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"
33 #include "pdeq.h" /* Fuer code placement */
38 #include "irbackedge_t.h"
44 #include "iredges_t.h"
47 /* Defined in iropt.c */
48 pset *new_identities (void);
49 void del_identities (pset *value_table);
50 void add_identities (pset *value_table, ir_node *node);
52 /*------------------------------------------------------------------*/
53 /* apply optimizations of iropt to all nodes. */
54 /*------------------------------------------------------------------*/
56 static void init_link (ir_node *n, void *env) {
57 set_irn_link(n, NULL);
60 #if 0 /* Old version. Avoids Ids.
61 This is not necessary: we do a post walk, and get_irn_n
62 removes ids anyways. So it's much cheaper to call the
63 optimization less often and use the exchange() algorithm. */
65 optimize_in_place_wrapper (ir_node *n, void *env) {
67 ir_node *optimized, *old;
69 irn_arity = get_irn_arity(n);
70 for (i = 0; i < irn_arity; i++) {
71 /* get_irn_n skips Id nodes, so comparison old != optimized does not
72 show all optimizations. Therefore always set new predecessor. */
73 old = get_irn_intra_n(n, i);
74 optimized = optimize_in_place_2(old);
75 set_irn_n(n, i, optimized);
78 if (get_irn_op(n) == op_Block) {
79 optimized = optimize_in_place_2(n);
80 if (optimized != n) exchange (n, optimized);
85 optimize_in_place_wrapper (ir_node *n, void *env) {
86 ir_node *optimized = optimize_in_place_2(n);
87 if (optimized != n) exchange (n, optimized);
92 static INLINE void do_local_optimize(ir_node *n) {
93 /* Handle graph state */
94 assert(get_irg_phase_state(current_ir_graph) != phase_building);
96 if (get_opt_global_cse())
97 set_irg_pinned(current_ir_graph, op_pin_state_floats);
98 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
99 set_irg_outs_inconsistent(current_ir_graph);
100 if (get_irg_dom_state(current_ir_graph) == dom_consistent)
101 set_irg_dom_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 set_Block_dead(block);
131 local_optimize_graph (ir_graph *irg) {
132 ir_graph *rem = current_ir_graph;
133 current_ir_graph = irg;
135 if (get_irg_dom_state(current_ir_graph) == dom_consistent)
136 irg_block_walk_graph(irg, NULL, kill_dead_blocks, NULL);
138 do_local_optimize(irg->end);
140 current_ir_graph = rem;
144 /*------------------------------------------------------------------*/
145 /* Routines for dead node elimination / copying garbage collection */
146 /* of the obstack. */
147 /*------------------------------------------------------------------*/
150 * Remember the new node in the old node by using a field all nodes have.
153 set_new_node (ir_node *old, ir_node *new)
159 * Get this new node, before the old node is forgotten.
161 static INLINE ir_node *
162 get_new_node (ir_node * n) {
167 * We use the block_visited flag to mark that we have computed the
168 * number of useful predecessors for this block.
169 * Further we encode the new arity in this flag in the old blocks.
170 * Remembering the arity is useful, as it saves a lot of pointer
171 * accesses. This function is called for all Phi and Block nodes
175 compute_new_arity(ir_node *b) {
176 int i, res, irn_arity;
179 irg_v = get_irg_block_visited(current_ir_graph);
180 block_v = get_Block_block_visited(b);
181 if (block_v >= irg_v) {
182 /* we computed the number of preds for this block and saved it in the
184 return block_v - irg_v;
186 /* compute the number of good predecessors */
187 res = irn_arity = get_irn_arity(b);
188 for (i = 0; i < irn_arity; i++)
189 if (get_irn_opcode(get_irn_n(b, i)) == iro_Bad) res--;
190 /* save it in the flag. */
191 set_Block_block_visited(b, irg_v + res);
196 /* TODO: add an ir_op operation */
197 static INLINE void new_backedge_info(ir_node *n) {
198 switch(get_irn_opcode(n)) {
200 n->attr.block.cg_backedge = NULL;
201 n->attr.block.backedge = new_backedge_arr(current_ir_graph->obst, get_irn_arity(n));
204 n->attr.phi_backedge = new_backedge_arr(current_ir_graph->obst, get_irn_arity(n));
207 n->attr.filter.backedge = new_backedge_arr(current_ir_graph->obst, get_irn_arity(n));
214 * Copies the node to the new obstack. The Ins of the new node point to
215 * the predecessors on the old obstack. For block/phi nodes not all
216 * predecessors might be copied. n->link points to the new node.
217 * For Phi and Block nodes the function allocates in-arrays with an arity
218 * only for useful predecessors. The arity is determined by counting
219 * the non-bad predecessors of the block.
221 * @param n The node to be copied
222 * @param env if non-NULL, the node number attribute will be copied to the new node
224 * Note: Also used for loop unrolling.
226 static void copy_node(ir_node *n, void *env) {
229 ir_op *op = get_irn_op(n);
230 int copy_node_nr = env != NULL;
232 /* The end node looses it's flexible in array. This doesn't matter,
233 as dead node elimination builds End by hand, inlineing doesn't use
235 /* assert(op == op_End || ((_ARR_DESCR(n->in))->cookie != ARR_F_MAGIC)); */
238 /* node copied already */
240 } else if (op == op_Block) {
242 new_arity = compute_new_arity(n);
243 n->attr.block.graph_arr = NULL;
245 block = get_nodes_block(n);
247 new_arity = compute_new_arity(block);
249 new_arity = get_irn_arity(n);
252 nn = new_ir_node(get_irn_dbg_info(n),
259 /* Copy the attributes. These might point to additional data. If this
260 was allocated on the old obstack the pointers now are dangling. This
261 frees e.g. the memory of the graph_arr allocated in new_immBlock. */
262 copy_node_attr(n, nn);
263 new_backedge_info(nn);
267 /* for easier debugging, we want to copy the node numbers too */
268 nn->node_nr = n->node_nr;
276 * Copies new predecessors of old node to new node remembered in link.
277 * Spare the Bad predecessors of Phi and Block nodes.
280 copy_preds (ir_node *n, void *env) {
284 nn = get_new_node(n);
286 /* printf("\n old node: "); DDMSG2(n);
287 printf(" new node: "); DDMSG2(nn);
288 printf(" arities: old: %d, new: %d\n", get_irn_arity(n), get_irn_arity(nn)); */
291 /* Don't copy Bad nodes. */
293 irn_arity = get_irn_arity(n);
294 for (i = 0; i < irn_arity; i++)
295 if (! is_Bad(get_irn_n(n, i))) {
296 set_irn_n (nn, j, get_new_node(get_irn_n(n, i)));
297 /*if (is_backedge(n, i)) set_backedge(nn, j);*/
300 /* repair the block visited flag from above misuse. Repair it in both
301 graphs so that the old one can still be used. */
302 set_Block_block_visited(nn, 0);
303 set_Block_block_visited(n, 0);
304 /* Local optimization could not merge two subsequent blocks if
305 in array contained Bads. Now it's possible.
306 We don't call optimize_in_place as it requires
307 that the fields in ir_graph are set properly. */
308 if ((get_opt_control_flow_straightening()) &&
309 (get_Block_n_cfgpreds(nn) == 1) &&
310 (get_irn_op(get_Block_cfgpred(nn, 0)) == op_Jmp)) {
311 ir_node *old = get_nodes_block(get_Block_cfgpred(nn, 0));
313 /* Jmp jumps into the block it is in -- deal self cycle. */
314 assert(is_Bad(get_new_node(get_irg_bad(current_ir_graph))));
315 exchange(nn, get_new_node(get_irg_bad(current_ir_graph)));
320 } else if (get_irn_op(n) == op_Phi) {
321 /* Don't copy node if corresponding predecessor in block is Bad.
