2 * Copyright (C) 1995-2008 University of Karlsruhe. All right reserved.
4 * This file is part of libFirm.
6 * This file may be distributed and/or modified under the terms of the
7 * GNU General Public License version 2 as published by the Free Software
8 * Foundation and appearing in the file LICENSE.GPL included in the
9 * packaging of this file.
11 * Licensees holding valid libFirm Professional Edition licenses may use
12 * this file in accordance with the libFirm Commercial License.
13 * Agreement provided with the Software.
15 * This file is provided AS IS with NO WARRANTY OF ANY KIND, INCLUDING THE
16 * WARRANTY OF DESIGN, MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22 * @brief Dead node elimination and Procedure Inlining.
23 * @author Michael Beck, Goetz Lindenmaier
32 #include "irgraph_t.h"
35 #include "iroptimize.h"
52 #include "irbackedge_t.h"
53 #include "opt_inline_t.h"
58 #include "analyze_irg_args.h"
59 #include "iredges_t.h"
63 #include "iropt_dbg.h"
65 DEBUG_ONLY(static firm_dbg_module_t *dbg;)
67 /*------------------------------------------------------------------*/
68 /* Routines for dead node elimination / copying garbage collection */
70 /*------------------------------------------------------------------*/
73 * Remember the new node in the old node by using a field all nodes have.
75 #define set_new_node(oldn, newn) set_irn_link(oldn, newn)
78 * Get this new node, before the old node is forgotten.
80 #define get_new_node(oldn) get_irn_link(oldn)
83 * Check if a new node was set.
85 #define has_new_node(n) (get_new_node(n) != NULL)
88 * We use the block_visited flag to mark that we have computed the
89 * number of useful predecessors for this block.
90 * Further we encode the new arity in this flag in the old blocks.
91 * Remembering the arity is useful, as it saves a lot of pointer
92 * accesses. This function is called for all Phi and Block nodes
96 compute_new_arity(ir_node *b) {
97 int i, res, irn_arity;
100 irg_v = get_irg_block_visited(current_ir_graph);
101 block_v = get_Block_block_visited(b);
102 if (block_v >= irg_v) {
103 /* we computed the number of preds for this block and saved it in the
105 return block_v - irg_v;
107 /* compute the number of good predecessors */
108 res = irn_arity = get_irn_arity(b);
109 for (i = 0; i < irn_arity; i++)
110 if (is_Bad(get_irn_n(b, i))) res--;
111 /* save it in the flag. */
112 set_Block_block_visited(b, irg_v + res);
118 * Copies the node to the new obstack. The Ins of the new node point to
119 * the predecessors on the old obstack. For block/phi nodes not all
120 * predecessors might be copied. n->link points to the new node.
121 * For Phi and Block nodes the function allocates in-arrays with an arity
122 * only for useful predecessors. The arity is determined by counting
123 * the non-bad predecessors of the block.
125 * @param n The node to be copied
126 * @param env if non-NULL, the node number attribute will be copied to the new node
128 * Note: Also used for loop unrolling.
130 static void copy_node(ir_node *n, void *env) {
133 ir_op *op = get_irn_op(n);
137 /* node copied already */
139 } else if (op == op_Block) {
141 new_arity = compute_new_arity(n);
142 n->attr.block.graph_arr = NULL;
144 block = get_nodes_block(n);
146 new_arity = compute_new_arity(block);
148 new_arity = get_irn_arity(n);
151 nn = new_ir_node(get_irn_dbg_info(n),
158 /* Copy the attributes. These might point to additional data. If this
159 was allocated on the old obstack the pointers now are dangling. This
160 frees e.g. the memory of the graph_arr allocated in new_immBlock. */
161 if (op == op_Block) {
162 /* we cannot allow blocks WITHOUT macroblock input */
163 set_Block_MacroBlock(nn, get_Block_MacroBlock(n));
165 copy_node_attr(n, nn);
168 /* for easier debugging, we want to copy the node numbers too */
169 nn->node_nr = n->node_nr;
173 hook_dead_node_elim_subst(current_ir_graph, n, nn);
177 * Copies new predecessors of old node to new node remembered in link.
178 * Spare the Bad predecessors of Phi and Block nodes.
180 static void copy_preds(ir_node *n, void *env) {
185 nn = get_new_node(n);
188 /* copy the macro block header */
189 ir_node *mbh = get_Block_MacroBlock(n);
192 /* this block is a macroblock header */
193 set_Block_MacroBlock(nn, nn);
195 /* get the macro block header */
196 ir_node *nmbh = get_new_node(mbh);
197 assert(nmbh != NULL);
198 set_Block_MacroBlock(nn, nmbh);
201 /* Don't copy Bad nodes. */
203 irn_arity = get_irn_arity(n);
204 for (i = 0; i < irn_arity; i++) {
205 if (! is_Bad(get_irn_n(n, i))) {
206 ir_node *pred = get_irn_n(n, i);
207 set_irn_n(nn, j, get_new_node(pred));
211 /* repair the block visited flag from above misuse. Repair it in both
212 graphs so that the old one can still be used. */
213 set_Block_block_visited(nn, 0);
214 set_Block_block_visited(n, 0);
215 /* Local optimization could not merge two subsequent blocks if
216 in array contained Bads. Now it's possible.
217 We don't call optimize_in_place as it requires
218 that the fields in ir_graph are set properly. */
219 if (!has_Block_label(nn) &&
220 get_opt_control_flow_straightening() &&
221 get_Block_n_cfgpreds(nn) == 1 &&
222 is_Jmp(get_Block_cfgpred(nn, 0))) {
223 ir_node *old = get_nodes_block(get_Block_cfgpred(nn, 0));
225 /* Jmp jumps into the block it is in -- deal self cycle. */
226 assert(is_Bad(get_new_node(get_irg_bad(current_ir_graph))));
227 exchange(nn, get_new_node(get_irg_bad(current_ir_graph)));
232 } else if (is_Phi(n) && get_irn_arity(n) > 0) {
233 /* Don't copy node if corresponding predecessor in block is Bad.
234 The Block itself should not be Bad. */
235 block = get_nodes_block(n);
236 set_nodes_block(nn, get_new_node(block));
238 irn_arity = get_irn_arity(n);
239 for (i = 0; i < irn_arity; i++) {
240 if (! is_Bad(get_irn_n(block, i))) {
241 ir_node *pred = get_irn_n(n, i);
242 set_irn_n(nn, j, get_new_node(pred));
243 /*if (is_backedge(n, i)) set_backedge(nn, j);*/
247 /* If the pre walker reached this Phi after the post walker visited the
248 block block_visited is > 0. */
249 set_Block_block_visited(get_nodes_block(n), 0);
250 /* Compacting the Phi's ins might generate Phis with only one
252 if (get_irn_arity(nn) == 1)
253 exchange(nn, get_irn_n(nn, 0));
255 irn_arity = get_irn_arity(n);
256 for (i = -1; i < irn_arity; i++)
257 set_irn_n(nn, i, get_new_node(get_irn_n(n, i)));
259 /* Now the new node is complete. We can add it to the hash table for CSE.
260 @@@ inlining aborts if we identify End. Why? */
262 add_identities(current_ir_graph->value_table, nn);
266 * Copies the graph recursively, compacts the keep-alives of the end node.
268 * @param irg the graph to be copied
269 * @param copy_node_nr If non-zero, the node number will be copied
271 static void copy_graph(ir_graph *irg, int copy_node_nr) {
272 ir_node *oe, *ne, *ob, *nb, *om, *nm; /* old end, new end, old bad, new bad, old NoMem, new NoMem */
273 ir_node *ka; /* keep alive */
277 /* Some nodes must be copied by hand, sigh */
278 vfl = get_irg_visited(irg);
279 set_irg_visited(irg, vfl + 1);
281 oe = get_irg_end(irg);
282 mark_irn_visited(oe);
283 /* copy the end node by hand, allocate dynamic in array! */
284 ne = new_ir_node(get_irn_dbg_info(oe),
291 /* Copy the attributes. Well, there might be some in the future... */
292 copy_node_attr(oe, ne);
293 set_new_node(oe, ne);
295 /* copy the Bad node */
296 ob = get_irg_bad(irg);
297 mark_irn_visited(ob);
298 nb = new_ir_node(get_irn_dbg_info(ob),
305 copy_node_attr(ob, nb);
306 set_new_node(ob, nb);
308 /* copy the NoMem node */
309 om = get_irg_no_mem(irg);
310 mark_irn_visited(om);
311 nm = new_ir_node(get_irn_dbg_info(om),
318 copy_node_attr(om, nm);
319 set_new_node(om, nm);
321 /* copy the live nodes */
322 set_irg_visited(irg, vfl);
323 irg_walk(get_nodes_block(oe), copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
325 /* Note: from yet, the visited flag of the graph is equal to vfl + 1 */
327 /* visit the anchors as well */
328 for (i = get_irg_n_anchors(irg) - 1; i >= 0; --i) {
329 ir_node *n = get_irg_anchor(irg, i);
331 if (n && (get_irn_visited(n) <= vfl)) {
332 set_irg_visited(irg, vfl);
333 irg_walk(n, copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
337 /* copy_preds for the end node ... */
338 set_nodes_block(ne, get_new_node(get_nodes_block(oe)));
340 /*- ... and now the keep alives. -*/
341 /* First pick the not marked block nodes and walk them. We must pick these
342 first as else we will oversee blocks reachable from Phis. */
343 irn_arity = get_End_n_keepalives(oe);
344 for (i = 0; i < irn_arity; i++) {
345 ka = get_End_keepalive(oe, i);
347 if (get_irn_visited(ka) <= vfl) {
348 /* We must keep the block alive and copy everything reachable */
349 set_irg_visited(irg, vfl);
350 irg_walk(ka, copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
352 add_End_keepalive(ne, get_new_node(ka));
356 /* Now pick other nodes. Here we will keep all! */
357 irn_arity = get_End_n_keepalives(oe);
358 for (i = 0; i < irn_arity; i++) {
359 ka = get_End_keepalive(oe, i);
361 if (get_irn_visited(ka) <= vfl) {
362 /* We didn't copy the node yet. */
363 set_irg_visited(irg, vfl);
364 irg_walk(ka, copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
366 add_End_keepalive(ne, get_new_node(ka));
370 /* start block sometimes only reached after keep alives */
371 set_nodes_block(nb, get_new_node(get_nodes_block(ob)));
372 set_nodes_block(nm, get_new_node(get_nodes_block(om)));
376 * Copies the graph reachable from current_ir_graph->end to the obstack
377 * in current_ir_graph and fixes the environment.
