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"
54 #include "irbackedge_t.h"
55 #include "opt_inline_t.h"
60 #include "analyze_irg_args.h"
61 #include "iredges_t.h"
65 #include "iropt_dbg.h"
67 DEBUG_ONLY(static firm_dbg_module_t *dbg;)
69 /*------------------------------------------------------------------*/
70 /* Routines for dead node elimination / copying garbage collection */
72 /*------------------------------------------------------------------*/
75 * Remember the new node in the old node by using a field all nodes have.
77 #define set_new_node(oldn, newn) set_irn_link(oldn, newn)
80 * Get this new node, before the old node is forgotten.
82 #define get_new_node(oldn) get_irn_link(oldn)
85 * Check if a new node was set.
87 #define has_new_node(n) (get_new_node(n) != NULL)
90 * We use the block_visited flag to mark that we have computed the
91 * number of useful predecessors for this block.
92 * Further we encode the new arity in this flag in the old blocks.
93 * Remembering the arity is useful, as it saves a lot of pointer
94 * accesses. This function is called for all Phi and Block nodes
98 compute_new_arity(ir_node *b) {
99 int i, res, irn_arity;
102 irg_v = get_irg_block_visited(current_ir_graph);
103 block_v = get_Block_block_visited(b);
104 if (block_v >= irg_v) {
105 /* we computed the number of preds for this block and saved it in the
107 return block_v - irg_v;
109 /* compute the number of good predecessors */
110 res = irn_arity = get_irn_arity(b);
111 for (i = 0; i < irn_arity; i++)
112 if (is_Bad(get_irn_n(b, i))) res--;
113 /* save it in the flag. */
114 set_Block_block_visited(b, irg_v + res);
120 * Copies the node to the new obstack. The Ins of the new node point to
121 * the predecessors on the old obstack. For block/phi nodes not all
122 * predecessors might be copied. n->link points to the new node.
123 * For Phi and Block nodes the function allocates in-arrays with an arity
124 * only for useful predecessors. The arity is determined by counting
125 * the non-bad predecessors of the block.
127 * @param n The node to be copied
128 * @param env if non-NULL, the node number attribute will be copied to the new node
130 * Note: Also used for loop unrolling.
132 static void copy_node(ir_node *n, void *env) {
135 ir_op *op = get_irn_op(n);
139 /* node copied already */
141 } else if (op == op_Block) {
143 new_arity = compute_new_arity(n);
144 n->attr.block.graph_arr = NULL;
146 block = get_nodes_block(n);
148 new_arity = compute_new_arity(block);
150 new_arity = get_irn_arity(n);
153 nn = new_ir_node(get_irn_dbg_info(n),
160 /* Copy the attributes. These might point to additional data. If this
161 was allocated on the old obstack the pointers now are dangling. This
162 frees e.g. the memory of the graph_arr allocated in new_immBlock. */
163 if (op == op_Block) {
164 /* we cannot allow blocks WITHOUT macroblock input */
165 set_Block_MacroBlock(nn, get_Block_MacroBlock(n));
167 copy_node_attr(n, nn);
171 int copy_node_nr = env != NULL;
173 /* for easier debugging, we want to copy the node numbers too */
174 nn->node_nr = n->node_nr;
180 hook_dead_node_elim_subst(current_ir_graph, n, nn);
184 * Copies new predecessors of old node to new node remembered in link.
185 * Spare the Bad predecessors of Phi and Block nodes.
187 static void copy_preds(ir_node *n, void *env) {
192 nn = get_new_node(n);
195 /* copy the macro block header */
196 ir_node *mbh = get_Block_MacroBlock(n);
199 /* this block is a macroblock header */
200 set_Block_MacroBlock(nn, nn);
202 /* get the macro block header */
203 ir_node *nmbh = get_new_node(mbh);
204 assert(nmbh != NULL);
205 set_Block_MacroBlock(nn, nmbh);
208 /* Don't copy Bad nodes. */
210 irn_arity = get_irn_arity(n);
211 for (i = 0; i < irn_arity; i++) {
212 if (! is_Bad(get_irn_n(n, i))) {
213 ir_node *pred = get_irn_n(n, i);
214 set_irn_n(nn, j, get_new_node(pred));
218 /* repair the block visited flag from above misuse. Repair it in both
219 graphs so that the old one can still be used. */
220 set_Block_block_visited(nn, 0);
221 set_Block_block_visited(n, 0);
222 /* Local optimization could not merge two subsequent blocks if
223 in array contained Bads. Now it's possible.
224 We don't call optimize_in_place as it requires
225 that the fields in ir_graph are set properly. */
226 if ((get_opt_control_flow_straightening()) &&
227 (get_Block_n_cfgpreds(nn) == 1) &&
228 is_Jmp(get_Block_cfgpred(nn, 0))) {
229 ir_node *old = get_nodes_block(get_Block_cfgpred(nn, 0));
231 /* Jmp jumps into the block it is in -- deal self cycle. */
232 assert(is_Bad(get_new_node(get_irg_bad(current_ir_graph))));
233 exchange(nn, get_new_node(get_irg_bad(current_ir_graph)));
238 } else if (is_Phi(n) && get_irn_arity(n) > 0) {
239 /* Don't copy node if corresponding predecessor in block is Bad.
240 The Block itself should not be Bad. */
241 block = get_nodes_block(n);
242 set_nodes_block(nn, get_new_node(block));
244 irn_arity = get_irn_arity(n);
245 for (i = 0; i < irn_arity; i++) {
246 if (! is_Bad(get_irn_n(block, i))) {
247 ir_node *pred = get_irn_n(n, i);
248 set_irn_n(nn, j, get_new_node(pred));
249 /*if (is_backedge(n, i)) set_backedge(nn, j);*/
253 /* If the pre walker reached this Phi after the post walker visited the
254 block block_visited is > 0. */
255 set_Block_block_visited(get_nodes_block(n), 0);
256 /* Compacting the Phi's ins might generate Phis with only one
258 if (get_irn_arity(nn) == 1)
259 exchange(nn, get_irn_n(nn, 0));
261 irn_arity = get_irn_arity(n);
262 for (i = -1; i < irn_arity; i++)
263 set_irn_n(nn, i, get_new_node(get_irn_n(n, i)));
265 /* Now the new node is complete. We can add it to the hash table for CSE.
266 @@@ inlining aborts if we identify End. Why? */
268 add_identities(current_ir_graph->value_table, nn);
272 * Copies the graph recursively, compacts the keep-alives of the end node.
274 * @param irg the graph to be copied
275 * @param copy_node_nr If non-zero, the node number will be copied
277 static void copy_graph(ir_graph *irg, int copy_node_nr) {
278 ir_node *oe, *ne, *ob, *nb, *om, *nm; /* old end, new end, old bad, new bad, old NoMem, new NoMem */
279 ir_node *ka; /* keep alive */
283 /* Some nodes must be copied by hand, sigh */
284 vfl = get_irg_visited(irg);
285 set_irg_visited(irg, vfl + 1);
287 oe = get_irg_end(irg);
288 mark_irn_visited(oe);
289 /* copy the end node by hand, allocate dynamic in array! */
290 ne = new_ir_node(get_irn_dbg_info(oe),
297 /* Copy the attributes. Well, there might be some in the future... */
298 copy_node_attr(oe, ne);
299 set_new_node(oe, ne);
301 /* copy the Bad node */
302 ob = get_irg_bad(irg);
303 mark_irn_visited(ob);
304 nb = new_ir_node(get_irn_dbg_info(ob),
311 copy_node_attr(ob, nb);
312 set_new_node(ob, nb);
314 /* copy the NoMem node */
315 om = get_irg_no_mem(irg);
316 mark_irn_visited(om);
317 nm = new_ir_node(get_irn_dbg_info(om),
324 copy_node_attr(om, nm);
325 set_new_node(om, nm);
327 /* copy the live nodes */
328 set_irg_visited(irg, vfl);
329 irg_walk(get_nodes_block(oe), copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
331 /* Note: from yet, the visited flag of the graph is equal to vfl + 1 */
333 /* visit the anchors as well */
334 for (i = get_irg_n_anchors(irg) - 1; i >= 0; --i) {
335 ir_node *n = get_irg_anchor(irg, i);
337 if (n && (get_irn_visited(n) <= vfl)) {
338 set_irg_visited(irg, vfl);
339 irg_walk(n, copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
343 /* copy_preds for the end node ... */
344 set_nodes_block(ne, get_new_node(get_nodes_block(oe)));
346 /*- ... and now the keep alives. -*/
347 /* First pick the not marked block nodes and walk them. We must pick these
348 first as else we will oversee blocks reachable from Phis. */
349 irn_arity = get_End_n_keepalives(oe);
350 for (i = 0; i < irn_arity; i++) {
351 ka = get_End_keepalive(oe, i);
353 if (get_irn_visited(ka) <= vfl) {
354 /* We must keep the block alive and copy everything reachable */
355 set_irg_visited(irg, vfl);
356 irg_walk(ka, copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
358 add_End_keepalive(ne, get_new_node(ka));
362 /* Now pick other nodes. Here we will keep all! */
363 irn_arity = get_End_n_keepalives(oe);
364 for (i = 0; i < irn_arity; i++) {
365 ka = get_End_keepalive(oe, i);
367 if (get_irn_visited(ka) <= vfl) {
368 /* We didn't copy the node yet. */
369 set_irg_visited(irg, vfl);
370 irg_walk(ka, copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
372 add_End_keepalive(ne, get_new_node(ka));
376 /* start block sometimes only reached after keep alives */
377 set_nodes_block(nb, get_new_node(get_nodes_block(ob)));
378 set_nodes_block(nm, get_new_node(get_nodes_block(om)));
382 * Copies the graph reachable from current_ir_graph->end to the obstack
383 * in current_ir_graph and fixes the environment.