322 The Block itself should not be Bad. */
323 block = get_nodes_block(n);
324 set_irn_n (nn, -1, get_new_node(block));
326 irn_arity = get_irn_arity(n);
327 for (i = 0; i < irn_arity; i++)
328 if (! is_Bad(get_irn_n(block, i))) {
329 set_irn_n (nn, j, get_new_node(get_irn_n(n, i)));
330 /*if (is_backedge(n, i)) set_backedge(nn, j);*/
333 /* If the pre walker reached this Phi after the post walker visited the
334 block block_visited is > 0. */
335 set_Block_block_visited(get_nodes_block(n), 0);
336 /* Compacting the Phi's ins might generate Phis with only one
338 if (get_irn_arity(nn) == 1)
339 exchange(nn, get_irn_n(nn, 0));
341 irn_arity = get_irn_arity(n);
342 for (i = -1; i < irn_arity; i++)
343 set_irn_n (nn, i, get_new_node(get_irn_n(n, i)));
345 /* Now the new node is complete. We can add it to the hash table for CSE.
346 @@@ inlinening aborts if we identify End. Why? */
347 if (get_irn_op(nn) != op_End)
348 add_identities (current_ir_graph->value_table, nn);
352 * Copies the graph recursively, compacts the keepalive of the end node.
354 * @param copy_node_nr If non-zero, the node number will be copied
357 copy_graph (int copy_node_nr) {
358 ir_node *oe, *ne, *ob, *nb, *om, *nm; /* old end, new end, old bad, new bad, old NoMem, new NoMem */
359 ir_node *ka; /* keep alive */
362 oe = get_irg_end(current_ir_graph);
363 /* copy the end node by hand, allocate dynamic in array! */
364 ne = new_ir_node(get_irn_dbg_info(oe),
371 /* Copy the attributes. Well, there might be some in the future... */
372 copy_node_attr(oe, ne);
373 set_new_node(oe, ne);
375 /* copy the Bad node */
376 ob = get_irg_bad(current_ir_graph);
377 nb = new_ir_node(get_irn_dbg_info(ob),
384 set_new_node(ob, nb);
386 /* copy the NoMem node */
387 om = get_irg_no_mem(current_ir_graph);
388 nm = new_ir_node(get_irn_dbg_info(om),
395 set_new_node(om, nm);
397 /* copy the live nodes */
398 irg_walk(get_nodes_block(oe), copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
399 /* copy_preds for the end node ... */
400 set_nodes_block(ne, get_new_node(get_nodes_block(oe)));
402 /*- ... and now the keep alives. -*/
403 /* First pick the not marked block nodes and walk them. We must pick these
404 first as else we will oversee blocks reachable from Phis. */
405 irn_arity = get_irn_arity(oe);
406 for (i = 0; i < irn_arity; i++) {
407 ka = get_irn_intra_n(oe, i);
408 if ((get_irn_op(ka) == op_Block) &&
409 (get_irn_visited(ka) < get_irg_visited(current_ir_graph))) {
410 /* We must keep the block alive and copy everything reachable */
411 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
412 irg_walk(ka, copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
413 add_End_keepalive(ne, get_new_node(ka));
417 /* Now pick the Phis. Here we will keep all! */
418 irn_arity = get_irn_arity(oe);
419 for (i = 0; i < irn_arity; i++) {
420 ka = get_irn_intra_n(oe, i);
421 if ((get_irn_op(ka) == op_Phi)) {
422 if (get_irn_visited(ka) < get_irg_visited(current_ir_graph)) {
423 /* We didn't copy the Phi yet. */
424 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
425 irg_walk(ka, copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
427 add_End_keepalive(ne, get_new_node(ka));
431 /* start block sometimes only reached after keep alives */
432 set_nodes_block(nb, get_new_node(get_nodes_block(ob)));
433 set_nodes_block(nm, get_new_node(get_nodes_block(om)));
437 * Copies the graph reachable from current_ir_graph->end to the obstack
438 * in current_ir_graph and fixes the environment.
439 * Then fixes the fields in current_ir_graph containing nodes of the
442 * @param copy_node_nr If non-zero, the node number will be copied
445 copy_graph_env (int copy_node_nr) {
447 /* Not all nodes remembered in current_ir_graph might be reachable
448 from the end node. Assure their link is set to NULL, so that
449 we can test whether new nodes have been computed. */
450 set_irn_link(get_irg_frame (current_ir_graph), NULL);
451 set_irn_link(get_irg_globals (current_ir_graph), NULL);
452 set_irn_link(get_irg_args (current_ir_graph), NULL);
453 set_irn_link(get_irg_initial_mem(current_ir_graph), NULL);
454 set_irn_link(get_irg_no_mem (current_ir_graph), NULL);
456 /* we use the block walk flag for removing Bads from Blocks ins. */
457 inc_irg_block_visited(current_ir_graph);
460 copy_graph(copy_node_nr);
462 /* fix the fields in current_ir_graph */
463 old_end = get_irg_end(current_ir_graph);
464 set_irg_end (current_ir_graph, get_new_node(old_end));
465 set_irg_end_except (current_ir_graph, get_irg_end(current_ir_graph));
466 set_irg_end_reg (current_ir_graph, get_irg_end(current_ir_graph));
468 set_irg_end_block (current_ir_graph, get_new_node(get_irg_end_block(current_ir_graph)));
469 if (get_irn_link(get_irg_frame(current_ir_graph)) == NULL) {
470 copy_node (get_irg_frame(current_ir_graph), INT_TO_PTR(copy_node_nr));
471 copy_preds(get_irg_frame(current_ir_graph), NULL);
473 if (get_irn_link(get_irg_globals(current_ir_graph)) == NULL) {
474 copy_node (get_irg_globals(current_ir_graph), INT_TO_PTR(copy_node_nr));
475 copy_preds(get_irg_globals(current_ir_graph), NULL);
477 if (get_irn_link(get_irg_initial_mem(current_ir_graph)) == NULL) {
478 copy_node (get_irg_initial_mem(current_ir_graph), INT_TO_PTR(copy_node_nr));
479 copy_preds(get_irg_initial_mem(current_ir_graph), NULL);
481 if (get_irn_link(get_irg_args(current_ir_graph)) == NULL) {
482 copy_node (get_irg_args(current_ir_graph), INT_TO_PTR(copy_node_nr));
483 copy_preds(get_irg_args(current_ir_graph), NULL);
485 set_irg_start (current_ir_graph, get_new_node(get_irg_start(current_ir_graph)));
487 set_irg_start_block(current_ir_graph,
488 get_new_node(get_irg_start_block(current_ir_graph)));
489 set_irg_frame (current_ir_graph, get_new_node(get_irg_frame(current_ir_graph)));
490 set_irg_globals (current_ir_graph, get_new_node(get_irg_globals(current_ir_graph)));
491 set_irg_initial_mem(current_ir_graph, get_new_node(get_irg_initial_mem(current_ir_graph)));
492 set_irg_args (current_ir_graph, get_new_node(get_irg_args(current_ir_graph)));
494 if (get_irn_link(get_irg_bad(current_ir_graph)) == NULL) {
495 copy_node(get_irg_bad(current_ir_graph), INT_TO_PTR(copy_node_nr));
496 copy_preds(get_irg_bad(current_ir_graph), NULL);
498 set_irg_bad(current_ir_graph, get_new_node(get_irg_bad(current_ir_graph)));
500 if (get_irn_link(get_irg_no_mem(current_ir_graph)) == NULL) {
501 copy_node(get_irg_no_mem(current_ir_graph), INT_TO_PTR(copy_node_nr));
502 copy_preds(get_irg_no_mem(current_ir_graph), NULL);
504 set_irg_no_mem(current_ir_graph, get_new_node(get_irg_no_mem(current_ir_graph)));
508 * Copies all reachable nodes to a new obstack. Removes bad inputs
509 * from block nodes and the corresponding inputs from Phi nodes.
510 * Merges single exit blocks with single entry blocks and removes
512 * Adds all new nodes to a new hash table for CSE. Does not
513 * perform CSE, so the hash table might contain common subexpressions.