378 * Then fixes the fields in current_ir_graph containing nodes of the
381 * @param copy_node_nr If non-zero, the node number will be copied
384 copy_graph_env(int copy_node_nr) {
385 ir_graph *irg = current_ir_graph;
386 ir_node *old_end, *new_anchor;
389 /* remove end_except and end_reg nodes */
390 old_end = get_irg_end(irg);
391 set_irg_end_except (irg, old_end);
392 set_irg_end_reg (irg, old_end);
394 /* Not all nodes remembered in irg might be reachable
395 from the end node. Assure their link is set to NULL, so that
396 we can test whether new nodes have been computed. */
397 for (i = get_irg_n_anchors(irg) - 1; i >= 0; --i) {
398 ir_node *n = get_irg_anchor(irg, i);
400 set_new_node(n, NULL);
402 /* we use the block walk flag for removing Bads from Blocks ins. */
403 inc_irg_block_visited(irg);
406 copy_graph(irg, copy_node_nr);
409 old_end = get_irg_end(irg);
410 new_anchor = new_Anchor(irg);
412 for (i = get_irg_n_anchors(irg) - 1; i >= 0; --i) {
413 ir_node *n = get_irg_anchor(irg, i);
415 set_irn_n(new_anchor, i, get_new_node(n));
418 irg->anchor = new_anchor;
420 /* ensure the new anchor is placed in the endblock */
421 set_nodes_block(new_anchor, get_irg_end_block(irg));
425 * Copies all reachable nodes to a new obstack. Removes bad inputs
426 * from block nodes and the corresponding inputs from Phi nodes.
427 * Merges single exit blocks with single entry blocks and removes
429 * Adds all new nodes to a new hash table for CSE. Does not
430 * perform CSE, so the hash table might contain common subexpressions.
432 void dead_node_elimination(ir_graph *irg) {
434 #ifdef INTERPROCEDURAL_VIEW
435 int rem_ipview = get_interprocedural_view();
437 struct obstack *graveyard_obst = NULL;
438 struct obstack *rebirth_obst = NULL;
440 edges_deactivate(irg);
442 /* inform statistics that we started a dead-node elimination run */
443 hook_dead_node_elim(irg, 1);
445 /* Remember external state of current_ir_graph. */
446 rem = current_ir_graph;
447 current_ir_graph = irg;
448 #ifdef INTERPROCEDURAL_VIEW
449 set_interprocedural_view(0);
452 assert(get_irg_phase_state(irg) != phase_building);
454 /* Handle graph state */
455 free_callee_info(irg);
459 /* @@@ so far we loose loops when copying */
460 free_loop_information(irg);
462 set_irg_doms_inconsistent(irg);
464 /* A quiet place, where the old obstack can rest in peace,
465 until it will be cremated. */
466 graveyard_obst = irg->obst;
468 /* A new obstack, where the reachable nodes will be copied to. */
469 rebirth_obst = XMALLOC(struct obstack);
470 irg->obst = rebirth_obst;
471 obstack_init(irg->obst);
472 irg->last_node_idx = 0;
474 /* We also need a new value table for CSE */
475 del_identities(irg->value_table);
476 irg->value_table = new_identities();
478 /* Copy the graph from the old to the new obstack */
479 copy_graph_env(/*copy_node_nr=*/1);
481 /* Free memory from old unoptimized obstack */
482 obstack_free(graveyard_obst, 0); /* First empty the obstack ... */
483 xfree(graveyard_obst); /* ... then free it. */
485 /* inform statistics that the run is over */
486 hook_dead_node_elim(irg, 0);
488 current_ir_graph = rem;
489 #ifdef INTERPROCEDURAL_VIEW
490 set_interprocedural_view(rem_ipview);
495 * Relink bad predecessors of a block and store the old in array to the
496 * link field. This function is called by relink_bad_predecessors().
497 * The array of link field starts with the block operand at position 0.
498 * If block has bad predecessors, create a new in array without bad preds.
499 * Otherwise let in array untouched.
501 static void relink_bad_block_predecessors(ir_node *n, void *env) {
502 ir_node **new_in, *irn;
503 int i, new_irn_n, old_irn_arity, new_irn_arity = 0;
506 /* if link field of block is NULL, look for bad predecessors otherwise
507 this is already done */
508 if (is_Block(n) && get_irn_link(n) == NULL) {
509 /* save old predecessors in link field (position 0 is the block operand)*/
510 set_irn_link(n, get_irn_in(n));
512 /* count predecessors without bad nodes */
513 old_irn_arity = get_irn_arity(n);
514 for (i = 0; i < old_irn_arity; i++)
515 if (!is_Bad(get_irn_n(n, i)))
518 /* arity changing: set new predecessors without bad nodes */
519 if (new_irn_arity < old_irn_arity) {
520 /* Get new predecessor array. We do not resize the array, as we must
521 keep the old one to update Phis. */
522 new_in = NEW_ARR_D(ir_node *, current_ir_graph->obst, (new_irn_arity+1));
524 /* set new predecessors in array */
527 for (i = 0; i < old_irn_arity; i++) {
528 irn = get_irn_n(n, i);
530 new_in[new_irn_n] = irn;
531 is_backedge(n, i) ? set_backedge(n, new_irn_n-1) : set_not_backedge(n, new_irn_n-1);
535 /* ARR_SETLEN(int, n->attr.block.backedge, new_irn_arity); */
536 ARR_SHRINKLEN(n->attr.block.backedge, new_irn_arity);
538 } /* ir node has bad predecessors */
539 } /* Block is not relinked */
543 * Relinks Bad predecessors from Blocks and Phis called by walker
544 * remove_bad_predecesors(). If n is a Block, call
545 * relink_bad_block_redecessors(). If n is a Phi-node, call also the relinking
546 * function of Phi's Block. If this block has bad predecessors, relink preds
549 static void relink_bad_predecessors(ir_node *n, void *env) {
550 ir_node *block, **old_in;
551 int i, old_irn_arity, new_irn_arity;
553 /* relink bad predecessors of a block */
555 relink_bad_block_predecessors(n, env);
557 /* If Phi node relink its block and its predecessors */
559 /* Relink predecessors of phi's block */
560 block = get_nodes_block(n);
561 if (get_irn_link(block) == NULL)
562 relink_bad_block_predecessors(block, env);
564 old_in = (ir_node **)get_irn_link(block); /* Of Phi's Block */
565 old_irn_arity = ARR_LEN(old_in);
567 /* Relink Phi predecessors if count of predecessors changed */
568 if (old_irn_arity != ARR_LEN(get_irn_in(block))) {
569 /* set new predecessors in array
570 n->in[0] remains the same block */
572 for(i = 1; i < old_irn_arity; i++)
573 if (!is_Bad(old_in[i])) {
574 n->in[new_irn_arity] = n->in[i];
575 is_backedge(n, i) ? set_backedge(n, new_irn_arity) : set_not_backedge(n, new_irn_arity);
579 ARR_SETLEN(ir_node *, n->in, new_irn_arity);
580 ARR_SETLEN(int, n->attr.phi.u.backedge, new_irn_arity);
582 } /* n is a Phi node */
586 * Removes Bad Bad predecessors from Blocks and the corresponding
587 * inputs to Phi nodes as in dead_node_elimination but without
589 * On walking up set the link field to NULL, on walking down call
590 * relink_bad_predecessors() (This function stores the old in array
591 * to the link field and sets a new in array if arity of predecessors
594 void remove_bad_predecessors(ir_graph *irg) {
595 panic("Fix backedge handling first");
596 irg_walk_graph(irg, firm_clear_link, relink_bad_predecessors, NULL);
603 __)|_| | \_/ | \_/(/_ |_/\__|__
605 The following stuff implements a facility that automatically patches
606 registered ir_node pointers to the new node when a dead node elimination occurs.
