. */
4590 basic_block
4591 fallthru_bb_of_jump (rtx jump)
4592 {
4593 if (!JUMP_P (jump))
4594 return NULL;
4595
4596 if (!any_condjump_p (jump))
4597 return NULL;
4598
4599 /* A basic block that ends with a conditional jump may still have one successor
4600 (and be followed by a barrier), we are not interested. */
4601 if (single_succ_p (BLOCK_FOR_INSN (jump)))
4602 return NULL;
4603
4604 return FALLTHRU_EDGE (BLOCK_FOR_INSN (jump))->dest;
4605 }
4606
4607 /* Remove all notes from BB. */
4608 static void
4609 init_bb (basic_block bb)
4610 {
4611 remove_notes (bb_note (bb), BB_END (bb));
4612 BB_NOTE_LIST (bb) = note_list;
4613 }
4614
4615 void
4616 sel_init_bbs (bb_vec_t bbs)
4617 {
4618 const struct sched_scan_info_def ssi =
4619 {
4620 extend_bb_info, /* extend_bb */
4621 init_bb, /* init_bb */
4622 NULL, /* extend_insn */
4623 NULL /* init_insn */
4624 };
4625
4626 sched_scan (&ssi, bbs);
4627 }
4628
4629 /* Restore notes for the whole region. */
4630 static void
4631 sel_restore_notes (void)
4632 {
4633 int bb;
4634 insn_t insn;
4635
4636 for (bb = 0; bb < current_nr_blocks; bb++)
4637 {
4638 basic_block first, last;
4639
4640 first = EBB_FIRST_BB (bb);
4641 last = EBB_LAST_BB (bb)->next_bb;
4642
4643 do
4644 {
4645 note_list = BB_NOTE_LIST (first);
4646 restore_other_notes (NULL, first);
4647 BB_NOTE_LIST (first) = NULL_RTX;
4648
4649 FOR_BB_INSNS (first, insn)
4650 if (NONDEBUG_INSN_P (insn))
4651 reemit_notes (insn);
4652
4653 first = first->next_bb;
4654 }
4655 while (first != last);
4656 }
4657 }
4658
4659 /* Free per-bb data structures. */
4660 void
4661 sel_finish_bbs (void)
4662 {
4663 sel_restore_notes ();
4664
4665 /* Remove current loop preheader from this loop. */
4666 if (current_loop_nest)
4667 sel_remove_loop_preheader ();
4668
4669 finish_region_bb_info ();
4670 }
4671
4672 /* Return true if INSN has a single successor of type FLAGS. */
4673 bool
4674 sel_insn_has_single_succ_p (insn_t insn, int flags)
4675 {
4676 insn_t succ;
4677 succ_iterator si;
4678 bool first_p = true;
4679
4680 FOR_EACH_SUCC_1 (succ, si, insn, flags)
4681 {
4682 if (first_p)
4683 first_p = false;
4684 else
4685 return false;
4686 }
4687
4688 return true;
4689 }
4690
4691 /* Allocate successor's info. */
4692 static struct succs_info *
4693 alloc_succs_info (void)
4694 {
4695 if (succs_info_pool.top == succs_info_pool.max_top)
4696 {
4697 int i;
4698
4699 if (++succs_info_pool.max_top >= succs_info_pool.size)
4700 gcc_unreachable ();
4701
4702 i = ++succs_info_pool.top;
4703 succs_info_pool.stack[i].succs_ok.create (10);
4704 succs_info_pool.stack[i].succs_other.create (10);
4705 succs_info_pool.stack[i].probs_ok.create (10);
4706 }
4707 else
4708 succs_info_pool.top++;
4709
4710 return &succs_info_pool.stack[succs_info_pool.top];
4711 }
4712
4713 /* Free successor's info. */
4714 void
4715 free_succs_info (struct succs_info * sinfo)
4716 {
4717 gcc_assert (succs_info_pool.top >= 0
4718 && &succs_info_pool.stack[succs_info_pool.top] == sinfo);
4719 succs_info_pool.top--;
4720
4721 /* Clear stale info. */
4722 sinfo->succs_ok.block_remove (0, sinfo->succs_ok.length ());
4723 sinfo->succs_other.block_remove (0, sinfo->succs_other.length ());
4724 sinfo->probs_ok.block_remove (0, sinfo->probs_ok.length ());
4725 sinfo->all_prob = 0;
4726 sinfo->succs_ok_n = 0;
4727 sinfo->all_succs_n = 0;
4728 }
4729
4730 /* Compute successor info for INSN. FLAGS are the flags passed
4731 to the FOR_EACH_SUCC_1 iterator. */
4732 struct succs_info *
4733 compute_succs_info (insn_t insn, short flags)
4734 {
4735 succ_iterator si;
4736 insn_t succ;
4737 struct succs_info *sinfo = alloc_succs_info ();
4738
4739 /* Traverse *all* successors and decide what to do with each. */
4740 FOR_EACH_SUCC_1 (succ, si, insn, SUCCS_ALL)
4741 {
4742 /* FIXME: this doesn't work for skipping to loop exits, as we don't
4743 perform code motion through inner loops. */
4744 short current_flags = si.current_flags & ~SUCCS_SKIP_TO_LOOP_EXITS;
4745
4746 if (current_flags & flags)
4747 {
4748 sinfo->succs_ok.safe_push (succ);
4749 sinfo->probs_ok.safe_push (
4750 /* FIXME: Improve calculation when skipping
4751 inner loop to exits. */
4752 si.bb_end ? si.e1->probability : REG_BR_PROB_BASE);
4753 sinfo->succs_ok_n++;
4754 }
4755 else
4756 sinfo->succs_other.safe_push (succ);
4757
4758 /* Compute all_prob. */
4759 if (!si.bb_end)
4760 sinfo->all_prob = REG_BR_PROB_BASE;
4761 else
4762 sinfo->all_prob += si.e1->probability;
4763
4764 sinfo->all_succs_n++;
4765 }
4766
4767 return sinfo;
4768 }
4769
4770 /* Return the predecessors of BB in PREDS and their number in N.
4771 Empty blocks are skipped. SIZE is used to allocate PREDS. */
4772 static void
4773 cfg_preds_1 (basic_block bb, insn_t **preds, int *n, int *size)
4774 {
4775 edge e;
4776 edge_iterator ei;
4777
4778 gcc_assert (BLOCK_TO_BB (bb->index) != 0);
4779
4780 FOR_EACH_EDGE (e, ei, bb->preds)
4781 {
4782 basic_block pred_bb = e->src;
4783 insn_t bb_end = BB_END (pred_bb);
4784
4785 if (!in_current_region_p (pred_bb))
4786 {
4787 gcc_assert (flag_sel_sched_pipelining_outer_loops
4788 && current_loop_nest);
4789 continue;
4790 }
4791
4792 if (sel_bb_empty_p (pred_bb))
4793 cfg_preds_1 (pred_bb, preds, n, size);
4794 else
4795 {
4796 if (*n == *size)
4797 *preds = XRESIZEVEC (insn_t, *preds,
4798 (*size = 2 * *size + 1));
4799 (*preds)[(*n)++] = bb_end;
4800 }
4801 }
4802
4803 gcc_assert (*n != 0
4804 || (flag_sel_sched_pipelining_outer_loops
4805 && current_loop_nest));
4806 }
4807
4808 /* Find all predecessors of BB and record them in PREDS and their number
4809 in N. Empty blocks are skipped, and only normal (forward in-region)
4810 edges are processed. */
4811 static void
4812 cfg_preds (basic_block bb, insn_t **preds, int *n)
4813 {
4814 int size = 0;
4815
4816 *preds = NULL;
4817 *n = 0;
4818 cfg_preds_1 (bb, preds, n, &size);
4819 }
4820
4821 /* Returns true if we are moving INSN through join point. */
4822 bool
4823 sel_num_cfg_preds_gt_1 (insn_t insn)
4824 {
4825 basic_block bb;
4826
4827 if (!sel_bb_head_p (insn) || INSN_BB (insn) == 0)
4828 return false;
4829
4830 bb = BLOCK_FOR_INSN (insn);
4831
4832 while (1)
4833 {
4834 if (EDGE_COUNT (bb->preds) > 1)
4835 return true;
4836
4837 gcc_assert (EDGE_PRED (bb, 0)->dest == bb);
4838 bb = EDGE_PRED (bb, 0)->src;
4839
4840 if (!sel_bb_empty_p (bb))
4841 break;
4842 }
4843
4844 return false;
4845 }
4846
4847 /* Returns true when BB should be the end of an ebb. Adapted from the
4848 code in sched-ebb.c. */
4849 bool
4850 bb_ends_ebb_p (basic_block bb)
4851 {
4852 basic_block next_bb = bb_next_bb (bb);
4853 edge e;
4854
4855 if (next_bb == EXIT_BLOCK_PTR_FOR_FN (cfun)
4856 || bitmap_bit_p (forced_ebb_heads, next_bb->index)
4857 || (LABEL_P (BB_HEAD (next_bb))
4858 /* NB: LABEL_NUSES () is not maintained outside of jump.c.
