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
34 #include "irgraph_t.h"
37 #include "iroptimize.h"
44 #include "adt/array.h"
48 #include "adt/xmalloc.h"
52 #include "irbackedge_t.h"
53 #include "opt_inline_t.h"
58 #include "analyze_irg_args.h"
59 #include "iredges_t.h"
64 DEBUG_ONLY(static firm_dbg_module_t *dbg;)
66 /*------------------------------------------------------------------*/
67 /* Routines for dead node elimination / copying garbage collection */
69 /*------------------------------------------------------------------*/
72 * Remember the new node in the old node by using a field all nodes have.
74 #define set_new_node(oldn, newn) set_irn_link(oldn, newn)
77 * Get this new node, before the old node is forgotten.
79 #define get_new_node(oldn) get_irn_link(oldn)
82 * Check if a new node was set.
84 #define has_new_node(n) (get_new_node(n) != NULL)
87 * We use the block_visited flag to mark that we have computed the
88 * number of useful predecessors for this block.
89 * Further we encode the new arity in this flag in the old blocks.
90 * Remembering the arity is useful, as it saves a lot of pointer
91 * accesses. This function is called for all Phi and Block nodes
95 compute_new_arity(ir_node *b) {
96 int i, res, irn_arity;
99 irg_v = get_irg_block_visited(current_ir_graph);
100 block_v = get_Block_block_visited(b);
101 if (block_v >= irg_v) {
102 /* we computed the number of preds for this block and saved it in the
104 return block_v - irg_v;
106 /* compute the number of good predecessors */
107 res = irn_arity = get_irn_arity(b);
108 for (i = 0; i < irn_arity; i++)
109 if (is_Bad(get_irn_n(b, i))) res--;
110 /* save it in the flag. */
111 set_Block_block_visited(b, irg_v + res);
117 * Copies the node to the new obstack. The Ins of the new node point to
118 * the predecessors on the old obstack. For block/phi nodes not all
119 * predecessors might be copied. n->link points to the new node.
120 * For Phi and Block nodes the function allocates in-arrays with an arity
121 * only for useful predecessors. The arity is determined by counting
122 * the non-bad predecessors of the block.
124 * @param n The node to be copied
125 * @param env if non-NULL, the node number attribute will be copied to the new node
127 * Note: Also used for loop unrolling.
129 static void copy_node(ir_node *n, void *env) {
132 ir_op *op = get_irn_op(n);
136 /* node copied already */
138 } else if (op == op_Block) {
140 new_arity = compute_new_arity(n);
141 n->attr.block.graph_arr = NULL;
143 block = get_nodes_block(n);
145 new_arity = compute_new_arity(block);
147 new_arity = get_irn_arity(n);
150 nn = new_ir_node(get_irn_dbg_info(n),
157 /* Copy the attributes. These might point to additional data. If this
158 was allocated on the old obstack the pointers now are dangling. This
159 frees e.g. the memory of the graph_arr allocated in new_immBlock. */
160 if (op == op_Block) {
161 /* we cannot allow blocks WITHOUT macroblock input */
162 set_Block_MacroBlock(nn, get_Block_MacroBlock(n));
164 copy_node_attr(n, nn);
168 int copy_node_nr = env != NULL;
170 /* for easier debugging, we want to copy the node numbers too */
171 nn->node_nr = n->node_nr;
177 hook_dead_node_elim_subst(current_ir_graph, n, nn);
181 * Copies new predecessors of old node to new node remembered in link.
182 * Spare the Bad predecessors of Phi and Block nodes.
184 static void copy_preds(ir_node *n, void *env) {
189 nn = get_new_node(n);
192 /* copy the macro block header */
193 ir_node *mbh = get_Block_MacroBlock(n);
196 /* this block is a macroblock header */
197 set_Block_MacroBlock(nn, nn);
199 /* get the macro block header */
200 ir_node *nmbh = get_new_node(mbh);
201 assert(nmbh != NULL);
202 set_Block_MacroBlock(nn, nmbh);
205 /* Don't copy Bad nodes. */
207 irn_arity = get_irn_arity(n);
208 for (i = 0; i < irn_arity; i++) {
209 if (! is_Bad(get_irn_n(n, i))) {
210 ir_node *pred = get_irn_n(n, i);
211 set_irn_n(nn, j, get_new_node(pred));
215 /* repair the block visited flag from above misuse. Repair it in both
216 graphs so that the old one can still be used. */
217 set_Block_block_visited(nn, 0);
218 set_Block_block_visited(n, 0);
219 /* Local optimization could not merge two subsequent blocks if
220 in array contained Bads. Now it's possible.
221 We don't call optimize_in_place as it requires
222 that the fields in ir_graph are set properly. */
223 if ((get_opt_control_flow_straightening()) &&
224 (get_Block_n_cfgpreds(nn) == 1) &&
225 is_Jmp(get_Block_cfgpred(nn, 0))) {
226 ir_node *old = get_nodes_block(get_Block_cfgpred(nn, 0));
228 /* Jmp jumps into the block it is in -- deal self cycle. */
229 assert(is_Bad(get_new_node(get_irg_bad(current_ir_graph))));
230 exchange(nn, get_new_node(get_irg_bad(current_ir_graph)));
235 } else if (is_Phi(n) && get_irn_arity(n) > 0) {
236 /* Don't copy node if corresponding predecessor in block is Bad.
237 The Block itself should not be Bad. */
238 block = get_nodes_block(n);
239 set_nodes_block(nn, get_new_node(block));
241 irn_arity = get_irn_arity(n);
242 for (i = 0; i < irn_arity; i++) {
243 if (! is_Bad(get_irn_n(block, i))) {
244 ir_node *pred = get_irn_n(n, i);
245 set_irn_n(nn, j, get_new_node(pred));
246 /*if (is_backedge(n, i)) set_backedge(nn, j);*/
250 /* If the pre walker reached this Phi after the post walker visited the
251 block block_visited is > 0. */
252 set_Block_block_visited(get_nodes_block(n), 0);
253 /* Compacting the Phi's ins might generate Phis with only one
255 if (get_irn_arity(nn) == 1)
256 exchange(nn, get_irn_n(nn, 0));
258 irn_arity = get_irn_arity(n);
259 for (i = -1; i < irn_arity; i++)
260 set_irn_n(nn, i, get_new_node(get_irn_n(n, i)));
262 /* Now the new node is complete. We can add it to the hash table for CSE.
263 @@@ inlining aborts if we identify End. Why? */
265 add_identities(current_ir_graph->value_table, nn);
269 * Copies the graph recursively, compacts the keep-alives of the end node.