609 struct _survive_dce_t {
613 hook_entry_t dead_node_elim;
614 hook_entry_t dead_node_elim_subst;
617 typedef struct _survive_dce_list_t {
618 struct _survive_dce_list_t *next;
620 } survive_dce_list_t;
622 static void dead_node_hook(void *context, ir_graph *irg, int start) {
623 survive_dce_t *sd = context;
626 /* Create a new map before the dead node elimination is performed. */
628 sd->new_places = pmap_create_ex(pmap_count(sd->places));
630 /* Patch back all nodes if dead node elimination is over and something is to be done. */
631 pmap_destroy(sd->places);
632 sd->places = sd->new_places;
633 sd->new_places = NULL;
638 * Hook called when dead node elimination replaces old by nw.
640 static void dead_node_subst_hook(void *context, ir_graph *irg, ir_node *old, ir_node *nw) {
641 survive_dce_t *sd = context;
642 survive_dce_list_t *list = pmap_get(sd->places, old);
645 /* If the node is to be patched back, write the new address to all registered locations. */
647 survive_dce_list_t *p;
649 for (p = list; p; p = p->next)
652 pmap_insert(sd->new_places, nw, list);
657 * Make a new Survive DCE environment.
659 survive_dce_t *new_survive_dce(void) {
660 survive_dce_t *res = XMALLOC(survive_dce_t);
661 obstack_init(&res->obst);
662 res->places = pmap_create();
663 res->new_places = NULL;
665 res->dead_node_elim.hook._hook_dead_node_elim = dead_node_hook;
666 res->dead_node_elim.context = res;
667 res->dead_node_elim.next = NULL;
669 res->dead_node_elim_subst.hook._hook_dead_node_elim_subst = dead_node_subst_hook;
670 res->dead_node_elim_subst.context = res;
671 res->dead_node_elim_subst.next = NULL;
673 register_hook(hook_dead_node_elim, &res->dead_node_elim);
674 register_hook(hook_dead_node_elim_subst, &res->dead_node_elim_subst);
679 * Free a Survive DCE environment.
681 void free_survive_dce(survive_dce_t *sd) {
682 obstack_free(&sd->obst, NULL);
683 pmap_destroy(sd->places);
684 unregister_hook(hook_dead_node_elim, &sd->dead_node_elim);
685 unregister_hook(hook_dead_node_elim_subst, &sd->dead_node_elim_subst);
690 * Register a node pointer to be patched upon DCE.
691 * When DCE occurs, the node pointer specified by @p place will be
692 * patched to the new address of the node it is pointing to.
694 * @param sd The Survive DCE environment.
695 * @param place The address of the node pointer.
697 void survive_dce_register_irn(survive_dce_t *sd, ir_node **place) {
698 if (*place != NULL) {
699 ir_node *irn = *place;
700 survive_dce_list_t *curr = pmap_get(sd->places, irn);
701 survive_dce_list_t *nw = obstack_alloc(&sd->obst, sizeof(nw[0]));
706 pmap_insert(sd->places, irn, nw);
710 /*--------------------------------------------------------------------*/
711 /* Functionality for inlining */
712 /*--------------------------------------------------------------------*/
715 * Copy node for inlineing. Updates attributes that change when
716 * inlineing but not for dead node elimination.
718 * Copies the node by calling copy_node() and then updates the entity if
719 * it's a local one. env must be a pointer of the frame type of the
720 * inlined procedure. The new entities must be in the link field of
723 static void copy_node_inline(ir_node *n, void *env) {
725 ir_type *frame_tp = (ir_type *)env;
729 nn = get_new_node (n);
731 if (get_entity_owner(get_Sel_entity(n)) == frame_tp) {
732 set_Sel_entity(nn, get_entity_link(get_Sel_entity(n)));
734 } else if (is_Block(n)) {
735 nn = get_new_node (n);
736 nn->attr.block.irg = current_ir_graph;
741 * Copies new predecessors of old node and move constants to
744 static void copy_preds_inline(ir_node *n, void *env) {
748 nn = skip_Id(get_new_node(n));
749 if (is_irn_constlike(nn)) {
750 /* move Constants into the start block */
751 set_nodes_block(nn, get_irg_start_block(current_ir_graph));
753 n = identify_remember(current_ir_graph->value_table, nn);
762 * Walker: checks if P_value_arg_base is used.
764 static void find_addr(ir_node *node, void *env) {
765 int *allow_inline = env;
767 ir_graph *irg = current_ir_graph;
768 if (get_Sel_ptr(node) == get_irg_frame(irg)) {
769 /* access to frame */
770 ir_entity *ent = get_Sel_entity(node);
771 if (get_entity_owner(ent) != get_irg_frame_type(irg)) {
772 /* access to value_type */
776 } else if (is_Alloc(node) && get_Alloc_where(node) == stack_alloc) {
778 * Refuse to inline alloca call unless user explicitly forced so as this
779 * may change program's memory overhead drastically when the function
780 * using alloca is called in loop. In GCC present in SPEC2000 inlining
781 * into schedule_block cause it to require 2GB of ram instead of 256MB.
783 * Sorrily this is true with our implementation also.
784 * Moreover, we cannot differentiate between alloca() and VLA yet, so this
785 * disables inlining of functions using VLA (with are completely save).
788 * - add a flag to the Alloc node for "real" alloca() calls
789 * - add a new Stack-Restore node at the end of a function using alloca()
796 * Check if we can inline a given call.
797 * Currently, we cannot inline two cases:
798 * - call with compound arguments
799 * - graphs that take the address of a parameter
801 * check these conditions here
803 static int can_inline(ir_node *call, ir_graph *called_graph) {
804 ir_type *call_type = get_Call_type(call);
805 int params, ress, i, res;
806 assert(is_Method_type(call_type));
808 params = get_method_n_params(call_type);
809 ress = get_method_n_ress(call_type);
811 /* check parameters for compound arguments */
812 for (i = 0; i < params; ++i) {
813 ir_type *p_type = get_method_param_type(call_type, i);
815 if (is_compound_type(p_type))
819 /* check results for compound arguments */
820 for (i = 0; i < ress; ++i) {
821 ir_type *r_type = get_method_res_type(call_type, i);
823 if (is_compound_type(r_type))
828 irg_walk_graph(called_graph, find_addr, NULL, &res);
834 exc_handler = 0, /**< There is a handler. */
835 exc_to_end = 1, /**< Branches to End. */
836 exc_no_handler = 2 /**< Exception handling not represented. */
839 /* Inlines a method at the given call site. */
840 int inline_method(ir_node *call, ir_graph *called_graph) {
842 ir_node *post_call, *post_bl;
843 ir_node *in[pn_Start_max];
844 ir_node *end, *end_bl, *block;
849 int arity, n_ret, n_exc, n_res, i, n, j, rem_opt, irn_arity, n_params;
850 enum exc_mode exc_handling;
851 ir_type *called_frame, *curr_frame, *mtp, *ctp;
854 irg_inline_property prop = get_irg_inline_property(called_graph);
855 unsigned long visited;
857 if (prop == irg_inline_forbidden)
860 ent = get_irg_entity(called_graph);
862 mtp = get_entity_type(ent);
863 ctp = get_Call_type(call);
864 if (get_method_n_params(mtp) > get_method_n_params(ctp)) {
865 /* this is a bad feature of C: without a prototype, we can
866 * call a function with less parameters than needed. Currently
867 * we don't support this, although we could use Unknown than. */
871 /* Argh, compiling C has some bad consequences:
872 * It is implementation dependent what happens in that case.
873 * We support inlining, if the bitsize of the types matches AND
874 * the same arithmetic is used. */
875 n_params = get_method_n_params(mtp);
876 for (i = n_params - 1; i >= 0; --i) {
877 ir_type *param_tp = get_method_param_type(mtp, i);
878 ir_type *arg_tp = get_method_param_type(ctp, i);
880 if (param_tp != arg_tp) {
881 ir_mode *pmode = get_type_mode(param_tp);
882 ir_mode *amode = get_type_mode(arg_tp);
884 if (pmode == NULL || amode == NULL)
886 if (get_mode_size_bits(pmode) != get_mode_size_bits(amode))
888 if (get_mode_arithmetic(pmode) != get_mode_arithmetic(amode))
890 /* otherwise we can simply "reinterpret" the bits */
894 irg = get_irn_irg(call);
897 * We cannot inline a recursive call. The graph must be copied before
898 * the call the inline_method() using create_irg_copy().
900 if (called_graph == irg)
904 * currently, we cannot inline two cases:
905 * - call with compound arguments
906 * - graphs that take the address of a parameter
908 if (! can_inline(call, called_graph))
911 rem = current_ir_graph;
912 current_ir_graph = irg;
914 DB((dbg, LEVEL_1, "Inlining %+F(%+F) into %+F\n", call, called_graph, irg));
916 /* -- Turn off optimizations, this can cause problems when allocating new nodes. -- */
917 rem_opt = get_opt_optimize();
920 /* Handle graph state */
921 assert(get_irg_phase_state(irg) != phase_building);
922 assert(get_irg_pinned(irg) == op_pin_state_pinned);
923 assert(get_irg_pinned(called_graph) == op_pin_state_pinned);
924 set_irg_outs_inconsistent(irg);
925 set_irg_extblk_inconsistent(irg);
926 set_irg_doms_inconsistent(irg);
927 set_irg_loopinfo_inconsistent(irg);
928 set_irg_callee_info_state(irg, irg_callee_info_inconsistent);
929 set_irg_entity_usage_state(irg, ir_entity_usage_not_computed);
931 /* -- Check preconditions -- */
932 assert(is_Call(call));
934 /* here we know we WILL inline, so inform the statistics */
935 hook_inline(call, called_graph);
937 /* -- Decide how to handle exception control flow: Is there a handler
938 for the Call node, or do we branch directly to End on an exception?