384 * Then fixes the fields in current_ir_graph containing nodes of the
387 * @param copy_node_nr If non-zero, the node number will be copied
390 copy_graph_env(int copy_node_nr) {
391 ir_graph *irg = current_ir_graph;
392 ir_node *old_end, *new_anchor;
395 /* remove end_except and end_reg nodes */
396 old_end = get_irg_end(irg);
397 set_irg_end_except (irg, old_end);
398 set_irg_end_reg (irg, old_end);
400 /* Not all nodes remembered in irg might be reachable
401 from the end node. Assure their link is set to NULL, so that
402 we can test whether new nodes have been computed. */
403 for (i = get_irg_n_anchors(irg) - 1; i >= 0; --i) {
404 ir_node *n = get_irg_anchor(irg, i);
406 set_new_node(n, NULL);
408 /* we use the block walk flag for removing Bads from Blocks ins. */
409 inc_irg_block_visited(irg);
412 copy_graph(irg, copy_node_nr);
415 old_end = get_irg_end(irg);
416 new_anchor = new_Anchor(irg);
418 for (i = get_irg_n_anchors(irg) - 1; i >= 0; --i) {
419 ir_node *n = get_irg_anchor(irg, i);
421 set_irn_n(new_anchor, i, get_new_node(n));
424 irg->anchor = new_anchor;
426 /* ensure the new anchor is placed in the endblock */
427 set_nodes_block(new_anchor, get_irg_end_block(irg));
431 * Copies all reachable nodes to a new obstack. Removes bad inputs
432 * from block nodes and the corresponding inputs from Phi nodes.
433 * Merges single exit blocks with single entry blocks and removes
435 * Adds all new nodes to a new hash table for CSE. Does not
436 * perform CSE, so the hash table might contain common subexpressions.
438 void dead_node_elimination(ir_graph *irg) {
440 #ifdef INTERPROCEDURAL_VIEW
441 int rem_ipview = get_interprocedural_view();
443 struct obstack *graveyard_obst = NULL;
444 struct obstack *rebirth_obst = NULL;
446 edges_deactivate(irg);
448 /* inform statistics that we started a dead-node elimination run */
449 hook_dead_node_elim(irg, 1);
451 /* Remember external state of current_ir_graph. */
452 rem = current_ir_graph;
453 current_ir_graph = irg;
454 #ifdef INTERPROCEDURAL_VIEW
455 set_interprocedural_view(0);
458 assert(get_irg_phase_state(irg) != phase_building);
460 /* Handle graph state */
461 free_callee_info(irg);
465 /* @@@ so far we loose loops when copying */
466 free_loop_information(irg);
468 set_irg_doms_inconsistent(irg);
470 /* A quiet place, where the old obstack can rest in peace,
471 until it will be cremated. */
472 graveyard_obst = irg->obst;
474 /* A new obstack, where the reachable nodes will be copied to. */
475 rebirth_obst = XMALLOC(struct obstack);
476 irg->obst = rebirth_obst;
477 obstack_init(irg->obst);
478 irg->last_node_idx = 0;
480 /* We also need a new value table for CSE */
481 del_identities(irg->value_table);
482 irg->value_table = new_identities();
484 /* Copy the graph from the old to the new obstack */
485 copy_graph_env(/*copy_node_nr=*/1);
487 /* Free memory from old unoptimized obstack */
488 obstack_free(graveyard_obst, 0); /* First empty the obstack ... */
489 xfree(graveyard_obst); /* ... then free it. */
491 /* inform statistics that the run is over */
492 hook_dead_node_elim(irg, 0);
494 current_ir_graph = rem;
495 #ifdef INTERPROCEDURAL_VIEW
496 set_interprocedural_view(rem_ipview);
501 * Relink bad predecessors of a block and store the old in array to the
502 * link field. This function is called by relink_bad_predecessors().
503 * The array of link field starts with the block operand at position 0.
504 * If block has bad predecessors, create a new in array without bad preds.
505 * Otherwise let in array untouched.
507 static void relink_bad_block_predecessors(ir_node *n, void *env) {
508 ir_node **new_in, *irn;
509 int i, new_irn_n, old_irn_arity, new_irn_arity = 0;
512 /* if link field of block is NULL, look for bad predecessors otherwise
513 this is already done */
514 if (is_Block(n) && get_irn_link(n) == NULL) {
515 /* save old predecessors in link field (position 0 is the block operand)*/
516 set_irn_link(n, get_irn_in(n));
518 /* count predecessors without bad nodes */
519 old_irn_arity = get_irn_arity(n);
520 for (i = 0; i < old_irn_arity; i++)
521 if (!is_Bad(get_irn_n(n, i)))
524 /* arity changing: set new predecessors without bad nodes */
525 if (new_irn_arity < old_irn_arity) {
526 /* Get new predecessor array. We do not resize the array, as we must
527 keep the old one to update Phis. */
528 new_in = NEW_ARR_D(ir_node *, current_ir_graph->obst, (new_irn_arity+1));
530 /* set new predecessors in array */
533 for (i = 0; i < old_irn_arity; i++) {
534 irn = get_irn_n(n, i);
536 new_in[new_irn_n] = irn;
537 is_backedge(n, i) ? set_backedge(n, new_irn_n-1) : set_not_backedge(n, new_irn_n-1);
541 /* ARR_SETLEN(int, n->attr.block.backedge, new_irn_arity); */
542 ARR_SHRINKLEN(n->attr.block.backedge, new_irn_arity);
544 } /* ir node has bad predecessors */
545 } /* Block is not relinked */
549 * Relinks Bad predecessors from Blocks and Phis called by walker
550 * remove_bad_predecesors(). If n is a Block, call
551 * relink_bad_block_redecessors(). If n is a Phi-node, call also the relinking
552 * function of Phi's Block. If this block has bad predecessors, relink preds
555 static void relink_bad_predecessors(ir_node *n, void *env) {
556 ir_node *block, **old_in;
557 int i, old_irn_arity, new_irn_arity;
559 /* relink bad predecessors of a block */
561 relink_bad_block_predecessors(n, env);
563 /* If Phi node relink its block and its predecessors */
565 /* Relink predecessors of phi's block */
566 block = get_nodes_block(n);
567 if (get_irn_link(block) == NULL)
568 relink_bad_block_predecessors(block, env);
570 old_in = (ir_node **)get_irn_link(block); /* Of Phi's Block */
571 old_irn_arity = ARR_LEN(old_in);
573 /* Relink Phi predecessors if count of predecessors changed */
574 if (old_irn_arity != ARR_LEN(get_irn_in(block))) {
575 /* set new predecessors in array
576 n->in[0] remains the same block */
578 for(i = 1; i < old_irn_arity; i++)
579 if (!is_Bad(old_in[i])) {
580 n->in[new_irn_arity] = n->in[i];
581 is_backedge(n, i) ? set_backedge(n, new_irn_arity) : set_not_backedge(n, new_irn_arity);
585 ARR_SETLEN(ir_node *, n->in, new_irn_arity);
586 ARR_SETLEN(int, n->attr.phi.u.backedge, new_irn_arity);
588 } /* n is a Phi node */
592 * Removes Bad Bad predecessors from Blocks and the corresponding
593 * inputs to Phi nodes as in dead_node_elimination but without
595 * On walking up set the link field to NULL, on walking down call
596 * relink_bad_predecessors() (This function stores the old in array
597 * to the link field and sets a new in array if arity of predecessors
600 void remove_bad_predecessors(ir_graph *irg) {
601 panic("Fix backedge handling first");
602 irg_walk_graph(irg, firm_clear_link, relink_bad_predecessors, NULL);
609 __)|_| | \_/ | \_/(/_ |_/\__|__
611 The following stuff implements a facility that automatically patches
612 registered ir_node pointers to the new node when a dead node elimination occurs.