271 * @param irg the graph to be copied
272 * @param copy_node_nr If non-zero, the node number will be copied
274 static void copy_graph(ir_graph *irg, int copy_node_nr) {
275 ir_node *oe, *ne, *ob, *nb, *om, *nm; /* old end, new end, old bad, new bad, old NoMem, new NoMem */
276 ir_node *ka; /* keep alive */
280 /* Some nodes must be copied by hand, sigh */
281 vfl = get_irg_visited(irg);
282 set_irg_visited(irg, vfl + 1);
284 oe = get_irg_end(irg);
285 mark_irn_visited(oe);
286 /* copy the end node by hand, allocate dynamic in array! */
287 ne = new_ir_node(get_irn_dbg_info(oe),
294 /* Copy the attributes. Well, there might be some in the future... */
295 copy_node_attr(oe, ne);
296 set_new_node(oe, ne);
298 /* copy the Bad node */
299 ob = get_irg_bad(irg);
300 mark_irn_visited(ob);
301 nb = new_ir_node(get_irn_dbg_info(ob),
308 copy_node_attr(ob, nb);
309 set_new_node(ob, nb);
311 /* copy the NoMem node */
312 om = get_irg_no_mem(irg);
313 mark_irn_visited(om);
314 nm = new_ir_node(get_irn_dbg_info(om),
321 copy_node_attr(om, nm);
322 set_new_node(om, nm);
324 /* copy the live nodes */
325 set_irg_visited(irg, vfl);
326 irg_walk(get_nodes_block(oe), copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
328 /* Note: from yet, the visited flag of the graph is equal to vfl + 1 */
330 /* visit the anchors as well */
331 for (i = get_irg_n_anchors(irg) - 1; i >= 0; --i) {
332 ir_node *n = get_irg_anchor(irg, i);
334 if (n && (get_irn_visited(n) <= vfl)) {
335 set_irg_visited(irg, vfl);
336 irg_walk(n, copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
340 /* copy_preds for the end node ... */
341 set_nodes_block(ne, get_new_node(get_nodes_block(oe)));
343 /*- ... and now the keep alives. -*/
344 /* First pick the not marked block nodes and walk them. We must pick these
345 first as else we will oversee blocks reachable from Phis. */
346 irn_arity = get_End_n_keepalives(oe);
347 for (i = 0; i < irn_arity; i++) {
348 ka = get_End_keepalive(oe, i);
350 if (get_irn_visited(ka) <= vfl) {
351 /* We must keep the block alive and copy everything reachable */
352 set_irg_visited(irg, vfl);
353 irg_walk(ka, copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
355 add_End_keepalive(ne, get_new_node(ka));
359 /* Now pick other nodes. Here we will keep all! */
360 irn_arity = get_End_n_keepalives(oe);
361 for (i = 0; i < irn_arity; i++) {
362 ka = get_End_keepalive(oe, i);
364 if (get_irn_visited(ka) <= vfl) {
365 /* We didn't copy the node yet. */
366 set_irg_visited(irg, vfl);
367 irg_walk(ka, copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
369 add_End_keepalive(ne, get_new_node(ka));
373 /* start block sometimes only reached after keep alives */
374 set_nodes_block(nb, get_new_node(get_nodes_block(ob)));
375 set_nodes_block(nm, get_new_node(get_nodes_block(om)));
379 * Copies the graph reachable from current_ir_graph->end to the obstack
380 * in current_ir_graph and fixes the environment.
381 * Then fixes the fields in current_ir_graph containing nodes of the
384 * @param copy_node_nr If non-zero, the node number will be copied
387 copy_graph_env(int copy_node_nr) {
388 ir_graph *irg = current_ir_graph;
389 ir_node *old_end, *new_anchor;
392 /* remove end_except and end_reg nodes */
393 old_end = get_irg_end(irg);
394 set_irg_end_except (irg, old_end);
395 set_irg_end_reg (irg, old_end);
397 /* Not all nodes remembered in irg might be reachable
398 from the end node. Assure their link is set to NULL, so that
399 we can test whether new nodes have been computed. */
400 for (i = get_irg_n_anchors(irg) - 1; i >= 0; --i) {
401 ir_node *n = get_irg_anchor(irg, i);
403 set_new_node(n, NULL);
405 /* we use the block walk flag for removing Bads from Blocks ins. */
406 inc_irg_block_visited(irg);
409 copy_graph(irg, copy_node_nr);
412 old_end = get_irg_end(irg);
413 new_anchor = new_Anchor(irg);
415 for (i = get_irg_n_anchors(irg) - 1; i >= 0; --i) {
416 ir_node *n = get_irg_anchor(irg, i);
418 set_irn_n(new_anchor, i, get_new_node(n));
421 irg->anchor = new_anchor;
423 /* ensure the new anchor is placed in the endblock */
424 set_nodes_block(new_anchor, get_irg_end_block(irg));
428 * Copies all reachable nodes to a new obstack. Removes bad inputs
429 * from block nodes and the corresponding inputs from Phi nodes.
430 * Merges single exit blocks with single entry blocks and removes
432 * Adds all new nodes to a new hash table for CSE. Does not
433 * perform CSE, so the hash table might contain common subexpressions.
435 void dead_node_elimination(ir_graph *irg) {
437 #ifdef INTERPROCEDURAL_VIEW
438 int rem_ipview = get_interprocedural_view();
440 struct obstack *graveyard_obst = NULL;
441 struct obstack *rebirth_obst = NULL;
443 edges_deactivate(irg);
445 /* inform statistics that we started a dead-node elimination run */
446 hook_dead_node_elim(irg, 1);
448 /* Remember external state of current_ir_graph. */
449 rem = current_ir_graph;
450 current_ir_graph = irg;
451 #ifdef INTERPROCEDURAL_VIEW
452 set_interprocedural_view(0);
455 assert(get_irg_phase_state(irg) != phase_building);
457 /* Handle graph state */
458 free_callee_info(irg);
462 /* @@@ so far we loose loops when copying */
463 free_loop_information(irg);
465 set_irg_doms_inconsistent(irg);
467 /* A quiet place, where the old obstack can rest in peace,
468 until it will be cremated. */
469 graveyard_obst = irg->obst;
471 /* A new obstack, where the reachable nodes will be copied to. */
472 rebirth_obst = xmalloc(sizeof(*rebirth_obst));
473 irg->obst = rebirth_obst;
474 obstack_init(irg->obst);
475 irg->last_node_idx = 0;
477 /* We also need a new value table for CSE */
478 del_identities(irg->value_table);
479 irg->value_table = new_identities();
481 /* Copy the graph from the old to the new obstack */
482 copy_graph_env(/*copy_node_nr=*/1);
484 /* Free memory from old unoptimized obstack */
485 obstack_free(graveyard_obst, 0); /* First empty the obstack ... */
486 xfree(graveyard_obst); /* ... then free it. */
488 /* inform statistics that the run is over */
489 hook_dead_node_elim(irg, 0);
491 current_ir_graph = rem;
492 #ifdef INTERPROCEDURAL_VIEW
493 set_interprocedural_view(rem_ipview);
498 * Relink bad predecessors of a block and store the old in array to the
499 * link field. This function is called by relink_bad_predecessors().
500 * The array of link field starts with the block operand at position 0.
501 * If block has bad predecessors, create a new in array without bad preds.
502 * Otherwise let in array untouched.
504 static void relink_bad_block_predecessors(ir_node *n, void *env) {
505 ir_node **new_in, *irn;
506 int i, new_irn_n, old_irn_arity, new_irn_arity = 0;
509 /* if link field of block is NULL, look for bad predecessors otherwise
510 this is already done */
511 if (is_Block(n) && get_irn_link(n) == NULL) {
512 /* save old predecessors in link field (position 0 is the block operand)*/
513 set_irn_link(n, get_irn_in(n));
515 /* count predecessors without bad nodes */
516 old_irn_arity = get_irn_arity(n);
517 for (i = 0; i < old_irn_arity; i++)
518 if (!is_Bad(get_irn_n(n, i)))
521 /* arity changing: set new predecessors without bad nodes */
522 if (new_irn_arity < old_irn_arity) {
523 /* Get new predecessor array. We do not resize the array, as we must
524 keep the old one to update Phis. */
525 new_in = NEW_ARR_D(ir_node *, current_ir_graph->obst, (new_irn_arity+1));
527 /* set new predecessors in array */
530 for (i = 0; i < old_irn_arity; i++) {
531 irn = get_irn_n(n, i);
533 new_in[new_irn_n] = irn;
534 is_backedge(n, i) ? set_backedge(n, new_irn_n-1) : set_not_backedge(n, new_irn_n-1);
538 /* ARR_SETLEN(int, n->attr.block.backedge, new_irn_arity); */
539 ARR_SHRINKLEN(n->attr.block.backedge, new_irn_arity);
541 } /* ir node has bad predecessors */
542 } /* Block is not relinked */
546 * Relinks Bad predecessors from Blocks and Phis called by walker
547 * remove_bad_predecesors(). If n is a Block, call
548 * relink_bad_block_redecessors(). If n is a Phi-node, call also the relinking
549 * function of Phi's Block. If this block has bad predecessors, relink preds
552 static void relink_bad_predecessors(ir_node *n, void *env) {
553 ir_node *block, **old_in;
554 int i, old_irn_arity, new_irn_arity;
556 /* relink bad predecessors of a block */
558 relink_bad_block_predecessors(n, env);
560 /* If Phi node relink its block and its predecessors */
562 /* Relink predecessors of phi's block */
563 block = get_nodes_block(n);
564 if (get_irn_link(block) == NULL)
565 relink_bad_block_predecessors(block, env);
567 old_in = (ir_node **)get_irn_link(block); /* Of Phi's Block */
568 old_irn_arity = ARR_LEN(old_in);
570 /* Relink Phi predecessors if count of predecessors changed */
571 if (old_irn_arity != ARR_LEN(get_irn_in(block))) {
572 /* set new predecessors in array
573 n->in[0] remains the same block */
575 for(i = 1; i < old_irn_arity; i++)
576 if (!is_Bad(old_in[i])) {
577 n->in[new_irn_arity] = n->in[i];
578 is_backedge(n, i) ? set_backedge(n, new_irn_arity) : set_not_backedge(n, new_irn_arity);
582 ARR_SETLEN(ir_node *, n->in, new_irn_arity);
583 ARR_SETLEN(int, n->attr.phi.u.backedge, new_irn_arity);
585 } /* n is a Phi node */
589 * Removes Bad Bad predecessors from Blocks and the corresponding
590 * inputs to Phi nodes as in dead_node_elimination but without
592 * On walking up set the link field to NULL, on walking down call
593 * relink_bad_predecessors() (This function stores the old in array
594 * to the link field and sets a new in array if arity of predecessors
597 void remove_bad_predecessors(ir_graph *irg) {
598 panic("Fix backedge handling first");
599 irg_walk_graph(irg, firm_clear_link, relink_bad_predecessors, NULL);
606 __)|_| | \_/ | \_/(/_ |_/\__|__
608 The following stuff implements a facility that automatically patches
609 registered ir_node pointers to the new node when a dead node elimination occurs.