4859 Work around that. */
4860 && !single_pred_p (next_bb)))
4861 return true;
4862
4863 if (!in_current_region_p (next_bb))
4864 return true;
4865
4866 e = find_fallthru_edge (bb->succs);
4867 if (e)
4868 {
4869 gcc_assert (e->dest == next_bb);
4870
4871 return false;
4872 }
4873
4874 return true;
4875 }
4876
4877 /* Returns true when INSN and SUCC are in the same EBB, given that SUCC is a
4878 successor of INSN. */
4879 bool
4880 in_same_ebb_p (insn_t insn, insn_t succ)
4881 {
4882 basic_block ptr = BLOCK_FOR_INSN (insn);
4883
4884 for (;;)
4885 {
4886 if (ptr == BLOCK_FOR_INSN (succ))
4887 return true;
4888
4889 if (bb_ends_ebb_p (ptr))
4890 return false;
4891
4892 ptr = bb_next_bb (ptr);
4893 }
4894
4895 gcc_unreachable ();
4896 return false;
4897 }
4898
4899 /* Recomputes the reverse topological order for the function and
4900 saves it in REV_TOP_ORDER_INDEX. REV_TOP_ORDER_INDEX_LEN is also
4901 modified appropriately. */
4902 static void
4903 recompute_rev_top_order (void)
4904 {
4905 int *postorder;
4906 int n_blocks, i;
4907
4908 if (!rev_top_order_index
4909 || rev_top_order_index_len < last_basic_block_for_fn (cfun))
4910 {
4911 rev_top_order_index_len = last_basic_block_for_fn (cfun);
4912 rev_top_order_index = XRESIZEVEC (int, rev_top_order_index,
4913 rev_top_order_index_len);
4914 }
4915
4916 postorder = XNEWVEC (int, n_basic_blocks_for_fn (cfun));
4917
4918 n_blocks = post_order_compute (postorder, true, false);
4919 gcc_assert (n_basic_blocks_for_fn (cfun) == n_blocks);
4920
4921 /* Build reverse function: for each basic block with BB->INDEX == K
4922 rev_top_order_index[K] is it's reverse topological sort number. */
4923 for (i = 0; i < n_blocks; i++)
4924 {
4925 gcc_assert (postorder[i] < rev_top_order_index_len);
4926 rev_top_order_index[postorder[i]] = i;
4927 }
4928
4929 free (postorder);
4930 }
4931
4932 /* Clear all flags from insns in BB that could spoil its rescheduling. */
4933 void
4934 clear_outdated_rtx_info (basic_block bb)
4935 {
4936 rtx insn;
4937
4938 FOR_BB_INSNS (bb, insn)
4939 if (INSN_P (insn))
4940 {
4941 SCHED_GROUP_P (insn) = 0;
4942 INSN_AFTER_STALL_P (insn) = 0;
4943 INSN_SCHED_TIMES (insn) = 0;
4944 EXPR_PRIORITY_ADJ (INSN_EXPR (insn)) = 0;
4945
4946 /* We cannot use the changed caches, as previously we could ignore
4947 the LHS dependence due to enabled renaming and transform
4948 the expression, and currently we'll be unable to do this. */
4949 htab_empty (INSN_TRANSFORMED_INSNS (insn));
4950 }
4951 }
4952
4953 /* Add BB_NOTE to the pool of available basic block notes. */
4954 static void
4955 return_bb_to_pool (basic_block bb)
4956 {
4957 rtx note = bb_note (bb);
4958
4959 gcc_assert (NOTE_BASIC_BLOCK (note) == bb
4960 && bb->aux == NULL);
4961
4962 /* It turns out that current cfg infrastructure does not support
4963 reuse of basic blocks. Don't bother for now. */
4964 /*bb_note_pool.safe_push (note);*/
4965 }
4966
4967 /* Get a bb_note from pool or return NULL_RTX if pool is empty. */
4968 static rtx
4969 get_bb_note_from_pool (void)
4970 {
4971 if (bb_note_pool.is_empty ())
4972 return NULL_RTX;
4973 else
4974 {
4975 rtx note = bb_note_pool.pop ();
4976
4977 PREV_INSN (note) = NULL_RTX;
4978 NEXT_INSN (note) = NULL_RTX;
4979
4980 return note;
4981 }
4982 }
4983
4984 /* Free bb_note_pool. */
4985 void
4986 free_bb_note_pool (void)
4987 {
4988 bb_note_pool.release ();
4989 }
4990
4991 /* Setup scheduler pool and successor structure. */
4992 void
4993 alloc_sched_pools (void)
4994 {
4995 int succs_size;
4996
4997 succs_size = MAX_WS + 1;
4998 succs_info_pool.stack = XCNEWVEC (struct succs_info, succs_size);
4999 succs_info_pool.size = succs_size;
5000 succs_info_pool.top = -1;
5001 succs_info_pool.max_top = -1;
5002
5003 sched_lists_pool = create_alloc_pool ("sel-sched-lists",
5004 sizeof (struct _list_node), 500);
5005 }
5006
5007 /* Free the pools. */
5008 void
5009 free_sched_pools (void)
5010 {
5011 int i;
5012
5013 free_alloc_pool (sched_lists_pool);
5014 gcc_assert (succs_info_pool.top == -1);
5015 for (i = 0; i <= succs_info_pool.max_top; i++)
5016 {
5017 succs_info_pool.stack[i].succs_ok.release ();
5018 succs_info_pool.stack[i].succs_other.release ();
5019 succs_info_pool.stack[i].probs_ok.release ();
5020 }
5021 free (succs_info_pool.stack);
5022 }
5023 \f
5024
5025 /* Returns a position in RGN where BB can be inserted retaining
5026 topological order. */
5027 static int
5028 find_place_to_insert_bb (basic_block bb, int rgn)
5029 {
5030 bool has_preds_outside_rgn = false;
5031 edge e;
5032 edge_iterator ei;
5033
5034 /* Find whether we have preds outside the region. */
5035 FOR_EACH_EDGE (e, ei, bb->preds)
5036 if (!in_current_region_p (e->src))
5037 {
5038 has_preds_outside_rgn = true;
5039 break;
5040 }
5041
5042 /* Recompute the top order -- needed when we have > 1 pred
5043 and in case we don't have preds outside. */
5044 if (flag_sel_sched_pipelining_outer_loops
5045 && (has_preds_outside_rgn || EDGE_COUNT (bb->preds) > 1))
5046 {
5047 int i, bbi = bb->index, cur_bbi;
5048
5049 recompute_rev_top_order ();
5050 for (i = RGN_NR_BLOCKS (rgn) - 1; i >= 0; i--)
5051 {
5052 cur_bbi = BB_TO_BLOCK (i);
5053 if (rev_top_order_index[bbi]
5054 < rev_top_order_index[cur_bbi])
5055 break;
5056 }
5057
5058 /* We skipped the right block, so we increase i. We accommodate
5059 it for increasing by step later, so we decrease i. */
5060 return (i + 1) - 1;
5061 }
5062 else if (has_preds_outside_rgn)
5063 {
5064 /* This is the case when we generate an extra empty block
5065 to serve as region head during pipelining. */
5066 e = EDGE_SUCC (bb, 0);
5067 gcc_assert (EDGE_COUNT (bb->succs) == 1
5068 && in_current_region_p (EDGE_SUCC (bb, 0)->dest)
5069 && (BLOCK_TO_BB (e->dest->index) == 0));
5070 return -1;
5071 }
5072
5073 /* We don't have preds outside the region. We should have
5074 the only pred, because the multiple preds case comes from
5075 the pipelining of outer loops, and that is handled above.