615 struct _survive_dce_t {
619 hook_entry_t dead_node_elim;
620 hook_entry_t dead_node_elim_subst;
623 typedef struct _survive_dce_list_t {
624 struct _survive_dce_list_t *next;
626 } survive_dce_list_t;
628 static void dead_node_hook(void *context, ir_graph *irg, int start) {
629 survive_dce_t *sd = context;
632 /* Create a new map before the dead node elimination is performed. */
634 sd->new_places = pmap_create_ex(pmap_count(sd->places));
636 /* Patch back all nodes if dead node elimination is over and something is to be done. */
637 pmap_destroy(sd->places);
638 sd->places = sd->new_places;
639 sd->new_places = NULL;
644 * Hook called when dead node elimination replaces old by nw.
646 static void dead_node_subst_hook(void *context, ir_graph *irg, ir_node *old, ir_node *nw) {
647 survive_dce_t *sd = context;
648 survive_dce_list_t *list = pmap_get(sd->places, old);
651 /* If the node is to be patched back, write the new address to all registered locations. */
653 survive_dce_list_t *p;
655 for (p = list; p; p = p->next)
658 pmap_insert(sd->new_places, nw, list);
663 * Make a new Survive DCE environment.
665 survive_dce_t *new_survive_dce(void) {
666 survive_dce_t *res = XMALLOC(survive_dce_t);
667 obstack_init(&res->obst);
668 res->places = pmap_create();
669 res->new_places = NULL;
671 res->dead_node_elim.hook._hook_dead_node_elim = dead_node_hook;
672 res->dead_node_elim.context = res;
673 res->dead_node_elim.next = NULL;
675 res->dead_node_elim_subst.hook._hook_dead_node_elim_subst = dead_node_subst_hook;
676 res->dead_node_elim_subst.context = res;
677 res->dead_node_elim_subst.next = NULL;
679 register_hook(hook_dead_node_elim, &res->dead_node_elim);
680 register_hook(hook_dead_node_elim_subst, &res->dead_node_elim_subst);
685 * Free a Survive DCE environment.
687 void free_survive_dce(survive_dce_t *sd) {
688 obstack_free(&sd->obst, NULL);
689 pmap_destroy(sd->places);
690 unregister_hook(hook_dead_node_elim, &sd->dead_node_elim);
691 unregister_hook(hook_dead_node_elim_subst, &sd->dead_node_elim_subst);
696 * Register a node pointer to be patched upon DCE.
697 * When DCE occurs, the node pointer specified by @p place will be
698 * patched to the new address of the node it is pointing to.
700 * @param sd The Survive DCE environment.
701 * @param place The address of the node pointer.
703 void survive_dce_register_irn(survive_dce_t *sd, ir_node **place) {
704 if (*place != NULL) {
705 ir_node *irn = *place;
706 survive_dce_list_t *curr = pmap_get(sd->places, irn);
707 survive_dce_list_t *nw = obstack_alloc(&sd->obst, sizeof(nw[0]));
712 pmap_insert(sd->places, irn, nw);
716 /*--------------------------------------------------------------------*/
717 /* Functionality for inlining */
718 /*--------------------------------------------------------------------*/
721 * Copy node for inlineing. Updates attributes that change when
722 * inlineing but not for dead node elimination.
724 * Copies the node by calling copy_node() and then updates the entity if
725 * it's a local one. env must be a pointer of the frame type of the
726 * inlined procedure. The new entities must be in the link field of
729 static void copy_node_inline(ir_node *n, void *env) {
731 ir_type *frame_tp = (ir_type *)env;
735 nn = get_new_node (n);
737 if (get_entity_owner(get_Sel_entity(n)) == frame_tp) {
738 set_Sel_entity(nn, get_entity_link(get_Sel_entity(n)));
740 } else if (is_Block(n)) {
741 nn = get_new_node (n);
742 nn->attr.block.irg = current_ir_graph;
747 * Copies new predecessors of old node and move constants to
750 static void copy_preds_inline(ir_node *n, void *env) {
754 nn = skip_Id(get_new_node(n));
755 if (is_irn_constlike(nn)) {
756 /* move Constants into the start block */
757 set_nodes_block(nn, get_irg_start_block(current_ir_graph));
759 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 * 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 can call a function with less
866 parameters than needed. Currently we don't support this, although it would be
867 to use Unknown than. */
871 /* Argh, compiling C has some bad consequences:
872 the call type AND the method type might be different.
873 It is implementation defendant what happens in that case.
874 We support inlining, if the bitsize of the types matches AND
875 the same arithmetic is used. */
876 n_params = get_method_n_params(mtp);
877 for (i = n_params - 1; i >= 0; --i) {
878 ir_type *param_tp = get_method_param_type(mtp, i);
879 ir_type *arg_tp = get_method_param_type(ctp, i);
881 if (param_tp != arg_tp) {
882 ir_mode *pmode = get_type_mode(param_tp);
883 ir_mode *amode = get_type_mode(arg_tp);
885 if (pmode == NULL || amode == NULL)
887 if (get_mode_size_bits(pmode) != get_mode_size_bits(amode))
889 if (get_mode_arithmetic(pmode) != get_mode_arithmetic(amode))
891 /* otherwise we can simply "reinterpret" the bits */
895 irg = get_irn_irg(call);
898 * We cannot inline a recursive call. The graph must be copied before
899 * the call the inline_method() using create_irg_copy().
901 if (called_graph == irg)
905 * currently, we cannot inline two cases:
906 * - call with compound arguments
907 * - graphs that take the address of a parameter
909 if (! can_inline(call, called_graph))
912 rem = current_ir_graph;
913 current_ir_graph = irg;
915 DB((dbg, LEVEL_1, "Inlining %+F(%+F) into %+F\n", call, called_graph, irg));
917 /* -- Turn off optimizations, this can cause problems when allocating new nodes. -- */
918 rem_opt = get_opt_optimize();
921 /* Handle graph state */
922 assert(get_irg_phase_state(irg) != phase_building);
923 assert(get_irg_pinned(irg) == op_pin_state_pinned);
924 assert(get_irg_pinned(called_graph) == op_pin_state_pinned);
925 set_irg_outs_inconsistent(irg);
926 set_irg_extblk_inconsistent(irg);
927 set_irg_doms_inconsistent(irg);
928 set_irg_loopinfo_inconsistent(irg);
929 set_irg_callee_info_state(irg, irg_callee_info_inconsistent);
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 /* in[pn_Start_P_value_arg_base] = ??? */
983 assert(pn_Start_P_value_arg_base == pn_Start_max - 1 && "pn_Start_P_value_arg_base not supported, fix");
984 pre_call = new_Tuple(pn_Start_max - 1, in);
988 The new block gets the ins of the old block, pre_call and all its
989 predecessors and all Phi nodes. -- */
990 part_block(pre_call);
992 /* -- Prepare state for dead node elimination -- */
993 /* Visited flags in calling irg must be >= flag in called irg.