612 struct _survive_dce_t {
616 hook_entry_t dead_node_elim;
617 hook_entry_t dead_node_elim_subst;
620 typedef struct _survive_dce_list_t {
621 struct _survive_dce_list_t *next;
623 } survive_dce_list_t;
625 static void dead_node_hook(void *context, ir_graph *irg, int start) {
626 survive_dce_t *sd = context;
629 /* Create a new map before the dead node elimination is performed. */
631 sd->new_places = pmap_create_ex(pmap_count(sd->places));
633 /* Patch back all nodes if dead node elimination is over and something is to be done. */
634 pmap_destroy(sd->places);
635 sd->places = sd->new_places;
636 sd->new_places = NULL;
641 * Hook called when dead node elimination replaces old by nw.
643 static void dead_node_subst_hook(void *context, ir_graph *irg, ir_node *old, ir_node *nw) {
644 survive_dce_t *sd = context;
645 survive_dce_list_t *list = pmap_get(sd->places, old);
648 /* If the node is to be patched back, write the new address to all registered locations. */
650 survive_dce_list_t *p;
652 for (p = list; p; p = p->next)
655 pmap_insert(sd->new_places, nw, list);
660 * Make a new Survive DCE environment.
662 survive_dce_t *new_survive_dce(void) {
663 survive_dce_t *res = xmalloc(sizeof(res[0]));
664 obstack_init(&res->obst);
665 res->places = pmap_create();
666 res->new_places = NULL;
668 res->dead_node_elim.hook._hook_dead_node_elim = dead_node_hook;
669 res->dead_node_elim.context = res;
670 res->dead_node_elim.next = NULL;
672 res->dead_node_elim_subst.hook._hook_dead_node_elim_subst = dead_node_subst_hook;
673 res->dead_node_elim_subst.context = res;
674 res->dead_node_elim_subst.next = NULL;
676 #ifndef FIRM_ENABLE_HOOKS
677 assert(0 && "need hooks enabled");
680 register_hook(hook_dead_node_elim, &res->dead_node_elim);
681 register_hook(hook_dead_node_elim_subst, &res->dead_node_elim_subst);
686 * Free a Survive DCE environment.
688 void free_survive_dce(survive_dce_t *sd) {
689 obstack_free(&sd->obst, NULL);
690 pmap_destroy(sd->places);
691 unregister_hook(hook_dead_node_elim, &sd->dead_node_elim);
692 unregister_hook(hook_dead_node_elim_subst, &sd->dead_node_elim_subst);
697 * Register a node pointer to be patched upon DCE.
698 * When DCE occurs, the node pointer specified by @p place will be
699 * patched to the new address of the node it is pointing to.
701 * @param sd The Survive DCE environment.
702 * @param place The address of the node pointer.
704 void survive_dce_register_irn(survive_dce_t *sd, ir_node **place) {
705 if (*place != NULL) {
706 ir_node *irn = *place;
707 survive_dce_list_t *curr = pmap_get(sd->places, irn);
708 survive_dce_list_t *nw = obstack_alloc(&sd->obst, sizeof(nw[0]));
713 pmap_insert(sd->places, irn, nw);
717 /*--------------------------------------------------------------------*/
718 /* Functionality for inlining */
719 /*--------------------------------------------------------------------*/
722 * Copy node for inlineing. Updates attributes that change when
723 * inlineing but not for dead node elimination.
725 * Copies the node by calling copy_node() and then updates the entity if
726 * it's a local one. env must be a pointer of the frame type of the
727 * inlined procedure. The new entities must be in the link field of
730 static void copy_node_inline(ir_node *n, void *env) {
732 ir_type *frame_tp = (ir_type *)env;
736 nn = get_new_node (n);
738 if (get_entity_owner(get_Sel_entity(n)) == frame_tp) {
739 set_Sel_entity(nn, get_entity_link(get_Sel_entity(n)));
741 } else if (is_Block(n)) {
742 nn = get_new_node (n);
743 nn->attr.block.irg = current_ir_graph;
748 * Copies new predecessors of old node and move constants to
751 static void copy_preds_inline(ir_node *n, void *env) {
755 nn = skip_Id(get_new_node(n));
756 if (is_irn_constlike(nn)) {
757 /* move Constants into the start block */
758 set_nodes_block(nn, get_irg_start_block(current_ir_graph));
760 n = identify_remember(current_ir_graph->value_table, nn);
768 * Walker: checks if P_value_arg_base is used.
770 static void find_addr(ir_node *node, void *env) {
771 int *allow_inline = env;
773 is_Start(get_Proj_pred(node)) &&
774 get_Proj_proj(node) == pn_Start_P_value_arg_base) {
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 * Sorryly 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);
856 if (prop == irg_inline_forbidden)
859 ent = get_irg_entity(called_graph);
861 mtp = get_entity_type(ent);
862 ctp = get_Call_type(call);
863 if (get_method_n_params(mtp) > get_method_n_params(ctp)) {
864 /* this is a bad feature of C: without a prototype, we can can call a function with less
865 parameters than needed. Currently we don't support this, although it would be
866 to use Unknown than. */
870 /* Argh, compiling C has some bad consequences:
871 the call type AND the method type might be different.
872 It is implementation defendant 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);
930 /* -- Check preconditions -- */
931 assert(is_Call(call));
933 /* here we know we WILL inline, so inform the statistics */
934 hook_inline(call, called_graph);
936 /* -- Decide how to handle exception control flow: Is there a handler
937 for the Call node, or do we branch directly to End on an exception?
939 0 There is a handler.
941 2 Exception handling not represented in Firm. -- */
943 ir_node *proj, *Mproj = NULL, *Xproj = NULL;
944 for (proj = get_irn_link(call); proj; proj = get_irn_link(proj)) {
945 long proj_nr = get_Proj_proj(proj);
946 if (proj_nr == pn_Call_X_except) Xproj = proj;
947 if (proj_nr == pn_Call_M_except) Mproj = proj;
949 if (Mproj) { assert(Xproj); exc_handling = exc_handler; } /* Mproj */
950 else if (Xproj) { exc_handling = exc_to_end; } /* !Mproj && Xproj */
951 else { exc_handling = exc_no_handler; } /* !Mproj && !Xproj */
954 /* create the argument tuple */
955 NEW_ARR_A(ir_type *, args_in, n_params);
957 block = get_nodes_block(call);
958 for (i = n_params - 1; i >= 0; --i) {
959 ir_node *arg = get_Call_param(call, i);
960 ir_type *param_tp = get_method_param_type(mtp, i);
961 ir_mode *mode = get_type_mode(param_tp);
963 if (mode != get_irn_mode(arg)) {
964 arg = new_r_Conv(irg, block, arg, mode);
970 the procedure and later replaces the Start node of the called graph.
