3 * File name: ir/ir/irgopt.c
4 * Purpose: Optimizations for a whole ir graph, i.e., a procedure.
5 * Author: Christian Schaefer, Goetz Lindenmaier
6 * Modified by: Sebastian Felis
9 * Copyright: (c) 1998-2003 Universität Karlsruhe
10 * Licence: This file protected by GPL - GNU GENERAL PUBLIC LICENSE.
21 #include "irgraph_t.h"
34 #include "pdeq.h" /* Fuer code placement */
39 #include "irbackedge_t.h"
46 #include "iredges_t.h"
49 /* Defined in iropt.c */
50 pset *new_identities (void);
51 void del_identities (pset *value_table);
52 void add_identities (pset *value_table, ir_node *node);
54 /*------------------------------------------------------------------*/
55 /* apply optimizations of iropt to all nodes. */
56 /*------------------------------------------------------------------*/
58 static void init_link (ir_node *n, void *env) {
59 set_irn_link(n, NULL);
62 #if 0 /* Old version. Avoids Ids.
63 This is not necessary: we do a post walk, and get_irn_n
64 removes ids anyways. So it's much cheaper to call the
65 optimization less often and use the exchange() algorithm. */
67 optimize_in_place_wrapper (ir_node *n, void *env) {
69 ir_node *optimized, *old;
71 irn_arity = get_irn_arity(n);
72 for (i = 0; i < irn_arity; i++) {
73 /* get_irn_n skips Id nodes, so comparison old != optimized does not
74 show all optimizations. Therefore always set new predecessor. */
75 old = get_irn_intra_n(n, i);
76 optimized = optimize_in_place_2(old);
77 set_irn_n(n, i, optimized);
80 if (get_irn_op(n) == op_Block) {
81 optimized = optimize_in_place_2(n);
82 if (optimized != n) exchange (n, optimized);
87 optimize_in_place_wrapper (ir_node *n, void *env) {
88 ir_node *optimized = optimize_in_place_2(n);
89 if (optimized != n) exchange (n, optimized);
94 static INLINE void do_local_optimize(ir_node *n) {
95 /* Handle graph state */
96 assert(get_irg_phase_state(current_ir_graph) != phase_building);
98 if (get_opt_global_cse())
99 set_irg_pinned(current_ir_graph, op_pin_state_floats);
100 set_irg_outs_inconsistent(current_ir_graph);
101 set_irg_doms_inconsistent(current_ir_graph);
102 set_irg_loopinfo_inconsistent(current_ir_graph);
104 /* Clean the value_table in irg for the CSE. */
105 del_identities(current_ir_graph->value_table);
106 current_ir_graph->value_table = new_identities();
108 /* walk over the graph */
109 irg_walk(n, init_link, optimize_in_place_wrapper, NULL);
112 void local_optimize_node(ir_node *n) {
113 ir_graph *rem = current_ir_graph;
114 current_ir_graph = get_irn_irg(n);
116 do_local_optimize(n);
118 current_ir_graph = rem;
122 * Block-Walker: uses dominance depth to mark dead blocks.
124 static void kill_dead_blocks(ir_node *block, void *env)
126 if (get_Block_dom_depth(block) < 0)
127 if (block != get_irg_end_block(current_ir_graph)) {
128 /* we don't want that the end block of graphs with
129 endless loops is marked bad (although it is of course */
130 set_Block_dead(block);
135 local_optimize_graph (ir_graph *irg) {
136 ir_graph *rem = current_ir_graph;
137 current_ir_graph = irg;
139 if (get_irg_dom_state(current_ir_graph) == dom_consistent)
140 irg_block_walk_graph(irg, NULL, kill_dead_blocks, NULL);
142 do_local_optimize(get_irg_end(irg));
144 current_ir_graph = rem;
148 /*------------------------------------------------------------------*/
149 /* Routines for dead node elimination / copying garbage collection */
150 /* of the obstack. */
151 /*------------------------------------------------------------------*/
154 * Remember the new node in the old node by using a field all nodes have.
156 #define set_new_node(oldn, newn) set_irn_link(oldn, newn)
159 * Get this new node, before the old node is forgotten.
161 #define get_new_node(oldn) get_irn_link(oldn)
164 * Check if a new node was set.
166 #define has_new_node(n) (get_new_node(n) != NULL)
169 * We use the block_visited flag to mark that we have computed the
170 * number of useful predecessors for this block.
171 * Further we encode the new arity in this flag in the old blocks.
172 * Remembering the arity is useful, as it saves a lot of pointer
173 * accesses. This function is called for all Phi and Block nodes
177 compute_new_arity(ir_node *b) {
178 int i, res, irn_arity;
181 irg_v = get_irg_block_visited(current_ir_graph);
182 block_v = get_Block_block_visited(b);
183 if (block_v >= irg_v) {
184 /* we computed the number of preds for this block and saved it in the
186 return block_v - irg_v;
188 /* compute the number of good predecessors */
189 res = irn_arity = get_irn_arity(b);
190 for (i = 0; i < irn_arity; i++)
191 if (get_irn_opcode(get_irn_n(b, i)) == iro_Bad) res--;
192 /* save it in the flag. */
193 set_Block_block_visited(b, irg_v + res);
199 * Copies the node to the new obstack. The Ins of the new node point to
200 * the predecessors on the old obstack. For block/phi nodes not all
201 * predecessors might be copied. n->link points to the new node.
202 * For Phi and Block nodes the function allocates in-arrays with an arity
203 * only for useful predecessors. The arity is determined by counting
204 * the non-bad predecessors of the block.
206 * @param n The node to be copied
207 * @param env if non-NULL, the node number attribute will be copied to the new node
209 * Note: Also used for loop unrolling.
211 static void copy_node(ir_node *n, void *env) {
214 ir_op *op = get_irn_op(n);
215 int copy_node_nr = env != NULL;
217 /* The end node looses it's flexible in array. This doesn't matter,
218 as dead node elimination builds End by hand, inlineing doesn't use
220 /* assert(op == op_End || ((_ARR_DESCR(n->in))->cookie != ARR_F_MAGIC)); */
223 /* node copied already */
225 } else if (op == op_Block) {
227 new_arity = compute_new_arity(n);
228 n->attr.block.graph_arr = NULL;
230 block = get_nodes_block(n);
232 new_arity = compute_new_arity(block);
234 new_arity = get_irn_arity(n);
237 nn = new_ir_node(get_irn_dbg_info(n),
244 /* Copy the attributes. These might point to additional data. If this
245 was allocated on the old obstack the pointers now are dangling. This
246 frees e.g. the memory of the graph_arr allocated in new_immBlock. */
247 copy_node_attr(n, nn);
248 new_backedge_info(nn);
252 /* for easier debugging, we want to copy the node numbers too */
253 nn->node_nr = n->node_nr;
258 hook_dead_node_elim_subst(current_ir_graph, n, nn);
262 * Copies new predecessors of old node to new node remembered in link.
263 * Spare the Bad predecessors of Phi and Block nodes.
266 copy_preds (ir_node *n, void *env) {
270 nn = get_new_node(n);
272 /* printf("\n old node: "); DDMSG2(n);
273 printf(" new node: "); DDMSG2(nn);
274 printf(" arities: old: %d, new: %d\n", get_irn_arity(n), get_irn_arity(nn)); */
277 /* Don't copy Bad nodes. */
279 irn_arity = get_irn_arity(n);
280 for (i = 0; i < irn_arity; i++)
281 if (! is_Bad(get_irn_n(n, i))) {
282 set_irn_n (nn, j, get_new_node(get_irn_n(n, i)));
283 /*if (is_backedge(n, i)) set_backedge(nn, j);*/
286 /* repair the block visited flag from above misuse. Repair it in both
287 graphs so that the old one can still be used. */
288 set_Block_block_visited(nn, 0);
289 set_Block_block_visited(n, 0);
290 /* Local optimization could not merge two subsequent blocks if
291 in array contained Bads. Now it's possible.
292 We don't call optimize_in_place as it requires
293 that the fields in ir_graph are set properly. */
294 if ((get_opt_control_flow_straightening()) &&
295 (get_Block_n_cfgpreds(nn) == 1) &&
296 (get_irn_op(get_Block_cfgpred(nn, 0)) == op_Jmp)) {
297 ir_node *old = get_nodes_block(get_Block_cfgpred(nn, 0));
299 /* Jmp jumps into the block it is in -- deal self cycle. */
300 assert(is_Bad(get_new_node(get_irg_bad(current_ir_graph))));
301 exchange(nn, get_new_node(get_irg_bad(current_ir_graph)));
306 } else if (get_irn_op(n) == op_Phi) {
307 /* Don't copy node if corresponding predecessor in block is Bad.