994 Else walker and arity computation will not work. */
995 if (get_irg_visited(irg) <= get_irg_visited(called_graph))
996 set_irg_visited(irg, get_irg_visited(called_graph) + 1);
997 if (get_irg_block_visited(irg) < get_irg_block_visited(called_graph))
998 set_irg_block_visited(irg, get_irg_block_visited(called_graph));
999 visited = get_irg_visited(irg);
1001 /* Set pre_call as new Start node in link field of the start node of
1002 calling graph and pre_calls block as new block for the start block
1004 Further mark these nodes so that they are not visited by the
1006 set_irn_link(get_irg_start(called_graph), pre_call);
1007 set_irn_visited(get_irg_start(called_graph), visited);
1008 set_irn_link(get_irg_start_block(called_graph), get_nodes_block(pre_call));
1009 set_irn_visited(get_irg_start_block(called_graph), visited);
1011 set_irn_link(get_irg_bad(called_graph), get_irg_bad(current_ir_graph));
1012 set_irn_visited(get_irg_bad(called_graph), visited);
1014 set_irn_link(get_irg_no_mem(called_graph), get_irg_no_mem(current_ir_graph));
1015 set_irn_visited(get_irg_no_mem(called_graph), visited);
1017 /* Initialize for compaction of in arrays */
1018 inc_irg_block_visited(irg);
1020 /* -- Replicate local entities of the called_graph -- */
1021 /* copy the entities. */
1022 called_frame = get_irg_frame_type(called_graph);
1023 curr_frame = get_irg_frame_type(irg);
1024 for (i = 0, n = get_class_n_members(called_frame); i < n; ++i) {
1025 ir_entity *new_ent, *old_ent;
1026 old_ent = get_class_member(called_frame, i);
1027 new_ent = copy_entity_own(old_ent, curr_frame);
1028 set_entity_link(old_ent, new_ent);
1031 /* visited is > than that of called graph. With this trick visited will
1032 remain unchanged so that an outer walker, e.g., searching the call nodes
1033 to inline, calling this inline will not visit the inlined nodes. */
1034 set_irg_visited(irg, get_irg_visited(irg)-1);
1036 /* -- Performing dead node elimination inlines the graph -- */
1037 /* Copies the nodes to the obstack of current_ir_graph. Updates links to new
1039 irg_walk(get_irg_end(called_graph), copy_node_inline, copy_preds_inline,
1040 get_irg_frame_type(called_graph));
1042 /* Repair called_graph */
1043 set_irg_visited(called_graph, get_irg_visited(irg));
1044 set_irg_block_visited(called_graph, get_irg_block_visited(irg));
1045 set_Block_block_visited(get_irg_start_block(called_graph), 0);
1047 /* -- Merge the end of the inlined procedure with the call site -- */
1048 /* We will turn the old Call node into a Tuple with the following
1051 0: Phi of all Memories of Return statements.
1052 1: Jmp from new Block that merges the control flow from all exception
1053 predecessors of the old end block.
1054 2: Tuple of all arguments.
1055 3: Phi of Exception memories.
1056 In case the old Call directly branches to End on an exception we don't
1057 need the block merging all exceptions nor the Phi of the exception
1061 /* -- Precompute some values -- */
1062 end_bl = get_new_node(get_irg_end_block(called_graph));
1063 end = get_new_node(get_irg_end(called_graph));
1064 arity = get_Block_n_cfgpreds(end_bl); /* arity = n_exc + n_ret */
1065 n_res = get_method_n_ress(get_Call_type(call));
1067 res_pred = XMALLOCN(ir_node*, n_res);
1068 cf_pred = XMALLOCN(ir_node*, arity);
1070 set_irg_current_block(irg, post_bl); /* just to make sure */
1072 /* -- archive keepalives -- */
1073 irn_arity = get_irn_arity(end);
1074 for (i = 0; i < irn_arity; i++) {
1075 ir_node *ka = get_End_keepalive(end, i);
1077 add_End_keepalive(get_irg_end(irg), ka);
1080 /* The new end node will die. We need not free as the in array is on the obstack:
1081 copy_node() only generated 'D' arrays. */
1083 /* -- Replace Return nodes by Jump nodes. -- */
1085 for (i = 0; i < arity; i++) {
1087 ret = get_Block_cfgpred(end_bl, i);
1088 if (is_Return(ret)) {
1089 cf_pred[n_ret] = new_r_Jmp(irg, get_nodes_block(ret));
1093 set_irn_in(post_bl, n_ret, cf_pred);
1095 /* -- Build a Tuple for all results of the method.
1096 Add Phi node if there was more than one Return. -- */
1097 turn_into_tuple(post_call, pn_Call_max);
1098 /* First the Memory-Phi */
1100 for (i = 0; i < arity; i++) {
1101 ret = get_Block_cfgpred(end_bl, i);
1102 if (is_Return(ret)) {
1103 cf_pred[n_ret] = get_Return_mem(ret);
1107 phi = new_Phi(n_ret, cf_pred, mode_M);
1108 set_Tuple_pred(call, pn_Call_M_regular, phi);
1109 /* Conserve Phi-list for further inlinings -- but might be optimized */
1110 if (get_nodes_block(phi) == post_bl) {
1111 set_irn_link(phi, get_irn_link(post_bl));
1112 set_irn_link(post_bl, phi);
1114 /* Now the real results */
1116 for (j = 0; j < n_res; j++) {
1118 for (i = 0; i < arity; i++) {
1119 ret = get_Block_cfgpred(end_bl, i);
1120 if (is_Return(ret)) {
1121 cf_pred[n_ret] = get_Return_res(ret, j);
1126 phi = new_Phi(n_ret, cf_pred, get_irn_mode(cf_pred[0]));
1130 /* Conserve Phi-list for further inlinings -- but might be optimized */
1131 if (get_nodes_block(phi) == post_bl) {
1132 set_Phi_next(phi, get_Block_phis(post_bl));
1133 set_Block_phis(post_bl, phi);
1136 set_Tuple_pred(call, pn_Call_T_result, new_Tuple(n_res, res_pred));
1138 set_Tuple_pred(call, pn_Call_T_result, new_Bad());
1140 /* handle the regular call */
1141 set_Tuple_pred(call, pn_Call_X_regular, new_Jmp());
1143 /* For now, we cannot inline calls with value_base */
1144 set_Tuple_pred(call, pn_Call_P_value_res_base, new_Bad());
1146 /* Finally the exception control flow.
1147 We have two (three) possible situations:
1148 First if the Call branches to an exception handler: We need to add a Phi node to
1149 collect the memory containing the exception objects. Further we need
1150 to add another block to get a correct representation of this Phi. To
1151 this block we add a Jmp that resolves into the X output of the Call
1152 when the Call is turned into a tuple.
1153 Second the Call branches to End, the exception is not handled. Just
1154 add all inlined exception branches to the End node.
1155 Third: there is no Exception edge at all. Handle as case two. */
1156 if (exc_handling == exc_handler) {
1158 for (i = 0; i < arity; i++) {
1160 ret = get_Block_cfgpred(end_bl, i);
1161 irn = skip_Proj(ret);
1162 if (is_fragile_op(irn) || is_Raise(irn)) {
1163 cf_pred[n_exc] = ret;
1168 new_Block(n_exc, cf_pred); /* watch it: current_block is changed! */
1169 set_Tuple_pred(call, pn_Call_X_except, new_Jmp());
1170 /* The Phi for the memories with the exception objects */
1172 for (i = 0; i < arity; i++) {
1174 ret = skip_Proj(get_Block_cfgpred(end_bl, i));
1176 cf_pred[n_exc] = new_r_Proj(irg, get_nodes_block(ret), ret, mode_M, 3);
1178 } else if (is_fragile_op(ret)) {
1179 /* We rely that all cfops have the memory output at the same position. */
1180 cf_pred[n_exc] = new_r_Proj(irg, get_nodes_block(ret), ret, mode_M, 0);
1182 } else if (is_Raise(ret)) {
1183 cf_pred[n_exc] = new_r_Proj(irg, get_nodes_block(ret), ret, mode_M, 1);
1187 set_Tuple_pred(call, pn_Call_M_except, new_Phi(n_exc, cf_pred, mode_M));
1189 set_Tuple_pred(call, pn_Call_X_except, new_Bad());
1190 set_Tuple_pred(call, pn_Call_M_except, new_Bad());
1193 ir_node *main_end_bl;
1194 int main_end_bl_arity;
1195 ir_node **end_preds;
1197 /* assert(exc_handling == 1 || no exceptions. ) */
1199 for (i = 0; i < arity; i++) {
1200 ir_node *ret = get_Block_cfgpred(end_bl, i);
1201 ir_node *irn = skip_Proj(ret);
1203 if (is_fragile_op(irn) || is_Raise(irn)) {
1204 cf_pred[n_exc] = ret;
1208 main_end_bl = get_irg_end_block(irg);
1209 main_end_bl_arity = get_irn_arity(main_end_bl);
1210 end_preds = XMALLOCN(ir_node*, n_exc + main_end_bl_arity);
1212 for (i = 0; i < main_end_bl_arity; ++i)
1213 end_preds[i] = get_irn_n(main_end_bl, i);
1214 for (i = 0; i < n_exc; ++i)
1215 end_preds[main_end_bl_arity + i] = cf_pred[i];
1216 set_irn_in(main_end_bl, n_exc + main_end_bl_arity, end_preds);
1217 set_Tuple_pred(call, pn_Call_X_except, new_Bad());
1218 set_Tuple_pred(call, pn_Call_M_except, new_Bad());
1224 /* -- Turn CSE back on. -- */
1225 set_optimize(rem_opt);
1226 current_ir_graph = rem;
1231 /********************************************************************/
1232 /* Apply inlining to small methods. */
1233 /********************************************************************/
1235 static struct obstack temp_obst;
1237 /** Represents a possible inlinable call in a graph. */
1238 typedef struct _call_entry {
1239 ir_node *call; /**< The Call node. */
1240 ir_graph *callee; /**< The callee IR-graph. */
1241 list_head list; /**< List head for linking the next one. */
1242 int loop_depth; /**< The loop depth of this call. */
1243 int benefice; /**< The calculated benefice of this call. */
1244 unsigned local_adr:1; /**< Set if this call gets an address of a local variable. */
1245 unsigned all_const:1; /**< Set if this call has only constant parameters. */
1249 * environment for inlining small irgs
1251 typedef struct _inline_env_t {
1252 struct obstack obst; /**< An obstack where call_entries are allocated on. */
1253 list_head calls; /**< The call entry list. */
1257 * Returns the irg called from a Call node. If the irg is not
1258 * known, NULL is returned.