971 Post_call is the old Call node and collects the results of the called
972 graph. Both will end up being a tuple. -- */
973 post_bl = get_nodes_block(call);
974 set_irg_current_block(irg, post_bl);
975 /* XxMxPxPxPxT of Start + parameter of Call */
976 in[pn_Start_X_initial_exec] = new_Jmp();
977 in[pn_Start_M] = get_Call_mem(call);
978 in[pn_Start_P_frame_base] = get_irg_frame(irg);
979 in[pn_Start_P_tls] = get_irg_tls(irg);
980 in[pn_Start_T_args] = new_Tuple(n_params, args_in);
981 /* in[pn_Start_P_value_arg_base] = ??? */
982 assert(pn_Start_P_value_arg_base == pn_Start_max - 1 && "pn_Start_P_value_arg_base not supported, fix");
983 pre_call = new_Tuple(pn_Start_max - 1, in);
987 The new block gets the ins of the old block, pre_call and all its
988 predecessors and all Phi nodes. -- */
989 part_block(pre_call);
991 /* -- Prepare state for dead node elimination -- */
992 /* Visited flags in calling irg must be >= flag in called irg.
993 Else walker and arity computation will not work. */
994 if (get_irg_visited(irg) <= get_irg_visited(called_graph))
995 set_irg_visited(irg, get_irg_visited(called_graph) + 1);
996 if (get_irg_block_visited(irg) < get_irg_block_visited(called_graph))
997 set_irg_block_visited(irg, get_irg_block_visited(called_graph));
998 /* Set pre_call as new Start node in link field of the start node of
999 calling graph and pre_calls block as new block for the start block
1001 Further mark these nodes so that they are not visited by the
1003 set_irn_link(get_irg_start(called_graph), pre_call);
1004 set_irn_visited(get_irg_start(called_graph), get_irg_visited(irg));
1005 set_irn_link(get_irg_start_block(called_graph), get_nodes_block(pre_call));
1006 set_irn_visited(get_irg_start_block(called_graph), get_irg_visited(irg));
1007 set_irn_link(get_irg_bad(called_graph), get_irg_bad(irg));
1008 set_irn_visited(get_irg_bad(called_graph), get_irg_visited(irg));
1010 /* Initialize for compaction of in arrays */
1011 inc_irg_block_visited(irg);
1013 /* -- Replicate local entities of the called_graph -- */
1014 /* copy the entities. */
1015 called_frame = get_irg_frame_type(called_graph);
1016 curr_frame = get_irg_frame_type(irg);
1017 for (i = 0, n = get_class_n_members(called_frame); i < n; ++i) {
1018 ir_entity *new_ent, *old_ent;
1019 old_ent = get_class_member(called_frame, i);
1020 new_ent = copy_entity_own(old_ent, curr_frame);
1021 set_entity_link(old_ent, new_ent);
1024 /* visited is > than that of called graph. With this trick visited will
1025 remain unchanged so that an outer walker, e.g., searching the call nodes
1026 to inline, calling this inline will not visit the inlined nodes. */
1027 set_irg_visited(irg, get_irg_visited(irg)-1);
1029 /* -- Performing dead node elimination inlines the graph -- */
1030 /* Copies the nodes to the obstack of current_ir_graph. Updates links to new
1032 irg_walk(get_irg_end(called_graph), copy_node_inline, copy_preds_inline,
1033 get_irg_frame_type(called_graph));
1035 /* Repair called_graph */
1036 set_irg_visited(called_graph, get_irg_visited(irg));
1037 set_irg_block_visited(called_graph, get_irg_block_visited(irg));
1038 set_Block_block_visited(get_irg_start_block(called_graph), 0);
1040 /* -- Merge the end of the inlined procedure with the call site -- */
1041 /* We will turn the old Call node into a Tuple with the following
1044 0: Phi of all Memories of Return statements.
1045 1: Jmp from new Block that merges the control flow from all exception
1046 predecessors of the old end block.
1047 2: Tuple of all arguments.
1048 3: Phi of Exception memories.
1049 In case the old Call directly branches to End on an exception we don't
1050 need the block merging all exceptions nor the Phi of the exception
1054 /* -- Precompute some values -- */
1055 end_bl = get_new_node(get_irg_end_block(called_graph));
1056 end = get_new_node(get_irg_end(called_graph));
1057 arity = get_irn_arity(end_bl); /* arity = n_exc + n_ret */
1058 n_res = get_method_n_ress(get_Call_type(call));
1060 res_pred = xmalloc(n_res * sizeof(*res_pred));
1061 cf_pred = xmalloc(arity * sizeof(*res_pred));
1063 set_irg_current_block(irg, post_bl); /* just to make sure */
1065 /* -- archive keepalives -- */
1066 irn_arity = get_irn_arity(end);
1067 for (i = 0; i < irn_arity; i++) {
1068 ir_node *ka = get_End_keepalive(end, i);
1070 add_End_keepalive(get_irg_end(irg), ka);
1073 /* The new end node will die. We need not free as the in array is on the obstack:
1074 copy_node() only generated 'D' arrays. */
1076 /* -- Replace Return nodes by Jump nodes. -- */
1078 for (i = 0; i < arity; i++) {
1080 ret = get_irn_n(end_bl, i);
1081 if (is_Return(ret)) {
1082 cf_pred[n_ret] = new_r_Jmp(irg, get_nodes_block(ret));
1086 set_irn_in(post_bl, n_ret, cf_pred);
1088 /* -- Build a Tuple for all results of the method.
1089 Add Phi node if there was more than one Return. -- */
1090 turn_into_tuple(post_call, pn_Call_max);
1091 /* First the Memory-Phi */
1093 for (i = 0; i < arity; i++) {
1094 ret = get_irn_n(end_bl, i);
1095 if (is_Return(ret)) {
1096 cf_pred[n_ret] = get_Return_mem(ret);
1100 phi = new_Phi(n_ret, cf_pred, mode_M);
1101 set_Tuple_pred(call, pn_Call_M_regular, phi);
1102 /* Conserve Phi-list for further inlinings -- but might be optimized */
1103 if (get_nodes_block(phi) == post_bl) {
1104 set_irn_link(phi, get_irn_link(post_bl));
1105 set_irn_link(post_bl, phi);
1107 /* Now the real results */
1109 for (j = 0; j < n_res; j++) {
1111 for (i = 0; i < arity; i++) {
1112 ret = get_irn_n(end_bl, i);
1113 if (is_Return(ret)) {
1114 cf_pred[n_ret] = get_Return_res(ret, j);
1119 phi = new_Phi(n_ret, cf_pred, get_irn_mode(cf_pred[0]));
1123 /* Conserve Phi-list for further inlinings -- but might be optimized */
1124 if (get_nodes_block(phi) == post_bl) {
1125 set_Phi_next(phi, get_Block_phis(post_bl));
1126 set_Block_phis(post_bl, phi);
1129 set_Tuple_pred(call, pn_Call_T_result, new_Tuple(n_res, res_pred));
1131 set_Tuple_pred(call, pn_Call_T_result, new_Bad());
1133 /* handle the regular call */
1134 set_Tuple_pred(call, pn_Call_X_regular, new_Jmp());
1136 /* For now, we cannot inline calls with value_base */
1137 set_Tuple_pred(call, pn_Call_P_value_res_base, new_Bad());
1139 /* Finally the exception control flow.
1140 We have two (three) possible situations:
1141 First if the Call branches to an exception handler: We need to add a Phi node to
1142 collect the memory containing the exception objects. Further we need
1143 to add another block to get a correct representation of this Phi. To
1144 this block we add a Jmp that resolves into the X output of the Call
1145 when the Call is turned into a tuple.