308 The Block itself should not be Bad. */
309 block = get_nodes_block(n);
310 set_irn_n (nn, -1, get_new_node(block));
312 irn_arity = get_irn_arity(n);
313 for (i = 0; i < irn_arity; i++)
314 if (! is_Bad(get_irn_n(block, i))) {
315 set_irn_n (nn, j, get_new_node(get_irn_n(n, i)));
316 /*if (is_backedge(n, i)) set_backedge(nn, j);*/
319 /* If the pre walker reached this Phi after the post walker visited the
320 block block_visited is > 0. */
321 set_Block_block_visited(get_nodes_block(n), 0);
322 /* Compacting the Phi's ins might generate Phis with only one
324 if (get_irn_arity(nn) == 1)
325 exchange(nn, get_irn_n(nn, 0));
327 irn_arity = get_irn_arity(n);
328 for (i = -1; i < irn_arity; i++)
329 set_irn_n (nn, i, get_new_node(get_irn_n(n, i)));
331 /* Now the new node is complete. We can add it to the hash table for CSE.
332 @@@ inlining aborts if we identify End. Why? */
333 if (get_irn_op(nn) != op_End)
334 add_identities (current_ir_graph->value_table, nn);
338 * Copies the graph recursively, compacts the keep-alives of the end node.
340 * @param irg the graph to be copied
341 * @param copy_node_nr If non-zero, the node number will be copied
343 static void copy_graph(ir_graph *irg, int copy_node_nr) {
344 ir_node *oe, *ne, *ob, *nb, *om, *nm; /* old end, new end, old bad, new bad, old NoMem, new NoMem */
345 ir_node *ka; /* keep alive */
349 /* Some nodes must be copied by hand, sigh */
350 vfl = get_irg_visited(irg);
351 set_irg_visited(irg, vfl + 1);
353 oe = get_irg_end(irg);
354 mark_irn_visited(oe);
355 /* copy the end node by hand, allocate dynamic in array! */
356 ne = new_ir_node(get_irn_dbg_info(oe),
363 /* Copy the attributes. Well, there might be some in the future... */
364 copy_node_attr(oe, ne);
365 set_new_node(oe, ne);
367 /* copy the Bad node */
368 ob = get_irg_bad(irg);
369 mark_irn_visited(ob);
370 nb = new_ir_node(get_irn_dbg_info(ob),
377 copy_node_attr(ob, nb);
378 set_new_node(ob, nb);
380 /* copy the NoMem node */
381 om = get_irg_no_mem(irg);
382 mark_irn_visited(om);
383 nm = new_ir_node(get_irn_dbg_info(om),
390 copy_node_attr(om, nm);
391 set_new_node(om, nm);
393 /* copy the live nodes */
394 set_irg_visited(irg, vfl);
395 irg_walk(get_nodes_block(oe), copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
397 /* Note: from yet, the visited flag of the graph is equal to vfl + 1 */
399 /* visit the anchors as well */
400 for (i = anchor_max - 1; i >= 0; --i) {
401 ir_node *n = irg->anchors[i];
403 if (n && (get_irn_visited(n) <= vfl)) {
404 set_irg_visited(irg, vfl);
405 irg_walk(n, copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
409 /* copy_preds for the end node ... */
410 set_nodes_block(ne, get_new_node(get_nodes_block(oe)));
412 /*- ... and now the keep alives. -*/
413 /* First pick the not marked block nodes and walk them. We must pick these
414 first as else we will oversee blocks reachable from Phis. */
415 irn_arity = get_irn_arity(oe);
416 for (i = 0; i < irn_arity; i++) {
417 ka = get_irn_intra_n(oe, i);
419 (get_irn_visited(ka) <= vfl)) {
420 /* We must keep the block alive and copy everything reachable */
421 set_irg_visited(irg, vfl);
422 irg_walk(ka, copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
423 add_End_keepalive(ne, get_new_node(ka));
427 /* Now pick other nodes. Here we will keep all! */
428 irn_arity = get_irn_arity(oe);
429 for (i = 0; i < irn_arity; i++) {
430 ka = get_irn_intra_n(oe, i);
432 if (get_irn_visited(ka) <= vfl) {
433 /* We didn't copy the node yet. */
434 set_irg_visited(irg, vfl);
435 irg_walk(ka, copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
437 add_End_keepalive(ne, get_new_node(ka));
441 /* start block sometimes only reached after keep alives */
442 set_nodes_block(nb, get_new_node(get_nodes_block(ob)));
443 set_nodes_block(nm, get_new_node(get_nodes_block(om)));
447 * Copies the graph reachable from current_ir_graph->end to the obstack
448 * in current_ir_graph and fixes the environment.
449 * Then fixes the fields in current_ir_graph containing nodes of the
452 * @param copy_node_nr If non-zero, the node number will be copied
455 copy_graph_env (int copy_node_nr) {
456 ir_graph *irg = current_ir_graph;
457 ir_node *old_end, *n;
460 /* remove end_except and end_reg nodes */
461 old_end = get_irg_end(irg);
462 set_irg_end_except (irg, old_end);
463 set_irg_end_reg (irg, old_end);
465 /* Not all nodes remembered in irg might be reachable
466 from the end node. Assure their link is set to NULL, so that
467 we can test whether new nodes have been computed. */
468 for (i = anchor_max - 1; i >= 0; --i)
470 set_new_node(irg->anchors[i], NULL);
472 /* we use the block walk flag for removing Bads from Blocks ins. */
473 inc_irg_block_visited(irg);
476 copy_graph(irg, copy_node_nr);
478 /* fix the fields in irg */
479 old_end = get_irg_end(irg);
480 for (i = anchor_max - 1; i >= 0; --i) {
483 irg->anchors[i] = get_new_node(n);
489 * Copies all reachable nodes to a new obstack. Removes bad inputs
490 * from block nodes and the corresponding inputs from Phi nodes.
491 * Merges single exit blocks with single entry blocks and removes
493 * Adds all new nodes to a new hash table for CSE. Does not
494 * perform CSE, so the hash table might contain common subexpressions.
497 dead_node_elimination(ir_graph *irg) {
499 int rem_ipview = get_interprocedural_view();
500 struct obstack *graveyard_obst = NULL;
501 struct obstack *rebirth_obst = NULL;
503 if (get_opt_optimize() && get_opt_dead_node_elimination()) {
504 assert(! edges_activated(irg) && "dead node elimination requires disabled edges");
506 /* inform statistics that we started a dead-node elimination run */
507 hook_dead_node_elim(irg, 1);
509 /* Remember external state of current_ir_graph. */
510 rem = current_ir_graph;
511 current_ir_graph = irg;
512 set_interprocedural_view(0);
514 assert(get_irg_phase_state(current_ir_graph) != phase_building);
516 /* Handle graph state */
517 free_callee_info(current_ir_graph);
518 free_irg_outs(current_ir_graph);
521 /* @@@ so far we loose loops when copying */
522 free_loop_information(current_ir_graph);
524 set_irg_doms_inconsistent(irg);
526 /* A quiet place, where the old obstack can rest in peace,
527 until it will be cremated. */
528 graveyard_obst = irg->obst;
530 /* A new obstack, where the reachable nodes will be copied to. */
531 rebirth_obst = xmalloc (sizeof(*rebirth_obst));
532 current_ir_graph->obst = rebirth_obst;
533 obstack_init (current_ir_graph->obst);
534 current_ir_graph->last_node_idx = 0;
536 /* We also need a new hash table for cse */
537 del_identities (irg->value_table);
538 irg->value_table = new_identities ();
540 /* Copy the graph from the old to the new obstack */
543 /* Free memory from old unoptimized obstack */
544 obstack_free(graveyard_obst, 0); /* First empty the obstack ... */
545 xfree (graveyard_obst); /* ... then free it. */
547 /* inform statistics that the run is over */
548 hook_dead_node_elim(irg, 0);
550 current_ir_graph = rem;
551 set_interprocedural_view(rem_ipview);
556 * Relink bad predecessors of a block and store the old in array to the
557 * link field. This function is called by relink_bad_predecessors().
558 * The array of link field starts with the block operand at position 0.
559 * If block has bad predecessors, create a new in array without bad preds.
560 * Otherwise let in array untouched.