5076 Just take the bbi of this single pred. */
5077 if (EDGE_COUNT (bb->succs) > 0)
5078 {
5079 int pred_bbi;
5080
5081 gcc_assert (EDGE_COUNT (bb->preds) == 1);
5082
5083 pred_bbi = EDGE_PRED (bb, 0)->src->index;
5084 return BLOCK_TO_BB (pred_bbi);
5085 }
5086 else
5087 /* BB has no successors. It is safe to put it in the end. */
5088 return current_nr_blocks - 1;
5089 }
5090
5091 /* Deletes an empty basic block freeing its data. */
5092 static void
5093 delete_and_free_basic_block (basic_block bb)
5094 {
5095 gcc_assert (sel_bb_empty_p (bb));
5096
5097 if (BB_LV_SET (bb))
5098 free_lv_set (bb);
5099
5100 bitmap_clear_bit (blocks_to_reschedule, bb->index);
5101
5102 /* Can't assert av_set properties because we use sel_aremove_bb
5103 when removing loop preheader from the region. At the point of
5104 removing the preheader we already have deallocated sel_region_bb_info. */
5105 gcc_assert (BB_LV_SET (bb) == NULL
5106 && !BB_LV_SET_VALID_P (bb)
5107 && BB_AV_LEVEL (bb) == 0
5108 && BB_AV_SET (bb) == NULL);
5109
5110 delete_basic_block (bb);
5111 }
5112
5113 /* Add BB to the current region and update the region data. */
5114 static void
5115 add_block_to_current_region (basic_block bb)
5116 {
5117 int i, pos, bbi = -2, rgn;
5118
5119 rgn = CONTAINING_RGN (BB_TO_BLOCK (0));
5120 bbi = find_place_to_insert_bb (bb, rgn);
5121 bbi += 1;
5122 pos = RGN_BLOCKS (rgn) + bbi;
5123
5124 gcc_assert (RGN_HAS_REAL_EBB (rgn) == 0
5125 && ebb_head[bbi] == pos);
5126
5127 /* Make a place for the new block. */
5128 extend_regions ();
5129
5130 for (i = RGN_BLOCKS (rgn + 1) - 1; i >= pos; i--)
5131 BLOCK_TO_BB (rgn_bb_table[i])++;
5132
5133 memmove (rgn_bb_table + pos + 1,
5134 rgn_bb_table + pos,
5135 (RGN_BLOCKS (nr_regions) - pos) * sizeof (*rgn_bb_table));
5136
5137 /* Initialize data for BB. */
5138 rgn_bb_table[pos] = bb->index;
5139 BLOCK_TO_BB (bb->index) = bbi;
5140 CONTAINING_RGN (bb->index) = rgn;
5141
5142 RGN_NR_BLOCKS (rgn)++;
5143
5144 for (i = rgn + 1; i <= nr_regions; i++)
5145 RGN_BLOCKS (i)++;
5146 }
5147
5148 /* Remove BB from the current region and update the region data. */
5149 static void
5150 remove_bb_from_region (basic_block bb)
5151 {
5152 int i, pos, bbi = -2, rgn;
5153
5154 rgn = CONTAINING_RGN (BB_TO_BLOCK (0));
5155 bbi = BLOCK_TO_BB (bb->index);
5156 pos = RGN_BLOCKS (rgn) + bbi;
5157
5158 gcc_assert (RGN_HAS_REAL_EBB (rgn) == 0
5159 && ebb_head[bbi] == pos);
5160
5161 for (i = RGN_BLOCKS (rgn + 1) - 1; i >= pos; i--)
5162 BLOCK_TO_BB (rgn_bb_table[i])--;
5163
5164 memmove (rgn_bb_table + pos,
5165 rgn_bb_table + pos + 1,
5166 (RGN_BLOCKS (nr_regions) - pos) * sizeof (*rgn_bb_table));
5167
5168 RGN_NR_BLOCKS (rgn)--;
5169 for (i = rgn + 1; i <= nr_regions; i++)
5170 RGN_BLOCKS (i)--;
5171 }
5172
5173 /* Add BB to the current region and update all data. If BB is NULL, add all
5174 blocks from last_added_blocks vector. */
5175 static void
5176 sel_add_bb (basic_block bb)
5177 {
5178 /* Extend luids so that new notes will receive zero luids. */
5179 sched_extend_luids ();
5180 sched_init_bbs ();
5181 sel_init_bbs (last_added_blocks);
5182
5183 /* When bb is passed explicitly, the vector should contain
5184 the only element that equals to bb; otherwise, the vector
5185 should not be NULL. */
5186 gcc_assert (last_added_blocks.exists ());
5187
5188 if (bb != NULL)
5189 {
5190 gcc_assert (last_added_blocks.length () == 1
5191 && last_added_blocks[0] == bb);
5192 add_block_to_current_region (bb);
5193
5194 /* We associate creating/deleting data sets with the first insn
5195 appearing / disappearing in the bb. */
5196 if (!sel_bb_empty_p (bb) && BB_LV_SET (bb) == NULL)
5197 create_initial_data_sets (bb);
5198
5199 last_added_blocks.release ();
5200 }
5201 else
5202 /* BB is NULL - process LAST_ADDED_BLOCKS instead. */
5203 {
5204 int i;
5205 basic_block temp_bb = NULL;
5206
5207 for (i = 0;
5208 last_added_blocks.iterate (i, &bb); i++)
5209 {
5210 add_block_to_current_region (bb);
5211 temp_bb = bb;
5212 }
5213
5214 /* We need to fetch at least one bb so we know the region
5215 to update. */
5216 gcc_assert (temp_bb != NULL);
5217 bb = temp_bb;
5218
5219 last_added_blocks.release ();
5220 }
5221
5222 rgn_setup_region (CONTAINING_RGN (bb->index));
5223 }
5224
5225 /* Remove BB from the current region and update all data.
5226 If REMOVE_FROM_CFG_PBB is true, also remove the block cfom cfg. */
5227 static void
5228 sel_remove_bb (basic_block bb, bool remove_from_cfg_p)
5229 {
5230 unsigned idx = bb->index;
5231
5232 gcc_assert (bb != NULL && BB_NOTE_LIST (bb) == NULL_RTX);
5233
5234 remove_bb_from_region (bb);
5235 return_bb_to_pool (bb);
5236 bitmap_clear_bit (blocks_to_reschedule, idx);
5237
5238 if (remove_from_cfg_p)
5239 {
5240 basic_block succ = single_succ (bb);
5241 delete_and_free_basic_block (bb);
5242 set_immediate_dominator (CDI_DOMINATORS, succ,
5243 recompute_dominator (CDI_DOMINATORS, succ));
5244 }
5245
5246 rgn_setup_region (CONTAINING_RGN (idx));
5247 }
5248
5249 /* Concatenate info of EMPTY_BB to info of MERGE_BB. */
5250 static void
5251 move_bb_info (basic_block merge_bb, basic_block empty_bb)
5252 {
5253 if (in_current_region_p (merge_bb))
5254 concat_note_lists (BB_NOTE_LIST (empty_bb),
5255 &BB_NOTE_LIST (merge_bb));
5256 BB_NOTE_LIST (empty_bb) = NULL_RTX;
5257
5258 }
5259
5260 /* Remove EMPTY_BB. If REMOVE_FROM_CFG_P is false, remove EMPTY_BB from
5261 region, but keep it in CFG. */
5262 static void
5263 remove_empty_bb (basic_block empty_bb, bool remove_from_cfg_p)
5264 {
5265 /* The block should contain just a note or a label.
5266 We try to check whether it is unused below. */
5267 gcc_assert (BB_HEAD (empty_bb) == BB_END (empty_bb)
5268 || LABEL_P (BB_HEAD (empty_bb)));
5269
5270 /* If basic block has predecessors or successors, redirect them. */
5271 if (remove_from_cfg_p
5272 && (EDGE_COUNT (empty_bb->preds) > 0
5273 || EDGE_COUNT (empty_bb->succs) > 0))
5274 {
5275 basic_block pred;
5276 basic_block succ;
5277
5278 /* We need to init PRED and SUCC before redirecting edges. */
5279 if (EDGE_COUNT (empty_bb->preds) > 0)
5280 {
5281 edge e;
5282
5283 gcc_assert (EDGE_COUNT (empty_bb->preds) == 1);
5284
5285 e = EDGE_PRED (empty_bb, 0);
5286 gcc_assert (e->src == empty_bb->prev_bb
5287 && (e->flags & EDGE_FALLTHRU));
5288
5289 pred = empty_bb->prev_bb;
5290 }
5291 else
5292 pred = NULL;
5293
5294 if (EDGE_COUNT (empty_bb->succs) > 0)
5295 {
5296 /* We do not check fallthruness here as above, because
5297 after removing a jump the edge may actually be not fallthru. */
5298 gcc_assert (EDGE_COUNT (empty_bb->succs) == 1);
5299 succ = EDGE_SUCC (empty_bb, 0)->dest;
5300 }
5301 else
5302 succ = NULL;
5303
5304 if (EDGE_COUNT (empty_bb->preds) > 0 && succ != NULL)
5305 {
5306 edge e = EDGE_PRED (empty_bb, 0);
5307
5308 if (e->flags & EDGE_FALLTHRU)
5309 redirect_edge_succ_nodup (e, succ);
5310 else
5311 sel_redirect_edge_and_branch (EDGE_PRED (empty_bb, 0), succ);
5312 }
5313
5314 if (EDGE_COUNT (empty_bb->succs) > 0 && pred != NULL)
5315 {
5316 edge e = EDGE_SUCC (empty_bb, 0);
5317
5318 if (find_edge (pred, e->dest) == NULL)
5319 redirect_edge_pred (e, pred);
5320 }
5321 }
5322
5323 /* Finish removing. */
5324 sel_remove_bb (empty_bb, remove_from_cfg_p);
5325 }
5326
5327 /* An implementation of create_basic_block hook, which additionally updates
5328 per-bb data structures. */
5329 static basic_block
5330 sel_create_basic_block (void *headp, void *endp, basic_block after)
5331 {
5332 basic_block new_bb;
5333 insn_t new_bb_note;
5334
5335 gcc_assert (flag_sel_sched_pipelining_outer_loops
5336 || !last_added_blocks.exists ());
5337
5338 new_bb_note = get_bb_note_from_pool ();
5339
5340 if (new_bb_note == NULL_RTX)
5341 new_bb = orig_cfg_hooks.create_basic_block (headp, endp, after);
5342 else
5343 {
5344 new_bb = create_basic_block_structure ((rtx) headp, (rtx) endp,
5345 new_bb_note, after);
5346 new_bb->aux = NULL;
5347 }
5348
5349 last_added_blocks.safe_push (new_bb);
5350
5351 return new_bb;
5352 }
5353
5354 /* Implement sched_init_only_bb (). */
5355 static void
5356 sel_init_only_bb (basic_block bb, basic_block after)
5357 {
5358 gcc_assert (after == NULL);
5359
5360 extend_regions ();
5361 rgn_make_new_region_out_of_new_block (bb);
5362 }
5363
5364 /* Update the latch when we've splitted or merged it from FROM block to TO.