1146 Second the Call branches to End, the exception is not handled. Just
1147 add all inlined exception branches to the End node.
1148 Third: there is no Exception edge at all. Handle as case two. */
1149 if (exc_handling == exc_handler) {
1151 for (i = 0; i < arity; i++) {
1153 ret = get_irn_n(end_bl, i);
1154 irn = skip_Proj(ret);
1155 if (is_fragile_op(irn) || is_Raise(irn)) {
1156 cf_pred[n_exc] = ret;
1161 new_Block(n_exc, cf_pred); /* watch it: current_block is changed! */
1162 set_Tuple_pred(call, pn_Call_X_except, new_Jmp());
1163 /* The Phi for the memories with the exception objects */
1165 for (i = 0; i < arity; i++) {
1167 ret = skip_Proj(get_irn_n(end_bl, i));
1169 cf_pred[n_exc] = new_r_Proj(irg, get_nodes_block(ret), ret, mode_M, 3);
1171 } else if (is_fragile_op(ret)) {
1172 /* We rely that all cfops have the memory output at the same position. */
1173 cf_pred[n_exc] = new_r_Proj(irg, get_nodes_block(ret), ret, mode_M, 0);
1175 } else if (is_Raise(ret)) {
1176 cf_pred[n_exc] = new_r_Proj(irg, get_nodes_block(ret), ret, mode_M, 1);
1180 set_Tuple_pred(call, pn_Call_M_except, new_Phi(n_exc, cf_pred, mode_M));
1182 set_Tuple_pred(call, pn_Call_X_except, new_Bad());
1183 set_Tuple_pred(call, pn_Call_M_except, new_Bad());
1186 ir_node *main_end_bl;
1187 int main_end_bl_arity;
1188 ir_node **end_preds;
1190 /* assert(exc_handling == 1 || no exceptions. ) */
1192 for (i = 0; i < arity; i++) {
1193 ir_node *ret = get_irn_n(end_bl, i);
1194 ir_node *irn = skip_Proj(ret);
1196 if (is_fragile_op(irn) || is_Raise(irn)) {
1197 cf_pred[n_exc] = ret;
1201 main_end_bl = get_irg_end_block(irg);
1202 main_end_bl_arity = get_irn_arity(main_end_bl);
1203 end_preds = xmalloc((n_exc + main_end_bl_arity) * sizeof(*end_preds));
1205 for (i = 0; i < main_end_bl_arity; ++i)
1206 end_preds[i] = get_irn_n(main_end_bl, i);
1207 for (i = 0; i < n_exc; ++i)
1208 end_preds[main_end_bl_arity + i] = cf_pred[i];
1209 set_irn_in(main_end_bl, n_exc + main_end_bl_arity, end_preds);
1210 set_Tuple_pred(call, pn_Call_X_except, new_Bad());
1211 set_Tuple_pred(call, pn_Call_M_except, new_Bad());
1217 /* -- Turn CSE back on. -- */
1218 set_optimize(rem_opt);
1219 current_ir_graph = rem;
1224 /********************************************************************/
1225 /* Apply inlineing to small methods. */
1226 /********************************************************************/
1228 static struct obstack temp_obst;
1230 /** Represents a possible inlinable call in a graph. */
1231 typedef struct _call_entry call_entry;
1232 struct _call_entry {
1233 ir_node *call; /**< the Call node */
1234 ir_graph *callee; /**< the callee IR-graph called here */
1235 call_entry *next; /**< for linking the next one */
1236 int loop_depth; /**< the loop depth of this call */
1240 * environment for inlining small irgs
1242 typedef struct _inline_env_t {
1243 struct obstack obst; /**< an obstack where call_entries are allocated on. */
1244 call_entry *head; /**< the head of the call entry list */
1245 call_entry *tail; /**< the tail of the call entry list */
1249 * Returns the irg called from a Call node. If the irg is not
1250 * known, NULL is returned.
1252 * @param call the call node
1254 static ir_graph *get_call_called_irg(ir_node *call) {
1257 addr = get_Call_ptr(call);
1258 if (is_Global(addr)) {
1259 ir_entity *ent = get_Global_entity(addr);
1260 return get_entity_irg(ent);
1267 * Walker: Collect all calls to known graphs inside a graph.
1269 static void collect_calls(ir_node *call, void *env) {
1270 if (is_Call(call)) {
1271 ir_graph *called_irg = get_call_called_irg(call);
1273 if (called_irg != NULL) {
1274 /* The Call node calls a locally defined method. Remember to inline. */
1275 inline_env_t *ienv = env;
1276 call_entry *entry = obstack_alloc(&ienv->obst, sizeof(*entry));
1278 entry->callee = called_irg;
1280 entry->loop_depth = 0;
1282 if (ienv->tail == NULL)
1285 ienv->tail->next = entry;
1292 * Inlines all small methods at call sites where the called address comes
1293 * from a Const node that references the entity representing the called
1295 * The size argument is a rough measure for the code size of the method:
1296 * Methods where the obstack containing the firm graph is smaller than
1299 void inline_small_irgs(ir_graph *irg, int size) {
1300 ir_graph *rem = current_ir_graph;
1304 current_ir_graph = irg;
1305 /* Handle graph state */
1306 assert(get_irg_phase_state(irg) != phase_building);
1307 free_callee_info(irg);
1309 /* Find Call nodes to inline.
1310 (We can not inline during a walk of the graph, as inlineing the same
1311 method several times changes the visited flag of the walked graph:
1312 after the first inlineing visited of the callee equals visited of
1313 the caller. With the next inlineing both are increased.) */
1314 obstack_init(&env.obst);
1315 env.head = env.tail = NULL;
1316 irg_walk_graph(irg, NULL, collect_calls, &env);
1318 if (env.head != NULL) {
1319 /* There are calls to inline */
1320 collect_phiprojs(irg);
1321 for (entry = env.head; entry != NULL; entry = entry->next) {
1322 ir_graph *callee = entry->callee;
1323 if (((_obstack_memory_used(callee->obst) - (int)obstack_room(callee->obst)) < size) ||
1324 (get_irg_inline_property(callee) >= irg_inline_forced)) {
1325 inline_method(entry->call, callee);
1329 obstack_free(&env.obst, NULL);
1330 current_ir_graph = rem;
1334 * Environment for inlining irgs.
1337 int n_nodes; /**< Number of nodes in graph except Id, Tuple, Proj, Start, End. */
1338 int n_blocks; /**< Number of Blocks in graph without Start and End block. */
1339 int n_nodes_orig; /**< for statistics */
1340 int n_call_nodes; /**< Number of Call nodes in the graph. */
1341 int n_call_nodes_orig; /**< for statistics */
1342 int n_callers; /**< Number of known graphs that call this graphs. */
1343 int n_callers_orig; /**< for statistics */
1344 unsigned got_inline:1; /**< Set, if at least one call inside this graph was inlined. */
1345 unsigned local_vars:1; /**< Set, if a inlined function gets the address of an inlined variable. */
1346 unsigned recursive:1; /**< Set, if this function is self recursive. */
1347 call_entry *call_head; /**< The head of the list of all call nodes in this graph. */
1348 call_entry *call_tail; /**< The tail of the list of all call nodes in this graph .*/
1349 unsigned *local_weights; /**< Once allocated, the beneficial weight for transmitting local addresses. */
1353 * Allocate a new environment for inlining.
1355 static inline_irg_env *alloc_inline_irg_env(void) {
1356 inline_irg_env *env = obstack_alloc(&temp_obst, sizeof(*env));
1357 env->n_nodes = -2; /* do not count count Start, End */
1358 env->n_blocks = -2; /* do not count count Start, End Block */
1359 env->n_nodes_orig = -2; /* do not count Start, End */
1360 env->call_head = NULL;
1361 env->call_tail = NULL;
1362 env->n_call_nodes = 0;
1363 env->n_call_nodes_orig = 0;
1365 env->n_callers_orig = 0;
1366 env->got_inline = 0;
1367 env->local_vars = 0;
1369 env->local_weights = NULL;
1373 typedef struct walker_env {
1374 inline_irg_env *x; /**< the inline environment */
1375 call_entry *last_call; /**< points to the last inserted call */
1376 char ignore_runtime; /**< the ignore runtime flag */
1377 char ignore_callers; /**< if set, do change callers data */
1381 * post-walker: collect all calls in the inline-environment
1382 * of a graph and sum some statistics.