562 static void relink_bad_block_predecessors(ir_node *n, void *env) {
563 ir_node **new_in, *irn;
564 int i, new_irn_n, old_irn_arity, new_irn_arity = 0;
566 /* if link field of block is NULL, look for bad predecessors otherwise
567 this is already done */
568 if (get_irn_op(n) == op_Block &&
569 get_irn_link(n) == NULL) {
571 /* save old predecessors in link field (position 0 is the block operand)*/
572 set_irn_link(n, get_irn_in(n));
574 /* count predecessors without bad nodes */
575 old_irn_arity = get_irn_arity(n);
576 for (i = 0; i < old_irn_arity; i++)
577 if (!is_Bad(get_irn_n(n, i))) new_irn_arity++;
579 /* arity changing: set new predecessors without bad nodes */
580 if (new_irn_arity < old_irn_arity) {
581 /* Get new predecessor array. We do not resize the array, as we must
582 keep the old one to update Phis. */
583 new_in = NEW_ARR_D (ir_node *, current_ir_graph->obst, (new_irn_arity+1));
585 /* set new predecessors in array */
588 for (i = 0; i < old_irn_arity; i++) {
589 irn = get_irn_n(n, i);
591 new_in[new_irn_n] = irn;
592 is_backedge(n, i) ? set_backedge(n, new_irn_n-1) : set_not_backedge(n, new_irn_n-1);
596 //ARR_SETLEN(int, n->attr.block.backedge, new_irn_arity);
597 ARR_SHRINKLEN(n->attr.block.backedge, new_irn_arity);
600 } /* ir node has bad predecessors */
602 } /* Block is not relinked */
606 * Relinks Bad predecessors from Blocks and Phis called by walker
607 * remove_bad_predecesors(). If n is a Block, call
608 * relink_bad_block_redecessors(). If n is a Phi-node, call also the relinking
609 * function of Phi's Block. If this block has bad predecessors, relink preds
612 static void relink_bad_predecessors(ir_node *n, void *env) {
613 ir_node *block, **old_in;
614 int i, old_irn_arity, new_irn_arity;
616 /* relink bad predecessors of a block */
617 if (get_irn_op(n) == op_Block)
618 relink_bad_block_predecessors(n, env);
620 /* If Phi node relink its block and its predecessors */
621 if (get_irn_op(n) == op_Phi) {
623 /* Relink predecessors of phi's block */
624 block = get_nodes_block(n);
625 if (get_irn_link(block) == NULL)
626 relink_bad_block_predecessors(block, env);
628 old_in = (ir_node **)get_irn_link(block); /* Of Phi's Block */
629 old_irn_arity = ARR_LEN(old_in);
631 /* Relink Phi predecessors if count of predecessors changed */
632 if (old_irn_arity != ARR_LEN(get_irn_in(block))) {
633 /* set new predecessors in array
634 n->in[0] remains the same block */
636 for(i = 1; i < old_irn_arity; i++)
637 if (!is_Bad((ir_node *)old_in[i])) {
638 n->in[new_irn_arity] = n->in[i];
639 is_backedge(n, i) ? set_backedge(n, new_irn_arity) : set_not_backedge(n, new_irn_arity);
643 ARR_SETLEN(ir_node *, n->in, new_irn_arity);
644 ARR_SETLEN(int, n->attr.phi_backedge, new_irn_arity);
647 } /* n is a Phi node */
651 * Removes Bad Bad predecessors from Blocks and the corresponding
652 * inputs to Phi nodes as in dead_node_elimination but without
654 * On walking up set the link field to NULL, on walking down call
655 * relink_bad_predecessors() (This function stores the old in array
656 * to the link field and sets a new in array if arity of predecessors
659 void remove_bad_predecessors(ir_graph *irg) {
660 irg_walk_graph(irg, init_link, relink_bad_predecessors, NULL);
667 __)|_| | \_/ | \_/(/_ |_/\__|__
669 The following stuff implements a facility that automatically patches
670 registered ir_node pointers to the new node when a dead node elimination occurs.
673 struct _survive_dce_t {
677 hook_entry_t dead_node_elim;
678 hook_entry_t dead_node_elim_subst;
681 typedef struct _survive_dce_list_t {
682 struct _survive_dce_list_t *next;
684 } survive_dce_list_t;
686 static void dead_node_hook(void *context, ir_graph *irg, int start)
688 survive_dce_t *sd = context;
690 /* Create a new map before the dead node elimination is performed. */
692 sd->new_places = pmap_create_ex(pmap_count(sd->places));
695 /* Patch back all nodes if dead node elimination is over and something is to be done. */
697 pmap_destroy(sd->places);
698 sd->places = sd->new_places;
699 sd->new_places = NULL;
704 * Hook called when dead node elimination replaces old by nw.
706 static void dead_node_subst_hook(void *context, ir_graph *irg, ir_node *old, ir_node *nw)
708 survive_dce_t *sd = context;
709 survive_dce_list_t *list = pmap_get(sd->places, old);
711 /* If the node is to be patched back, write the new address to all registered locations. */
713 survive_dce_list_t *p;
715 for(p = list; p; p = p->next)
718 pmap_insert(sd->new_places, nw, list);
723 * Make a new Survive DCE environment.
725 survive_dce_t *new_survive_dce(void)
727 survive_dce_t *res = xmalloc(sizeof(res[0]));
728 obstack_init(&res->obst);
729 res->places = pmap_create();
730 res->new_places = NULL;
732 res->dead_node_elim.hook._hook_dead_node_elim = dead_node_hook;
733 res->dead_node_elim.context = res;
734 res->dead_node_elim.next = NULL;
736 res->dead_node_elim_subst.hook._hook_dead_node_elim_subst = dead_node_subst_hook;
737 res->dead_node_elim_subst.context = res;
738 res->dead_node_elim_subst.next = NULL;
740 register_hook(hook_dead_node_elim, &res->dead_node_elim);
741 register_hook(hook_dead_node_elim_subst, &res->dead_node_elim_subst);
746 * Free a Survive DCE environment.
748 void free_survive_dce(survive_dce_t *sd)
750 obstack_free(&sd->obst, NULL);
751 pmap_destroy(sd->places);
752 unregister_hook(hook_dead_node_elim, &sd->dead_node_elim);
753 unregister_hook(hook_dead_node_elim_subst, &sd->dead_node_elim_subst);
758 * Register a node pointer to be patched upon DCE.
759 * When DCE occurs, the node pointer specified by @p place will be
760 * patched to the new address of the node it is pointing to.
762 * @param sd The Survive DCE environment.
763 * @param place The address of the node pointer.
765 void survive_dce_register_irn(survive_dce_t *sd, ir_node **place)
768 ir_node *irn = *place;
769 survive_dce_list_t *curr = pmap_get(sd->places, irn);
770 survive_dce_list_t *nw = obstack_alloc(&sd->obst, sizeof(nw));
775 pmap_insert(sd->places, irn, nw);
779 /*--------------------------------------------------------------------*/
780 /* Functionality for inlining */
781 /*--------------------------------------------------------------------*/
784 * Copy node for inlineing. Updates attributes that change when
785 * inlineing but not for dead node elimination.