5365 This should be checked for all outer loops, too. */
5366 static void
5367 change_loops_latches (basic_block from, basic_block to)
5368 {
5369 gcc_assert (from != to);
5370
5371 if (current_loop_nest)
5372 {
5373 struct loop *loop;
5374
5375 for (loop = current_loop_nest; loop; loop = loop_outer (loop))
5376 if (considered_for_pipelining_p (loop) && loop->latch == from)
5377 {
5378 gcc_assert (loop == current_loop_nest);
5379 loop->latch = to;
5380 gcc_assert (loop_latch_edge (loop));
5381 }
5382 }
5383 }
5384
5385 /* Splits BB on two basic blocks, adding it to the region and extending
5386 per-bb data structures. Returns the newly created bb. */
5387 static basic_block
5388 sel_split_block (basic_block bb, rtx after)
5389 {
5390 basic_block new_bb;
5391 insn_t insn;
5392
5393 new_bb = sched_split_block_1 (bb, after);
5394 sel_add_bb (new_bb);
5395
5396 /* This should be called after sel_add_bb, because this uses
5397 CONTAINING_RGN for the new block, which is not yet initialized.
5398 FIXME: this function may be a no-op now. */
5399 change_loops_latches (bb, new_bb);
5400
5401 /* Update ORIG_BB_INDEX for insns moved into the new block. */
5402 FOR_BB_INSNS (new_bb, insn)
5403 if (INSN_P (insn))
5404 EXPR_ORIG_BB_INDEX (INSN_EXPR (insn)) = new_bb->index;
5405
5406 if (sel_bb_empty_p (bb))
5407 {
5408 gcc_assert (!sel_bb_empty_p (new_bb));
5409
5410 /* NEW_BB has data sets that need to be updated and BB holds
5411 data sets that should be removed. Exchange these data sets
5412 so that we won't lose BB's valid data sets. */
5413 exchange_data_sets (new_bb, bb);
5414 free_data_sets (bb);
5415 }
5416
5417 if (!sel_bb_empty_p (new_bb)
5418 && bitmap_bit_p (blocks_to_reschedule, bb->index))
5419 bitmap_set_bit (blocks_to_reschedule, new_bb->index);
5420
5421 return new_bb;
5422 }
5423
5424 /* If BB ends with a jump insn whose ID is bigger then PREV_MAX_UID, return it.
5425 Otherwise returns NULL. */
5426 static rtx
5427 check_for_new_jump (basic_block bb, int prev_max_uid)
5428 {
5429 rtx end;
5430
5431 end = sel_bb_end (bb);
5432 if (end && INSN_UID (end) >= prev_max_uid)
5433 return end;
5434 return NULL;
5435 }
5436
5437 /* Look for a new jump either in FROM_BB block or in newly created JUMP_BB block.
5438 New means having UID at least equal to PREV_MAX_UID. */
5439 static rtx
5440 find_new_jump (basic_block from, basic_block jump_bb, int prev_max_uid)
5441 {
5442 rtx jump;
5443
5444 /* Return immediately if no new insns were emitted. */
5445 if (get_max_uid () == prev_max_uid)
5446 return NULL;
5447
5448 /* Now check both blocks for new jumps. It will ever be only one. */
5449 if ((jump = check_for_new_jump (from, prev_max_uid)))
5450 return jump;
5451
5452 if (jump_bb != NULL
5453 && (jump = check_for_new_jump (jump_bb, prev_max_uid)))
5454 return jump;
5455 return NULL;
5456 }
5457
5458 /* Splits E and adds the newly created basic block to the current region.
5459 Returns this basic block. */
5460 basic_block
5461 sel_split_edge (edge e)
5462 {
5463 basic_block new_bb, src, other_bb = NULL;
5464 int prev_max_uid;
5465 rtx jump;
5466
5467 src = e->src;
5468 prev_max_uid = get_max_uid ();
5469 new_bb = split_edge (e);
5470
5471 if (flag_sel_sched_pipelining_outer_loops
5472 && current_loop_nest)
5473 {
5474 int i;
5475 basic_block bb;
5476
5477 /* Some of the basic blocks might not have been added to the loop.
5478 Add them here, until this is fixed in force_fallthru. */
5479 for (i = 0;
5480 last_added_blocks.iterate (i, &bb); i++)
5481 if (!bb->loop_father)
5482 {
5483 add_bb_to_loop (bb, e->dest->loop_father);
5484
5485 gcc_assert (!other_bb && (new_bb->index != bb->index));
5486 other_bb = bb;
5487 }
5488 }
5489
5490 /* Add all last_added_blocks to the region. */
5491 sel_add_bb (NULL);
5492
5493 jump = find_new_jump (src, new_bb, prev_max_uid);
5494 if (jump)
5495 sel_init_new_insn (jump, INSN_INIT_TODO_LUID | INSN_INIT_TODO_SIMPLEJUMP);
5496
5497 /* Put the correct lv set on this block. */
5498 if (other_bb && !sel_bb_empty_p (other_bb))
5499 compute_live (sel_bb_head (other_bb));
5500
5501 return new_bb;
5502 }
5503
5504 /* Implement sched_create_empty_bb (). */
5505 static basic_block
5506 sel_create_empty_bb (basic_block after)
5507 {
5508 basic_block new_bb;
5509
5510 new_bb = sched_create_empty_bb_1 (after);
5511
5512 /* We'll explicitly initialize NEW_BB via sel_init_only_bb () a bit
5513 later. */
5514 gcc_assert (last_added_blocks.length () == 1
5515 && last_added_blocks[0] == new_bb);
5516
5517 last_added_blocks.release ();
5518 return new_bb;
5519 }
5520
5521 /* Implement sched_create_recovery_block. ORIG_INSN is where block
5522 will be splitted to insert a check. */
5523 basic_block
5524 sel_create_recovery_block (insn_t orig_insn)
5525 {
5526 basic_block first_bb, second_bb, recovery_block;
5527 basic_block before_recovery = NULL;
5528 rtx jump;
5529
5530 first_bb = BLOCK_FOR_INSN (orig_insn);
5531 if (sel_bb_end_p (orig_insn))
5532 {
5533 /* Avoid introducing an empty block while splitting. */
5534 gcc_assert (single_succ_p (first_bb));
5535 second_bb = single_succ (first_bb);
5536 }
5537 else
5538 second_bb = sched_split_block (first_bb, orig_insn);
5539
5540 recovery_block = sched_create_recovery_block (&before_recovery);
5541 if (before_recovery)
5542 copy_lv_set_from (before_recovery, EXIT_BLOCK_PTR_FOR_FN (cfun));
5543
5544 gcc_assert (sel_bb_empty_p (recovery_block));
5545 sched_create_recovery_edges (first_bb, recovery_block, second_bb);
5546 if (current_loops != NULL)
5547 add_bb_to_loop (recovery_block, first_bb->loop_father);
5548
5549 sel_add_bb (recovery_block);
5550
5551 jump = BB_END (recovery_block);
5552 gcc_assert (sel_bb_head (recovery_block) == jump);
5553 sel_init_new_insn (jump, INSN_INIT_TODO_LUID | INSN_INIT_TODO_SIMPLEJUMP);
5554
5555 return recovery_block;
5556 }
5557
5558 /* Merge basic block B into basic block A. */
5559 static void
5560 sel_merge_blocks (basic_block a, basic_block b)
5561 {
5562 gcc_assert (sel_bb_empty_p (b)
5563 && EDGE_COUNT (b->preds) == 1
5564 && EDGE_PRED (b, 0)->src == b->prev_bb);
5565
5566 move_bb_info (b->prev_bb, b);
5567 remove_empty_bb (b, false);
5568 merge_blocks (a, b);
5569 change_loops_latches (b, a);
5570 }
5571
5572 /* A wrapper for redirect_edge_and_branch_force, which also initializes
5573 data structures for possibly created bb and insns. Returns the newly
5574 added bb or NULL, when a bb was not needed. */
5575 void
5576 sel_redirect_edge_and_branch_force (edge e, basic_block to)
5577 {
5578 basic_block jump_bb, src, orig_dest = e->dest;
5579 int prev_max_uid;