1384 static void collect_calls2(ir_node *call, void *ctx) {
1386 inline_irg_env *x = env->x;
1387 ir_opcode code = get_irn_opcode(call);
1391 /* count meaningful nodes in irg */
1392 if (code != iro_Proj && code != iro_Tuple && code != iro_Sync) {
1393 if (code != iro_Block) {
1401 if (code != iro_Call) return;
1403 /* check, if it's a runtime call */
1404 if (env->ignore_runtime) {
1405 ir_node *symc = get_Call_ptr(call);
1407 if (is_Global(symc)) {
1408 ir_entity *ent = get_Global_entity(symc);
1410 if (get_entity_additional_properties(ent) & mtp_property_runtime)
1415 /* collect all call nodes */
1417 ++x->n_call_nodes_orig;
1419 callee = get_call_called_irg(call);
1420 if (callee != NULL) {
1421 if (! env->ignore_callers) {
1422 inline_irg_env *callee_env = get_irg_link(callee);
1423 /* count all static callers */
1424 ++callee_env->n_callers;
1425 ++callee_env->n_callers_orig;
1427 if (callee == current_ir_graph)
1430 /* link it in the list of possible inlinable entries */
1431 entry = obstack_alloc(&temp_obst, sizeof(*entry));
1433 entry->callee = callee;
1435 entry->loop_depth = get_irn_loop(get_nodes_block(call))->depth;
1437 /* note: we use call_tail here as a pointer to the last inserted */
1438 if (x->call_head == NULL) {
1439 x->call_head = entry;
1441 if (entry->loop_depth == env->last_call->loop_depth) {
1442 /* same depth as the last one, enqueue after it */
1443 entry->next = env->last_call->next;
1444 env->last_call->next = entry;
1445 } else if (entry->loop_depth > x->call_head->loop_depth) {
1447 entry->next = x->call_head;
1448 x->call_head = entry;
1450 /* search the insertion point */
1453 for (p = x->call_head; p->next != NULL; p = p->next)
1454 if (entry->loop_depth > p->next->loop_depth)
1456 entry->next = p->next;
1460 env->last_call = entry;
1461 if (entry->next == NULL) {
1462 /* keep tail up to date */
1463 x->call_tail = entry;
1469 * Returns TRUE if the number of callers is 0 in the irg's environment,
1470 * hence this irg is a leave.
1472 INLINE static int is_leave(ir_graph *irg) {
1473 inline_irg_env *env = get_irg_link(irg);
1474 return env->n_call_nodes == 0;
1478 * Returns TRUE if the number of nodes in the callee is
1479 * smaller then size in the irg's environment.
1481 INLINE static int is_smaller(ir_graph *callee, int size) {
1482 inline_irg_env *env = get_irg_link(callee);
1483 return env->n_nodes < size;
1487 * Append the nodes of the list src to the nodes of the list in environment dst.
1489 static void append_call_list(inline_irg_env *dst, call_entry *src) {
1490 call_entry *entry, *nentry;
1492 /* Note that the src list points to Call nodes in the inlined graph, but
1493 we need Call nodes in our graph. Luckily the inliner leaves this information
1494 in the link field. */
1495 for (entry = src; entry != NULL; entry = entry->next) {
1496 nentry = obstack_alloc(&temp_obst, sizeof(*nentry));
1497 nentry->call = get_irn_link(entry->call);
1498 nentry->callee = entry->callee;
1499 nentry->next = NULL;
1500 nentry->loop_depth = entry->loop_depth;
1501 dst->call_tail->next = nentry;
1502 dst->call_tail = nentry;
1507 * Add the nodes of the list src in front to the nodes of the list dst.
1509 static call_entry *replace_entry_by_call_list(call_entry *dst, call_entry *src) {
1510 call_entry *entry, *nentry, *head, *tail;
1512 /* Note that the src list points to Call nodes in the inlined graph, but
1513 we need Call nodes in our graph. Luckily the inliner leaves this information
1514 in the link field. */
1516 for (entry = src; entry != NULL; entry = entry->next) {
1517 nentry = obstack_alloc(&temp_obst, sizeof(*nentry));
1518 nentry->call = get_irn_link(entry->call);
1519 nentry->callee = entry->callee;
1520 nentry->next = NULL;
1521 nentry->loop_depth = entry->loop_depth + dst->loop_depth;
1525 tail->next = nentry;
1528 /* skip the head of dst */
1530 tail->next = dst->next;
1538 * Inlines small leave methods at call sites where the called address comes
1539 * from a Const node that references the entity representing the called
1541 * The size argument is a rough measure for the code size of the method:
1542 * Methods where the obstack containing the firm graph is smaller than
1545 void inline_leave_functions(int maxsize, int leavesize, int size, int ignore_runtime) {
1546 inline_irg_env *env;
1552 call_entry *entry, *tail;
1553 const call_entry *centry;
1554 pmap *copied_graphs;
1555 pmap_entry *pm_entry;
1557 rem = current_ir_graph;
1558 obstack_init(&temp_obst);
1560 /* a map for the copied graphs, used to inline recursive calls */
1561 copied_graphs = pmap_create();
1563 /* extend all irgs by a temporary data structure for inlining. */
1564 n_irgs = get_irp_n_irgs();
1565 for (i = 0; i < n_irgs; ++i)
1566 set_irg_link(get_irp_irg(i), alloc_inline_irg_env());
1568 /* Precompute information in temporary data structure. */
1569 wenv.ignore_runtime = ignore_runtime;
1570 wenv.ignore_callers = 0;
1571 for (i = 0; i < n_irgs; ++i) {
1572 ir_graph *irg = get_irp_irg(i);
1574 assert(get_irg_phase_state(irg) != phase_building);
1575 free_callee_info(irg);
1577 assure_cf_loop(irg);
1578 wenv.x = get_irg_link(irg);
1579 irg_walk_graph(irg, NULL, collect_calls2, &wenv);
1582 /* -- and now inline. -- */
1584 /* Inline leaves recursively -- we might construct new leaves. */
1588 for (i = 0; i < n_irgs; ++i) {
1590 int phiproj_computed = 0;
1592 current_ir_graph = get_irp_irg(i);
1593 env = (inline_irg_env *)get_irg_link(current_ir_graph);
1596 for (entry = env->call_head; entry != NULL; entry = entry->next) {
1599 if (env->n_nodes > maxsize) break;
1602 callee = entry->callee;
1604 if (is_leave(callee) && (
1605 is_smaller(callee, leavesize) || (get_irg_inline_property(callee) >= irg_inline_forced))) {
1606 if (!phiproj_computed) {
1607 phiproj_computed = 1;
1608 collect_phiprojs(current_ir_graph);
1610 did_inline = inline_method(call, callee);
1613 inline_irg_env *callee_env = (inline_irg_env *)get_irg_link(callee);
1615 /* was inlined, must be recomputed */
1616 phiproj_computed = 0;
1618 /* Do some statistics */
1619 env->got_inline = 1;
1620 --env->n_call_nodes;
1621 env->n_nodes += callee_env->n_nodes;
1622 --callee_env->n_callers;
1624 /* remove this call from the list */
1626 tail->next = entry->next;
1628 env->call_head = entry->next;
1634 env->call_tail = tail;
1636 } while (did_inline);
1638 /* inline other small functions. */
1639 for (i = 0; i < n_irgs; ++i) {
1641 int phiproj_computed = 0;
1643 current_ir_graph = get_irp_irg(i);
1644 env = (inline_irg_env *)get_irg_link(current_ir_graph);
1646 /* note that the list of possible calls is updated during the process */
1648 for (entry = env->call_head; entry != NULL; entry = entry->next) {
1653 callee = entry->callee;
1655 e = pmap_find(copied_graphs, callee);
1658 * Remap callee if we have a copy.