940 0 There is a handler.
942 2 Exception handling not represented in Firm. -- */
944 ir_node *proj, *Mproj = NULL, *Xproj = NULL;
945 for (proj = get_irn_link(call); proj; proj = get_irn_link(proj)) {
946 long proj_nr = get_Proj_proj(proj);
947 if (proj_nr == pn_Call_X_except) Xproj = proj;
948 if (proj_nr == pn_Call_M_except) Mproj = proj;
950 if (Mproj) { assert(Xproj); exc_handling = exc_handler; } /* Mproj */
951 else if (Xproj) { exc_handling = exc_to_end; } /* !Mproj && Xproj */
952 else { exc_handling = exc_no_handler; } /* !Mproj && !Xproj */
955 /* create the argument tuple */
956 NEW_ARR_A(ir_type *, args_in, n_params);
958 block = get_nodes_block(call);
959 for (i = n_params - 1; i >= 0; --i) {
960 ir_node *arg = get_Call_param(call, i);
961 ir_type *param_tp = get_method_param_type(mtp, i);
962 ir_mode *mode = get_type_mode(param_tp);
964 if (mode != get_irn_mode(arg)) {
965 arg = new_r_Conv(irg, block, arg, mode);
971 the procedure and later replaces the Start node of the called graph.
972 Post_call is the old Call node and collects the results of the called
973 graph. Both will end up being a tuple. -- */
974 post_bl = get_nodes_block(call);
975 set_irg_current_block(irg, post_bl);
976 /* XxMxPxPxPxT of Start + parameter of Call */
977 in[pn_Start_X_initial_exec] = new_Jmp();
978 in[pn_Start_M] = get_Call_mem(call);
979 in[pn_Start_P_frame_base] = get_irg_frame(irg);
980 in[pn_Start_P_tls] = get_irg_tls(irg);
981 in[pn_Start_T_args] = new_Tuple(n_params, args_in);
982 pre_call = new_Tuple(pn_Start_max, in);
986 The new block gets the ins of the old block, pre_call and all its
987 predecessors and all Phi nodes. -- */
988 part_block(pre_call);
990 /* -- Prepare state for dead node elimination -- */
991 /* Visited flags in calling irg must be >= flag in called irg.
992 Else walker and arity computation will not work. */
993 if (get_irg_visited(irg) <= get_irg_visited(called_graph))
994 set_irg_visited(irg, get_irg_visited(called_graph) + 1);
995 if (get_irg_block_visited(irg) < get_irg_block_visited(called_graph))
996 set_irg_block_visited(irg, get_irg_block_visited(called_graph));
997 visited = get_irg_visited(irg);
999 /* Set pre_call as new Start node in link field of the start node of
1000 calling graph and pre_calls block as new block for the start block
1002 Further mark these nodes so that they are not visited by the
1004 set_irn_link(get_irg_start(called_graph), pre_call);
1005 set_irn_visited(get_irg_start(called_graph), visited);
1006 set_irn_link(get_irg_start_block(called_graph), get_nodes_block(pre_call));
1007 set_irn_visited(get_irg_start_block(called_graph), visited);
1009 set_irn_link(get_irg_bad(called_graph), get_irg_bad(current_ir_graph));
1010 set_irn_visited(get_irg_bad(called_graph), visited);
1012 set_irn_link(get_irg_no_mem(called_graph), get_irg_no_mem(current_ir_graph));
1013 set_irn_visited(get_irg_no_mem(called_graph), visited);
1015 /* Initialize for compaction of in arrays */
1016 inc_irg_block_visited(irg);
1018 /* -- Replicate local entities of the called_graph -- */
1019 /* copy the entities. */
1020 irp_reserve_resources(irp, IR_RESOURCE_ENTITY_LINK);
1021 called_frame = get_irg_frame_type(called_graph);
1022 curr_frame = get_irg_frame_type(irg);
1023 for (i = 0, n = get_class_n_members(called_frame); i < n; ++i) {
1024 ir_entity *new_ent, *old_ent;
1025 old_ent = get_class_member(called_frame, i);
1026 new_ent = copy_entity_own(old_ent, curr_frame);
1027 set_entity_link(old_ent, new_ent);
1030 /* visited is > than that of called graph. With this trick visited will
1031 remain unchanged so that an outer walker, e.g., searching the call nodes
1032 to inline, calling this inline will not visit the inlined nodes. */
1033 set_irg_visited(irg, get_irg_visited(irg)-1);
1035 /* -- Performing dead node elimination inlines the graph -- */
1036 /* Copies the nodes to the obstack of current_ir_graph. Updates links to new
1038 irg_walk(get_irg_end(called_graph), copy_node_inline, copy_preds_inline,
1039 get_irg_frame_type(called_graph));
1041 irp_free_resources(irp, IR_RESOURCE_ENTITY_LINK);
1043 /* Repair called_graph */
1044 set_irg_visited(called_graph, get_irg_visited(irg));
1045 set_irg_block_visited(called_graph, get_irg_block_visited(irg));
1046 set_Block_block_visited(get_irg_start_block(called_graph), 0);
1048 /* -- Merge the end of the inlined procedure with the call site -- */
1049 /* We will turn the old Call node into a Tuple with the following
1052 0: Phi of all Memories of Return statements.
1053 1: Jmp from new Block that merges the control flow from all exception
1054 predecessors of the old end block.
1055 2: Tuple of all arguments.
1056 3: Phi of Exception memories.
1057 In case the old Call directly branches to End on an exception we don't
1058 need the block merging all exceptions nor the Phi of the exception
1062 /* -- Precompute some values -- */
1063 end_bl = get_new_node(get_irg_end_block(called_graph));
1064 end = get_new_node(get_irg_end(called_graph));
1065 arity = get_Block_n_cfgpreds(end_bl); /* arity = n_exc + n_ret */
1066 n_res = get_method_n_ress(get_Call_type(call));
1068 res_pred = XMALLOCN(ir_node*, n_res);
1069 cf_pred = XMALLOCN(ir_node*, arity);
1071 set_irg_current_block(irg, post_bl); /* just to make sure */
1073 /* -- archive keepalives -- */
1074 irn_arity = get_irn_arity(end);
1075 for (i = 0; i < irn_arity; i++) {
1076 ir_node *ka = get_End_keepalive(end, i);
1078 add_End_keepalive(get_irg_end(irg), ka);
1081 /* The new end node will die. We need not free as the in array is on the obstack:
1082 copy_node() only generated 'D' arrays. */
1084 /* -- Replace Return nodes by Jump nodes. -- */
1086 for (i = 0; i < arity; i++) {
1088 ret = get_Block_cfgpred(end_bl, i);
1089 if (is_Return(ret)) {
1090 cf_pred[n_ret] = new_r_Jmp(irg, get_nodes_block(ret));
1094 set_irn_in(post_bl, n_ret, cf_pred);
1096 /* -- Build a Tuple for all results of the method.
1097 Add Phi node if there was more than one Return. -- */
1098 turn_into_tuple(post_call, pn_Call_max);
1099 /* First the Memory-Phi */
1101 for (i = 0; i < arity; i++) {
1102 ret = get_Block_cfgpred(end_bl, i);
1103 if (is_Return(ret)) {
1104 cf_pred[n_ret] = get_Return_mem(ret);
1108 phi = new_Phi(n_ret, cf_pred, mode_M);
1109 set_Tuple_pred(call, pn_Call_M_regular, phi);
1110 /* Conserve Phi-list for further inlinings -- but might be optimized */
1111 if (get_nodes_block(phi) == post_bl) {
1112 set_irn_link(phi, get_irn_link(post_bl));
1113 set_irn_link(post_bl, phi);
1115 /* Now the real results */
1117 for (j = 0; j < n_res; j++) {
1119 for (i = 0; i < arity; i++) {
1120 ret = get_Block_cfgpred(end_bl, i);
1121 if (is_Return(ret)) {
1122 cf_pred[n_ret] = get_Return_res(ret, j);
1127 phi = new_Phi(n_ret, cf_pred, get_irn_mode(cf_pred[0]));
1131 /* Conserve Phi-list for further inlinings -- but might be optimized */
1132 if (get_nodes_block(phi) == post_bl) {
1133 set_Phi_next(phi, get_Block_phis(post_bl));
1134 set_Block_phis(post_bl, phi);
1137 set_Tuple_pred(call, pn_Call_T_result, new_Tuple(n_res, res_pred));
1139 set_Tuple_pred(call, pn_Call_T_result, new_Bad());
1141 /* handle the regular call */
1142 set_Tuple_pred(call, pn_Call_X_regular, new_Jmp());
1144 /* For now, we cannot inline calls with value_base */
1145 set_Tuple_pred(call, pn_Call_P_value_res_base, new_Bad());
1147 /* Finally the exception control flow.
1148 We have two (three) possible situations:
1149 First if the Call branches to an exception handler: We need to add a Phi node to
1150 collect the memory containing the exception objects. Further we need
1151 to add another block to get a correct representation of this Phi. To
1152 this block we add a Jmp that resolves into the X output of the Call
1153 when the Call is turned into a tuple.