1260 * @param call the call node
1262 static ir_graph *get_call_called_irg(ir_node *call) {
1265 addr = get_Call_ptr(call);
1266 if (is_Global(addr)) {
1267 ir_entity *ent = get_Global_entity(addr);
1268 return get_entity_irg(ent);
1275 * Walker: Collect all calls to known graphs inside a graph.
1277 static void collect_calls(ir_node *call, void *env) {
1279 if (is_Call(call)) {
1280 ir_graph *called_irg = get_call_called_irg(call);
1282 if (called_irg != NULL) {
1283 /* The Call node calls a locally defined method. Remember to inline. */
1284 inline_env_t *ienv = env;
1285 call_entry *entry = obstack_alloc(&ienv->obst, sizeof(*entry));
1287 entry->callee = called_irg;
1288 entry->loop_depth = 0;
1289 entry->benefice = 0;
1290 entry->local_adr = 0;
1291 entry->all_const = 0;
1293 list_add_tail(&entry->list, &ienv->calls);
1299 * Inlines all small methods at call sites where the called address comes
1300 * from a Const node that references the entity representing the called
1302 * The size argument is a rough measure for the code size of the method:
1303 * Methods where the obstack containing the firm graph is smaller than
1306 void inline_small_irgs(ir_graph *irg, int size) {
1307 ir_graph *rem = current_ir_graph;
1311 current_ir_graph = irg;
1312 /* Handle graph state */
1313 assert(get_irg_phase_state(irg) != phase_building);
1314 free_callee_info(irg);
1316 /* Find Call nodes to inline.
1317 (We can not inline during a walk of the graph, as inlining the same
1318 method several times changes the visited flag of the walked graph:
1319 after the first inlining visited of the callee equals visited of
1320 the caller. With the next inlining both are increased.) */
1321 obstack_init(&env.obst);
1322 INIT_LIST_HEAD(&env.calls);
1323 irg_walk_graph(irg, NULL, collect_calls, &env);
1325 if (! list_empty(&env.calls)) {
1326 /* There are calls to inline */
1327 collect_phiprojs(irg);
1329 list_for_each_entry(call_entry, entry, &env.calls, list) {
1330 ir_graph *callee = entry->callee;
1331 irg_inline_property prop = get_irg_inline_property(callee);
1333 if (prop == irg_inline_forbidden || get_irg_additional_properties(callee) & mtp_property_weak) {
1334 /* do not inline forbidden / weak graphs */
1338 if (prop >= irg_inline_forced ||
1339 _obstack_memory_used(callee->obst) - (int)obstack_room(callee->obst) < size) {
1340 inline_method(entry->call, callee);
1344 obstack_free(&env.obst, NULL);
1345 current_ir_graph = rem;
1349 * Environment for inlining irgs.
1352 list_head calls; /**< List of of all call nodes in this graph. */
1353 unsigned *local_weights; /**< Once allocated, the beneficial weight for transmitting local addresses. */
1354 unsigned n_nodes; /**< Number of nodes in graph except Id, Tuple, Proj, Start, End. */
1355 unsigned n_blocks; /**< Number of Blocks in graph without Start and End block. */
1356 unsigned n_nodes_orig; /**< for statistics */
1357 unsigned n_call_nodes; /**< Number of Call nodes in the graph. */
1358 unsigned n_call_nodes_orig; /**< for statistics */
1359 unsigned n_callers; /**< Number of known graphs that call this graphs. */
1360 unsigned n_callers_orig; /**< for statistics */
1361 unsigned got_inline:1; /**< Set, if at least one call inside this graph was inlined. */
1362 unsigned local_vars:1; /**< Set, if an inlined function got the address of a local variable. */
1363 unsigned recursive:1; /**< Set, if this function is self recursive. */
1367 * Allocate a new environment for inlining.
1369 static inline_irg_env *alloc_inline_irg_env(void) {
1370 inline_irg_env *env = obstack_alloc(&temp_obst, sizeof(*env));
1371 INIT_LIST_HEAD(&env->calls);
1372 env->local_weights = NULL;
1373 env->n_nodes = -2; /* do not count count Start, End */
1374 env->n_blocks = -2; /* do not count count Start, End Block */
1375 env->n_nodes_orig = -2; /* do not count Start, End */
1376 env->n_call_nodes = 0;
1377 env->n_call_nodes_orig = 0;
1379 env->n_callers_orig = 0;
1380 env->got_inline = 0;
1381 env->local_vars = 0;
1386 typedef struct walker_env {
1387 inline_irg_env *x; /**< the inline environment */
1388 char ignore_runtime; /**< the ignore runtime flag */
1389 char ignore_callers; /**< if set, do change callers data */
1393 * post-walker: collect all calls in the inline-environment
1394 * of a graph and sum some statistics.
1396 static void collect_calls2(ir_node *call, void *ctx) {
1398 inline_irg_env *x = env->x;
1399 ir_opcode code = get_irn_opcode(call);
1403 /* count meaningful nodes in irg */
1404 if (code != iro_Proj && code != iro_Tuple && code != iro_Sync) {
1405 if (code != iro_Block) {
1413 if (code != iro_Call) return;
1415 /* check, if it's a runtime call */
1416 if (env->ignore_runtime) {
1417 ir_node *symc = get_Call_ptr(call);
1419 if (is_Global(symc)) {
1420 ir_entity *ent = get_Global_entity(symc);
1422 if (get_entity_additional_properties(ent) & mtp_property_runtime)
1427 /* collect all call nodes */
1429 ++x->n_call_nodes_orig;
1431 callee = get_call_called_irg(call);
1432 if (callee != NULL) {
1433 if (! env->ignore_callers) {
1434 inline_irg_env *callee_env = get_irg_link(callee);
1435 /* count all static callers */
1436 ++callee_env->n_callers;
1437 ++callee_env->n_callers_orig;
1439 if (callee == current_ir_graph)
1442 /* link it in the list of possible inlinable entries */
1443 entry = obstack_alloc(&temp_obst, sizeof(*entry));
1445 entry->callee = callee;
1446 entry->loop_depth = get_irn_loop(get_nodes_block(call))->depth;
1447 entry->benefice = 0;
1448 entry->local_adr = 0;
1449 entry->all_const = 0;
1451 list_add_tail(&entry->list, &x->calls);
1456 * Returns TRUE if the number of callers is 0 in the irg's environment,
1457 * hence this irg is a leave.