1659 * FIXME: Should we do this only for recursive Calls ?
1664 if (((is_smaller(callee, size) && (env->n_nodes < maxsize)) || /* small function */
1665 (get_irg_inline_property(callee) >= irg_inline_forced))) {
1666 if (current_ir_graph == callee) {
1668 * Recursive call: we cannot directly inline because we cannot walk
1669 * the graph and change it. So we have to make a copy of the graph
1673 inline_irg_env *callee_env;
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 /* allocate new environment */
1687 callee_env = alloc_inline_irg_env();
1688 set_irg_link(copy, callee_env);
1690 assure_cf_loop(copy);
1691 wenv.x = callee_env;
1692 wenv.ignore_callers = 1;
1693 irg_walk_graph(copy, NULL, collect_calls2, &wenv);
1696 * Enter the entity of the original graph. This is needed
1697 * for inline_method(). However, note that ent->irg still points
1698 * to callee, NOT to copy.
1700 set_irg_entity(copy, get_irg_entity(callee));
1702 pmap_insert(copied_graphs, callee, copy);
1705 /* we have only one caller: the original graph */
1706 callee_env->n_callers = 1;
1707 callee_env->n_callers_orig = 1;
1709 if (! phiproj_computed) {
1710 phiproj_computed = 1;
1711 collect_phiprojs(current_ir_graph);
1713 did_inline = inline_method(call, callee);
1715 inline_irg_env *callee_env = (inline_irg_env *)get_irg_link(callee);
1717 /* was inlined, must be recomputed */
1718 phiproj_computed = 0;
1720 /* callee was inline. Append it's call list. */
1721 env->got_inline = 1;
1722 --env->n_call_nodes;
1723 append_call_list(env, callee_env->call_head);
1724 env->n_call_nodes += callee_env->n_call_nodes;
1725 env->n_nodes += callee_env->n_nodes;
1726 --callee_env->n_callers;
1728 /* after we have inlined callee, all called methods inside callee
1729 are now called once more */
1730 for (centry = callee_env->call_head; centry != NULL; centry = centry->next) {
1731 inline_irg_env *penv = get_irg_link(centry->callee);
1735 /* remove this call from the list */
1737 tail->next = entry->next;
1739 env->call_head = entry->next;
1745 env->call_tail = tail;
1748 for (i = 0; i < n_irgs; ++i) {
1749 irg = get_irp_irg(i);
1750 env = (inline_irg_env *)get_irg_link(irg);
1752 if (env->got_inline) {
1753 optimize_graph_df(irg);
1756 if (env->got_inline || (env->n_callers_orig != env->n_callers)) {
1757 DB((dbg, LEVEL_1, "Nodes:%3d ->%3d, calls:%3d ->%3d, callers:%3d ->%3d, -- %s\n",
1758 env->n_nodes_orig, env->n_nodes, env->n_call_nodes_orig, env->n_call_nodes,
1759 env->n_callers_orig, env->n_callers,
1760 get_entity_name(get_irg_entity(irg))));
1764 /* kill the copied graphs: we don't need them anymore */
1765 foreach_pmap(copied_graphs, pm_entry) {
1766 ir_graph *copy = pm_entry->value;
1768 /* reset the entity, otherwise it will be deleted in the next step ... */
1769 set_irg_entity(copy, NULL);
1770 free_ir_graph(copy);
1772 pmap_destroy(copied_graphs);
1774 obstack_free(&temp_obst, NULL);
1775 current_ir_graph = rem;
1779 * Calculate the parameter weights for transmitting the address of a local variable.
1781 static unsigned calc_method_local_weight(ir_node *arg) {
1783 unsigned v, weight = 0;
1785 for (i = get_irn_n_outs(arg) - 1; i >= 0; --i) {
1786 ir_node *succ = get_irn_out(arg, i);
1788 switch (get_irn_opcode(succ)) {
1791 /* Loads and Store can be removed */
1795 /* check if all args are constant */
1796 for (j = get_Sel_n_indexs(succ) - 1; j >= 0; --j) {
1797 ir_node *idx = get_Sel_index(succ, j);
1798 if (! is_Const(idx))
1801 /* Check users on this Sel. Note: if a 0 is returned here, there was
1802 some unsupported node. */
1803 v = calc_method_local_weight(succ);
1806 /* we can kill one Sel with constant indexes, this is cheap */
1810 /* when looking backward we might find Id nodes */
1811 weight += calc_method_local_weight(succ);
1814 /* unoptimized tuple */
1815 for (j = get_Tuple_n_preds(succ) - 1; j >= 0; --j) {
1816 ir_node *pred = get_Tuple_pred(succ, j);
1818 /* look for Proj(j) */
1819 for (k = get_irn_n_outs(succ) - 1; k >= 0; --k) {
1820 ir_node *succ_succ = get_irn_out(succ, k);
1821 if (is_Proj(succ_succ)) {
1822 if (get_Proj_proj(succ_succ) == j) {
1824 weight += calc_method_local_weight(succ_succ);
1827 /* this should NOT happen */
1835 /* any other node: unsupported yet or bad. */
1843 * Calculate the parameter weights for transmitting the address of a local variable.
1845 static void analyze_irg_local_weights(inline_irg_env *env, ir_graph *irg) {
1846 ir_entity *ent = get_irg_entity(irg);
1848 int nparams, i, proj_nr;
1849 ir_node *irg_args, *arg;
1851 mtp = get_entity_type(ent);
1852 nparams = get_method_n_params(mtp);
1854 /* allocate a new array. currently used as 'analysed' flag */
1855 env->local_weights = NEW_ARR_D(unsigned, &temp_obst, nparams);
1857 /* If the method haven't parameters we have nothing to do. */
1861 assure_irg_outs(irg);
1862 irg_args = get_irg_args(irg);
1863 for (i = get_irn_n_outs(irg_args) - 1; i >= 0; --i) {
1864 arg = get_irn_out(irg_args, i);
1865 proj_nr = get_Proj_proj(arg);
1866 env->local_weights[proj_nr] = calc_method_local_weight(arg);
1871 * Calculate the benefice for transmitting an local variable address.
1872 * After inlining, the local variable might be transformed into a
1873 * SSA variable by scalar_replacement().
1875 static unsigned get_method_local_adress_weight(ir_graph *callee, int pos) {
1876 inline_irg_env *env = get_irg_link(callee);
1878 if (env->local_weights != NULL) {
1879 if (pos < ARR_LEN(env->local_weights))
1880 return env->local_weights[pos];
1884 analyze_irg_local_weights(env, callee);
1886 if (pos < ARR_LEN(env->local_weights))
1887 return env->local_weights[pos];
1892 * Calculate a benefice value for inlining the given call.