516 dead_node_elimination(ir_graph *irg) {
518 int rem_ipview = get_interprocedural_view();
519 struct obstack *graveyard_obst = NULL;
520 struct obstack *rebirth_obst = NULL;
522 edges_init_graph(irg);
524 /* inform statistics that we started a dead-node elimination run */
525 hook_dead_node_elim(irg, 1);
527 /* Remember external state of current_ir_graph. */
528 rem = current_ir_graph;
529 current_ir_graph = irg;
530 set_interprocedural_view(0);
532 /* Handle graph state */
533 assert(get_irg_phase_state(current_ir_graph) != phase_building);
534 free_callee_info(current_ir_graph);
535 free_irg_outs(current_ir_graph);
537 /* @@@ so far we loose loops when copying */
538 free_loop_information(current_ir_graph);
540 if (get_opt_optimize() && get_opt_dead_node_elimination()) {
542 /* A quiet place, where the old obstack can rest in peace,
543 until it will be cremated. */
544 graveyard_obst = irg->obst;
546 /* A new obstack, where the reachable nodes will be copied to. */
547 rebirth_obst = xmalloc (sizeof(*rebirth_obst));
548 current_ir_graph->obst = rebirth_obst;
549 obstack_init (current_ir_graph->obst);
551 /* We also need a new hash table for cse */
552 del_identities (irg->value_table);
553 irg->value_table = new_identities ();
555 /* Copy the graph from the old to the new obstack */
558 /* Free memory from old unoptimized obstack */
559 obstack_free(graveyard_obst, 0); /* First empty the obstack ... */
560 xfree (graveyard_obst); /* ... then free it. */
563 /* inform statistics that the run is over */
564 hook_dead_node_elim(irg, 0);
566 current_ir_graph = rem;
567 set_interprocedural_view(rem_ipview);
569 /* reset the dominator's state */
570 if(get_irg_dom_state(irg) != dom_none)
571 set_irg_dom_inconsistent(irg);
575 * Relink bad predecessors of a block and store the old in array to the
576 * link field. This function is called by relink_bad_predecessors().
577 * The array of link field starts with the block operand at position 0.
578 * If block has bad predecessors, create a new in array without bad preds.
579 * Otherwise let in array untouched.
581 static void relink_bad_block_predecessors(ir_node *n, void *env) {
582 ir_node **new_in, *irn;
583 int i, new_irn_n, old_irn_arity, new_irn_arity = 0;
585 /* if link field of block is NULL, look for bad predecessors otherwise
586 this is already done */
587 if (get_irn_op(n) == op_Block &&
588 get_irn_link(n) == NULL) {
590 /* save old predecessors in link field (position 0 is the block operand)*/
591 set_irn_link(n, get_irn_in(n));
593 /* count predecessors without bad nodes */
594 old_irn_arity = get_irn_arity(n);
595 for (i = 0; i < old_irn_arity; i++)
596 if (!is_Bad(get_irn_n(n, i))) new_irn_arity++;
598 /* arity changing: set new predecessors without bad nodes */
599 if (new_irn_arity < old_irn_arity) {
600 /* Get new predecessor array. We do not resize the array, as we must
601 keep the old one to update Phis. */
602 new_in = NEW_ARR_D (ir_node *, current_ir_graph->obst, (new_irn_arity+1));
604 /* set new predecessors in array */
607 for (i = 0; i < old_irn_arity; i++) {
608 irn = get_irn_n(n, i);
610 new_in[new_irn_n] = irn;
611 is_backedge(n, i) ? set_backedge(n, new_irn_n-1) : set_not_backedge(n, new_irn_n-1);
615 //ARR_SETLEN(int, n->attr.block.backedge, new_irn_arity);
616 ARR_SHRINKLEN(n->attr.block.backedge, new_irn_arity);
619 } /* ir node has bad predecessors */
621 } /* Block is not relinked */
625 * Relinks Bad predecessors from Blocks and Phis called by walker
626 * remove_bad_predecesors(). If n is a Block, call
627 * relink_bad_block_redecessors(). If n is a Phi-node, call also the relinking
628 * function of Phi's Block. If this block has bad predecessors, relink preds
631 static void relink_bad_predecessors(ir_node *n, void *env) {
632 ir_node *block, **old_in;
633 int i, old_irn_arity, new_irn_arity;
635 /* relink bad predecessors of a block */
636 if (get_irn_op(n) == op_Block)
637 relink_bad_block_predecessors(n, env);
639 /* If Phi node relink its block and its predecessors */
640 if (get_irn_op(n) == op_Phi) {
642 /* Relink predecessors of phi's block */
643 block = get_nodes_block(n);
644 if (get_irn_link(block) == NULL)
645 relink_bad_block_predecessors(block, env);
647 old_in = (ir_node **)get_irn_link(block); /* Of Phi's Block */
648 old_irn_arity = ARR_LEN(old_in);
650 /* Relink Phi predecessors if count of predecessors changed */
651 if (old_irn_arity != ARR_LEN(get_irn_in(block))) {
652 /* set new predecessors in array
653 n->in[0] remains the same block */
655 for(i = 1; i < old_irn_arity; i++)
656 if (!is_Bad((ir_node *)old_in[i])) {
657 n->in[new_irn_arity] = n->in[i];
658 is_backedge(n, i) ? set_backedge(n, new_irn_arity) : set_not_backedge(n, new_irn_arity);
662 ARR_SETLEN(ir_node *, n->in, new_irn_arity);
663 ARR_SETLEN(int, n->attr.phi_backedge, new_irn_arity);
666 } /* n is a Phi node */
670 * Removes Bad Bad predecessors from Blocks and the corresponding
671 * inputs to Phi nodes as in dead_node_elimination but without
673 * On walking up set the link field to NULL, on walking down call
674 * relink_bad_predecessors() (This function stores the old in array
675 * to the link field and sets a new in array if arity of predecessors
678 void remove_bad_predecessors(ir_graph *irg) {
679 irg_walk_graph(irg, init_link, relink_bad_predecessors, NULL);
683 /*--------------------------------------------------------------------*/
684 /* Functionality for inlining */
685 /*--------------------------------------------------------------------*/
688 * Copy node for inlineing. Updates attributes that change when
689 * inlineing but not for dead node elimination.
691 * Copies the node by calling copy_node() and then updates the entity if
692 * it's a local one. env must be a pointer of the frame type of the
693 * inlined procedure. The new entities must be in the link field of
697 copy_node_inline (ir_node *n, void *env) {
699 type *frame_tp = (type *)env;
702 if (get_irn_op(n) == op_Sel) {
703 new = get_new_node (n);
704 assert(get_irn_op(new) == op_Sel);
705 if (get_entity_owner(get_Sel_entity(n)) == frame_tp) {
706 set_Sel_entity(new, get_entity_link(get_Sel_entity(n)));
708 } else if (get_irn_op(n) == op_Block) {
709 new = get_new_node (n);
710 new->attr.block.irg = current_ir_graph;
714 static void find_addr(ir_node *node, void *env)
716 if (get_irn_opcode(node) == iro_Proj) {
717 if (get_Proj_proj(node) == pn_Start_P_value_arg_base)
723 * currently, we cannot inline two cases:
724 * - call with compound arguments
725 * - graphs that take the address of a parameter
727 * check these conditions here
729 static int can_inline(ir_node *call, ir_graph *called_graph)
731 type *call_type = get_Call_type(call);
732 int params, ress, i, res;
733 assert(is_Method_type(call_type));
735 params = get_method_n_params(call_type);
736 ress = get_method_n_ress(call_type);
739 for (i = 0; i < params; ++i) {
740 type *p_type = get_method_param_type(call_type, i);
742 if (is_compound_type(p_type))
747 for (i = 0; i < ress; ++i) {
748 type *r_type = get_method_res_type(call_type, i);
750 if (is_compound_type(r_type))
755 irg_walk_graph(called_graph, find_addr, NULL, &res);
760 int inline_method(ir_node *call, ir_graph *called_graph) {
762 ir_node *post_call, *post_bl;
764 ir_node *end, *end_bl;
768 int arity, n_ret, n_exc, n_res, i, j, rem_opt, irn_arity;
771 irg_inline_property prop = get_irg_inline_property(called_graph);
773 if ( (prop != irg_inline_forced) &&
774 (!get_opt_optimize() || !get_opt_inline() || (prop == irg_inline_forbidden))) return 0;
776 /* Do not inline variadic functions. */
777 if (get_method_variadicity(get_entity_type(get_irg_entity(called_graph))) == variadicity_variadic)
780 assert(get_method_n_params(get_entity_type(get_irg_entity(called_graph))) ==
781 get_method_n_params(get_Call_type(call)));
784 * currently, we cannot inline two cases:
785 * - call with compound arguments
786 * - graphs that take the address of a parameter
788 if (! can_inline(call, called_graph))
791 /* -- Turn off optimizations, this can cause problems when allocating new nodes. -- */
792 rem_opt = get_opt_optimize();
795 /* Handle graph state */
796 assert(get_irg_phase_state(current_ir_graph) != phase_building);
797 assert(get_irg_pinned(current_ir_graph) == op_pin_state_pinned);
798 assert(get_irg_pinned(called_graph) == op_pin_state_pinned);
799 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
800 set_irg_outs_inconsistent(current_ir_graph);
801 set_irg_loopinfo_inconsistent(current_ir_graph);
802 set_irg_callee_info_state(current_ir_graph, irg_callee_info_inconsistent);
804 /* -- Check preconditions -- */
805 assert(get_irn_op(call) == op_Call);
806 /* @@@ does not work for InterfaceIII.java after cgana
807 assert(get_Call_type(call) == get_entity_type(get_irg_entity(called_graph)));
808 assert(smaller_type(get_entity_type(get_irg_entity(called_graph)),
809 get_Call_type(call)));
811 assert(get_type_tpop(get_Call_type(call)) == type_method);
812 if (called_graph == current_ir_graph) {
813 set_optimize(rem_opt);
817 /* here we know we WILL inline, so inform the statistics */
818 hook_inline(call, called_graph);
820 /* -- Decide how to handle exception control flow: Is there a handler
821 for the Call node, or do we branch directly to End on an exception?