787 * Copies the node by calling copy_node() and then updates the entity if
788 * it's a local one. env must be a pointer of the frame type of the
789 * inlined procedure. The new entities must be in the link field of
793 copy_node_inline (ir_node *n, void *env) {
795 ir_type *frame_tp = (ir_type *)env;
798 if (get_irn_op(n) == op_Sel) {
799 nn = get_new_node (n);
801 if (get_entity_owner(get_Sel_entity(n)) == frame_tp) {
802 set_Sel_entity(nn, get_entity_link(get_Sel_entity(n)));
804 } else if (get_irn_op(n) == op_Block) {
805 nn = get_new_node (n);
806 nn->attr.block.irg = current_ir_graph;
810 static void find_addr(ir_node *node, void *env)
812 if (get_irn_opcode(node) == iro_Proj) {
813 if (get_Proj_proj(node) == pn_Start_P_value_arg_base)
819 * currently, we cannot inline two cases:
820 * - call with compound arguments
821 * - graphs that take the address of a parameter
823 * check these conditions here
825 static int can_inline(ir_node *call, ir_graph *called_graph)
827 ir_type *call_type = get_Call_type(call);
828 int params, ress, i, res;
829 assert(is_Method_type(call_type));
831 params = get_method_n_params(call_type);
832 ress = get_method_n_ress(call_type);
835 for (i = 0; i < params; ++i) {
836 ir_type *p_type = get_method_param_type(call_type, i);
838 if (is_compound_type(p_type))
843 for (i = 0; i < ress; ++i) {
844 ir_type *r_type = get_method_res_type(call_type, i);
846 if (is_compound_type(r_type))
851 irg_walk_graph(called_graph, find_addr, NULL, &res);
856 int inline_method(ir_node *call, ir_graph *called_graph) {
858 ir_node *post_call, *post_bl;
860 ir_node *end, *end_bl;
864 int arity, n_ret, n_exc, n_res, i, j, rem_opt, irn_arity;
866 ir_type *called_frame;
867 irg_inline_property prop = get_irg_inline_property(called_graph);
869 if ( (prop != irg_inline_forced) &&
870 (!get_opt_optimize() || !get_opt_inline() || (prop == irg_inline_forbidden))) return 0;
872 /* Do not inline variadic functions. */
873 if (get_method_variadicity(get_entity_type(get_irg_entity(called_graph))) == variadicity_variadic)
876 assert(get_method_n_params(get_entity_type(get_irg_entity(called_graph))) ==
877 get_method_n_params(get_Call_type(call)));
880 * currently, we cannot inline two cases:
881 * - call with compound arguments
882 * - graphs that take the address of a parameter
884 if (! can_inline(call, called_graph))
887 /* -- Turn off optimizations, this can cause problems when allocating new nodes. -- */
888 rem_opt = get_opt_optimize();
891 /* Handle graph state */
892 assert(get_irg_phase_state(current_ir_graph) != phase_building);
893 assert(get_irg_pinned(current_ir_graph) == op_pin_state_pinned);
894 assert(get_irg_pinned(called_graph) == op_pin_state_pinned);
895 set_irg_outs_inconsistent(current_ir_graph);
896 set_irg_extblk_inconsistent(current_ir_graph);
897 set_irg_doms_inconsistent(current_ir_graph);
898 set_irg_loopinfo_inconsistent(current_ir_graph);
899 set_irg_callee_info_state(current_ir_graph, irg_callee_info_inconsistent);
901 /* -- Check preconditions -- */
902 assert(is_Call(call));
903 /* @@@ does not work for InterfaceIII.java after cgana
904 assert(get_Call_type(call) == get_entity_type(get_irg_entity(called_graph)));
905 assert(smaller_type(get_entity_type(get_irg_entity(called_graph)),
906 get_Call_type(call)));
908 assert(get_type_tpop(get_Call_type(call)) == type_method);
909 if (called_graph == current_ir_graph) {
910 set_optimize(rem_opt);
914 /* here we know we WILL inline, so inform the statistics */
915 hook_inline(call, called_graph);
917 /* -- Decide how to handle exception control flow: Is there a handler
918 for the Call node, or do we branch directly to End on an exception?
920 0 There is a handler.
922 2 Exception handling not represented in Firm. -- */
924 ir_node *proj, *Mproj = NULL, *Xproj = NULL;
925 for (proj = (ir_node *)get_irn_link(call); proj; proj = (ir_node *)get_irn_link(proj)) {
926 assert(get_irn_op(proj) == op_Proj);
927 if (get_Proj_proj(proj) == pn_Call_X_except) Xproj = proj;
928 if (get_Proj_proj(proj) == pn_Call_M_except) Mproj = proj;
930 if (Mproj) { assert(Xproj); exc_handling = 0; } /* Mproj */
931 else if (Xproj) { exc_handling = 1; } /* !Mproj && Xproj */
932 else { exc_handling = 2; } /* !Mproj && !Xproj */
937 the procedure and later replaces the Start node of the called graph.
938 Post_call is the old Call node and collects the results of the called
939 graph. Both will end up being a tuple. -- */
940 post_bl = get_nodes_block(call);
941 set_irg_current_block(current_ir_graph, post_bl);
942 /* XxMxPxP of Start + parameter of Call */
943 in[pn_Start_X_initial_exec] = new_Jmp();
944 in[pn_Start_M] = get_Call_mem(call);
945 in[pn_Start_P_frame_base] = get_irg_frame(current_ir_graph);
946 in[pn_Start_P_globals] = get_irg_globals(current_ir_graph);
947 in[pn_Start_T_args] = new_Tuple(get_Call_n_params(call), get_Call_param_arr(call));
948 /* in[pn_Start_P_value_arg_base] = ??? */
949 pre_call = new_Tuple(5, in);
953 The new block gets the ins of the old block, pre_call and all its
954 predecessors and all Phi nodes. -- */
955 part_block(pre_call);
957 /* -- Prepare state for dead node elimination -- */
958 /* Visited flags in calling irg must be >= flag in called irg.
959 Else walker and arity computation will not work. */
960 if (get_irg_visited(current_ir_graph) <= get_irg_visited(called_graph))
961 set_irg_visited(current_ir_graph, get_irg_visited(called_graph)+1);
962 if (get_irg_block_visited(current_ir_graph)< get_irg_block_visited(called_graph))
963 set_irg_block_visited(current_ir_graph, get_irg_block_visited(called_graph));
964 /* Set pre_call as new Start node in link field of the start node of
965 calling graph and pre_calls block as new block for the start block
967 Further mark these nodes so that they are not visited by the
969 set_irn_link(get_irg_start(called_graph), pre_call);
970 set_irn_visited(get_irg_start(called_graph), get_irg_visited(current_ir_graph));
971 set_irn_link(get_irg_start_block(called_graph), get_nodes_block(pre_call));
972 set_irn_visited(get_irg_start_block(called_graph), get_irg_visited(current_ir_graph));
973 set_irn_link(get_irg_bad(called_graph), get_irg_bad(current_ir_graph));
974 set_irn_visited(get_irg_bad(called_graph), get_irg_visited(current_ir_graph));
976 /* Initialize for compaction of in arrays */
977 inc_irg_block_visited(current_ir_graph);
979 /* -- Replicate local entities of the called_graph -- */
980 /* copy the entities. */
981 called_frame = get_irg_frame_type(called_graph);
982 for (i = 0; i < get_class_n_members(called_frame); i++) {
983 entity *new_ent, *old_ent;
984 old_ent = get_class_member(called_frame, i);
985 new_ent = copy_entity_own(old_ent, get_cur_frame_type());
986 set_entity_link(old_ent, new_ent);
989 /* visited is > than that of called graph. With this trick visited will
990 remain unchanged so that an outer walker, e.g., searching the call nodes
991 to inline, calling this inline will not visit the inlined nodes. */
992 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
994 /* -- Performing dead node elimination inlines the graph -- */
995 /* Copies the nodes to the obstack of current_ir_graph. Updates links to new
997 /* @@@ endless loops are not copied!! -- they should be, I think... */
998 irg_walk(get_irg_end(called_graph), copy_node_inline, copy_preds,
999 get_irg_frame_type(called_graph));
1001 /* Repair called_graph */
1002 set_irg_visited(called_graph, get_irg_visited(current_ir_graph));
1003 set_irg_block_visited(called_graph, get_irg_block_visited(current_ir_graph));
1004 set_Block_block_visited(get_irg_start_block(called_graph), 0);
1006 /* -- Merge the end of the inlined procedure with the call site -- */
1007 /* We will turn the old Call node into a Tuple with the following
1010 0: Phi of all Memories of Return statements.
1011 1: Jmp from new Block that merges the control flow from all exception
1012 predecessors of the old end block.
1013 2: Tuple of all arguments.
1014 3: Phi of Exception memories.
1015 In case the old Call directly branches to End on an exception we don't
1016 need the block merging all exceptions nor the Phi of the exception
1020 /* -- Precompute some values -- */
1021 end_bl = get_new_node(get_irg_end_block(called_graph));
1022 end = get_new_node(get_irg_end(called_graph));
1023 arity = get_irn_arity(end_bl); /* arity = n_exc + n_ret */
1024 n_res = get_method_n_ress(get_Call_type(call));
1026 res_pred = xmalloc (n_res * sizeof(*res_pred));
1027 cf_pred = xmalloc (arity * sizeof(*res_pred));
1029 set_irg_current_block(current_ir_graph, post_bl); /* just to make sure */
1031 /* -- archive keepalives -- */
1032 irn_arity = get_irn_arity(end);
1033 for (i = 0; i < irn_arity; i++)
1034 add_End_keepalive(get_irg_end(current_ir_graph), get_irn_n(end, i));
1036 /* The new end node will die. We need not free as the in array is on the obstack:
1037 copy_node() only generated 'D' arrays. */
1039 /* -- Replace Return nodes by Jump nodes. -- */
1041 for (i = 0; i < arity; i++) {
1043 ret = get_irn_n(end_bl, i);
1044 if (is_Return(ret)) {
1045 cf_pred[n_ret] = new_r_Jmp(current_ir_graph, get_nodes_block(ret));
1049 set_irn_in(post_bl, n_ret, cf_pred);
1051 /* -- Build a Tuple for all results of the method.