5580 rtx jump;
5581
5582 /* This function is now used only for bookkeeping code creation, where
5583 we'll never get the single pred of orig_dest block and thus will not
5584 hit unreachable blocks when updating dominator info. */
5585 gcc_assert (!sel_bb_empty_p (e->src)
5586 && !single_pred_p (orig_dest));
5587 src = e->src;
5588 prev_max_uid = get_max_uid ();
5589 jump_bb = redirect_edge_and_branch_force (e, to);
5590
5591 if (jump_bb != NULL)
5592 sel_add_bb (jump_bb);
5593
5594 /* This function could not be used to spoil the loop structure by now,
5595 thus we don't care to update anything. But check it to be sure. */
5596 if (current_loop_nest
5597 && pipelining_p)
5598 gcc_assert (loop_latch_edge (current_loop_nest));
5599
5600 jump = find_new_jump (src, jump_bb, prev_max_uid);
5601 if (jump)
5602 sel_init_new_insn (jump, INSN_INIT_TODO_LUID | INSN_INIT_TODO_SIMPLEJUMP);
5603 set_immediate_dominator (CDI_DOMINATORS, to,
5604 recompute_dominator (CDI_DOMINATORS, to));
5605 set_immediate_dominator (CDI_DOMINATORS, orig_dest,
5606 recompute_dominator (CDI_DOMINATORS, orig_dest));
5607 }
5608
5609 /* A wrapper for redirect_edge_and_branch. Return TRUE if blocks connected by
5610 redirected edge are in reverse topological order. */
5611 bool
5612 sel_redirect_edge_and_branch (edge e, basic_block to)
5613 {
5614 bool latch_edge_p;
5615 basic_block src, orig_dest = e->dest;
5616 int prev_max_uid;
5617 rtx jump;
5618 edge redirected;
5619 bool recompute_toporder_p = false;
5620 bool maybe_unreachable = single_pred_p (orig_dest);
5621
5622 latch_edge_p = (pipelining_p
5623 && current_loop_nest
5624 && e == loop_latch_edge (current_loop_nest));
5625
5626 src = e->src;
5627 prev_max_uid = get_max_uid ();
5628
5629 redirected = redirect_edge_and_branch (e, to);
5630
5631 gcc_assert (redirected && !last_added_blocks.exists ());
5632
5633 /* When we've redirected a latch edge, update the header. */
5634 if (latch_edge_p)
5635 {
5636 current_loop_nest->header = to;
5637 gcc_assert (loop_latch_edge (current_loop_nest));
5638 }
5639
5640 /* In rare situations, the topological relation between the blocks connected
5641 by the redirected edge can change (see PR42245 for an example). Update
5642 block_to_bb/bb_to_block. */
5643 if (CONTAINING_RGN (e->src->index) == CONTAINING_RGN (to->index)
5644 && BLOCK_TO_BB (e->src->index) > BLOCK_TO_BB (to->index))
5645 recompute_toporder_p = true;
5646
5647 jump = find_new_jump (src, NULL, prev_max_uid);
5648 if (jump)
5649 sel_init_new_insn (jump, INSN_INIT_TODO_LUID | INSN_INIT_TODO_SIMPLEJUMP);
5650
5651 /* Only update dominator info when we don't have unreachable blocks.
5652 Otherwise we'll update in maybe_tidy_empty_bb. */
5653 if (!maybe_unreachable)
5654 {
5655 set_immediate_dominator (CDI_DOMINATORS, to,
5656 recompute_dominator (CDI_DOMINATORS, to));
5657 set_immediate_dominator (CDI_DOMINATORS, orig_dest,
5658 recompute_dominator (CDI_DOMINATORS, orig_dest));
5659 }
5660 return recompute_toporder_p;
5661 }
5662
5663 /* This variable holds the cfg hooks used by the selective scheduler. */
5664 static struct cfg_hooks sel_cfg_hooks;
5665
5666 /* Register sel-sched cfg hooks. */
5667 void
5668 sel_register_cfg_hooks (void)
5669 {
5670 sched_split_block = sel_split_block;
5671
5672 orig_cfg_hooks = get_cfg_hooks ();
5673 sel_cfg_hooks = orig_cfg_hooks;
5674
5675 sel_cfg_hooks.create_basic_block = sel_create_basic_block;
5676
5677 set_cfg_hooks (sel_cfg_hooks);
5678
5679 sched_init_only_bb = sel_init_only_bb;
5680 sched_split_block = sel_split_block;
5681 sched_create_empty_bb = sel_create_empty_bb;
5682 }
5683
5684 /* Unregister sel-sched cfg hooks. */
5685 void
5686 sel_unregister_cfg_hooks (void)
5687 {
5688 sched_create_empty_bb = NULL;
5689 sched_split_block = NULL;
5690 sched_init_only_bb = NULL;
5691
5692 set_cfg_hooks (orig_cfg_hooks);
5693 }
5694 \f
5695
5696 /* Emit an insn rtx based on PATTERN. If a jump insn is wanted,
5697 LABEL is where this jump should be directed. */
5698 rtx
5699 create_insn_rtx_from_pattern (rtx pattern, rtx label)
5700 {
5701 rtx insn_rtx;
5702
5703 gcc_assert (!INSN_P (pattern));
5704
5705 start_sequence ();
5706
5707 if (label == NULL_RTX)
5708 insn_rtx = emit_insn (pattern);
5709 else if (DEBUG_INSN_P (label))
5710 insn_rtx = emit_debug_insn (pattern);
5711 else
5712 {
5713 insn_rtx = emit_jump_insn (pattern);
5714 JUMP_LABEL (insn_rtx) = label;
5715 ++LABEL_NUSES (label);
5716 }
5717
5718 end_sequence ();
5719
5720 sched_extend_luids ();
5721 sched_extend_target ();
5722 sched_deps_init (false);
5723
5724 /* Initialize INSN_CODE now. */
5725 recog_memoized (insn_rtx);
5726 return insn_rtx;
5727 }
5728
5729 /* Create a new vinsn for INSN_RTX. FORCE_UNIQUE_P is true when the vinsn
5730 must not be clonable. */
5731 vinsn_t
5732 create_vinsn_from_insn_rtx (rtx insn_rtx, bool force_unique_p)
5733 {
5734 gcc_assert (INSN_P (insn_rtx) && !INSN_IN_STREAM_P (insn_rtx));
5735
5736 /* If VINSN_TYPE is not USE, retain its uniqueness. */
5737 return vinsn_create (insn_rtx, force_unique_p);
5738 }
5739
5740 /* Create a copy of INSN_RTX. */
5741 rtx
5742 create_copy_of_insn_rtx (rtx insn_rtx)
5743 {
5744 rtx res, link;
5745
5746 if (DEBUG_INSN_P (insn_rtx))
5747 return create_insn_rtx_from_pattern (copy_rtx (PATTERN (insn_rtx)),
5748 insn_rtx);
5749
5750 gcc_assert (NONJUMP_INSN_P (insn_rtx));
5751
5752 res = create_insn_rtx_from_pattern (copy_rtx (PATTERN (insn_rtx)),
5753 NULL_RTX);
5754
5755 /* Copy all REG_NOTES except REG_EQUAL/REG_EQUIV and REG_LABEL_OPERAND
5756 since mark_jump_label will make them. REG_LABEL_TARGETs are created
5757 there too, but are supposed to be sticky, so we copy them. */
5758 for (link = REG_NOTES (insn_rtx); link; link = XEXP (link, 1))
5759 if (REG_NOTE_KIND (link) != REG_LABEL_OPERAND
5760 && REG_NOTE_KIND (link) != REG_EQUAL
5761 && REG_NOTE_KIND (link) != REG_EQUIV)
5762 {
5763 if (GET_CODE (link) == EXPR_LIST)
5764 add_reg_note (res, REG_NOTE_KIND (link),
5765 copy_insn_1 (XEXP (link, 0)));
5766 else
5767 add_reg_note (res, REG_NOTE_KIND (link), XEXP (link, 0));
5768 }
5769
5770 return res;
5771 }
5772
5773 /* Change vinsn field of EXPR to hold NEW_VINSN. */
5774 void
5775 change_vinsn_in_expr (expr_t expr, vinsn_t new_vinsn)
5776 {
5777 vinsn_detach (EXPR_VINSN (expr));
5778
5779 EXPR_VINSN (expr) = new_vinsn;
5780 vinsn_attach (new_vinsn);
5781 }
5782
5783 /* Helpers for global init. */
5784 /* This structure is used to be able to call existing bundling mechanism
5785 and calculate insn priorities. */
5786 static struct haifa_sched_info sched_sel_haifa_sched_info =