1459 INLINE static int is_leave(ir_graph *irg) {
1460 inline_irg_env *env = get_irg_link(irg);
1461 return env->n_call_nodes == 0;
1465 * Returns TRUE if the number of nodes in the callee is
1466 * smaller then size in the irg's environment.
1468 INLINE static int is_smaller(ir_graph *callee, unsigned size) {
1469 inline_irg_env *env = get_irg_link(callee);
1470 return env->n_nodes < size;
1474 * Duplicate a call entry.
1476 * @param entry the original entry to duplicate
1477 * @param new_call the new call node
1478 * @param loop_depth_delta
1479 * delta value for the loop depth
1481 static call_entry *duplicate_call_entry(const call_entry *entry,
1482 ir_node *new_call, int loop_depth_delta) {
1483 call_entry *nentry = obstack_alloc(&temp_obst, sizeof(*nentry));
1484 nentry->call = new_call;
1485 nentry->callee = entry->callee;
1486 nentry->benefice = entry->benefice;
1487 nentry->loop_depth = entry->loop_depth + loop_depth_delta;
1488 nentry->local_adr = entry->local_adr;
1489 nentry->all_const = entry->all_const;
1495 * Append all call nodes of the source environment to the nodes of in the destination
1498 * @param dst destination environment
1499 * @param src source environment
1500 * @param loop_depth the loop depth of the call that is replaced by the src list
1502 static void append_call_list(inline_irg_env *dst, inline_irg_env *src, int loop_depth) {
1503 call_entry *entry, *nentry;
1505 /* Note that the src list points to Call nodes in the inlined graph, but
1506 we need Call nodes in our graph. Luckily the inliner leaves this information
1507 in the link field. */
1508 list_for_each_entry(call_entry, entry, &src->calls, list) {
1509 nentry = duplicate_call_entry(entry, get_irn_link(entry->call), loop_depth);
1510 list_add_tail(&nentry->list, &dst->calls);
1512 dst->n_call_nodes += src->n_call_nodes;
1513 dst->n_nodes += src->n_nodes;
1517 * Inlines small leave methods at call sites where the called address comes
1518 * from a Const node that references the entity representing the called
1520 * The size argument is a rough measure for the code size of the method:
1521 * Methods where the obstack containing the firm graph is smaller than
1524 void inline_leave_functions(unsigned maxsize, unsigned leavesize,
1525 unsigned size, int ignore_runtime)
1527 inline_irg_env *env;
1533 call_entry *entry, *next;
1534 const call_entry *centry;
1535 pmap *copied_graphs;
1536 pmap_entry *pm_entry;
1538 rem = current_ir_graph;
1539 obstack_init(&temp_obst);
1541 /* a map for the copied graphs, used to inline recursive calls */
1542 copied_graphs = pmap_create();
1544 /* extend all irgs by a temporary data structure for inlining. */
1545 n_irgs = get_irp_n_irgs();
1546 for (i = 0; i < n_irgs; ++i)
1547 set_irg_link(get_irp_irg(i), alloc_inline_irg_env());
1549 /* Pre-compute information in temporary data structure. */
1550 wenv.ignore_runtime = ignore_runtime;
1551 wenv.ignore_callers = 0;
1552 for (i = 0; i < n_irgs; ++i) {
1553 ir_graph *irg = get_irp_irg(i);
1555 assert(get_irg_phase_state(irg) != phase_building);
1556 free_callee_info(irg);
1558 assure_cf_loop(irg);
1559 wenv.x = get_irg_link(irg);
1560 irg_walk_graph(irg, NULL, collect_calls2, &wenv);
1563 /* -- and now inline. -- */
1565 /* Inline leaves recursively -- we might construct new leaves. */
1569 for (i = 0; i < n_irgs; ++i) {
1571 int phiproj_computed = 0;
1573 current_ir_graph = get_irp_irg(i);
1574 env = get_irg_link(current_ir_graph);
1576 list_for_each_entry_safe(call_entry, entry, next, &env->calls, list) {
1578 irg_inline_property prop;
1580 if (env->n_nodes > maxsize)
1584 callee = entry->callee;
1586 prop = get_irg_inline_property(callee);
1587 if (prop == irg_inline_forbidden || get_irg_additional_properties(callee) & mtp_property_weak) {
1588 /* do not inline forbidden / weak graphs */
1592 if (is_leave(callee) && (
1593 is_smaller(callee, leavesize) || prop >= irg_inline_forced)) {
1594 if (!phiproj_computed) {
1595 phiproj_computed = 1;
1596 collect_phiprojs(current_ir_graph);
1598 did_inline = inline_method(call, callee);
1601 inline_irg_env *callee_env = get_irg_link(callee);
1603 /* call was inlined, Phi/Projs for current graph must be recomputed */
1604 phiproj_computed = 0;
1606 /* Do some statistics */
1607 env->got_inline = 1;
1608 --env->n_call_nodes;
1609 env->n_nodes += callee_env->n_nodes;
1610 --callee_env->n_callers;
1612 /* remove this call from the list */
1613 list_del(&entry->list);
1619 } while (did_inline);
1621 /* inline other small functions. */
1622 for (i = 0; i < n_irgs; ++i) {
1624 int phiproj_computed = 0;
1626 current_ir_graph = get_irp_irg(i);
1627 env = get_irg_link(current_ir_graph);
1629 /* note that the list of possible calls is updated during the process */
1630 list_for_each_entry_safe(call_entry, entry, next, &env->calls, list) {
1631 irg_inline_property prop;
1636 callee = entry->callee;
1638 prop = get_irg_inline_property(callee);
1639 if (prop == irg_inline_forbidden || get_irg_additional_properties(callee) & mtp_property_weak) {
1640 /* do not inline forbidden / weak graphs */
1644 e = pmap_find(copied_graphs, callee);
1647 * Remap callee if we have a copy.
1648 * FIXME: Should we do this only for recursive Calls ?
1653 if (prop >= irg_inline_forced ||
1654 (is_smaller(callee, size) && env->n_nodes < maxsize) /* small function */) {
1655 if (current_ir_graph == callee) {
1657 * Recursive call: we cannot directly inline because we cannot walk
1658 * the graph and change it. So we have to make a copy of the graph
1662 inline_irg_env *callee_env;
1666 * No copy yet, create one.
1667 * Note that recursive methods are never leaves, so it is sufficient
1668 * to test this condition here.
1670 copy = create_irg_copy(callee);
1672 /* create_irg_copy() destroys the Proj links, recompute them */
1673 phiproj_computed = 0;
1675 /* allocate new environment */
1676 callee_env = alloc_inline_irg_env();
1677 set_irg_link(copy, callee_env);
1679 assure_cf_loop(copy);
1680 wenv.x = callee_env;
1681 wenv.ignore_callers = 1;
1682 irg_walk_graph(copy, NULL, collect_calls2, &wenv);
1685 * Enter the entity of the original graph. This is needed
1686 * for inline_method(). However, note that ent->irg still points
1687 * to callee, NOT to copy.
1689 set_irg_entity(copy, get_irg_entity(callee));
1691 pmap_insert(copied_graphs, callee, copy);
1694 /* we have only one caller: the original graph */
1695 callee_env->n_callers = 1;
1696 callee_env->n_callers_orig = 1;
1698 if (! phiproj_computed) {
1699 phiproj_computed = 1;
1700 collect_phiprojs(current_ir_graph);
1702 did_inline = inline_method(call, callee);
1704 inline_irg_env *callee_env = (inline_irg_env *)get_irg_link(callee);
1706 /* call was inlined, Phi/Projs for current graph must be recomputed */
1707 phiproj_computed = 0;
1709 /* callee was inline. Append it's call list. */
1710 env->got_inline = 1;
1711 --env->n_call_nodes;
1712 append_call_list(env, callee_env, entry->loop_depth);
1713 --callee_env->n_callers;
1715 /* after we have inlined callee, all called methods inside callee
1716 are now called once more */
1717 list_for_each_entry(call_entry, centry, &callee_env->calls, list) {
1718 inline_irg_env *penv = get_irg_link(centry->callee);
1722 /* remove this call from the list */
1723 list_del(&entry->list);
1730 for (i = 0; i < n_irgs; ++i) {
1731 irg = get_irp_irg(i);
1732 env = get_irg_link(irg);
1734 if (env->got_inline) {
1735 optimize_graph_df(irg);
1738 if (env->got_inline || (env->n_callers_orig != env->n_callers)) {
1739 DB((dbg, LEVEL_1, "Nodes:%3d ->%3d, calls:%3d ->%3d, callers:%3d ->%3d, -- %s\n",
1740 env->n_nodes_orig, env->n_nodes, env->n_call_nodes_orig, env->n_call_nodes,
1741 env->n_callers_orig, env->n_callers,
1742 get_entity_name(get_irg_entity(irg))));
1746 /* kill the copied graphs: we don't need them anymore */
1747 foreach_pmap(copied_graphs, pm_entry) {
1748 ir_graph *copy = pm_entry->value;
1750 /* reset the entity, otherwise it will be deleted in the next step ... */
1751 set_irg_entity(copy, NULL);
1752 free_ir_graph(copy);
1754 pmap_destroy(copied_graphs);
1756 obstack_free(&temp_obst, NULL);
1757 current_ir_graph = rem;
1761 * Calculate the parameter weights for transmitting the address of a local variable.