1052 Add Phi node if there was more than one Return. -- */
1053 turn_into_tuple(post_call, 4);
1054 /* First the Memory-Phi */
1056 for (i = 0; i < arity; i++) {
1057 ret = get_irn_n(end_bl, i);
1058 if (is_Return(ret)) {
1059 cf_pred[n_ret] = get_Return_mem(ret);
1063 phi = new_Phi(n_ret, cf_pred, mode_M);
1064 set_Tuple_pred(call, pn_Call_M_regular, phi);
1065 /* Conserve Phi-list for further inlinings -- but might be optimized */
1066 if (get_nodes_block(phi) == post_bl) {
1067 set_irn_link(phi, get_irn_link(post_bl));
1068 set_irn_link(post_bl, phi);
1070 /* Now the real results */
1072 for (j = 0; j < n_res; j++) {
1074 for (i = 0; i < arity; i++) {
1075 ret = get_irn_n(end_bl, i);
1076 if (get_irn_op(ret) == op_Return) {
1077 cf_pred[n_ret] = get_Return_res(ret, j);
1082 phi = new_Phi(n_ret, cf_pred, get_irn_mode(cf_pred[0]));
1086 /* Conserve Phi-list for further inlinings -- but might be optimized */
1087 if (get_nodes_block(phi) == post_bl) {
1088 set_irn_link(phi, get_irn_link(post_bl));
1089 set_irn_link(post_bl, phi);
1092 set_Tuple_pred(call, pn_Call_T_result, new_Tuple(n_res, res_pred));
1094 set_Tuple_pred(call, pn_Call_T_result, new_Bad());
1096 /* Finally the exception control flow.
1097 We have two (three) possible situations:
1098 First if the Call branches to an exception handler: We need to add a Phi node to
1099 collect the memory containing the exception objects. Further we need
1100 to add another block to get a correct representation of this Phi. To
1101 this block we add a Jmp that resolves into the X output of the Call
1102 when the Call is turned into a tuple.
1103 Second the Call branches to End, the exception is not handled. Just
1104 add all inlined exception branches to the End node.
1105 Third: there is no Exception edge at all. Handle as case two. */
1106 if (exc_handling == 0) {
1108 for (i = 0; i < arity; i++) {
1110 ret = get_irn_n(end_bl, i);
1111 if (is_fragile_op(skip_Proj(ret)) || (get_irn_op(skip_Proj(ret)) == op_Raise)) {
1112 cf_pred[n_exc] = ret;
1117 new_Block(n_exc, cf_pred); /* watch it: current_block is changed! */
1118 set_Tuple_pred(call, pn_Call_X_except, new_Jmp());
1119 /* The Phi for the memories with the exception objects */
1121 for (i = 0; i < arity; i++) {
1123 ret = skip_Proj(get_irn_n(end_bl, i));
1125 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_block(ret), ret, mode_M, 3);
1127 } else if (is_fragile_op(ret)) {
1128 /* We rely that all cfops have the memory output at the same position. */
1129 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_block(ret), ret, mode_M, 0);
1131 } else if (get_irn_op(ret) == op_Raise) {
1132 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_block(ret), ret, mode_M, 1);
1136 set_Tuple_pred(call, pn_Call_M_except, new_Phi(n_exc, cf_pred, mode_M));
1138 set_Tuple_pred(call, pn_Call_X_except, new_Bad());
1139 set_Tuple_pred(call, pn_Call_M_except, new_Bad());
1142 ir_node *main_end_bl;
1143 int main_end_bl_arity;
1144 ir_node **end_preds;
1146 /* assert(exc_handling == 1 || no exceptions. ) */
1148 for (i = 0; i < arity; i++) {
1149 ir_node *ret = get_irn_n(end_bl, i);
1151 if (is_fragile_op(skip_Proj(ret)) || (get_irn_op(skip_Proj(ret)) == op_Raise)) {
1152 cf_pred[n_exc] = ret;
1156 main_end_bl = get_irg_end_block(current_ir_graph);
1157 main_end_bl_arity = get_irn_arity(main_end_bl);
1158 end_preds = xmalloc ((n_exc + main_end_bl_arity) * sizeof(*end_preds));
1160 for (i = 0; i < main_end_bl_arity; ++i)
1161 end_preds[i] = get_irn_n(main_end_bl, i);
1162 for (i = 0; i < n_exc; ++i)
1163 end_preds[main_end_bl_arity + i] = cf_pred[i];
1164 set_irn_in(main_end_bl, n_exc + main_end_bl_arity, end_preds);
1165 set_Tuple_pred(call, pn_Call_X_except, new_Bad());
1166 set_Tuple_pred(call, pn_Call_M_except, new_Bad());
1172 #if 0 /* old. now better, correcter, faster implementation. */
1174 /* -- If the exception control flow from the inlined Call directly
1175 branched to the end block we now have the following control
1176 flow predecessor pattern: ProjX -> Tuple -> Jmp. We must
1177 remove the Jmp along with it's empty block and add Jmp's
1178 predecessors as predecessors of this end block. No problem if
1179 there is no exception, because then branches Bad to End which
1181 @@@ can't we know this beforehand: by getting the Proj(1) from
1182 the Call link list and checking whether it goes to Proj. */
1183 /* find the problematic predecessor of the end block. */
1184 end_bl = get_irg_end_block(current_ir_graph);
1185 for (i = 0; i < get_Block_n_cfgpreds(end_bl); i++) {
1186 cf_op = get_Block_cfgpred(end_bl, i);
1187 if (get_irn_op(cf_op) == op_Proj) {
1188 cf_op = get_Proj_pred(cf_op);
1189 if ((get_irn_op(cf_op) == op_Tuple) && (cf_op == call)) {
1190 /* There are unoptimized tuples from inlineing before when no exc */
1191 assert(get_Proj_proj(get_Block_cfgpred(end_bl, i)) == pn_Call_X_except);
1192 cf_op = get_Tuple_pred(cf_op, pn_Call_X_except);
1193 assert(get_irn_op(cf_op) == op_Jmp);
1199 if (i < get_Block_n_cfgpreds(end_bl)) {
1200 bl = get_nodes_block(cf_op);
1201 arity = get_Block_n_cfgpreds(end_bl) + get_Block_n_cfgpreds(bl) - 1;
1202 cf_pred = xmalloc (arity * sizeof(*cf_pred));
1203 for (j = 0; j < i; j++)
1204 cf_pred[j] = get_Block_cfgpred(end_bl, j);
1205 for (j = j; j < i + get_Block_n_cfgpreds(bl); j++)
1206 cf_pred[j] = get_Block_cfgpred(bl, j-i);
1207 for (j = j; j < arity; j++)
1208 cf_pred[j] = get_Block_cfgpred(end_bl, j-get_Block_n_cfgpreds(bl) +1);
1209 set_irn_in(end_bl, arity, cf_pred);
1211 /* Remove the exception pred from post-call Tuple. */
1212 set_Tuple_pred(call, pn_Call_X_except, new_Bad());
1217 /* -- Turn CSE back on. -- */
1218 set_optimize(rem_opt);
1223 /********************************************************************/
1224 /* Apply inlineing to small methods. */
1225 /********************************************************************/
1227 /* It makes no sense to inline too many calls in one procedure. Anyways,
1228 I didn't get a version with NEW_ARR_F to run. */
1229 #define MAX_INLINE 1024
1232 * environment for inlining small irgs
1234 typedef struct _inline_env_t {
1236 ir_node *calls[MAX_INLINE];
1240 * Returns the irg called from a Call node. If the irg is not
1241 * known, NULL is returned.