5787 {
5788 NULL, /* init_ready_list */
5789 NULL, /* can_schedule_ready_p */
5790 NULL, /* schedule_more_p */
5791 NULL, /* new_ready */
5792 NULL, /* rgn_rank */
5793 sel_print_insn, /* rgn_print_insn */
5794 contributes_to_priority,
5795 NULL, /* insn_finishes_block_p */
5796
5797 NULL, NULL,
5798 NULL, NULL,
5799 0, 0,
5800
5801 NULL, /* add_remove_insn */
5802 NULL, /* begin_schedule_ready */
5803 NULL, /* begin_move_insn */
5804 NULL, /* advance_target_bb */
5805
5806 NULL,
5807 NULL,
5808
5809 SEL_SCHED | NEW_BBS
5810 };
5811
5812 /* Setup special insns used in the scheduler. */
5813 void
5814 setup_nop_and_exit_insns (void)
5815 {
5816 gcc_assert (nop_pattern == NULL_RTX
5817 && exit_insn == NULL_RTX);
5818
5819 nop_pattern = constm1_rtx;
5820
5821 start_sequence ();
5822 emit_insn (nop_pattern);
5823 exit_insn = get_insns ();
5824 end_sequence ();
5825 set_block_for_insn (exit_insn, EXIT_BLOCK_PTR_FOR_FN (cfun));
5826 }
5827
5828 /* Free special insns used in the scheduler. */
5829 void
5830 free_nop_and_exit_insns (void)
5831 {
5832 exit_insn = NULL_RTX;
5833 nop_pattern = NULL_RTX;
5834 }
5835
5836 /* Setup a special vinsn used in new insns initialization. */
5837 void
5838 setup_nop_vinsn (void)
5839 {
5840 nop_vinsn = vinsn_create (exit_insn, false);
5841 vinsn_attach (nop_vinsn);
5842 }
5843
5844 /* Free a special vinsn used in new insns initialization. */
5845 void
5846 free_nop_vinsn (void)
5847 {
5848 gcc_assert (VINSN_COUNT (nop_vinsn) == 1);
5849 vinsn_detach (nop_vinsn);
5850 nop_vinsn = NULL;
5851 }
5852
5853 /* Call a set_sched_flags hook. */
5854 void
5855 sel_set_sched_flags (void)
5856 {
5857 /* ??? This means that set_sched_flags were called, and we decided to
5858 support speculation. However, set_sched_flags also modifies flags
5859 on current_sched_info, doing this only at global init. And we
5860 sometimes change c_s_i later. So put the correct flags again. */
5861 if (spec_info && targetm.sched.set_sched_flags)
5862 targetm.sched.set_sched_flags (spec_info);
5863 }
5864
5865 /* Setup pointers to global sched info structures. */
5866 void
5867 sel_setup_sched_infos (void)
5868 {
5869 rgn_setup_common_sched_info ();
5870
5871 memcpy (&sel_common_sched_info, common_sched_info,
5872 sizeof (sel_common_sched_info));
5873
5874 sel_common_sched_info.fix_recovery_cfg = NULL;
5875 sel_common_sched_info.add_block = NULL;
5876 sel_common_sched_info.estimate_number_of_insns
5877 = sel_estimate_number_of_insns;
5878 sel_common_sched_info.luid_for_non_insn = sel_luid_for_non_insn;
5879 sel_common_sched_info.sched_pass_id = SCHED_SEL_PASS;
5880
5881 common_sched_info = &sel_common_sched_info;
5882
5883 current_sched_info = &sched_sel_haifa_sched_info;
5884 current_sched_info->sched_max_insns_priority =
5885 get_rgn_sched_max_insns_priority ();
5886
5887 sel_set_sched_flags ();
5888 }
5889 \f
5890
5891 /* Adds basic block BB to region RGN at the position *BB_ORD_INDEX,
5892 *BB_ORD_INDEX after that is increased. */
5893 static void
5894 sel_add_block_to_region (basic_block bb, int *bb_ord_index, int rgn)
5895 {
5896 RGN_NR_BLOCKS (rgn) += 1;
5897 RGN_DONT_CALC_DEPS (rgn) = 0;
5898 RGN_HAS_REAL_EBB (rgn) = 0;
5899 CONTAINING_RGN (bb->index) = rgn;
5900 BLOCK_TO_BB (bb->index) = *bb_ord_index;
5901 rgn_bb_table[RGN_BLOCKS (rgn) + *bb_ord_index] = bb->index;
5902 (*bb_ord_index)++;
5903
5904 /* FIXME: it is true only when not scheduling ebbs. */
5905 RGN_BLOCKS (rgn + 1) = RGN_BLOCKS (rgn) + RGN_NR_BLOCKS (rgn);
5906 }
5907
5908 /* Functions to support pipelining of outer loops. */
5909
5910 /* Creates a new empty region and returns it's number. */
5911 static int
5912 sel_create_new_region (void)
5913 {
5914 int new_rgn_number = nr_regions;
5915
5916 RGN_NR_BLOCKS (new_rgn_number) = 0;
5917
5918 /* FIXME: This will work only when EBBs are not created. */
5919 if (new_rgn_number != 0)
5920 RGN_BLOCKS (new_rgn_number) = RGN_BLOCKS (new_rgn_number - 1) +
5921 RGN_NR_BLOCKS (new_rgn_number - 1);
5922 else
5923 RGN_BLOCKS (new_rgn_number) = 0;
5924
5925 /* Set the blocks of the next region so the other functions may
5926 calculate the number of blocks in the region. */
5927 RGN_BLOCKS (new_rgn_number + 1) = RGN_BLOCKS (new_rgn_number) +
5928 RGN_NR_BLOCKS (new_rgn_number);
5929
5930 nr_regions++;
5931
5932 return new_rgn_number;
5933 }
5934
5935 /* If X has a smaller topological sort number than Y, returns -1;
5936 if greater, returns 1. */
5937 static int
5938 bb_top_order_comparator (const void *x, const void *y)
5939 {
5940 basic_block bb1 = *(const basic_block *) x;
5941 basic_block bb2 = *(const basic_block *) y;
5942
5943 gcc_assert (bb1 == bb2
5944 || rev_top_order_index[bb1->index]
5945 != rev_top_order_index[bb2->index]);
5946
5947 /* It's a reverse topological order in REV_TOP_ORDER_INDEX, so
5948 bbs with greater number should go earlier. */
5949 if (rev_top_order_index[bb1->index] > rev_top_order_index[bb2->index])
5950 return -1;
5951 else
5952 return 1;
5953 }
5954
5955 /* Create a region for LOOP and return its number. If we don't want
5956 to pipeline LOOP, return -1. */
5957 static int
5958 make_region_from_loop (struct loop *loop)
5959 {
5960 unsigned int i;
5961 int new_rgn_number = -1;
5962 struct loop *inner;
5963
5964 /* Basic block index, to be assigned to BLOCK_TO_BB. */
5965 int bb_ord_index = 0;
5966 basic_block *loop_blocks;
5967 basic_block preheader_block;
5968
5969 if (loop->num_nodes
5970 > (unsigned) PARAM_VALUE (PARAM_MAX_PIPELINE_REGION_BLOCKS))
5971 return -1;
5972
5973 /* Don't pipeline loops whose latch belongs to some of its inner loops. */
5974 for (inner = loop->inner; inner; inner = inner->inner)
5975 if (flow_bb_inside_loop_p (inner, loop->latch))
5976 return -1;
5977
5978 loop->ninsns = num_loop_insns (loop);
5979 if ((int) loop->ninsns > PARAM_VALUE (PARAM_MAX_PIPELINE_REGION_INSNS))
5980 return -1;
5981
5982 loop_blocks = get_loop_body_in_custom_order (loop, bb_top_order_comparator);
5983
5984 for (i = 0; i < loop->num_nodes; i++)
5985 if (loop_blocks[i]->flags & BB_IRREDUCIBLE_LOOP)
5986 {
5987 free (loop_blocks);
5988 return -1;
5989 }
5990
5991 preheader_block = loop_preheader_edge (loop)->src;
5992 gcc_assert (preheader_block);
5993 gcc_assert (loop_blocks[0] == loop->header);
5994
5995 new_rgn_number = sel_create_new_region ();
5996
5997 sel_add_block_to_region (preheader_block, &bb_ord_index, new_rgn_number);
5998 bitmap_set_bit (bbs_in_loop_rgns, preheader_block->index);
5999
6000 for (i = 0; i < loop->num_nodes; i++)
6001 {
6002 /* Add only those blocks that haven't been scheduled in the inner loop.