1154 Second the Call branches to End, the exception is not handled. Just
1155 add all inlined exception branches to the End node.
1156 Third: there is no Exception edge at all. Handle as case two. */
1157 if (exc_handling == exc_handler) {
1159 for (i = 0; i < arity; i++) {
1161 ret = get_Block_cfgpred(end_bl, i);
1162 irn = skip_Proj(ret);
1163 if (is_fragile_op(irn) || is_Raise(irn)) {
1164 cf_pred[n_exc] = ret;
1169 ir_node *block = new_Block(n_exc, cf_pred);
1170 set_cur_block(block);
1172 set_Tuple_pred(call, pn_Call_X_except, new_Jmp());
1173 /* The Phi for the memories with the exception objects */
1175 for (i = 0; i < arity; i++) {
1177 ret = skip_Proj(get_Block_cfgpred(end_bl, i));
1179 cf_pred[n_exc] = new_r_Proj(irg, get_nodes_block(ret), ret, mode_M, 3);
1181 } else if (is_fragile_op(ret)) {
1182 /* We rely that all cfops have the memory output at the same position. */
1183 cf_pred[n_exc] = new_r_Proj(irg, get_nodes_block(ret), ret, mode_M, 0);
1185 } else if (is_Raise(ret)) {
1186 cf_pred[n_exc] = new_r_Proj(irg, get_nodes_block(ret), ret, mode_M, 1);
1190 set_Tuple_pred(call, pn_Call_M_except, new_Phi(n_exc, cf_pred, mode_M));
1192 set_Tuple_pred(call, pn_Call_X_except, new_Bad());
1193 set_Tuple_pred(call, pn_Call_M_except, new_Bad());
1196 ir_node *main_end_bl;
1197 int main_end_bl_arity;
1198 ir_node **end_preds;
1200 /* assert(exc_handling == 1 || no exceptions. ) */
1202 for (i = 0; i < arity; i++) {
1203 ir_node *ret = get_Block_cfgpred(end_bl, i);
1204 ir_node *irn = skip_Proj(ret);
1206 if (is_fragile_op(irn) || is_Raise(irn)) {
1207 cf_pred[n_exc] = ret;
1211 main_end_bl = get_irg_end_block(irg);
1212 main_end_bl_arity = get_irn_arity(main_end_bl);
1213 end_preds = XMALLOCN(ir_node*, n_exc + main_end_bl_arity);
1215 for (i = 0; i < main_end_bl_arity; ++i)
1216 end_preds[i] = get_irn_n(main_end_bl, i);
1217 for (i = 0; i < n_exc; ++i)
1218 end_preds[main_end_bl_arity + i] = cf_pred[i];
1219 set_irn_in(main_end_bl, n_exc + main_end_bl_arity, end_preds);
1220 set_Tuple_pred(call, pn_Call_X_except, new_Bad());
1221 set_Tuple_pred(call, pn_Call_M_except, new_Bad());
1227 /* -- Turn CSE back on. -- */
1228 set_optimize(rem_opt);
1229 current_ir_graph = rem;
1234 /********************************************************************/
1235 /* Apply inlining to small methods. */
1236 /********************************************************************/
1238 static struct obstack temp_obst;
1240 /** Represents a possible inlinable call in a graph. */
1241 typedef struct _call_entry {
1242 ir_node *call; /**< The Call node. */
1243 ir_graph *callee; /**< The callee IR-graph. */
1244 list_head list; /**< List head for linking the next one. */
1245 int loop_depth; /**< The loop depth of this call. */
1246 int benefice; /**< The calculated benefice of this call. */
1247 unsigned local_adr:1; /**< Set if this call gets an address of a local variable. */
1248 unsigned all_const:1; /**< Set if this call has only constant parameters. */
1252 * environment for inlining small irgs
1254 typedef struct _inline_env_t {
1255 struct obstack obst; /**< An obstack where call_entries are allocated on. */
1256 list_head calls; /**< The call entry list. */
1260 * Returns the irg called from a Call node. If the irg is not
1261 * known, NULL is returned.
1263 * @param call the call node
1265 static ir_graph *get_call_called_irg(ir_node *call) {
1268 addr = get_Call_ptr(call);
1269 if (is_Global(addr)) {
1270 ir_entity *ent = get_Global_entity(addr);
1271 return get_entity_irg(ent);
1278 * Walker: Collect all calls to known graphs inside a graph.
1280 static void collect_calls(ir_node *call, void *env) {
1282 if (is_Call(call)) {
1283 ir_graph *called_irg = get_call_called_irg(call);
1285 if (called_irg != NULL) {
1286 /* The Call node calls a locally defined method. Remember to inline. */
1287 inline_env_t *ienv = env;
1288 call_entry *entry = obstack_alloc(&ienv->obst, sizeof(*entry));
1290 entry->callee = called_irg;
1291 entry->loop_depth = 0;
1292 entry->benefice = 0;
1293 entry->local_adr = 0;
1294 entry->all_const = 0;
1296 list_add_tail(&entry->list, &ienv->calls);
1302 * Inlines all small methods at call sites where the called address comes
1303 * from a Const node that references the entity representing the called
1305 * The size argument is a rough measure for the code size of the method:
1306 * Methods where the obstack containing the firm graph is smaller than
1309 void inline_small_irgs(ir_graph *irg, int size) {
1310 ir_graph *rem = current_ir_graph;
1314 current_ir_graph = irg;
1315 /* Handle graph state */
1316 assert(get_irg_phase_state(irg) != phase_building);
1317 free_callee_info(irg);
1319 /* Find Call nodes to inline.
1320 (We can not inline during a walk of the graph, as inlining the same
1321 method several times changes the visited flag of the walked graph:
1322 after the first inlining visited of the callee equals visited of
1323 the caller. With the next inlining both are increased.) */
1324 obstack_init(&env.obst);
1325 INIT_LIST_HEAD(&env.calls);
1326 irg_walk_graph(irg, NULL, collect_calls, &env);
1328 if (! list_empty(&env.calls)) {
1329 /* There are calls to inline */
1330 ir_reserve_resources(irg, IR_RESOURCE_IRN_LINK|IR_RESOURCE_PHI_LIST);
1331 collect_phiprojs(irg);
1333 list_for_each_entry(call_entry, entry, &env.calls, list) {
1334 ir_graph *callee = entry->callee;
1335 irg_inline_property prop = get_irg_inline_property(callee);
1337 if (prop == irg_inline_forbidden || get_irg_additional_properties(callee) & mtp_property_weak) {
1338 /* do not inline forbidden / weak graphs */
1342 if (prop >= irg_inline_forced ||
1343 _obstack_memory_used(callee->obst) - (int)obstack_room(callee->obst) < size) {
1344 inline_method(entry->call, callee);
1347 ir_free_resources(irg, IR_RESOURCE_IRN_LINK|IR_RESOURCE_PHI_LIST);
1349 obstack_free(&env.obst, NULL);
1350 current_ir_graph = rem;
1354 * Environment for inlining irgs.
1357 list_head calls; /**< List of of all call nodes in this graph. */
1358 unsigned *local_weights; /**< Once allocated, the beneficial weight for transmitting local addresses. */
1359 unsigned n_nodes; /**< Number of nodes in graph except Id, Tuple, Proj, Start, End. */
1360 unsigned n_blocks; /**< Number of Blocks in graph without Start and End block. */
1361 unsigned n_nodes_orig; /**< for statistics */
1362 unsigned n_call_nodes; /**< Number of Call nodes in the graph. */
1363 unsigned n_call_nodes_orig; /**< for statistics */
1364 unsigned n_callers; /**< Number of known graphs that call this graphs. */
1365 unsigned n_callers_orig; /**< for statistics */
1366 unsigned got_inline:1; /**< Set, if at least one call inside this graph was inlined. */
1367 unsigned local_vars:1; /**< Set, if an inlined function got the address of a local variable. */
1368 unsigned recursive:1; /**< Set, if this function is self recursive. */
1372 * Allocate a new environment for inlining.
1374 static inline_irg_env *alloc_inline_irg_env(void) {
1375 inline_irg_env *env = obstack_alloc(&temp_obst, sizeof(*env));
1376 INIT_LIST_HEAD(&env->calls);
1377 env->local_weights = NULL;
1378 env->n_nodes = -2; /* do not count count Start, End */
1379 env->n_blocks = -2; /* do not count count Start, End Block */
1380 env->n_nodes_orig = -2; /* do not count Start, End */
1381 env->n_call_nodes = 0;
1382 env->n_call_nodes_orig = 0;
1384 env->n_callers_orig = 0;
1385 env->got_inline = 0;
1386 env->local_vars = 0;
1391 typedef struct walker_env {
1392 inline_irg_env *x; /**< the inline environment */
1393 char ignore_runtime; /**< the ignore runtime flag */
1394 char ignore_callers; /**< if set, do change callers data */
1398 * post-walker: collect all calls in the inline-environment
1399 * of a graph and sum some statistics.