1894 * @param call the call node we have to inspect
1895 * @param callee the called graph
1896 * @param local_adr set after return if an address of a local variable is
1897 * transmitted as a parameter
1899 static int calc_inline_benefice(ir_node *call, ir_graph *callee, unsigned *local_adr) {
1900 ir_entity *ent = get_irg_entity(callee);
1904 int i, n_params, all_const;
1907 inline_irg_env *curr_env, *callee_env;
1909 if (get_entity_additional_properties(ent) & mtp_property_noreturn) {
1910 /* do NOT inline noreturn calls */
1914 /* costs for every passed parameter */
1915 n_params = get_Call_n_params(call);
1916 mtp = get_entity_type(ent);
1917 cc = get_method_calling_convention(mtp);
1918 if (cc & cc_reg_param) {
1919 /* register parameter, smaller costs for register parameters */
1920 int max_regs = cc & ~cc_bits;
1922 if (max_regs < n_params)
1923 weight += max_regs * 2 + (n_params - max_regs) * 5;
1925 weight += n_params * 2;
1927 /* parameters are passed an stack */
1928 weight += 5 * n_params;
1931 /* constant parameters improve the benefice */
1932 frame_ptr = get_irg_frame(current_ir_graph);
1934 for (i = 0; i < n_params; ++i) {
1935 ir_node *param = get_Call_param(call, i);
1937 if (is_Const(param)) {
1938 weight += get_method_param_weight(ent, i);
1941 if (is_SymConst(param))
1942 weight += get_method_param_weight(ent, i);
1943 else if (is_Sel(param) && get_Sel_ptr(param) == frame_ptr) {
1945 * An address of a local variable is transmitted. After inlining,
1946 * scalar_replacement might be able to remove the local variable,
1949 v = get_method_local_adress_weight(callee, i);
1957 callee_env = get_irg_link(callee);
1958 if (get_entity_visibility(ent) == visibility_local &&
1959 callee_env->n_callers_orig == 1 &&
1960 callee != current_ir_graph) {
1961 /* we are the only caller, give big bonus */
1965 /* do not inline big functions */
1966 weight -= callee_env->n_nodes;
1968 /* reduce the benefice if the current function is already big */
1969 curr_env = get_irg_link(current_ir_graph);
1970 weight -= curr_env->n_nodes / 50;
1972 /* give a bonus for functions with one block */
1973 if (callee_env->n_blocks == 1)
1974 weight = weight * 3 / 2;
1976 /* and one for small non-recursive functions: we want them to be inlined in mostly every case */
1977 else if (callee_env->n_nodes < 20 && !callee_env->recursive)
1980 /* and finally for leaves: they do not increase the register pressure
1981 because of callee safe registers */
1982 else if (callee_env->n_call_nodes == 0)
1986 * Reduce the weight for recursive function IFF not all arguments are const.
1987 * inlining recursive functions is rarely good.
1989 if (callee_env->recursive && !all_const)
1993 * All arguments constant is probably a good sign, give an extra bonus
2002 * Heuristic inliner. Calculates a benefice value for every call and inlines
2003 * those calls with a value higher than the threshold.
2005 void inline_functions(int maxsize, int inline_threshold) {
2006 inline_irg_env *env;
2011 call_entry *curr_call, **last_call;
2012 const call_entry *centry;
2013 pmap *copied_graphs;
2014 pmap_entry *pm_entry;
2016 rem = current_ir_graph;
2017 obstack_init(&temp_obst);
2019 /* a map for the copied graphs, used to inline recursive calls */
2020 copied_graphs = pmap_create();
2022 /* extend all irgs by a temporary data structure for inlining. */
2023 n_irgs = get_irp_n_irgs();
2024 for (i = 0; i < n_irgs; ++i)
2025 set_irg_link(get_irp_irg(i), alloc_inline_irg_env());
2027 /* Precompute information in temporary data structure. */
2028 wenv.ignore_runtime = 0;
2029 wenv.ignore_callers = 0;
2030 for (i = 0; i < n_irgs; ++i) {
2031 ir_graph *irg = get_irp_irg(i);
2033 assert(get_irg_phase_state(irg) != phase_building);
2034 free_callee_info(irg);
2036 wenv.x = get_irg_link(irg);
2037 wenv.last_call = NULL;
2038 assure_cf_loop(irg);
2039 irg_walk_graph(irg, NULL, collect_calls2, &wenv);
2042 /* -- and now inline. -- */
2043 for (i = 0; i < n_irgs; ++i) {
2044 int phiproj_computed = 0;
2046 ir_graph *irg = get_irp_irg(i);
2048 current_ir_graph = irg;
2049 env = get_irg_link(irg);
2051 /* note that the list of possible calls is updated during the process */
2052 last_call = &env->call_head;
2053 for (curr_call = env->call_head; curr_call != NULL;) {
2059 if (env->n_nodes > maxsize) break;
2061 call = curr_call->call;
2062 callee = curr_call->callee;
2064 e = pmap_find(copied_graphs, callee);
2067 * Remap callee if we have a copy.
2068 * FIXME: Should we do this only for recursive Calls ?
2073 /* calculate the benefice on the original call to prevent excessive inlining */
2075 benefice = calc_inline_benefice(call, callee, &local_adr);
2076 DB((dbg, LEVEL_2, "In %+F Call %+F has benefice %d\n", irg, callee, benefice));
2078 if (benefice > -inline_threshold ||
2079 (get_irg_inline_property(callee) >= irg_inline_forced)) {
2080 if (current_ir_graph == callee) {
2082 * Recursive call: we cannot directly inline because we cannot walk
2083 * the graph and change it. So we have to make a copy of the graph
2087 inline_irg_env *callee_env;
2091 * No copy yet, create one.
2092 * Note that recursive methods are never leaves, so it is sufficient
2093 * to test this condition here.
2095 copy = create_irg_copy(callee);
2097 /* create_irg_copy() destroys the Proj links, recompute them */
2098 phiproj_computed = 0;
2100 /* allocate new environment */
2101 callee_env = alloc_inline_irg_env();
2102 set_irg_link(copy, callee_env);
2104 assure_cf_loop(copy);
2105 wenv.x = callee_env;
2106 wenv.ignore_callers = 1;
2107 irg_walk_graph(copy, NULL, collect_calls2, &wenv);
2110 * Enter the entity of the original graph. This is needed
2111 * for inline_method(). However, note that ent->irg still points
2112 * to callee, NOT to copy.
2114 set_irg_entity(copy, get_irg_entity(callee));
2116 pmap_insert(copied_graphs, callee, copy);
2119 /* we have only one caller: the original graph */
2120 callee_env->n_callers = 1;
2121 callee_env->n_callers_orig = 1;
2123 if (! phiproj_computed) {
2124 phiproj_computed = 1;
2125 collect_phiprojs(current_ir_graph);
2127 did_inline = inline_method(call, callee);
2129 inline_irg_env *callee_env = (inline_irg_env *)get_irg_link(callee);
2131 /* was inlined, must be recomputed */
2132 phiproj_computed = 0;
2134 /* after we have inlined callee, all called methods inside callee
2135 are now called once more */
2136 for (centry = callee_env->call_head; centry != NULL; centry = centry->next) {
2137 inline_irg_env *penv = get_irg_link(centry->callee);
2141 /* callee was inline. Append it's call list. */
2142 env->got_inline = 1;
2144 env->local_vars = 1;
2145 --env->n_call_nodes;
2146 curr_call = replace_entry_by_call_list(curr_call, callee_env->call_head);
2147 env->n_call_nodes += callee_env->n_call_nodes;
2148 env->n_nodes += callee_env->n_nodes;
2149 --callee_env->n_callers;
2151 /* remove the current call entry from the list */
2152 *last_call = curr_call;
2156 last_call = &curr_call->next;
2157 curr_call = curr_call->next;
2160 if (env->got_inline) {
2161 /* this irg got calls inlined: optimize it */
2163 /* scalar replacement does not work well with Tuple nodes, so optimize them away */
2164 optimize_graph_df(irg);
2166 if (env->local_vars) {
2167 if (scalar_replacement_opt(irg)) {
2168 optimize_graph_df(irg);
2173 if (env->got_inline || (env->n_callers_orig != env->n_callers)) {
2174 DB((dbg, LEVEL_1, "Nodes:%3d ->%3d, calls:%3d ->%3d, callers:%3d ->%3d, -- %s\n",
2175 env->n_nodes_orig, env->n_nodes, env->n_call_nodes_orig, env->n_call_nodes,
2176 env->n_callers_orig, env->n_callers,
2177 get_entity_name(get_irg_entity(irg))));
2181 /* kill the copied graphs: we don't need them anymore */
2182 foreach_pmap(copied_graphs, pm_entry) {
2183 ir_graph *copy = pm_entry->value;
2185 /* reset the entity, otherwise it will be deleted in the next step ... */
2186 set_irg_entity(copy, NULL);
2187 free_ir_graph(copy);
2189 pmap_destroy(copied_graphs);
2191 obstack_free(&temp_obst, NULL);
2192 current_ir_graph = rem;
2195 void firm_init_inline(void) {
2196 FIRM_DBG_REGISTER(dbg, "firm.opt.inline");