1243 static ir_graph *get_call_called_irg(ir_node *call) {
1245 ir_graph *called_irg = NULL;
1247 assert(is_Call(call));
1249 addr = get_Call_ptr(call);
1250 if ((get_irn_op(addr) == op_SymConst) && (get_SymConst_kind (addr) == symconst_addr_ent)) {
1251 called_irg = get_entity_irg(get_SymConst_entity(addr));
1257 static void collect_calls(ir_node *call, void *env) {
1260 if (! is_Call(call)) return;
1262 addr = get_Call_ptr(call);
1264 if (get_irn_op(addr) == op_SymConst) {
1265 if (get_SymConst_kind(addr) == symconst_addr_ent) {
1266 ir_graph *called_irg = get_entity_irg(get_SymConst_entity(addr));
1267 inline_env_t *ienv = (inline_env_t *)env;
1268 if (called_irg && ienv->pos < MAX_INLINE) {
1269 /* The Call node calls a locally defined method. Remember to inline. */
1270 ienv->calls[ienv->pos++] = call;
1277 * Inlines all small methods at call sites where the called address comes
1278 * from a Const node that references the entity representing the called
1280 * The size argument is a rough measure for the code size of the method:
1281 * Methods where the obstack containing the firm graph is smaller than
1284 void inline_small_irgs(ir_graph *irg, int size) {
1286 ir_graph *rem = current_ir_graph;
1287 inline_env_t env /* = {0, NULL}*/;
1289 if (!(get_opt_optimize() && get_opt_inline())) return;
1291 current_ir_graph = irg;
1292 /* Handle graph state */
1293 assert(get_irg_phase_state(current_ir_graph) != phase_building);
1294 free_callee_info(current_ir_graph);
1296 /* Find Call nodes to inline.
1297 (We can not inline during a walk of the graph, as inlineing the same
1298 method several times changes the visited flag of the walked graph:
1299 after the first inlineing visited of the callee equals visited of
1300 the caller. With the next inlineing both are increased.) */
1302 irg_walk(get_irg_end(irg), NULL, collect_calls, &env);
1304 if ((env.pos > 0) && (env.pos < MAX_INLINE)) {
1305 /* There are calls to inline */
1306 collect_phiprojs(irg);
1307 for (i = 0; i < env.pos; i++) {
1309 callee = get_entity_irg(get_SymConst_entity(get_Call_ptr(env.calls[i])));
1310 if (((_obstack_memory_used(callee->obst) - (int)obstack_room(callee->obst)) < size) ||
1311 (get_irg_inline_property(callee) == irg_inline_forced)) {
1312 inline_method(env.calls[i], callee);
1317 current_ir_graph = rem;
1321 * Environment for inlining irgs.
1324 int n_nodes; /**< Nodes in graph except Id, Tuple, Proj, Start, End */
1325 int n_nodes_orig; /**< for statistics */
1326 eset *call_nodes; /**< All call nodes in this graph */
1328 int n_call_nodes_orig; /**< for statistics */
1329 int n_callers; /**< Number of known graphs that call this graphs. */
1330 int n_callers_orig; /**< for statistics */
1334 * Allocate a new environment for inlining.
1336 static inline_irg_env *new_inline_irg_env(void) {
1337 inline_irg_env *env = xmalloc(sizeof(*env));
1338 env->n_nodes = -2; /* do not count count Start, End */
1339 env->n_nodes_orig = -2; /* do not count Start, End */
1340 env->call_nodes = eset_create();
1341 env->n_call_nodes = 0;
1342 env->n_call_nodes_orig = 0;
1344 env->n_callers_orig = 0;
1349 * destroy an environment for inlining.
1351 static void free_inline_irg_env(inline_irg_env *env) {
1352 eset_destroy(env->call_nodes);
1357 * post-walker: collect all calls in the inline-environment
1358 * of a graph and sum some statistics.
1360 static void collect_calls2(ir_node *call, void *env) {
1361 inline_irg_env *x = (inline_irg_env *)env;
1362 ir_op *op = get_irn_op(call);
1365 /* count meaningful nodes in irg */
1366 if (op != op_Proj && op != op_Tuple && op != op_Sync) {
1371 if (op != op_Call) return;
1373 /* collect all call nodes */
1374 eset_insert(x->call_nodes, call);
1376 x->n_call_nodes_orig++;
1378 /* count all static callers */
1379 callee = get_call_called_irg(call);
1381 inline_irg_env *callee_env = get_irg_link(callee);
1382 callee_env->n_callers++;
1383 callee_env->n_callers_orig++;
1388 * Returns TRUE if the number of callers in 0 in the irg's environment,
1389 * hence this irg is a leave.
1391 INLINE static int is_leave(ir_graph *irg) {
1392 return (((inline_irg_env *)get_irg_link(irg))->n_call_nodes == 0);
1396 * Returns TRUE if the number of callers is smaller size in the irg's environment.
1398 INLINE static int is_smaller(ir_graph *callee, int size) {
1399 return (((inline_irg_env *)get_irg_link(callee))->n_nodes < size);
1404 * Inlines small leave methods at call sites where the called address comes
1405 * from a Const node that references the entity representing the called
1407 * The size argument is a rough measure for the code size of the method:
1408 * Methods where the obstack containing the firm graph is smaller than
1411 void inline_leave_functions(int maxsize, int leavesize, int size) {
1412 inline_irg_env *env;
1413 int i, n_irgs = get_irp_n_irgs();
1414 ir_graph *rem = current_ir_graph;
1417 if (!(get_opt_optimize() && get_opt_inline())) return;
1419 /* extend all irgs by a temporary data structure for inlining. */
1420 for (i = 0; i < n_irgs; ++i)
1421 set_irg_link(get_irp_irg(i), new_inline_irg_env());
1423 /* Precompute information in temporary data structure. */
1424 for (i = 0; i < n_irgs; ++i) {
1425 current_ir_graph = get_irp_irg(i);
1426 assert(get_irg_phase_state(current_ir_graph) != phase_building);
1427 free_callee_info(current_ir_graph);
1429 irg_walk(get_irg_end(current_ir_graph), NULL, collect_calls2,
1430 get_irg_link(current_ir_graph));
1433 /* -- and now inline. -- */
1435 /* Inline leaves recursively -- we might construct new leaves. */
1436 while (did_inline) {
1439 for (i = 0; i < n_irgs; ++i) {
1441 int phiproj_computed = 0;
1443 current_ir_graph = get_irp_irg(i);
1444 env = (inline_irg_env *)get_irg_link(current_ir_graph);
1446 for (call = eset_first(env->call_nodes); call; call = eset_next(env->call_nodes)) {
1449 if (get_irn_op(call) == op_Tuple) continue; /* We already have inlined this call. */
1450 callee = get_call_called_irg(call);
1452 if (env->n_nodes > maxsize) continue; // break;
1454 if (callee && (is_leave(callee) && is_smaller(callee, leavesize))) {
1455 if (!phiproj_computed) {
1456 phiproj_computed = 1;
1457 collect_phiprojs(current_ir_graph);
1459 did_inline = inline_method(call, callee);
1462 /* Do some statistics */
1463 inline_irg_env *callee_env = (inline_irg_env *)get_irg_link(callee);
1464 env->n_call_nodes --;
1465 env->n_nodes += callee_env->n_nodes;
1466 callee_env->n_callers--;
1473 /* inline other small functions. */
1474 for (i = 0; i < n_irgs; ++i) {
1477 int phiproj_computed = 0;
1479 current_ir_graph = get_irp_irg(i);
1480 env = (inline_irg_env *)get_irg_link(current_ir_graph);
1482 /* we can not walk and change a set, nor remove from it.
1484 walkset = env->call_nodes;
1485 env->call_nodes = eset_create();
1486 for (call = eset_first(walkset); call; call = eset_next(walkset)) {
1489 if (get_irn_op(call) == op_Tuple) continue; /* We already inlined. */
1490 callee = get_call_called_irg(call);
1493 ((is_smaller(callee, size) && (env->n_nodes < maxsize)) || /* small function */
1494 (get_irg_inline_property(callee) == irg_inline_forced))) {
1495 if (!phiproj_computed) {
1496 phiproj_computed = 1;
1497 collect_phiprojs(current_ir_graph);
1499 if (inline_method(call, callee)) {
1500 inline_irg_env *callee_env = (inline_irg_env *)get_irg_link(callee);
1501 env->n_call_nodes--;
1502 eset_insert_all(env->call_nodes, callee_env->call_nodes); /* @@@ ??? This are the wrong nodes !? Not the copied ones. */
1503 env->n_call_nodes += callee_env->n_call_nodes;
1504 env->n_nodes += callee_env->n_nodes;
1505 callee_env->n_callers--;
1508 eset_insert(env->call_nodes, call);
1511 eset_destroy(walkset);
1514 for (i = 0; i < n_irgs; ++i) {
1515 current_ir_graph = get_irp_irg(i);
1517 env = (inline_irg_env *)get_irg_link(current_ir_graph);
1518 if ((env->n_call_nodes_orig != env->n_call_nodes) ||
1519 (env->n_callers_orig != env->n_callers))
1520 printf("Nodes:%3d ->%3d, calls:%3d ->%3d, callers:%3d ->%3d, -- %s\n",
1521 env->n_nodes_orig, env->n_nodes, env->n_call_nodes_orig, env->n_call_nodes,
1522 env->n_callers_orig, env->n_callers,
1523 get_entity_name(get_irg_entity(current_ir_graph)));
1525 free_inline_irg_env((inline_irg_env *)get_irg_link(current_ir_graph));
1528 current_ir_graph = rem;
1531 /*******************************************************************/
1532 /* Code Placement. Pins all floating nodes to a block where they */
1533 /* will be executed only if needed. */
1534 /*******************************************************************/
1537 * Returns non-zero, is a block is not reachable from Start.