1401 static void collect_calls2(ir_node *call, void *ctx) {
1403 inline_irg_env *x = env->x;
1404 ir_opcode code = get_irn_opcode(call);
1408 /* count meaningful nodes in irg */
1409 if (code != iro_Proj && code != iro_Tuple && code != iro_Sync) {
1410 if (code != iro_Block) {
1418 if (code != iro_Call) return;
1420 /* check, if it's a runtime call */
1421 if (env->ignore_runtime) {
1422 ir_node *symc = get_Call_ptr(call);
1424 if (is_Global(symc)) {
1425 ir_entity *ent = get_Global_entity(symc);
1427 if (get_entity_additional_properties(ent) & mtp_property_runtime)
1432 /* collect all call nodes */
1434 ++x->n_call_nodes_orig;
1436 callee = get_call_called_irg(call);
1437 if (callee != NULL) {
1438 if (! env->ignore_callers) {
1439 inline_irg_env *callee_env = get_irg_link(callee);
1440 /* count all static callers */
1441 ++callee_env->n_callers;
1442 ++callee_env->n_callers_orig;
1444 if (callee == current_ir_graph)
1447 /* link it in the list of possible inlinable entries */
1448 entry = obstack_alloc(&temp_obst, sizeof(*entry));
1450 entry->callee = callee;
1451 entry->loop_depth = get_irn_loop(get_nodes_block(call))->depth;
1452 entry->benefice = 0;
1453 entry->local_adr = 0;
1454 entry->all_const = 0;
1456 list_add_tail(&entry->list, &x->calls);
1461 * Returns TRUE if the number of callers is 0 in the irg's environment,
1462 * hence this irg is a leave.
1464 inline static int is_leave(ir_graph *irg) {
1465 inline_irg_env *env = get_irg_link(irg);
1466 return env->n_call_nodes == 0;
1470 * Returns TRUE if the number of nodes in the callee is
1471 * smaller then size in the irg's environment.
1473 inline static int is_smaller(ir_graph *callee, unsigned size) {
1474 inline_irg_env *env = get_irg_link(callee);
1475 return env->n_nodes < size;
1479 * Duplicate a call entry.
1481 * @param entry the original entry to duplicate
1482 * @param new_call the new call node
1483 * @param loop_depth_delta
1484 * delta value for the loop depth
1486 static call_entry *duplicate_call_entry(const call_entry *entry,
1487 ir_node *new_call, int loop_depth_delta) {
1488 call_entry *nentry = obstack_alloc(&temp_obst, sizeof(*nentry));
1489 nentry->call = new_call;
1490 nentry->callee = entry->callee;
1491 nentry->benefice = entry->benefice;
1492 nentry->loop_depth = entry->loop_depth + loop_depth_delta;
1493 nentry->local_adr = entry->local_adr;
1494 nentry->all_const = entry->all_const;
1500 * Append all call nodes of the source environment to the nodes of in the destination
1503 * @param dst destination environment
1504 * @param src source environment
1505 * @param loop_depth the loop depth of the call that is replaced by the src list
1507 static void append_call_list(inline_irg_env *dst, inline_irg_env *src, int loop_depth) {
1508 call_entry *entry, *nentry;
1510 /* Note that the src list points to Call nodes in the inlined graph, but
1511 we need Call nodes in our graph. Luckily the inliner leaves this information
1512 in the link field. */
1513 list_for_each_entry(call_entry, entry, &src->calls, list) {
1514 nentry = duplicate_call_entry(entry, get_irn_link(entry->call), loop_depth);
1515 list_add_tail(&nentry->list, &dst->calls);
1517 dst->n_call_nodes += src->n_call_nodes;
1518 dst->n_nodes += src->n_nodes;
1522 * Inlines small leave methods at call sites where the called address comes
1523 * from a Const node that references the entity representing the called
1525 * The size argument is a rough measure for the code size of the method:
1526 * Methods where the obstack containing the firm graph is smaller than
1529 void inline_leave_functions(unsigned maxsize, unsigned leavesize,
1530 unsigned size, int ignore_runtime)
1532 inline_irg_env *env;
1538 call_entry *entry, *next;
1539 const call_entry *centry;
1540 pmap *copied_graphs;
1541 pmap_entry *pm_entry;
1543 rem = current_ir_graph;
1544 obstack_init(&temp_obst);
1546 /* a map for the copied graphs, used to inline recursive calls */
1547 copied_graphs = pmap_create();
1549 /* extend all irgs by a temporary data structure for inlining. */
1550 n_irgs = get_irp_n_irgs();
1551 for (i = 0; i < n_irgs; ++i)
1552 set_irg_link(get_irp_irg(i), alloc_inline_irg_env());
1554 /* Pre-compute information in temporary data structure. */
1555 wenv.ignore_runtime = ignore_runtime;
1556 wenv.ignore_callers = 0;
1557 for (i = 0; i < n_irgs; ++i) {
1558 ir_graph *irg = get_irp_irg(i);
1560 assert(get_irg_phase_state(irg) != phase_building);
1561 free_callee_info(irg);
1563 assure_cf_loop(irg);
1564 wenv.x = get_irg_link(irg);
1565 irg_walk_graph(irg, NULL, collect_calls2, &wenv);
1568 /* -- and now inline. -- */
1570 /* Inline leaves recursively -- we might construct new leaves. */
1574 for (i = 0; i < n_irgs; ++i) {
1576 int phiproj_computed = 0;
1578 current_ir_graph = get_irp_irg(i);
1579 env = get_irg_link(current_ir_graph);
1581 ir_reserve_resources(current_ir_graph, IR_RESOURCE_IRN_LINK|IR_RESOURCE_PHI_LIST);
1582 list_for_each_entry_safe(call_entry, entry, next, &env->calls, list) {
1584 irg_inline_property prop;
1586 if (env->n_nodes > maxsize)
1590 callee = entry->callee;
1592 prop = get_irg_inline_property(callee);
1593 if (prop == irg_inline_forbidden || get_irg_additional_properties(callee) & mtp_property_weak) {
1594 /* do not inline forbidden / weak graphs */
1598 if (is_leave(callee) && (
1599 is_smaller(callee, leavesize) || prop >= irg_inline_forced)) {
1600 if (!phiproj_computed) {
1601 phiproj_computed = 1;
1602 collect_phiprojs(current_ir_graph);
1604 did_inline = inline_method(call, callee);
1607 inline_irg_env *callee_env = get_irg_link(callee);
1609 /* call was inlined, Phi/Projs for current graph must be recomputed */
1610 phiproj_computed = 0;
1612 /* Do some statistics */
1613 env->got_inline = 1;
1614 --env->n_call_nodes;
1615 env->n_nodes += callee_env->n_nodes;
1616 --callee_env->n_callers;
1618 /* remove this call from the list */
1619 list_del(&entry->list);
1624 ir_free_resources(current_ir_graph, IR_RESOURCE_IRN_LINK|IR_RESOURCE_PHI_LIST);
1626 } while (did_inline);
1628 /* inline other small functions. */
1629 for (i = 0; i < n_irgs; ++i) {
1631 int phiproj_computed = 0;
1633 current_ir_graph = get_irp_irg(i);
1634 env = get_irg_link(current_ir_graph);
1636 ir_reserve_resources(current_ir_graph, IR_RESOURCE_IRN_LINK|IR_RESOURCE_PHI_LIST);
1638 /* note that the list of possible calls is updated during the process */
1639 list_for_each_entry_safe(call_entry, entry, next, &env->calls, list) {
1640 irg_inline_property prop;
1645 callee = entry->callee;
1647 prop = get_irg_inline_property(callee);
1648 if (prop == irg_inline_forbidden || get_irg_additional_properties(callee) & mtp_property_weak) {
1649 /* do not inline forbidden / weak graphs */
1653 e = pmap_find(copied_graphs, callee);
1656 * Remap callee if we have a copy.
1657 * FIXME: Should we do this only for recursive Calls ?
1662 if (prop >= irg_inline_forced ||
1663 (is_smaller(callee, size) && env->n_nodes < maxsize) /* small function */) {
1664 if (current_ir_graph == callee) {
1666 * Recursive call: we cannot directly inline because we cannot walk
1667 * the graph and change it. So we have to make a copy of the graph
1671 inline_irg_env *callee_env;
1674 ir_free_resources(current_ir_graph, IR_RESOURCE_IRN_LINK|IR_RESOURCE_PHI_LIST);
1677 * No copy yet, create one.
1678 * Note that recursive methods are never leaves, so it is sufficient
1679 * to test this condition here.
1681 copy = create_irg_copy(callee);
1683 /* create_irg_copy() destroys the Proj links, recompute them */
1684 phiproj_computed = 0;
1686 ir_reserve_resources(current_ir_graph, IR_RESOURCE_IRN_LINK|IR_RESOURCE_PHI_LIST);
1688 /* allocate new environment */
1689 callee_env = alloc_inline_irg_env();
1690 set_irg_link(copy, callee_env);
1692 assure_cf_loop(copy);
1693 wenv.x = callee_env;
1694 wenv.ignore_callers = 1;
1695 irg_walk_graph(copy, NULL, collect_calls2, &wenv);
1698 * Enter the entity of the original graph. This is needed
1699 * for inline_method(). However, note that ent->irg still points
1700 * to callee, NOT to copy.