1763 static unsigned calc_method_local_weight(ir_node *arg) {
1765 unsigned v, weight = 0;
1767 for (i = get_irn_n_outs(arg) - 1; i >= 0; --i) {
1768 ir_node *succ = get_irn_out(arg, i);
1770 switch (get_irn_opcode(succ)) {
1773 /* Loads and Store can be removed */
1777 /* check if all args are constant */
1778 for (j = get_Sel_n_indexs(succ) - 1; j >= 0; --j) {
1779 ir_node *idx = get_Sel_index(succ, j);
1780 if (! is_Const(idx))
1783 /* Check users on this Sel. Note: if a 0 is returned here, there was
1784 some unsupported node. */
1785 v = calc_method_local_weight(succ);
1788 /* we can kill one Sel with constant indexes, this is cheap */
1792 /* when looking backward we might find Id nodes */
1793 weight += calc_method_local_weight(succ);
1796 /* unoptimized tuple */
1797 for (j = get_Tuple_n_preds(succ) - 1; j >= 0; --j) {
1798 ir_node *pred = get_Tuple_pred(succ, j);
1800 /* look for Proj(j) */
1801 for (k = get_irn_n_outs(succ) - 1; k >= 0; --k) {
1802 ir_node *succ_succ = get_irn_out(succ, k);
1803 if (is_Proj(succ_succ)) {
1804 if (get_Proj_proj(succ_succ) == j) {
1806 weight += calc_method_local_weight(succ_succ);
1809 /* this should NOT happen */
1817 /* any other node: unsupported yet or bad. */
1825 * Calculate the parameter weights for transmitting the address of a local variable.
1827 static void analyze_irg_local_weights(inline_irg_env *env, ir_graph *irg) {
1828 ir_entity *ent = get_irg_entity(irg);
1830 int nparams, i, proj_nr;
1831 ir_node *irg_args, *arg;
1833 mtp = get_entity_type(ent);
1834 nparams = get_method_n_params(mtp);
1836 /* allocate a new array. currently used as 'analysed' flag */
1837 env->local_weights = NEW_ARR_D(unsigned, &temp_obst, nparams);
1839 /* If the method haven't parameters we have nothing to do. */
1843 assure_irg_outs(irg);
1844 irg_args = get_irg_args(irg);
1845 for (i = get_irn_n_outs(irg_args) - 1; i >= 0; --i) {
1846 arg = get_irn_out(irg_args, i);
1847 proj_nr = get_Proj_proj(arg);
1848 env->local_weights[proj_nr] = calc_method_local_weight(arg);
1853 * Calculate the benefice for transmitting an local variable address.
1854 * After inlining, the local variable might be transformed into a
1855 * SSA variable by scalar_replacement().
1857 static unsigned get_method_local_adress_weight(ir_graph *callee, int pos) {
1858 inline_irg_env *env = get_irg_link(callee);
1860 if (env->local_weights != NULL) {
1861 if (pos < ARR_LEN(env->local_weights))
1862 return env->local_weights[pos];
1866 analyze_irg_local_weights(env, callee);
1868 if (pos < ARR_LEN(env->local_weights))
1869 return env->local_weights[pos];
1874 * Calculate a benefice value for inlining the given call.
1876 * @param call the call node we have to inspect
1877 * @param callee the called graph
1879 static int calc_inline_benefice(call_entry *entry, ir_graph *callee)
1881 ir_node *call = entry->call;
1882 ir_entity *ent = get_irg_entity(callee);
1886 int i, n_params, all_const;
1888 irg_inline_property prop;
1890 inline_irg_env *callee_env;
1892 prop = get_irg_inline_property(callee);
1893 if (prop == irg_inline_forbidden) {
1894 DB((dbg, LEVEL_2, "In %+F Call to %+F: inlining forbidden\n",
1896 return entry->benefice = INT_MIN;
1899 if (get_irg_additional_properties(callee) & (mtp_property_noreturn | mtp_property_weak)) {
1900 DB((dbg, LEVEL_2, "In %+F Call to %+F: not inlining noreturn or weak\n",
1902 return entry->benefice = INT_MIN;
1905 /* costs for every passed parameter */
1906 n_params = get_Call_n_params(call);
1907 mtp = get_entity_type(ent);
1908 cc = get_method_calling_convention(mtp);
1909 if (cc & cc_reg_param) {
1910 /* register parameter, smaller costs for register parameters */
1911 int max_regs = cc & ~cc_bits;
1913 if (max_regs < n_params)
1914 weight += max_regs * 2 + (n_params - max_regs) * 5;
1916 weight += n_params * 2;
1918 /* parameters are passed an stack */
1919 weight += 5 * n_params;
1922 /* constant parameters improve the benefice */
1923 frame_ptr = get_irg_frame(current_ir_graph);
1925 for (i = 0; i < n_params; ++i) {
1926 ir_node *param = get_Call_param(call, i);
1928 if (is_Const(param)) {
1929 weight += get_method_param_weight(ent, i);
1932 if (is_SymConst(param))
1933 weight += get_method_param_weight(ent, i);
1934 else if (is_Sel(param) && get_Sel_ptr(param) == frame_ptr) {
1936 * An address of a local variable is transmitted. After
1937 * inlining, scalar_replacement might be able to remove the
1938 * local variable, so honor this.
1940 v = get_method_local_adress_weight(callee, i);
1943 entry->local_adr = 1;
1947 entry->all_const = all_const;
1949 callee_env = get_irg_link(callee);
1950 if (callee_env->n_callers == 1 &&
1951 callee != current_ir_graph &&
1952 get_entity_visibility(ent) == visibility_local) {
1956 /* give a bonus for functions with one block */
1957 if (callee_env->n_blocks == 1)
1958 weight = weight * 3 / 2;
1960 /* and one for small non-recursive functions: we want them to be inlined in mostly every case */
1961 if (callee_env->n_nodes < 30 && !callee_env->recursive)
1964 /* and finally for leaves: they do not increase the register pressure
1965 because of callee safe registers */
1966 if (callee_env->n_call_nodes == 0)
1969 /** it's important to inline inner loops first */
1970 if (entry->loop_depth > 30)
1971 weight += 30 * 1024;
1973 weight += entry->loop_depth * 1024;
1976 * All arguments constant is probably a good sign, give an extra bonus
1981 return entry->benefice = weight;
1984 static ir_graph **irgs;
1985 static int last_irg;
1988 * Callgraph walker, collect all visited graphs.
1990 static void callgraph_walker(ir_graph *irg, void *data) {
1992 irgs[last_irg++] = irg;
1996 * Creates an inline order for all graphs.
1998 * @return the list of graphs.
2000 static ir_graph **create_irg_list(void) {
2001 ir_entity **free_methods;
2003 int n_irgs = get_irp_n_irgs();
2005 cgana(&arr_len, &free_methods);
2006 xfree(free_methods);
2008 compute_callgraph();
2011 irgs = XMALLOCNZ(ir_graph*, n_irgs);
2013 callgraph_walk(NULL, callgraph_walker, NULL);
2014 assert(n_irgs == last_irg);
2020 * Push a call onto the priority list if its benefice is big enough.