823 0 There is a handler.
825 2 Exception handling not represented in Firm. -- */
827 ir_node *proj, *Mproj = NULL, *Xproj = NULL;
828 for (proj = (ir_node *)get_irn_link(call); proj; proj = (ir_node *)get_irn_link(proj)) {
829 assert(get_irn_op(proj) == op_Proj);
830 if (get_Proj_proj(proj) == pn_Call_X_except) Xproj = proj;
831 if (get_Proj_proj(proj) == pn_Call_M_except) Mproj = proj;
833 if (Mproj) { assert(Xproj); exc_handling = 0; } /* Mproj */
834 else if (Xproj) { exc_handling = 1; } /* !Mproj && Xproj */
835 else { exc_handling = 2; } /* !Mproj && !Xproj */
840 the procedure and later replaces the Start node of the called graph.
841 Post_call is the old Call node and collects the results of the called
842 graph. Both will end up being a tuple. -- */
843 post_bl = get_nodes_block(call);
844 set_irg_current_block(current_ir_graph, post_bl);
845 /* XxMxPxP of Start + parameter of Call */
846 in[pn_Start_X_initial_exec] = new_Jmp();
847 in[pn_Start_M] = get_Call_mem(call);
848 in[pn_Start_P_frame_base] = get_irg_frame(current_ir_graph);
849 in[pn_Start_P_globals] = get_irg_globals(current_ir_graph);
850 in[pn_Start_T_args] = new_Tuple(get_Call_n_params(call), get_Call_param_arr(call));
851 /* in[pn_Start_P_value_arg_base] = ??? */
852 pre_call = new_Tuple(5, in);
856 The new block gets the ins of the old block, pre_call and all its
857 predecessors and all Phi nodes. -- */
858 part_block(pre_call);
860 /* -- Prepare state for dead node elimination -- */
861 /* Visited flags in calling irg must be >= flag in called irg.
862 Else walker and arity computation will not work. */
863 if (get_irg_visited(current_ir_graph) <= get_irg_visited(called_graph))
864 set_irg_visited(current_ir_graph, get_irg_visited(called_graph)+1);
865 if (get_irg_block_visited(current_ir_graph)< get_irg_block_visited(called_graph))
866 set_irg_block_visited(current_ir_graph, get_irg_block_visited(called_graph));
867 /* Set pre_call as new Start node in link field of the start node of
868 calling graph and pre_calls block as new block for the start block
870 Further mark these nodes so that they are not visited by the
872 set_irn_link(get_irg_start(called_graph), pre_call);
873 set_irn_visited(get_irg_start(called_graph), get_irg_visited(current_ir_graph));
874 set_irn_link(get_irg_start_block(called_graph), get_nodes_block(pre_call));
875 set_irn_visited(get_irg_start_block(called_graph), get_irg_visited(current_ir_graph));
876 set_irn_link(get_irg_bad(called_graph), get_irg_bad(current_ir_graph));
877 set_irn_visited(get_irg_bad(called_graph), get_irg_visited(current_ir_graph));
879 /* Initialize for compaction of in arrays */
880 inc_irg_block_visited(current_ir_graph);
882 /* -- Replicate local entities of the called_graph -- */
883 /* copy the entities. */
884 called_frame = get_irg_frame_type(called_graph);
885 for (i = 0; i < get_class_n_members(called_frame); i++) {
886 entity *new_ent, *old_ent;
887 old_ent = get_class_member(called_frame, i);
888 new_ent = copy_entity_own(old_ent, get_cur_frame_type());
889 set_entity_link(old_ent, new_ent);
892 /* visited is > than that of called graph. With this trick visited will
893 remain unchanged so that an outer walker, e.g., searching the call nodes
894 to inline, calling this inline will not visit the inlined nodes. */
895 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
897 /* -- Performing dead node elimination inlines the graph -- */
898 /* Copies the nodes to the obstack of current_ir_graph. Updates links to new
900 /* @@@ endless loops are not copied!! -- they should be, I think... */
901 irg_walk(get_irg_end(called_graph), copy_node_inline, copy_preds,
902 get_irg_frame_type(called_graph));
904 /* Repair called_graph */
905 set_irg_visited(called_graph, get_irg_visited(current_ir_graph));
906 set_irg_block_visited(called_graph, get_irg_block_visited(current_ir_graph));
907 set_Block_block_visited(get_irg_start_block(called_graph), 0);
909 /* -- Merge the end of the inlined procedure with the call site -- */
910 /* We will turn the old Call node into a Tuple with the following
913 0: Phi of all Memories of Return statements.
914 1: Jmp from new Block that merges the control flow from all exception
915 predecessors of the old end block.
916 2: Tuple of all arguments.
917 3: Phi of Exception memories.
918 In case the old Call directly branches to End on an exception we don't
919 need the block merging all exceptions nor the Phi of the exception
923 /* -- Precompute some values -- */
924 end_bl = get_new_node(get_irg_end_block(called_graph));
925 end = get_new_node(get_irg_end(called_graph));
926 arity = get_irn_arity(end_bl); /* arity = n_exc + n_ret */
927 n_res = get_method_n_ress(get_Call_type(call));
929 res_pred = xmalloc (n_res * sizeof(*res_pred));
930 cf_pred = xmalloc (arity * sizeof(*res_pred));
932 set_irg_current_block(current_ir_graph, post_bl); /* just to make sure */
934 /* -- archive keepalives -- */
935 irn_arity = get_irn_arity(end);
936 for (i = 0; i < irn_arity; i++)
937 add_End_keepalive(get_irg_end(current_ir_graph), get_irn_n(end, i));
939 /* The new end node will die. We need not free as the in array is on the obstack:
940 copy_node() only generated 'D' arrays. */
942 /* -- Replace Return nodes by Jump nodes. -- */
944 for (i = 0; i < arity; i++) {
946 ret = get_irn_n(end_bl, i);
947 if (get_irn_op(ret) == op_Return) {
948 cf_pred[n_ret] = new_r_Jmp(current_ir_graph, get_nodes_block(ret));
952 set_irn_in(post_bl, n_ret, cf_pred);
954 /* -- Build a Tuple for all results of the method.
955 Add Phi node if there was more than one Return. -- */
956 turn_into_tuple(post_call, 4);
957 /* First the Memory-Phi */
959 for (i = 0; i < arity; i++) {
960 ret = get_irn_n(end_bl, i);
961 if (get_irn_op(ret) == op_Return) {
962 cf_pred[n_ret] = get_Return_mem(ret);
966 phi = new_Phi(n_ret, cf_pred, mode_M);
967 set_Tuple_pred(call, pn_Call_M_regular, phi);
968 /* Conserve Phi-list for further inlinings -- but might be optimized */
969 if (get_nodes_block(phi) == post_bl) {
970 set_irn_link(phi, get_irn_link(post_bl));
971 set_irn_link(post_bl, phi);
973 /* Now the real results */
975 for (j = 0; j < n_res; j++) {
977 for (i = 0; i < arity; i++) {
978 ret = get_irn_n(end_bl, i);
979 if (get_irn_op(ret) == op_Return) {
980 cf_pred[n_ret] = get_Return_res(ret, j);
985 phi = new_Phi(n_ret, cf_pred, get_irn_mode(cf_pred[0]));
989 /* Conserve Phi-list for further inlinings -- but might be optimized */
990 if (get_nodes_block(phi) == post_bl) {
991 set_irn_link(phi, get_irn_link(post_bl));
992 set_irn_link(post_bl, phi);
995 set_Tuple_pred(call, pn_Call_T_result, new_Tuple(n_res, res_pred));
997 set_Tuple_pred(call, pn_Call_T_result, new_Bad());
999 /* Finally the exception control flow.