6003 The exception is the basic blocks with bookkeeping code - they should
6004 be added to the region (and they actually don't belong to the loop
6005 body, but to the region containing that loop body). */
6006
6007 gcc_assert (new_rgn_number >= 0);
6008
6009 if (! bitmap_bit_p (bbs_in_loop_rgns, loop_blocks[i]->index))
6010 {
6011 sel_add_block_to_region (loop_blocks[i], &bb_ord_index,
6012 new_rgn_number);
6013 bitmap_set_bit (bbs_in_loop_rgns, loop_blocks[i]->index);
6014 }
6015 }
6016
6017 free (loop_blocks);
6018 MARK_LOOP_FOR_PIPELINING (loop);
6019
6020 return new_rgn_number;
6021 }
6022
6023 /* Create a new region from preheader blocks LOOP_BLOCKS. */
6024 void
6025 make_region_from_loop_preheader (vec<basic_block> *&loop_blocks)
6026 {
6027 unsigned int i;
6028 int new_rgn_number = -1;
6029 basic_block bb;
6030
6031 /* Basic block index, to be assigned to BLOCK_TO_BB. */
6032 int bb_ord_index = 0;
6033
6034 new_rgn_number = sel_create_new_region ();
6035
6036 FOR_EACH_VEC_ELT (*loop_blocks, i, bb)
6037 {
6038 gcc_assert (new_rgn_number >= 0);
6039
6040 sel_add_block_to_region (bb, &bb_ord_index, new_rgn_number);
6041 }
6042
6043 vec_free (loop_blocks);
6044 }
6045
6046
6047 /* Create region(s) from loop nest LOOP, such that inner loops will be
6048 pipelined before outer loops. Returns true when a region for LOOP
6049 is created. */
6050 static bool
6051 make_regions_from_loop_nest (struct loop *loop)
6052 {
6053 struct loop *cur_loop;
6054 int rgn_number;
6055
6056 /* Traverse all inner nodes of the loop. */
6057 for (cur_loop = loop->inner; cur_loop; cur_loop = cur_loop->next)
6058 if (! bitmap_bit_p (bbs_in_loop_rgns, cur_loop->header->index))
6059 return false;
6060
6061 /* At this moment all regular inner loops should have been pipelined.
6062 Try to create a region from this loop. */
6063 rgn_number = make_region_from_loop (loop);
6064
6065 if (rgn_number < 0)
6066 return false;
6067
6068 loop_nests.safe_push (loop);
6069 return true;
6070 }
6071
6072 /* Initalize data structures needed. */
6073 void
6074 sel_init_pipelining (void)
6075 {
6076 /* Collect loop information to be used in outer loops pipelining. */
6077 loop_optimizer_init (LOOPS_HAVE_PREHEADERS
6078 | LOOPS_HAVE_FALLTHRU_PREHEADERS
6079 | LOOPS_HAVE_RECORDED_EXITS
6080 | LOOPS_HAVE_MARKED_IRREDUCIBLE_REGIONS);
6081 current_loop_nest = NULL;
6082
6083 bbs_in_loop_rgns = sbitmap_alloc (last_basic_block_for_fn (cfun));
6084 bitmap_clear (bbs_in_loop_rgns);
6085
6086 recompute_rev_top_order ();
6087 }
6088
6089 /* Returns a struct loop for region RGN. */
6090 loop_p
6091 get_loop_nest_for_rgn (unsigned int rgn)
6092 {
6093 /* Regions created with extend_rgns don't have corresponding loop nests,
6094 because they don't represent loops. */
6095 if (rgn < loop_nests.length ())
6096 return loop_nests[rgn];
6097 else
6098 return NULL;
6099 }
6100
6101 /* True when LOOP was included into pipelining regions. */
6102 bool
6103 considered_for_pipelining_p (struct loop *loop)
6104 {
6105 if (loop_depth (loop) == 0)
6106 return false;
6107
6108 /* Now, the loop could be too large or irreducible. Check whether its
6109 region is in LOOP_NESTS.
6110 We determine the region number of LOOP as the region number of its
6111 latch. We can't use header here, because this header could be
6112 just removed preheader and it will give us the wrong region number.
6113 Latch can't be used because it could be in the inner loop too. */
6114 if (LOOP_MARKED_FOR_PIPELINING_P (loop))
6115 {
6116 int rgn = CONTAINING_RGN (loop->latch->index);
6117
6118 gcc_assert ((unsigned) rgn < loop_nests.length ());
6119 return true;
6120 }
6121
6122 return false;
6123 }
6124
6125 /* Makes regions from the rest of the blocks, after loops are chosen
6126 for pipelining. */
6127 static void
6128 make_regions_from_the_rest (void)
6129 {
6130 int cur_rgn_blocks;
6131 int *loop_hdr;
6132 int i;
6133
6134 basic_block bb;
6135 edge e;
6136 edge_iterator ei;
6137 int *degree;
6138
6139 /* Index in rgn_bb_table where to start allocating new regions. */
6140 cur_rgn_blocks = nr_regions ? RGN_BLOCKS (nr_regions) : 0;
6141
6142 /* Make regions from all the rest basic blocks - those that don't belong to
6143 any loop or belong to irreducible loops. Prepare the data structures
6144 for extend_rgns. */
6145
6146 /* LOOP_HDR[I] == -1 if I-th bb doesn't belong to any loop,
6147 LOOP_HDR[I] == LOOP_HDR[J] iff basic blocks I and J reside within the same
6148 loop. */
6149 loop_hdr = XNEWVEC (int, last_basic_block_for_fn (cfun));
6150 degree = XCNEWVEC (int, last_basic_block_for_fn (cfun));
6151
6152
6153 /* For each basic block that belongs to some loop assign the number
6154 of innermost loop it belongs to. */
6155 for (i = 0; i < last_basic_block_for_fn (cfun); i++)
6156 loop_hdr[i] = -1;
6157
6158 FOR_EACH_BB_FN (bb, cfun)
6159 {
6160 if (bb->loop_father && !bb->loop_father->num == 0
6161 && !(bb->flags & BB_IRREDUCIBLE_LOOP))
6162 loop_hdr[bb->index] = bb->loop_father->num;
6163 }
6164
6165 /* For each basic block degree is calculated as the number of incoming
6166 edges, that are going out of bbs that are not yet scheduled.
6167 The basic blocks that are scheduled have degree value of zero. */
6168 FOR_EACH_BB_FN (bb, cfun)
6169 {
6170 degree[bb->index] = 0;
6171
6172 if (!bitmap_bit_p (bbs_in_loop_rgns, bb->index))
6173 {
6174 FOR_EACH_EDGE (e, ei, bb->preds)
6175 if (!bitmap_bit_p (bbs_in_loop_rgns, e->src->index))
6176 degree[bb->index]++;
6177 }
6178 else
6179 degree[bb->index] = -1;
6180 }
6181
6182 extend_rgns (degree, &cur_rgn_blocks, bbs_in_loop_rgns, loop_hdr);
6183
6184 /* Any block that did not end up in a region is placed into a region
6185 by itself. */
6186 FOR_EACH_BB_FN (bb, cfun)
6187 if (degree[bb->index] >= 0)
6188 {
6189 rgn_bb_table[cur_rgn_blocks] = bb->index;
6190 RGN_NR_BLOCKS (nr_regions) = 1;
6191 RGN_BLOCKS (nr_regions) = cur_rgn_blocks++;
6192 RGN_DONT_CALC_DEPS (nr_regions) = 0;
6193 RGN_HAS_REAL_EBB (nr_regions) = 0;
6194 CONTAINING_RGN (bb->index) = nr_regions++;
6195 BLOCK_TO_BB (bb->index) = 0;
6196 }
6197
6198 free (degree);
6199 free (loop_hdr);
6200 }
6201
6202 /* Free data structures used in pipelining of loops. */
6203 void sel_finish_pipelining (void)
6204 {
6205 struct loop *loop;
6206
6207 /* Release aux fields so we don't free them later by mistake. */
6208 FOR_EACH_LOOP (loop, 0)
6209 loop->aux = NULL;
6210
6211 loop_optimizer_finalize ();
6212
6213 loop_nests.release ();
6214
6215 free (rev_top_order_index);
6216 rev_top_order_index = NULL;
6217 }
6218
6219 /* This function replaces the find_rgns when
6220 FLAG_SEL_SCHED_PIPELINING_OUTER_LOOPS is set. */
6221 void
6222 sel_find_rgns (void)
6223 {
6224 sel_init_pipelining ();
6225 extend_regions ();
6226
6227 if (current_loops)
6228 {
6229 loop_p loop;
6230
6231 FOR_EACH_LOOP (loop, (flag_sel_sched_pipelining_outer_loops
6232 ? LI_FROM_INNERMOST
6233 : LI_ONLY_INNERMOST))
6234 make_regions_from_loop_nest (loop);
6235 }
6236
6237 /* Make regions from all the rest basic blocks and schedule them.