1539 * @param block the block to test
1542 is_Block_unreachable(ir_node *block) {
1543 return is_Block_dead(block) || get_Block_dom_depth(block) < 0;
1547 * Find the earliest correct block for N. --- Place N into the
1548 * same Block as its dominance-deepest Input.
1550 * We have to avoid calls to get_nodes_block() here
1551 * because the graph is floating.
1553 * move_out_of_loops() expects that place_floats_early() have placed
1554 * all "living" nodes into a living block. That's why we must
1555 * move nodes in dead block with "live" successors into a valid
1557 * We move them just into the same block as it's successor (or
1558 * in case of a Phi into the effective use block). For Phi successors,
1559 * this may still be a dead block, but then there is no real use, as
1560 * the control flow will be dead later.
1563 place_floats_early(ir_node *n, pdeq *worklist)
1567 /* we must not run into an infinite loop */
1568 assert(irn_not_visited(n));
1569 mark_irn_visited(n);
1571 /* Place floating nodes. */
1572 if (get_irn_pinned(n) == op_pin_state_floats) {
1573 ir_node *curr_block = get_irn_n(n, -1);
1574 int in_dead_block = is_Block_unreachable(curr_block);
1576 ir_node *b = NULL; /* The block to place this node in */
1578 assert(get_irn_op(n) != op_Block);
1580 if ((get_irn_op(n) == op_Const) ||
1581 (get_irn_op(n) == op_SymConst) ||
1583 (get_irn_op(n) == op_Unknown)) {
1584 /* These nodes will not be placed by the loop below. */
1585 b = get_irg_start_block(current_ir_graph);
1589 /* find the block for this node. */
1590 irn_arity = get_irn_arity(n);
1591 for (i = 0; i < irn_arity; i++) {
1592 ir_node *pred = get_irn_n(n, i);
1593 ir_node *pred_block;
1595 if ((irn_not_visited(pred))
1596 && (get_irn_pinned(pred) == op_pin_state_floats)) {
1599 * If the current node is NOT in a dead block, but one of its
1600 * predecessors is, we must move the predecessor to a live block.
1601 * Such thing can happen, if global CSE chose a node from a dead block.
1602 * We move it simple to our block.
1603 * Note that neither Phi nor End nodes are floating, so we don't
1604 * need to handle them here.
1606 if (! in_dead_block) {
1607 if (get_irn_pinned(pred) == op_pin_state_floats &&
1608 is_Block_unreachable(get_irn_n(pred, -1)))
1609 set_nodes_block(pred, curr_block);
1611 place_floats_early(pred, worklist);
1615 * A node in the Bad block must stay in the bad block,
1616 * so don't compute a new block for it.
1621 /* Because all loops contain at least one op_pin_state_pinned node, now all
1622 our inputs are either op_pin_state_pinned or place_early() has already
1623 been finished on them. We do not have any unfinished inputs! */
1624 pred_block = get_irn_n(pred, -1);
1625 if ((!is_Block_dead(pred_block)) &&
1626 (get_Block_dom_depth(pred_block) > depth)) {
1628 depth = get_Block_dom_depth(pred_block);
1630 /* Avoid that the node is placed in the Start block */
1631 if ((depth == 1) && (get_Block_dom_depth(get_irn_n(n, -1)) > 1)) {
1632 b = get_Block_cfg_out(get_irg_start_block(current_ir_graph), 0);
1633 assert(b != get_irg_start_block(current_ir_graph));
1638 set_nodes_block(n, b);
1642 * Add predecessors of non floating nodes and non-floating predecessors
1643 * of floating nodes to worklist and fix their blocks if the are in dead block.
1645 irn_arity = get_irn_arity(n);
1647 if (get_irn_op(n) == op_End) {
1649 * Simplest case: End node. Predecessors are keep-alives,
1650 * no need to move out of dead block.
1652 for (i = -1; i < irn_arity; ++i) {
1653 ir_node *pred = get_irn_n(n, i);
1654 if (irn_not_visited(pred))
1655 pdeq_putr(worklist, pred);
1658 else if (is_Block(n)) {
1660 * Blocks: Predecessors are control flow, no need to move
1661 * them out of dead block.
1663 for (i = irn_arity - 1; i >= 0; --i) {
1664 ir_node *pred = get_irn_n(n, i);
1665 if (irn_not_visited(pred))
1666 pdeq_putr(worklist, pred);
1669 else if (is_Phi(n)) {
1671 ir_node *curr_block = get_irn_n(n, -1);
1672 int in_dead_block = is_Block_unreachable(curr_block);
1675 * Phi nodes: move nodes from dead blocks into the effective use
1676 * of the Phi-input if the Phi is not in a bad block.
1678 pred = get_irn_n(n, -1);
1679 if (irn_not_visited(pred))
1680 pdeq_putr(worklist, pred);
1682 for (i = irn_arity - 1; i >= 0; --i) {
1683 ir_node *pred = get_irn_n(n, i);
1685 if (irn_not_visited(pred)) {
1686 if (! in_dead_block &&
1687 get_irn_pinned(pred) == op_pin_state_floats &&
1688 is_Block_unreachable(get_irn_n(pred, -1))) {
1689 set_nodes_block(pred, get_Block_cfgpred_block(curr_block, i));
1691 pdeq_putr(worklist, pred);
1697 ir_node *curr_block = get_irn_n(n, -1);
1698 int in_dead_block = is_Block_unreachable(curr_block);
1701 * All other nodes: move nodes from dead blocks into the same block.
1703 pred = get_irn_n(n, -1);
1704 if (irn_not_visited(pred))
1705 pdeq_putr(worklist, pred);
1707 for (i = irn_arity - 1; i >= 0; --i) {
1708 ir_node *pred = get_irn_n(n, i);
1710 if (irn_not_visited(pred)) {
1711 if (! in_dead_block &&
1712 get_irn_pinned(pred) == op_pin_state_floats &&
1713 is_Block_unreachable(get_irn_n(pred, -1))) {
1714 set_nodes_block(pred, curr_block);
1716 pdeq_putr(worklist, pred);
1723 * Floating nodes form subgraphs that begin at nodes as Const, Load,
1724 * Start, Call and that end at op_pin_state_pinned nodes as Store, Call. Place_early
1725 * places all floating nodes reachable from its argument through floating
1726 * nodes and adds all beginnings at op_pin_state_pinned nodes to the worklist.
1728 static INLINE void place_early(pdeq *worklist) {
1730 inc_irg_visited(current_ir_graph);
1732 /* this inits the worklist */
1733 place_floats_early(get_irg_end(current_ir_graph), worklist);
1735 /* Work the content of the worklist. */
1736 while (!pdeq_empty(worklist)) {
1737 ir_node *n = pdeq_getl(worklist);
1738 if (irn_not_visited(n))
1739 place_floats_early(n, worklist);
1742 set_irg_outs_inconsistent(current_ir_graph);
1743 set_irg_pinned(current_ir_graph, op_pin_state_pinned);
1747 * Compute the deepest common ancestor of block and dca.
1749 static ir_node *calc_dca(ir_node *dca, ir_node *block)
1753 /* we do not want to place nodes in dead blocks */
1754 if (is_Block_dead(block))
1757 /* We found a first legal placement. */
1758 if (!dca) return block;
1760 /* Find a placement that is dominates both, dca and block. */
1761 while (get_Block_dom_depth(block) > get_Block_dom_depth(dca))
1762 block = get_Block_idom(block);
1764 while (get_Block_dom_depth(dca) > get_Block_dom_depth(block)) {
1765 dca = get_Block_idom(dca);
1768 while (block != dca)
1769 { block = get_Block_idom(block); dca = get_Block_idom(dca); }
1774 /** Deepest common dominance ancestor of DCA and CONSUMER of PRODUCER.
1775 * I.e., DCA is the block where we might place PRODUCER.
1776 * A data flow edge points from producer to consumer.