1000 We have two (three) possible situations:
1001 First if the Call branches to an exception handler: We need to add a Phi node to
1002 collect the memory containing the exception objects. Further we need
1003 to add another block to get a correct representation of this Phi. To
1004 this block we add a Jmp that resolves into the X output of the Call
1005 when the Call is turned into a tuple.
1006 Second the Call branches to End, the exception is not handled. Just
1007 add all inlined exception branches to the End node.
1008 Third: there is no Exception edge at all. Handle as case two. */
1009 if (exc_handling == 0) {
1011 for (i = 0; i < arity; i++) {
1013 ret = get_irn_n(end_bl, i);
1014 if (is_fragile_op(skip_Proj(ret)) || (get_irn_op(skip_Proj(ret)) == op_Raise)) {
1015 cf_pred[n_exc] = ret;
1020 new_Block(n_exc, cf_pred); /* watch it: current_block is changed! */
1021 set_Tuple_pred(call, pn_Call_X_except, new_Jmp());
1022 /* The Phi for the memories with the exception objects */
1024 for (i = 0; i < arity; i++) {
1026 ret = skip_Proj(get_irn_n(end_bl, i));
1027 if (get_irn_op(ret) == op_Call) {
1028 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_block(ret), ret, mode_M, 3);
1030 } else if (is_fragile_op(ret)) {
1031 /* We rely that all cfops have the memory output at the same position. */
1032 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_block(ret), ret, mode_M, 0);
1034 } else if (get_irn_op(ret) == op_Raise) {
1035 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_block(ret), ret, mode_M, 1);
1039 set_Tuple_pred(call, pn_Call_M_except, new_Phi(n_exc, cf_pred, mode_M));
1041 set_Tuple_pred(call, pn_Call_X_except, new_Bad());
1042 set_Tuple_pred(call, pn_Call_M_except, new_Bad());
1045 ir_node *main_end_bl;
1046 int main_end_bl_arity;
1047 ir_node **end_preds;
1049 /* assert(exc_handling == 1 || no exceptions. ) */
1051 for (i = 0; i < arity; i++) {
1052 ir_node *ret = get_irn_n(end_bl, i);
1054 if (is_fragile_op(skip_Proj(ret)) || (get_irn_op(skip_Proj(ret)) == op_Raise)) {
1055 cf_pred[n_exc] = ret;
1059 main_end_bl = get_irg_end_block(current_ir_graph);
1060 main_end_bl_arity = get_irn_arity(main_end_bl);
1061 end_preds = xmalloc ((n_exc + main_end_bl_arity) * sizeof(*end_preds));
1063 for (i = 0; i < main_end_bl_arity; ++i)
1064 end_preds[i] = get_irn_n(main_end_bl, i);
1065 for (i = 0; i < n_exc; ++i)
1066 end_preds[main_end_bl_arity + i] = cf_pred[i];
1067 set_irn_in(main_end_bl, n_exc + main_end_bl_arity, end_preds);
1068 set_Tuple_pred(call, pn_Call_X_except, new_Bad());
1069 set_Tuple_pred(call, pn_Call_M_except, new_Bad());
1075 #if 0 /* old. now better, correcter, faster implementation. */
1077 /* -- If the exception control flow from the inlined Call directly
1078 branched to the end block we now have the following control
1079 flow predecessor pattern: ProjX -> Tuple -> Jmp. We must
1080 remove the Jmp along with it's empty block and add Jmp's
1081 predecessors as predecessors of this end block. No problem if
1082 there is no exception, because then branches Bad to End which
1084 @@@ can't we know this beforehand: by getting the Proj(1) from
1085 the Call link list and checking whether it goes to Proj. */
1086 /* find the problematic predecessor of the end block. */
1087 end_bl = get_irg_end_block(current_ir_graph);
1088 for (i = 0; i < get_Block_n_cfgpreds(end_bl); i++) {
1089 cf_op = get_Block_cfgpred(end_bl, i);
1090 if (get_irn_op(cf_op) == op_Proj) {
1091 cf_op = get_Proj_pred(cf_op);
1092 if ((get_irn_op(cf_op) == op_Tuple) && (cf_op == call)) {
1093 /* There are unoptimized tuples from inlineing before when no exc */
1094 assert(get_Proj_proj(get_Block_cfgpred(end_bl, i)) == pn_Call_X_except);
1095 cf_op = get_Tuple_pred(cf_op, pn_Call_X_except);
1096 assert(get_irn_op(cf_op) == op_Jmp);
1102 if (i < get_Block_n_cfgpreds(end_bl)) {
1103 bl = get_nodes_block(cf_op);
1104 arity = get_Block_n_cfgpreds(end_bl) + get_Block_n_cfgpreds(bl) - 1;
1105 cf_pred = xmalloc (arity * sizeof(*cf_pred));
1106 for (j = 0; j < i; j++)
1107 cf_pred[j] = get_Block_cfgpred(end_bl, j);
1108 for (j = j; j < i + get_Block_n_cfgpreds(bl); j++)
1109 cf_pred[j] = get_Block_cfgpred(bl, j-i);
1110 for (j = j; j < arity; j++)
1111 cf_pred[j] = get_Block_cfgpred(end_bl, j-get_Block_n_cfgpreds(bl) +1);
1112 set_irn_in(end_bl, arity, cf_pred);
1114 /* Remove the exception pred from post-call Tuple. */
1115 set_Tuple_pred(call, pn_Call_X_except, new_Bad());
1120 /* -- Turn CSE back on. -- */
1121 set_optimize(rem_opt);
1126 /********************************************************************/
1127 /* Apply inlineing to small methods. */
1128 /********************************************************************/
1130 /* It makes no sense to inline too many calls in one procedure. Anyways,
1131 I didn't get a version with NEW_ARR_F to run. */
1132 #define MAX_INLINE 1024
1135 * environment for inlining small irgs
1137 typedef struct _inline_env_t {
1139 ir_node *calls[MAX_INLINE];
1143 * Returns the irg called from a Call node. If the irg is not
1144 * known, NULL is returned.
1146 static ir_graph *get_call_called_irg(ir_node *call) {
1148 ir_graph *called_irg = NULL;
1150 assert(get_irn_op(call) == op_Call);
1152 addr = get_Call_ptr(call);
1153 if ((get_irn_op(addr) == op_SymConst) && (get_SymConst_kind (addr) == symconst_addr_ent)) {
1154 called_irg = get_entity_irg(get_SymConst_entity(addr));
1160 static void collect_calls(ir_node *call, void *env) {
1163 if (get_irn_op(call) != op_Call) return;
1165 addr = get_Call_ptr(call);
1167 if (get_irn_op(addr) == op_SymConst) {
1168 if (get_SymConst_kind(addr) == symconst_addr_ent) {
1169 ir_graph *called_irg = get_entity_irg(get_SymConst_entity(addr));
1170 inline_env_t *ienv = (inline_env_t *)env;
1171 if (called_irg && ienv->pos < MAX_INLINE) {
1172 /* The Call node calls a locally defined method. Remember to inline. */
1173 ienv->calls[ienv->pos++] = call;
1180 * Inlines all small methods at call sites where the called address comes
1181 * from a Const node that references the entity representing the called
1183 * The size argument is a rough measure for the code size of the method:
1184 * Methods where the obstack containing the firm graph is smaller than
1187 void inline_small_irgs(ir_graph *irg, int size) {
1189 ir_graph *rem = current_ir_graph;
1190 inline_env_t env /* = {0, NULL}*/;
1192 if (!(get_opt_optimize() && get_opt_inline())) return;
1194 current_ir_graph = irg;
1195 /* Handle graph state */
1196 assert(get_irg_phase_state(current_ir_graph) != phase_building);
1197 free_callee_info(current_ir_graph);
1199 /* Find Call nodes to inline.