6238 These blocks include blocks that don't belong to any loop or belong
6239 to irreducible loops. */
6240 make_regions_from_the_rest ();
6241
6242 /* We don't need bbs_in_loop_rgns anymore. */
6243 sbitmap_free (bbs_in_loop_rgns);
6244 bbs_in_loop_rgns = NULL;
6245 }
6246
6247 /* Add the preheader blocks from previous loop to current region taking
6248 it from LOOP_PREHEADER_BLOCKS (current_loop_nest) and record them in *BBS.
6249 This function is only used with -fsel-sched-pipelining-outer-loops. */
6250 void
6251 sel_add_loop_preheaders (bb_vec_t *bbs)
6252 {
6253 int i;
6254 basic_block bb;
6255 vec<basic_block> *preheader_blocks
6256 = LOOP_PREHEADER_BLOCKS (current_loop_nest);
6257
6258 if (!preheader_blocks)
6259 return;
6260
6261 for (i = 0; preheader_blocks->iterate (i, &bb); i++)
6262 {
6263 bbs->safe_push (bb);
6264 last_added_blocks.safe_push (bb);
6265 sel_add_bb (bb);
6266 }
6267
6268 vec_free (preheader_blocks);
6269 }
6270
6271 /* While pipelining outer loops, returns TRUE if BB is a loop preheader.
6272 Please note that the function should also work when pipelining_p is
6273 false, because it is used when deciding whether we should or should
6274 not reschedule pipelined code. */
6275 bool
6276 sel_is_loop_preheader_p (basic_block bb)
6277 {
6278 if (current_loop_nest)
6279 {
6280 struct loop *outer;
6281
6282 if (preheader_removed)
6283 return false;
6284
6285 /* Preheader is the first block in the region. */
6286 if (BLOCK_TO_BB (bb->index) == 0)
6287 return true;
6288
6289 /* We used to find a preheader with the topological information.
6290 Check that the above code is equivalent to what we did before. */
6291
6292 if (in_current_region_p (current_loop_nest->header))
6293 gcc_assert (!(BLOCK_TO_BB (bb->index)
6294 < BLOCK_TO_BB (current_loop_nest->header->index)));
6295
6296 /* Support the situation when the latch block of outer loop
6297 could be from here. */
6298 for (outer = loop_outer (current_loop_nest);
6299 outer;
6300 outer = loop_outer (outer))
6301 if (considered_for_pipelining_p (outer) && outer->latch == bb)
6302 gcc_unreachable ();
6303 }
6304
6305 return false;
6306 }
6307
6308 /* Check whether JUMP_BB ends with a jump insn that leads only to DEST_BB and
6309 can be removed, making the corresponding edge fallthrough (assuming that
6310 all basic blocks between JUMP_BB and DEST_BB are empty). */
6311 static bool
6312 bb_has_removable_jump_to_p (basic_block jump_bb, basic_block dest_bb)
6313 {
6314 if (!onlyjump_p (BB_END (jump_bb))
6315 || tablejump_p (BB_END (jump_bb), NULL, NULL))
6316 return false;
6317
6318 /* Several outgoing edges, abnormal edge or destination of jump is
6319 not DEST_BB. */
6320 if (EDGE_COUNT (jump_bb->succs) != 1
6321 || EDGE_SUCC (jump_bb, 0)->flags & (EDGE_ABNORMAL | EDGE_CROSSING)
6322 || EDGE_SUCC (jump_bb, 0)->dest != dest_bb)
6323 return false;
6324
6325 /* If not anything of the upper. */
6326 return true;
6327 }
6328
6329 /* Removes the loop preheader from the current region and saves it in
6330 PREHEADER_BLOCKS of the father loop, so they will be added later to
6331 region that represents an outer loop. */
6332 static void
6333 sel_remove_loop_preheader (void)
6334 {
6335 int i, old_len;
6336 int cur_rgn = CONTAINING_RGN (BB_TO_BLOCK (0));
6337 basic_block bb;
6338 bool all_empty_p = true;
6339 vec<basic_block> *preheader_blocks
6340 = LOOP_PREHEADER_BLOCKS (loop_outer (current_loop_nest));
6341
6342 vec_check_alloc (preheader_blocks, 0);
6343
6344 gcc_assert (current_loop_nest);
6345 old_len = preheader_blocks->length ();
6346
6347 /* Add blocks that aren't within the current loop to PREHEADER_BLOCKS. */
6348 for (i = 0; i < RGN_NR_BLOCKS (cur_rgn); i++)
6349 {
6350 bb = BASIC_BLOCK_FOR_FN (cfun, BB_TO_BLOCK (i));
6351
6352 /* If the basic block belongs to region, but doesn't belong to
6353 corresponding loop, then it should be a preheader. */
6354 if (sel_is_loop_preheader_p (bb))
6355 {
6356 preheader_blocks->safe_push (bb);
6357 if (BB_END (bb) != bb_note (bb))
6358 all_empty_p = false;
6359 }
6360 }
6361
6362 /* Remove these blocks only after iterating over the whole region. */
6363 for (i = preheader_blocks->length () - 1; i >= old_len; i--)
6364 {
6365 bb = (*preheader_blocks)[i];
6366 sel_remove_bb (bb, false);
6367 }
6368
6369 if (!considered_for_pipelining_p (loop_outer (current_loop_nest)))
6370 {
6371 if (!all_empty_p)
6372 /* Immediately create new region from preheader. */
6373 make_region_from_loop_preheader (preheader_blocks);
6374 else
6375 {
6376 /* If all preheader blocks are empty - dont create new empty region.
6377 Instead, remove them completely. */
6378 FOR_EACH_VEC_ELT (*preheader_blocks, i, bb)
6379 {
6380 edge e;
6381 edge_iterator ei;
6382 basic_block prev_bb = bb->prev_bb, next_bb = bb->next_bb;
6383
6384 /* Redirect all incoming edges to next basic block. */
6385 for (ei = ei_start (bb->preds); (e = ei_safe_edge (ei)); )
6386 {
6387 if (! (e->flags & EDGE_FALLTHRU))
6388 redirect_edge_and_branch (e, bb->next_bb);
6389 else
6390 redirect_edge_succ (e, bb->next_bb);
6391 }
6392 gcc_assert (BB_NOTE_LIST (bb) == NULL);
6393 delete_and_free_basic_block (bb);
6394
6395 /* Check if after deleting preheader there is a nonconditional
6396 jump in PREV_BB that leads to the next basic block NEXT_BB.
6397 If it is so - delete this jump and clear data sets of its
6398 basic block if it becomes empty. */
6399 if (next_bb->prev_bb == prev_bb
6400 && prev_bb != ENTRY_BLOCK_PTR_FOR_FN (cfun)
6401 && bb_has_removable_jump_to_p (prev_bb, next_bb))
6402 {
6403 redirect_edge_and_branch (EDGE_SUCC (prev_bb, 0), next_bb);
6404 if (BB_END (prev_bb) == bb_note (prev_bb))
6405 free_data_sets (prev_bb);
6406 }
6407
6408 set_immediate_dominator (CDI_DOMINATORS, next_bb,
6409 recompute_dominator (CDI_DOMINATORS,
6410 next_bb));
6411 }
6412 }
6413 vec_free (preheader_blocks);
6414 }
6415 else
6416 /* Store preheader within the father's loop structure. */
6417 SET_LOOP_PREHEADER_BLOCKS (loop_outer (current_loop_nest),
6418 preheader_blocks);
6419 }
6420 #endif