1702 set_irg_entity(copy, get_irg_entity(callee));
1704 pmap_insert(copied_graphs, callee, copy);
1707 /* we have only one caller: the original graph */
1708 callee_env->n_callers = 1;
1709 callee_env->n_callers_orig = 1;
1711 if (! phiproj_computed) {
1712 phiproj_computed = 1;
1713 collect_phiprojs(current_ir_graph);
1715 did_inline = inline_method(call, callee);
1717 inline_irg_env *callee_env = (inline_irg_env *)get_irg_link(callee);
1719 /* call was inlined, Phi/Projs for current graph must be recomputed */
1720 phiproj_computed = 0;
1722 /* callee was inline. Append it's call list. */
1723 env->got_inline = 1;
1724 --env->n_call_nodes;
1725 append_call_list(env, callee_env, entry->loop_depth);
1726 --callee_env->n_callers;
1728 /* after we have inlined callee, all called methods inside callee
1729 are now called once more */
1730 list_for_each_entry(call_entry, centry, &callee_env->calls, list) {
1731 inline_irg_env *penv = get_irg_link(centry->callee);
1735 /* remove this call from the list */
1736 list_del(&entry->list);
1741 ir_free_resources(current_ir_graph, IR_RESOURCE_IRN_LINK|IR_RESOURCE_PHI_LIST);
1744 for (i = 0; i < n_irgs; ++i) {
1745 irg = get_irp_irg(i);
1746 env = get_irg_link(irg);
1748 if (env->got_inline) {
1749 optimize_graph_df(irg);
1752 if (env->got_inline || (env->n_callers_orig != env->n_callers)) {
1753 DB((dbg, LEVEL_1, "Nodes:%3d ->%3d, calls:%3d ->%3d, callers:%3d ->%3d, -- %s\n",
1754 env->n_nodes_orig, env->n_nodes, env->n_call_nodes_orig, env->n_call_nodes,
1755 env->n_callers_orig, env->n_callers,
1756 get_entity_name(get_irg_entity(irg))));
1760 /* kill the copied graphs: we don't need them anymore */
1761 foreach_pmap(copied_graphs, pm_entry) {
1762 ir_graph *copy = pm_entry->value;
1764 /* reset the entity, otherwise it will be deleted in the next step ... */
1765 set_irg_entity(copy, NULL);
1766 free_ir_graph(copy);
1768 pmap_destroy(copied_graphs);
1770 obstack_free(&temp_obst, NULL);
1771 current_ir_graph = rem;
1775 * Calculate the parameter weights for transmitting the address of a local variable.
1777 static unsigned calc_method_local_weight(ir_node *arg) {
1779 unsigned v, weight = 0;
1781 for (i = get_irn_n_outs(arg) - 1; i >= 0; --i) {
1782 ir_node *succ = get_irn_out(arg, i);
1784 switch (get_irn_opcode(succ)) {
1787 /* Loads and Store can be removed */
1791 /* check if all args are constant */
1792 for (j = get_Sel_n_indexs(succ) - 1; j >= 0; --j) {
1793 ir_node *idx = get_Sel_index(succ, j);
1794 if (! is_Const(idx))
1797 /* Check users on this Sel. Note: if a 0 is returned here, there was
1798 some unsupported node. */
1799 v = calc_method_local_weight(succ);
1802 /* we can kill one Sel with constant indexes, this is cheap */
1806 /* when looking backward we might find Id nodes */
1807 weight += calc_method_local_weight(succ);
1810 /* unoptimized tuple */
1811 for (j = get_Tuple_n_preds(succ) - 1; j >= 0; --j) {
1812 ir_node *pred = get_Tuple_pred(succ, j);
1814 /* look for Proj(j) */
1815 for (k = get_irn_n_outs(succ) - 1; k >= 0; --k) {
1816 ir_node *succ_succ = get_irn_out(succ, k);
1817 if (is_Proj(succ_succ)) {
1818 if (get_Proj_proj(succ_succ) == j) {
1820 weight += calc_method_local_weight(succ_succ);
1823 /* this should NOT happen */
1831 /* any other node: unsupported yet or bad. */
1839 * Calculate the parameter weights for transmitting the address of a local variable.
1841 static void analyze_irg_local_weights(inline_irg_env *env, ir_graph *irg) {
1842 ir_entity *ent = get_irg_entity(irg);
1844 int nparams, i, proj_nr;
1845 ir_node *irg_args, *arg;
1847 mtp = get_entity_type(ent);
1848 nparams = get_method_n_params(mtp);
1850 /* allocate a new array. currently used as 'analysed' flag */
1851 env->local_weights = NEW_ARR_D(unsigned, &temp_obst, nparams);
1853 /* If the method haven't parameters we have nothing to do. */
1857 assure_irg_outs(irg);
1858 irg_args = get_irg_args(irg);
1859 for (i = get_irn_n_outs(irg_args) - 1; i >= 0; --i) {
1860 arg = get_irn_out(irg_args, i);
1861 proj_nr = get_Proj_proj(arg);
1862 env->local_weights[proj_nr] = calc_method_local_weight(arg);
1867 * Calculate the benefice for transmitting an local variable address.
1868 * After inlining, the local variable might be transformed into a
1869 * SSA variable by scalar_replacement().
1871 static unsigned get_method_local_adress_weight(ir_graph *callee, int pos) {
1872 inline_irg_env *env = get_irg_link(callee);
1874 if (env->local_weights != NULL) {
1875 if (pos < ARR_LEN(env->local_weights))
1876 return env->local_weights[pos];
1880 analyze_irg_local_weights(env, callee);
1882 if (pos < ARR_LEN(env->local_weights))
1883 return env->local_weights[pos];
1888 * Calculate a benefice value for inlining the given call.
1890 * @param call the call node we have to inspect
1891 * @param callee the called graph
1893 static int calc_inline_benefice(call_entry *entry, ir_graph *callee)
1895 ir_node *call = entry->call;
1896 ir_entity *ent = get_irg_entity(callee);
1900 int i, n_params, all_const;
1902 irg_inline_property prop;
1904 inline_irg_env *callee_env;
1906 prop = get_irg_inline_property(callee);
1907 if (prop == irg_inline_forbidden) {
1908 DB((dbg, LEVEL_2, "In %+F Call to %+F: inlining forbidden\n",
1910 return entry->benefice = INT_MIN;
1913 if (get_irg_additional_properties(callee) & (mtp_property_noreturn | mtp_property_weak)) {
1914 DB((dbg, LEVEL_2, "In %+F Call to %+F: not inlining noreturn or weak\n",
1916 return entry->benefice = INT_MIN;
1919 /* costs for every passed parameter */
1920 n_params = get_Call_n_params(call);
1921 mtp = get_entity_type(ent);
1922 cc = get_method_calling_convention(mtp);
1923 if (cc & cc_reg_param) {
1924 /* register parameter, smaller costs for register parameters */
1925 int max_regs = cc & ~cc_bits;
1927 if (max_regs < n_params)
1928 weight += max_regs * 2 + (n_params - max_regs) * 5;
1930 weight += n_params * 2;
1932 /* parameters are passed an stack */
1933 weight += 5 * n_params;
1936 /* constant parameters improve the benefice */
1937 frame_ptr = get_irg_frame(current_ir_graph);
1939 for (i = 0; i < n_params; ++i) {
1940 ir_node *param = get_Call_param(call, i);
1942 if (is_Const(param)) {
1943 weight += get_method_param_weight(ent, i);
1946 if (is_SymConst(param))
1947 weight += get_method_param_weight(ent, i);
1948 else if (is_Sel(param) && get_Sel_ptr(param) == frame_ptr) {
1950 * An address of a local variable is transmitted. After
1951 * inlining, scalar_replacement might be able to remove the
1952 * local variable, so honor this.
1954 v = get_method_local_adress_weight(callee, i);
1957 entry->local_adr = 1;
1961 entry->all_const = all_const;
1963 callee_env = get_irg_link(callee);
1964 if (callee_env->n_callers == 1 &&
1965 callee != current_ir_graph &&
1966 get_entity_visibility(ent) == visibility_local) {
1970 /* give a bonus for functions with one block */
1971 if (callee_env->n_blocks == 1)
1972 weight = weight * 3 / 2;
1974 /* and one for small non-recursive functions: we want them to be inlined in mostly every case */
1975 if (callee_env->n_nodes < 30 && !callee_env->recursive)
1978 /* and finally for leaves: they do not increase the register pressure
1979 because of callee safe registers */
1980 if (callee_env->n_call_nodes == 0)
1983 /** it's important to inline inner loops first */
1984 if (entry->loop_depth > 30)
1985 weight += 30 * 1024;
1987 weight += entry->loop_depth * 1024;
1990 * All arguments constant is probably a good sign, give an extra bonus
1995 return entry->benefice = weight;
1998 static ir_graph **irgs;
1999 static int last_irg;
2002 * Callgraph walker, collect all visited graphs.
2004 static void callgraph_walker(ir_graph *irg, void *data) {
2006 irgs[last_irg++] = irg;
2010 * Creates an inline order for all graphs.
2012 * @return the list of graphs.
2014 static ir_graph **create_irg_list(void) {
2015 ir_entity **free_methods;
2017 int n_irgs = get_irp_n_irgs();
2019 cgana(&arr_len, &free_methods);
2020 xfree(free_methods);
2022 compute_callgraph();
2025 irgs = XMALLOCNZ(ir_graph*, n_irgs);
2027 callgraph_walk(NULL, callgraph_walker, NULL);
2028 assert(n_irgs == last_irg);
2034 * Push a call onto the priority list if its benefice is big enough.