1200 (We can not inline during a walk of the graph, as inlineing the same
1201 method several times changes the visited flag of the walked graph:
1202 after the first inlineing visited of the callee equals visited of
1203 the caller. With the next inlineing both are increased.) */
1205 irg_walk(get_irg_end(irg), NULL, collect_calls, &env);
1207 if ((env.pos > 0) && (env.pos < MAX_INLINE)) {
1208 /* There are calls to inline */
1209 collect_phiprojs(irg);
1210 for (i = 0; i < env.pos; i++) {
1212 callee = get_entity_irg(get_SymConst_entity(get_Call_ptr(env.calls[i])));
1213 if (((_obstack_memory_used(callee->obst) - (int)obstack_room(callee->obst)) < size) ||
1214 (get_irg_inline_property(callee) == irg_inline_forced)) {
1215 inline_method(env.calls[i], callee);
1220 current_ir_graph = rem;
1224 * Environment for inlining irgs.
1227 int n_nodes; /**< Nodes in graph except Id, Tuple, Proj, Start, End */
1228 int n_nodes_orig; /**< for statistics */
1229 eset *call_nodes; /**< All call nodes in this graph */
1231 int n_call_nodes_orig; /**< for statistics */
1232 int n_callers; /**< Number of known graphs that call this graphs. */
1233 int n_callers_orig; /**< for statistics */
1237 * Allocate a new environment for inlining.
1239 static inline_irg_env *new_inline_irg_env(void) {
1240 inline_irg_env *env = xmalloc(sizeof(*env));
1241 env->n_nodes = -2; /* do not count count Start, End */
1242 env->n_nodes_orig = -2; /* do not count Start, End */
1243 env->call_nodes = eset_create();
1244 env->n_call_nodes = 0;
1245 env->n_call_nodes_orig = 0;
1247 env->n_callers_orig = 0;
1252 * destroy an environment for inlining.
1254 static void free_inline_irg_env(inline_irg_env *env) {
1255 eset_destroy(env->call_nodes);
1260 * post-walker: collect all calls in the inline-environment
1261 * of a graph and sum some statistics.
1263 static void collect_calls2(ir_node *call, void *env) {
1264 inline_irg_env *x = (inline_irg_env *)env;
1265 ir_op *op = get_irn_op(call);
1268 /* count meaningful nodes in irg */
1269 if (op != op_Proj && op != op_Tuple && op != op_Sync) {
1274 if (op != op_Call) return;
1276 /* collect all call nodes */
1277 eset_insert(x->call_nodes, call);
1279 x->n_call_nodes_orig++;
1281 /* count all static callers */
1282 callee = get_call_called_irg(call);
1284 inline_irg_env *callee_env = get_irg_link(callee);
1285 callee_env->n_callers++;
1286 callee_env->n_callers_orig++;
1291 * Returns TRUE if the number of callers in 0 in the irg's environment,
1292 * hence this irg is a leave.
1294 INLINE static int is_leave(ir_graph *irg) {
1295 return (((inline_irg_env *)get_irg_link(irg))->n_call_nodes == 0);
1299 * Returns TRUE if the number of callers is smaller size in the irg's environment.
1301 INLINE static int is_smaller(ir_graph *callee, int size) {
1302 return (((inline_irg_env *)get_irg_link(callee))->n_nodes < size);
1307 * Inlines small leave methods at call sites where the called address comes
1308 * from a Const node that references the entity representing the called
1310 * The size argument is a rough measure for the code size of the method:
1311 * Methods where the obstack containing the firm graph is smaller than
1314 void inline_leave_functions(int maxsize, int leavesize, int size) {
1315 inline_irg_env *env;
1316 int i, n_irgs = get_irp_n_irgs();
1317 ir_graph *rem = current_ir_graph;
1320 if (!(get_opt_optimize() && get_opt_inline())) return;
1322 /* extend all irgs by a temporary data structure for inlining. */
1323 for (i = 0; i < n_irgs; ++i)
1324 set_irg_link(get_irp_irg(i), new_inline_irg_env());
1326 /* Precompute information in temporary data structure. */
1327 for (i = 0; i < n_irgs; ++i) {
1328 current_ir_graph = get_irp_irg(i);
1329 assert(get_irg_phase_state(current_ir_graph) != phase_building);
1330 free_callee_info(current_ir_graph);
1332 irg_walk(get_irg_end(current_ir_graph), NULL, collect_calls2,
1333 get_irg_link(current_ir_graph));
1336 /* -- and now inline. -- */
1338 /* Inline leaves recursively -- we might construct new leaves. */
1339 while (did_inline) {
1342 for (i = 0; i < n_irgs; ++i) {
1344 int phiproj_computed = 0;
1346 current_ir_graph = get_irp_irg(i);
1347 env = (inline_irg_env *)get_irg_link(current_ir_graph);
1349 for (call = eset_first(env->call_nodes); call; call = eset_next(env->call_nodes)) {
1352 if (get_irn_op(call) == op_Tuple) continue; /* We already have inlined this call. */
1353 callee = get_call_called_irg(call);
1355 if (env->n_nodes > maxsize) continue; // break;
1357 if (callee && (is_leave(callee) && is_smaller(callee, leavesize))) {
1358 if (!phiproj_computed) {
1359 phiproj_computed = 1;
1360 collect_phiprojs(current_ir_graph);
1362 did_inline = inline_method(call, callee);
1365 /* Do some statistics */
1366 inline_irg_env *callee_env = (inline_irg_env *)get_irg_link(callee);
1367 env->n_call_nodes --;
1368 env->n_nodes += callee_env->n_nodes;
1369 callee_env->n_callers--;
1376 /* inline other small functions. */
1377 for (i = 0; i < n_irgs; ++i) {
1380 int phiproj_computed = 0;
1382 current_ir_graph = get_irp_irg(i);
1383 env = (inline_irg_env *)get_irg_link(current_ir_graph);
1385 /* we can not walk and change a set, nor remove from it.
1387 walkset = env->call_nodes;
1388 env->call_nodes = eset_create();
1389 for (call = eset_first(walkset); call; call = eset_next(walkset)) {
1392 if (get_irn_op(call) == op_Tuple) continue; /* We already inlined. */
1393 callee = get_call_called_irg(call);
1396 ((is_smaller(callee, size) && (env->n_nodes < maxsize)) || /* small function */
1397 (get_irg_inline_property(callee) == irg_inline_forced))) {
1398 if (!phiproj_computed) {
1399 phiproj_computed = 1;
1400 collect_phiprojs(current_ir_graph);
1402 if (inline_method(call, callee)) {
1403 inline_irg_env *callee_env = (inline_irg_env *)get_irg_link(callee);
1404 env->n_call_nodes--;
1405 eset_insert_all(env->call_nodes, callee_env->call_nodes); /* @@@ ??? This are the wrong nodes !? Not the copied ones. */
1406 env->n_call_nodes += callee_env->n_call_nodes;
1407 env->n_nodes += callee_env->n_nodes;
1408 callee_env->n_callers--;
1411 eset_insert(env->call_nodes, call);
1414 eset_destroy(walkset);
1417 for (i = 0; i < n_irgs; ++i) {
1418 current_ir_graph = get_irp_irg(i);
1420 env = (inline_irg_env *)get_irg_link(current_ir_graph);
1421 if ((env->n_call_nodes_orig != env->n_call_nodes) ||
1422 (env->n_callers_orig != env->n_callers))
1423 printf("Nodes:%3d ->%3d, calls:%3d ->%3d, callers:%3d ->%3d, -- %s\n",
1424 env->n_nodes_orig, env->n_nodes, env->n_call_nodes_orig, env->n_call_nodes,
1425 env->n_callers_orig, env->n_callers,
1426 get_entity_name(get_irg_entity(current_ir_graph)));
1428 free_inline_irg_env((inline_irg_env *)get_irg_link(current_ir_graph));
1431 current_ir_graph = rem;
1434 /*******************************************************************/
1435 /* Code Placement. Pins all floating nodes to a block where they */
1436 /* will be executed only if needed. */
1437 /*******************************************************************/
1440 * Returns non-zero, is a block is not reachable from Start.