1779 consumer_dom_dca(ir_node *dca, ir_node *consumer, ir_node *producer)
1781 ir_node *block = NULL;
1783 /* Compute the latest block into which we can place a node so that it is
1785 if (get_irn_op(consumer) == op_Phi) {
1786 /* our consumer is a Phi-node, the effective use is in all those
1787 blocks through which the Phi-node reaches producer */
1789 ir_node *phi_block = get_nodes_block(consumer);
1790 irn_arity = get_irn_arity(consumer);
1792 for (i = 0; i < irn_arity; i++) {
1793 if (get_irn_n(consumer, i) == producer) {
1794 ir_node *new_block = get_nodes_block(get_Block_cfgpred(phi_block, i));
1796 if (! is_Block_unreachable(new_block))
1797 block = calc_dca(block, new_block);
1802 block = get_irn_n(producer, -1);
1805 assert(is_no_Block(consumer));
1806 block = get_nodes_block(consumer);
1809 /* Compute the deepest common ancestor of block and dca. */
1810 return calc_dca(dca, block);
1813 /* FIXME: the name clashes here with the function from ana/field_temperature.c
1815 static INLINE int get_irn_loop_depth(ir_node *n) {
1816 return get_loop_depth(get_irn_loop(n));
1820 * Move n to a block with less loop depth than it's current block. The
1821 * new block must be dominated by early.
1823 * @param n the node that should be moved
1824 * @param early the earliest block we can n move to
1827 move_out_of_loops (ir_node *n, ir_node *early)
1829 ir_node *best, *dca;
1833 /* Find the region deepest in the dominator tree dominating
1834 dca with the least loop nesting depth, but still dominated
1835 by our early placement. */
1836 dca = get_nodes_block(n);
1839 while (dca != early) {
1840 dca = get_Block_idom(dca);
1841 if (!dca || is_Bad(dca)) break; /* may be Bad if not reachable from Start */
1842 if (get_irn_loop_depth(dca) < get_irn_loop_depth(best)) {
1846 if (best != get_nodes_block(n)) {
1848 printf("Moving out of loop: "); DDMN(n);
1849 printf(" Outermost block: "); DDMN(early);
1850 printf(" Best block: "); DDMN(best);
1851 printf(" Innermost block: "); DDMN(get_nodes_block(n));
1853 set_nodes_block(n, best);
1858 * Find the latest legal block for N and place N into the
1859 * `optimal' Block between the latest and earliest legal block.
1860 * The `optimal' block is the dominance-deepest block of those
1861 * with the least loop-nesting-depth. This places N out of as many
1862 * loops as possible and then makes it as control dependent as
1866 place_floats_late(ir_node *n, pdeq *worklist)
1871 assert(irn_not_visited(n)); /* no multiple placement */
1873 mark_irn_visited(n);
1875 /* no need to place block nodes, control nodes are already placed. */
1876 if ((get_irn_op(n) != op_Block) &&
1878 (get_irn_mode(n) != mode_X)) {
1879 /* Remember the early_blk placement of this block to move it
1880 out of loop no further than the early_blk placement. */
1881 early_blk = get_irn_n(n, -1);
1884 * BEWARE: Here we also get code, that is live, but
1885 * was in a dead block. If the node is life, but because
1886 * of CSE in a dead block, we still might need it.
1889 /* Assure that our users are all placed, except the Phi-nodes.
1890 --- Each data flow cycle contains at least one Phi-node. We
1891 have to break the `user has to be placed before the
1892 producer' dependence cycle and the Phi-nodes are the
1893 place to do so, because we need to base our placement on the
1894 final region of our users, which is OK with Phi-nodes, as they
1895 are op_pin_state_pinned, and they never have to be placed after a
1896 producer of one of their inputs in the same block anyway. */
1897 for (i = get_irn_n_outs(n) - 1; i >= 0; --i) {
1898 ir_node *succ = get_irn_out(n, i);
1899 if (irn_not_visited(succ) && (get_irn_op(succ) != op_Phi))
1900 place_floats_late(succ, worklist);
1903 if (! is_Block_dead(early_blk)) {
1904 /* do only move things that where not dead */
1906 /* We have to determine the final block of this node... except for
1908 if ((get_irn_pinned(n) == op_pin_state_floats) &&
1909 (get_irn_op(n) != op_Const) &&
1910 (get_irn_op(n) != op_SymConst)) {
1911 ir_node *dca = NULL; /* deepest common ancestor in the
1912 dominator tree of all nodes'
1913 blocks depending on us; our final
1914 placement has to dominate DCA. */
1915 for (i = get_irn_n_outs(n) - 1; i >= 0; --i) {
1916 ir_node *succ = get_irn_out(n, i);
1919 if (get_irn_op(succ) == op_End) {
1921 * This consumer is the End node, a keep alive edge.
1922 * This is not a real consumer, so we ignore it
1927 /* ignore if succ is in dead code */
1928 succ_blk = get_irn_n(succ, -1);
1929 if (is_Block_unreachable(succ_blk))
1931 dca = consumer_dom_dca(dca, succ, n);
1934 set_nodes_block(n, dca);
1935 move_out_of_loops(n, early_blk);
1941 /* Add predecessors of all non-floating nodes on list. (Those of floating
1942 nodes are placed already and therefore are marked.) */
1943 for (i = 0; i < get_irn_n_outs(n); i++) {
1944 ir_node *succ = get_irn_out(n, i);
1945 if (irn_not_visited(get_irn_out(n, i))) {
1946 pdeq_putr(worklist, succ);
1951 static INLINE void place_late(pdeq *worklist) {
1953 inc_irg_visited(current_ir_graph);
1955 /* This fills the worklist initially. */
1956 place_floats_late(get_irg_start_block(current_ir_graph), worklist);
1958 /* And now empty the worklist again... */
1959 while (!pdeq_empty(worklist)) {
1960 ir_node *n = pdeq_getl(worklist);
1961 if (irn_not_visited(n))
1962 place_floats_late(n, worklist);
1966 void place_code(ir_graph *irg) {
1968 ir_graph *rem = current_ir_graph;
1970 current_ir_graph = irg;
1972 if (!(get_opt_optimize() && get_opt_global_cse())) return;
1974 /* Handle graph state */
1975 assert(get_irg_phase_state(irg) != phase_building);
1976 if (get_irg_dom_state(irg) != dom_consistent)
1979 if (1 || get_irg_loopinfo_state(irg) != loopinfo_consistent) {
1980 free_loop_information(irg);
1981 construct_backedges(irg);
1984 /* Place all floating nodes as early as possible. This guarantees
1985 a legal code placement. */
1986 worklist = new_pdeq();
1987 place_early(worklist);
1989 /* place_early() invalidates the outs, place_late needs them. */
1990 compute_irg_outs(irg);
1992 /* Now move the nodes down in the dominator tree. This reduces the
1993 unnecessary executions of the node. */
1994 place_late(worklist);
1996 set_irg_outs_inconsistent(current_ir_graph);
1997 set_irg_loopinfo_inconsistent(current_ir_graph);
1999 current_ir_graph = rem;
2003 * Called by walker of remove_critical_cf_edges().
2005 * Place an empty block to an edge between a blocks of multiple
2006 * predecessors and a block of multiple successors.
2009 * @param env Environment of walker. The changed field.
2011 static void walk_critical_cf_edges(ir_node *n, void *env) {
2013 ir_node *pre, *block, *jmp;
2016 /* Block has multiple predecessors */
2017 if (is_Block(n) && (get_irn_arity(n) > 1)) {
2018 if (n == get_irg_end_block(current_ir_graph))
2019 return; /* No use to add a block here. */
2021 arity = get_irn_arity(n);
2022 for (i=0; i<arity; i++) {
2023 pre = get_irn_n(n, i);
2024 /* Predecessor has multiple successors. Insert new control flow edge. */
2025 if (op_Raise != get_irn_op(skip_Proj(pre))) {
2026 /* set predecessor of new block */
2027 block = new_Block(1, &pre);
2028 /* insert new jmp node to new block */
2029 set_cur_block(block);
2032 /* set successor of new block */
2033 set_irn_n(n, i, jmp);
2035 } /* predecessor has multiple successors */
2036 } /* for all predecessors */
2037 } /* n is a block */
2040 void remove_critical_cf_edges(ir_graph *irg) {
2042 irg_walk_graph(irg, NULL, walk_critical_cf_edges, &changed);
2045 /* control flow changed */
2046 set_irg_outs_inconsistent(irg);
2047 set_irg_extblk_inconsistent(irg);
2048 set_irg_doms_inconsistent(current_ir_graph);
2049 set_irg_loopinfo_inconsistent(current_ir_graph);