2036 * @param pqueue the priority queue of calls
2037 * @param call the call entry
2038 * @param inlien_threshold
2039 * the threshold value
2041 static void maybe_push_call(pqueue_t *pqueue, call_entry *call,
2042 int inline_threshold)
2044 ir_graph *callee = call->callee;
2045 irg_inline_property prop = get_irg_inline_property(callee);
2046 int benefice = calc_inline_benefice(call, callee);
2048 DB((dbg, LEVEL_2, "In %+F Call %+F to %+F has benefice %d\n",
2049 get_irn_irg(call->call), call->call, callee, benefice));
2051 if (prop < irg_inline_forced && benefice < inline_threshold) {
2055 pqueue_put(pqueue, call, benefice);
2059 * Try to inline calls into a graph.
2061 * @param irg the graph into which we inline
2062 * @param maxsize do NOT inline if the size of irg gets
2063 * bigger than this amount
2064 * @param inline_threshold
2065 * threshold value for inline decision
2066 * @param copied_graphs
2067 * map containing copied of recursive graphs
2069 static void inline_into(ir_graph *irg, unsigned maxsize,
2070 int inline_threshold, pmap *copied_graphs)
2072 int phiproj_computed = 0;
2073 inline_irg_env *env = get_irg_link(irg);
2074 call_entry *curr_call;
2078 if (env->n_call_nodes == 0)
2081 if (env->n_nodes > maxsize) {
2082 DB((dbg, LEVEL_2, "%+F: too big (%d)\n", irg, env->n_nodes));
2086 current_ir_graph = irg;
2087 ir_reserve_resources(irg, IR_RESOURCE_IRN_LINK|IR_RESOURCE_PHI_LIST);
2089 /* put irgs into the pqueue */
2090 pqueue = new_pqueue();
2092 list_for_each_entry(call_entry, curr_call, &env->calls, list) {
2093 assert(is_Call(curr_call->call));
2094 maybe_push_call(pqueue, curr_call, inline_threshold);
2097 /* note that the list of possible calls is updated during the process */
2098 while (!pqueue_empty(pqueue)) {
2100 call_entry *curr_call = pqueue_pop_front(pqueue);
2101 ir_graph *callee = curr_call->callee;
2102 ir_node *call_node = curr_call->call;
2103 inline_irg_env *callee_env = get_irg_link(callee);
2104 irg_inline_property prop = get_irg_inline_property(callee);
2106 const call_entry *centry;
2109 if ((prop < irg_inline_forced) && env->n_nodes + callee_env->n_nodes > maxsize) {
2110 DB((dbg, LEVEL_2, "%+F: too big (%d) + %+F (%d)\n", irg,
2111 env->n_nodes, callee, callee_env->n_nodes));
2115 e = pmap_find(copied_graphs, callee);
2117 int benefice = curr_call->benefice;
2119 * Reduce the weight for recursive function IFF not all arguments are const.
2120 * inlining recursive functions is rarely good.
2122 if (!curr_call->all_const)
2124 if (benefice < inline_threshold)
2128 * Remap callee if we have a copy.
2131 callee_env = get_irg_link(callee);
2134 if (current_ir_graph == callee) {
2136 * Recursive call: we cannot directly inline because we cannot
2137 * walk the graph and change it. So we have to make a copy of
2140 int benefice = curr_call->benefice;
2144 * Reduce the weight for recursive function IFF not all arguments are const.
2145 * inlining recursive functions is rarely good.
2147 if (!curr_call->all_const)
2149 if (benefice < inline_threshold)
2152 ir_free_resources(irg, IR_RESOURCE_IRN_LINK|IR_RESOURCE_PHI_LIST);
2155 * No copy yet, create one.
2156 * Note that recursive methods are never leaves, so it is
2157 * sufficient to test this condition here.
2159 copy = create_irg_copy(callee);
2161 /* create_irg_copy() destroys the Proj links, recompute them */
2162 phiproj_computed = 0;
2164 ir_reserve_resources(irg, IR_RESOURCE_IRN_LINK|IR_RESOURCE_PHI_LIST);
2166 /* allocate a new environment */
2167 callee_env = alloc_inline_irg_env();
2168 set_irg_link(copy, callee_env);
2170 assure_cf_loop(copy);
2171 wenv.x = callee_env;
2172 wenv.ignore_callers = 1;
2173 irg_walk_graph(copy, NULL, collect_calls2, &wenv);
2176 * Enter the entity of the original graph. This is needed
2177 * for inline_method(). However, note that ent->irg still points
2178 * to callee, NOT to copy.
2180 set_irg_entity(copy, get_irg_entity(callee));
2182 pmap_insert(copied_graphs, callee, copy);
2185 /* we have only one caller: the original graph */
2186 callee_env->n_callers = 1;
2187 callee_env->n_callers_orig = 1;
2189 if (! phiproj_computed) {
2190 phiproj_computed = 1;
2191 collect_phiprojs(current_ir_graph);
2193 did_inline = inline_method(call_node, callee);
2197 /* call was inlined, Phi/Projs for current graph must be recomputed */
2198 phiproj_computed = 0;
2200 /* remove it from the caller list */
2201 list_del(&curr_call->list);
2203 /* callee was inline. Append it's call list. */
2204 env->got_inline = 1;
2205 if (curr_call->local_adr)
2206 env->local_vars = 1;
2207 --env->n_call_nodes;
2209 /* we just generate a bunch of new calls */
2210 loop_depth = curr_call->loop_depth;
2211 list_for_each_entry(call_entry, centry, &callee_env->calls, list) {
2212 inline_irg_env *penv = get_irg_link(centry->callee);
2214 call_entry *new_entry;
2216 /* after we have inlined callee, all called methods inside
2217 * callee are now called once more */
2220 /* Note that the src list points to Call nodes in the inlined graph,
2221 * but we need Call nodes in our graph. Luckily the inliner leaves
2222 * this information in the link field. */
2223 new_call = get_irn_link(centry->call);
2224 assert(is_Call(new_call));
2226 new_entry = duplicate_call_entry(centry, new_call, loop_depth);
2227 list_add_tail(&new_entry->list, &env->calls);
2228 maybe_push_call(pqueue, new_entry, inline_threshold);
2231 env->n_call_nodes += callee_env->n_call_nodes;
2232 env->n_nodes += callee_env->n_nodes;
2233 --callee_env->n_callers;
2235 ir_free_resources(irg, IR_RESOURCE_IRN_LINK|IR_RESOURCE_PHI_LIST);
2240 * Heuristic inliner. Calculates a benefice value for every call and inlines
2241 * those calls with a value higher than the threshold.
2243 void inline_functions(unsigned maxsize, int inline_threshold) {
2244 inline_irg_env *env;
2248 pmap *copied_graphs;
2249 pmap_entry *pm_entry;
2252 rem = current_ir_graph;
2253 obstack_init(&temp_obst);
2255 irgs = create_irg_list();
2257 /* a map for the copied graphs, used to inline recursive calls */
2258 copied_graphs = pmap_create();
2260 /* extend all irgs by a temporary data structure for inlining. */
2261 n_irgs = get_irp_n_irgs();
2262 for (i = 0; i < n_irgs; ++i)
2263 set_irg_link(irgs[i], alloc_inline_irg_env());
2265 /* Pre-compute information in temporary data structure. */
2266 wenv.ignore_runtime = 0;
2267 wenv.ignore_callers = 0;
2268 for (i = 0; i < n_irgs; ++i) {
2269 ir_graph *irg = irgs[i];
2271 free_callee_info(irg);
2273 wenv.x = get_irg_link(irg);
2274 assure_cf_loop(irg);
2275 irg_walk_graph(irg, NULL, collect_calls2, &wenv);
2278 /* -- and now inline. -- */
2279 for (i = 0; i < n_irgs; ++i) {
2280 ir_graph *irg = irgs[i];
2282 inline_into(irg, maxsize, inline_threshold, copied_graphs);
2285 for (i = 0; i < n_irgs; ++i) {
2286 ir_graph *irg = irgs[i];
2288 env = get_irg_link(irg);
2289 if (env->got_inline) {
2290 /* this irg got calls inlined: optimize it */
2291 if (get_opt_combo()) {
2292 if (env->local_vars) {
2293 scalar_replacement_opt(irg);
2297 if (env->local_vars) {
2298 if (scalar_replacement_opt(irg)) {
2299 optimize_graph_df(irg);
2305 if (env->got_inline || (env->n_callers_orig != env->n_callers)) {
2306 DB((dbg, LEVEL_1, "Nodes:%3d ->%3d, calls:%3d ->%3d, callers:%3d ->%3d, -- %s\n",
2307 env->n_nodes_orig, env->n_nodes, env->n_call_nodes_orig, env->n_call_nodes,
2308 env->n_callers_orig, env->n_callers,
2309 get_entity_name(get_irg_entity(irg))));
2313 /* kill the copied graphs: we don't need them anymore */
2314 foreach_pmap(copied_graphs, pm_entry) {
2315 ir_graph *copy = pm_entry->value;
2317 /* reset the entity, otherwise it will be deleted in the next step ... */
2318 set_irg_entity(copy, NULL);
2319 free_ir_graph(copy);
2321 pmap_destroy(copied_graphs);
2325 obstack_free(&temp_obst, NULL);
2326 current_ir_graph = rem;
2329 void firm_init_inline(void) {
2330 FIRM_DBG_REGISTER(dbg, "firm.opt.inline");