1442 * @param block the block to test
1445 is_Block_unreachable(ir_node *block) {
1446 return is_Block_dead(block) || get_Block_dom_depth(block) < 0;
1450 * Find the earliest correct block for N. --- Place N into the
1451 * same Block as its dominance-deepest Input.
1453 * We have to avoid calls to get_nodes_block() here
1454 * because the graph is floating.
1456 * move_out_of_loops() expects that place_floats_early() have placed
1457 * all "living" nodes into a living block. That's why we must
1458 * move nodes in dead block with "live" successors into a valid
1460 * We move them just into the same block as it's successor (or
1461 * in case of a Phi into the effective use block). For Phi successors,
1462 * this may still be a dead block, but then there is no real use, as
1463 * the control flow will be dead later.
1466 place_floats_early(ir_node *n, pdeq *worklist)
1470 /* we must not run into an infinite loop */
1471 assert(irn_not_visited(n));
1472 mark_irn_visited(n);
1474 /* Place floating nodes. */
1475 if (get_irn_pinned(n) == op_pin_state_floats) {
1476 ir_node *curr_block = get_irn_n(n, -1);
1477 int in_dead_block = is_Block_unreachable(curr_block);
1479 ir_node *b = NULL; /* The block to place this node in */
1481 assert(get_irn_op(n) != op_Block);
1483 if ((get_irn_op(n) == op_Const) ||
1484 (get_irn_op(n) == op_SymConst) ||
1486 (get_irn_op(n) == op_Unknown)) {
1487 /* These nodes will not be placed by the loop below. */
1488 b = get_irg_start_block(current_ir_graph);
1492 /* find the block for this node. */
1493 irn_arity = get_irn_arity(n);
1494 for (i = 0; i < irn_arity; i++) {
1495 ir_node *pred = get_irn_n(n, i);
1496 ir_node *pred_block;
1498 if ((irn_not_visited(pred))
1499 && (get_irn_pinned(pred) == op_pin_state_floats)) {
1502 * If the current node is NOT in a dead block, but one of its
1503 * predecessors is, we must move the predecessor to a live block.
1504 * Such thing can happen, if global CSE chose a node from a dead block.
1505 * We move it simple to our block.
1506 * Note that neither Phi nor End nodes are floating, so we don't
1507 * need to handle them here.
1509 if (! in_dead_block) {
1510 if (get_irn_pinned(pred) == op_pin_state_floats &&
1511 is_Block_unreachable(get_irn_n(pred, -1)))
1512 set_nodes_block(pred, curr_block);
1514 place_floats_early(pred, worklist);
1518 * A node in the Bad block must stay in the bad block,
1519 * so don't compute a new block for it.
1524 /* Because all loops contain at least one op_pin_state_pinned node, now all
1525 our inputs are either op_pin_state_pinned or place_early() has already
1526 been finished on them. We do not have any unfinished inputs! */
1527 pred_block = get_irn_n(pred, -1);
1528 if ((!is_Block_dead(pred_block)) &&
1529 (get_Block_dom_depth(pred_block) > depth)) {
1531 depth = get_Block_dom_depth(pred_block);
1533 /* Avoid that the node is placed in the Start block */
1534 if ((depth == 1) && (get_Block_dom_depth(get_irn_n(n, -1)) > 1)) {
1535 b = get_Block_cfg_out(get_irg_start_block(current_ir_graph), 0);
1536 assert(b != get_irg_start_block(current_ir_graph));
1541 set_nodes_block(n, b);
1545 * Add predecessors of non floating nodes and non-floating predecessors
1546 * of floating nodes to worklist and fix their blocks if the are in dead block.
1548 irn_arity = get_irn_arity(n);
1550 if (get_irn_op(n) == op_End) {
1552 * Simplest case: End node. Predecessors are keep-alives,
1553 * no need to move out of dead block.
1555 for (i = -1; i < irn_arity; ++i) {
1556 ir_node *pred = get_irn_n(n, i);
1557 if (irn_not_visited(pred))
1558 pdeq_putr(worklist, pred);
1561 else if (is_Block(n)) {
1563 * Blocks: Predecessors are control flow, no need to move
1564 * them out of dead block.
1566 for (i = irn_arity - 1; i >= 0; --i) {
1567 ir_node *pred = get_irn_n(n, i);
1568 if (irn_not_visited(pred))
1569 pdeq_putr(worklist, pred);
1572 else if (is_Phi(n)) {
1574 ir_node *curr_block = get_irn_n(n, -1);
1575 int in_dead_block = is_Block_unreachable(curr_block);
1578 * Phi nodes: move nodes from dead blocks into the effective use
1579 * of the Phi-input if the Phi is not in a bad block.
1581 pred = get_irn_n(n, -1);
1582 if (irn_not_visited(pred))
1583 pdeq_putr(worklist, pred);
1585 for (i = irn_arity - 1; i >= 0; --i) {
1586 ir_node *pred = get_irn_n(n, i);
1588 if (irn_not_visited(pred)) {
1589 if (! in_dead_block &&
1590 get_irn_pinned(pred) == op_pin_state_floats &&
1591 is_Block_unreachable(get_irn_n(pred, -1))) {
1592 set_nodes_block(pred, get_Block_cfgpred_block(curr_block, i));
1594 pdeq_putr(worklist, pred);
1600 ir_node *curr_block = get_irn_n(n, -1);
1601 int in_dead_block = is_Block_unreachable(curr_block);
1604 * All other nodes: move nodes from dead blocks into the same block.
1606 pred = get_irn_n(n, -1);
1607 if (irn_not_visited(pred))
1608 pdeq_putr(worklist, pred);
1610 for (i = irn_arity - 1; i >= 0; --i) {
1611 ir_node *pred = get_irn_n(n, i);
1613 if (irn_not_visited(pred)) {
1614 if (! in_dead_block &&
1615 get_irn_pinned(pred) == op_pin_state_floats &&
1616 is_Block_unreachable(get_irn_n(pred, -1))) {
1617 set_nodes_block(pred, curr_block);
1619 pdeq_putr(worklist, pred);
1626 * Floating nodes form subgraphs that begin at nodes as Const, Load,
1627 * Start, Call and that end at op_pin_state_pinned nodes as Store, Call. Place_early
1628 * places all floating nodes reachable from its argument through floating
1629 * nodes and adds all beginnings at op_pin_state_pinned nodes to the worklist.
1631 static INLINE void place_early(pdeq *worklist) {
1633 inc_irg_visited(current_ir_graph);
1635 /* this inits the worklist */
1636 place_floats_early(get_irg_end(current_ir_graph), worklist);
1638 /* Work the content of the worklist. */
1639 while (!pdeq_empty(worklist)) {
1640 ir_node *n = pdeq_getl(worklist);
1641 if (irn_not_visited(n))
1642 place_floats_early(n, worklist);
1645 set_irg_outs_inconsistent(current_ir_graph);
1646 set_irg_pinned(current_ir_graph, op_pin_state_pinned);
1650 * Compute the deepest common ancestor of block and dca.
1652 static ir_node *calc_dca(ir_node *dca, ir_node *block)
1656 /* we do not want to place nodes in dead blocks */
1657 if (is_Block_dead(block))
1660 /* We found a first legal placement. */
1661 if (!dca) return block;
1663 /* Find a placement that is dominates both, dca and block. */
1664 while (get_Block_dom_depth(block) > get_Block_dom_depth(dca))
1665 block = get_Block_idom(block);
1667 while (get_Block_dom_depth(dca) > get_Block_dom_depth(block)) {
1668 dca = get_Block_idom(dca);
1671 while (block != dca)
1672 { block = get_Block_idom(block); dca = get_Block_idom(dca); }
1677 /** Deepest common dominance ancestor of DCA and CONSUMER of PRODUCER.
1678 * I.e., DCA is the block where we might place PRODUCER.
1679 * A data flow edge points from producer to consumer.
1682 consumer_dom_dca(ir_node *dca, ir_node *consumer, ir_node *producer)
1684 ir_node *block = NULL;
1686 /* Compute the latest block into which we can place a node so that it is
1688 if (get_irn_op(consumer) == op_Phi) {
1689 /* our consumer is a Phi-node, the effective use is in all those
1690 blocks through which the Phi-node reaches producer */
1692 ir_node *phi_block = get_nodes_block(consumer);
1693 irn_arity = get_irn_arity(consumer);
1695 for (i = 0; i < irn_arity; i++) {
1696 if (get_irn_n(consumer, i) == producer) {
1697 ir_node *new_block = get_nodes_block(get_Block_cfgpred(phi_block, i));
1699 if (! is_Block_unreachable(new_block))
1700 block = calc_dca(block, new_block);
1705 block = get_irn_n(producer, -1);
1708 assert(is_no_Block(consumer));
1709 block = get_nodes_block(consumer);
1712 /* Compute the deepest common ancestor of block and dca. */
1713 return calc_dca(dca, block);
1716 /* FIXME: the name clashes here with the function from ana/field_temperature.c
1718 static INLINE int get_irn_loop_depth(ir_node *n) {
1719 return get_loop_depth(get_irn_loop(n));
1723 * Move n to a block with less loop depth than it's current block. The
1724 * new block must be dominated by early.
1726 * @param n the node that should be moved
1727 * @param early the earliest block we can n move to
1730 move_out_of_loops (ir_node *n, ir_node *early)
1732 ir_node *best, *dca;
1736 /* Find the region deepest in the dominator tree dominating
1737 dca with the least loop nesting depth, but still dominated
1738 by our early placement. */
1739 dca = get_nodes_block(n);
1742 while (dca != early) {
1743 dca = get_Block_idom(dca);
1744 if (!dca || is_Bad(dca)) break; /* may be Bad if not reachable from Start */
1745 if (get_irn_loop_depth(dca) < get_irn_loop_depth(best)) {
1749 if (best != get_nodes_block(n)) {
1751 printf("Moving out of loop: "); DDMN(n);
1752 printf(" Outermost block: "); DDMN(early);
1753 printf(" Best block: "); DDMN(best);
1754 printf(" Innermost block: "); DDMN(get_nodes_block(n));
1756 set_nodes_block(n, best);
1761 * Find the latest legal block for N and place N into the
1762 * `optimal' Block between the latest and earliest legal block.
1763 * The `optimal' block is the dominance-deepest block of those
1764 * with the least loop-nesting-depth. This places N out of as many
1765 * loops as possible and then makes it as control dependent as
1769 place_floats_late(ir_node *n, pdeq *worklist)
1774 assert(irn_not_visited(n)); /* no multiple placement */
1776 mark_irn_visited(n);
1778 /* no need to place block nodes, control nodes are already placed. */
1779 if ((get_irn_op(n) != op_Block) &&
1781 (get_irn_mode(n) != mode_X)) {
1782 /* Remember the early_blk placement of this block to move it
1783 out of loop no further than the early_blk placement. */
1784 early_blk = get_irn_n(n, -1);
1787 * BEWARE: Here we also get code, that is live, but
1788 * was in a dead block. If the node is life, but because
1789 * of CSE in a dead block, we still might need it.
1792 /* Assure that our users are all placed, except the Phi-nodes.
1793 --- Each data flow cycle contains at least one Phi-node. We
1794 have to break the `user has to be placed before the
1795 producer' dependence cycle and the Phi-nodes are the
1796 place to do so, because we need to base our placement on the
1797 final region of our users, which is OK with Phi-nodes, as they
1798 are op_pin_state_pinned, and they never have to be placed after a
1799 producer of one of their inputs in the same block anyway. */
1800 for (i = get_irn_n_outs(n) - 1; i >= 0; --i) {
1801 ir_node *succ = get_irn_out(n, i);
1802 if (irn_not_visited(succ) && (get_irn_op(succ) != op_Phi))
1803 place_floats_late(succ, worklist);
1806 if (! is_Block_dead(early_blk)) {
1807 /* do only move things that where not dead */
1809 /* We have to determine the final block of this node... except for
1811 if ((get_irn_pinned(n) == op_pin_state_floats) &&
1812 (get_irn_op(n) != op_Const) &&
1813 (get_irn_op(n) != op_SymConst)) {
1814 ir_node *dca = NULL; /* deepest common ancestor in the
1815 dominator tree of all nodes'
1816 blocks depending on us; our final
1817 placement has to dominate DCA. */
1818 for (i = get_irn_n_outs(n) - 1; i >= 0; --i) {
1819 ir_node *succ = get_irn_out(n, i);
1822 if (get_irn_op(succ) == op_End) {
1824 * This consumer is the End node, a keep alive edge.
1825 * This is not a real consumer, so we ignore it
1830 /* ignore if succ is in dead code */
1831 succ_blk = get_irn_n(succ, -1);
1832 if (is_Block_unreachable(succ_blk))
1834 dca = consumer_dom_dca(dca, succ, n);
1837 set_nodes_block(n, dca);
1838 move_out_of_loops(n, early_blk);
1844 /* Add predecessors of all non-floating nodes on list. (Those of floating
1845 nodes are placed already and therefore are marked.) */
1846 for (i = 0; i < get_irn_n_outs(n); i++) {
1847 ir_node *succ = get_irn_out(n, i);
1848 if (irn_not_visited(get_irn_out(n, i))) {
1849 pdeq_putr(worklist, succ);
1854 static INLINE void place_late(pdeq *worklist) {
1856 inc_irg_visited(current_ir_graph);
1858 /* This fills the worklist initially. */
1859 place_floats_late(get_irg_start_block(current_ir_graph), worklist);
1861 /* And now empty the worklist again... */
1862 while (!pdeq_empty(worklist)) {
1863 ir_node *n = pdeq_getl(worklist);
1864 if (irn_not_visited(n))
1865 place_floats_late(n, worklist);
1869 void place_code(ir_graph *irg) {
1871 ir_graph *rem = current_ir_graph;
1873 current_ir_graph = irg;
1875 if (!(get_opt_optimize() && get_opt_global_cse())) return;
1877 /* Handle graph state */
1878 assert(get_irg_phase_state(irg) != phase_building);
1879 if (get_irg_dom_state(irg) != dom_consistent)
1882 if (1 || get_irg_loopinfo_state(irg) != loopinfo_consistent) {
1883 free_loop_information(irg);
1884 construct_backedges(irg);
1887 /* Place all floating nodes as early as possible. This guarantees
1888 a legal code placement. */
1889 worklist = new_pdeq();
1890 place_early(worklist);
1892 /* place_early() invalidates the outs, place_late needs them. */
1893 compute_irg_outs(irg);
1895 /* Now move the nodes down in the dominator tree. This reduces the
1896 unnecessary executions of the node. */
1897 place_late(worklist);
1899 set_irg_outs_inconsistent(current_ir_graph);
1900 set_irg_loopinfo_inconsistent(current_ir_graph);
1902 current_ir_graph = rem;
1906 * Called by walker of remove_critical_cf_edges().
1908 * Place an empty block to an edge between a blocks of multiple
1909 * predecessors and a block of multiple successors.
1912 * @param env Environment of walker. This field is unused and has
1915 static void walk_critical_cf_edges(ir_node *n, void *env) {
1917 ir_node *pre, *block, **in, *jmp;
1919 /* Block has multiple predecessors */
1920 if ((op_Block == get_irn_op(n)) &&
1921 (get_irn_arity(n) > 1)) {
1922 arity = get_irn_arity(n);
1924 if (n == get_irg_end_block(current_ir_graph))
1925 return; /* No use to add a block here. */
1927 for (i=0; i<arity; i++) {
1928 pre = get_irn_n(n, i);
1929 /* Predecessor has multiple successors. Insert new flow edge */
1930 if ((NULL != pre) &&
1931 (op_Proj == get_irn_op(pre)) &&
1932 op_Raise != get_irn_op(skip_Proj(pre))) {
1934 /* set predecessor array for new block */
1935 in = NEW_ARR_D (ir_node *, current_ir_graph->obst, 1);
1936 /* set predecessor of new block */
1938 block = new_Block(1, in);
1939 /* insert new jmp node to new block */
1940 set_cur_block(block);
1943 /* set successor of new block */
1944 set_irn_n(n, i, jmp);
1946 } /* predecessor has multiple successors */
1947 } /* for all predecessors */
1948 } /* n is a block */
1951 void remove_critical_cf_edges(ir_graph *irg) {
1952 if (get_opt_critical_edges())
1953 irg_walk_graph(irg, NULL, walk_critical_cf_edges, NULL);