2022 * @param pqueue the priority queue of calls
2023 * @param call the call entry
2024 * @param inlien_threshold
2025 * the threshold value
2027 static void maybe_push_call(pqueue_t *pqueue, call_entry *call,
2028 int inline_threshold)
2030 ir_graph *callee = call->callee;
2031 irg_inline_property prop = get_irg_inline_property(callee);
2032 int benefice = calc_inline_benefice(call, callee);
2034 DB((dbg, LEVEL_2, "In %+F Call %+F to %+F has benefice %d\n",
2035 get_irn_irg(call->call), call->call, callee, benefice));
2037 if (prop < irg_inline_forced && benefice < inline_threshold) {
2041 pqueue_put(pqueue, call, benefice);
2045 * Try to inline calls into a graph.
2047 * @param irg the graph into which we inline
2048 * @param maxsize do NOT inline if the size of irg gets
2049 * bigger than this amount
2050 * @param inline_threshold
2051 * threshold value for inline decision
2052 * @param copied_graphs
2053 * map containing copied of recursive graphs
2055 static void inline_into(ir_graph *irg, unsigned maxsize,
2056 int inline_threshold, pmap *copied_graphs)
2058 int phiproj_computed = 0;
2059 inline_irg_env *env = get_irg_link(irg);
2060 call_entry *curr_call;
2064 if (env->n_call_nodes == 0)
2067 if (env->n_nodes > maxsize) {
2068 DB((dbg, LEVEL_2, "%+F: too big (%d)\n", irg, env->n_nodes));
2072 current_ir_graph = irg;
2074 /* put irgs into the pqueue */
2075 pqueue = new_pqueue();
2077 list_for_each_entry(call_entry, curr_call, &env->calls, list) {
2078 assert(is_Call(curr_call->call));
2079 maybe_push_call(pqueue, curr_call, inline_threshold);
2082 /* note that the list of possible calls is updated during the process */
2083 while (!pqueue_empty(pqueue)) {
2085 call_entry *curr_call = pqueue_pop_front(pqueue);
2086 ir_graph *callee = curr_call->callee;
2087 ir_node *call_node = curr_call->call;
2088 inline_irg_env *callee_env = get_irg_link(callee);
2089 irg_inline_property prop = get_irg_inline_property(callee);
2091 const call_entry *centry;
2094 if ((prop < irg_inline_forced) && env->n_nodes + callee_env->n_nodes > maxsize) {
2095 DB((dbg, LEVEL_2, "%+F: too big (%d) + %+F (%d)\n", irg,
2096 env->n_nodes, callee, callee_env->n_nodes));
2100 e = pmap_find(copied_graphs, callee);
2102 int benefice = curr_call->benefice;
2104 * Reduce the weight for recursive function IFF not all arguments are const.
2105 * inlining recursive functions is rarely good.
2107 if (!curr_call->all_const)
2109 if (benefice < inline_threshold)
2113 * Remap callee if we have a copy.
2116 callee_env = get_irg_link(callee);
2119 if (current_ir_graph == callee) {
2121 * Recursive call: we cannot directly inline because we cannot
2122 * walk the graph and change it. So we have to make a copy of
2125 int benefice = curr_call->benefice;
2129 * Reduce the weight for recursive function IFF not all arguments are const.
2130 * inlining recursive functions is rarely good.
2132 if (!curr_call->all_const)
2134 if (benefice < inline_threshold)
2138 * No copy yet, create one.
2139 * Note that recursive methods are never leaves, so it is
2140 * sufficient to test this condition here.
2142 copy = create_irg_copy(callee);
2144 /* create_irg_copy() destroys the Proj links, recompute them */
2145 phiproj_computed = 0;
2147 /* allocate a new environment */
2148 callee_env = alloc_inline_irg_env();
2149 set_irg_link(copy, callee_env);
2151 assure_cf_loop(copy);
2152 wenv.x = callee_env;
2153 wenv.ignore_callers = 1;
2154 irg_walk_graph(copy, NULL, collect_calls2, &wenv);
2157 * Enter the entity of the original graph. This is needed
2158 * for inline_method(). However, note that ent->irg still points
2159 * to callee, NOT to copy.
2161 set_irg_entity(copy, get_irg_entity(callee));
2163 pmap_insert(copied_graphs, callee, copy);
2166 /* we have only one caller: the original graph */
2167 callee_env->n_callers = 1;
2168 callee_env->n_callers_orig = 1;
2170 if (! phiproj_computed) {
2171 phiproj_computed = 1;
2172 collect_phiprojs(current_ir_graph);
2174 did_inline = inline_method(call_node, callee);
2178 /* call was inlined, Phi/Projs for current graph must be recomputed */
2179 phiproj_computed = 0;
2181 /* remove it from the caller list */
2182 list_del(&curr_call->list);
2184 /* callee was inline. Append it's call list. */
2185 env->got_inline = 1;
2186 if (curr_call->local_adr)
2187 env->local_vars = 1;
2188 --env->n_call_nodes;
2190 /* we just generate a bunch of new calls */
2191 loop_depth = curr_call->loop_depth;
2192 list_for_each_entry(call_entry, centry, &callee_env->calls, list) {
2193 inline_irg_env *penv = get_irg_link(centry->callee);
2195 call_entry *new_entry;
2197 /* after we have inlined callee, all called methods inside
2198 * callee are now called once more */
2201 /* Note that the src list points to Call nodes in the inlined graph,
2202 * but we need Call nodes in our graph. Luckily the inliner leaves
2203 * this information in the link field. */
2204 new_call = get_irn_link(centry->call);
2205 assert(is_Call(new_call));
2207 new_entry = duplicate_call_entry(centry, new_call, loop_depth);
2208 list_add_tail(&new_entry->list, &env->calls);
2209 maybe_push_call(pqueue, new_entry, inline_threshold);
2212 env->n_call_nodes += callee_env->n_call_nodes;
2213 env->n_nodes += callee_env->n_nodes;
2214 --callee_env->n_callers;
2221 * Heuristic inliner. Calculates a benefice value for every call and inlines
2222 * those calls with a value higher than the threshold.
2224 void inline_functions(unsigned maxsize, int inline_threshold) {
2225 inline_irg_env *env;
2229 pmap *copied_graphs;
2230 pmap_entry *pm_entry;
2233 rem = current_ir_graph;
2234 obstack_init(&temp_obst);
2236 irgs = create_irg_list();
2238 /* a map for the copied graphs, used to inline recursive calls */
2239 copied_graphs = pmap_create();
2241 /* extend all irgs by a temporary data structure for inlining. */
2242 n_irgs = get_irp_n_irgs();
2243 for (i = 0; i < n_irgs; ++i)
2244 set_irg_link(irgs[i], alloc_inline_irg_env());
2246 /* Pre-compute information in temporary data structure. */
2247 wenv.ignore_runtime = 0;
2248 wenv.ignore_callers = 0;
2249 for (i = 0; i < n_irgs; ++i) {
2250 ir_graph *irg = irgs[i];
2252 free_callee_info(irg);
2254 wenv.x = get_irg_link(irg);
2255 assure_cf_loop(irg);
2256 irg_walk_graph(irg, NULL, collect_calls2, &wenv);
2259 /* -- and now inline. -- */
2260 for (i = 0; i < n_irgs; ++i) {
2261 ir_graph *irg = irgs[i];
2263 inline_into(irg, maxsize, inline_threshold, copied_graphs);
2266 for (i = 0; i < n_irgs; ++i) {
2267 ir_graph *irg = irgs[i];
2269 env = get_irg_link(irg);
2270 if (env->got_inline) {
2271 /* this irg got calls inlined: optimize it */
2272 if (get_opt_combo()) {
2273 if (env->local_vars) {
2274 scalar_replacement_opt(irg);
2278 if (env->local_vars) {
2279 if (scalar_replacement_opt(irg)) {
2280 optimize_graph_df(irg);
2286 if (env->got_inline || (env->n_callers_orig != env->n_callers)) {
2287 DB((dbg, LEVEL_1, "Nodes:%3d ->%3d, calls:%3d ->%3d, callers:%3d ->%3d, -- %s\n",
2288 env->n_nodes_orig, env->n_nodes, env->n_call_nodes_orig, env->n_call_nodes,
2289 env->n_callers_orig, env->n_callers,
2290 get_entity_name(get_irg_entity(irg))));
2294 /* kill the copied graphs: we don't need them anymore */
2295 foreach_pmap(copied_graphs, pm_entry) {
2296 ir_graph *copy = pm_entry->value;
2298 /* reset the entity, otherwise it will be deleted in the next step ... */
2299 set_irg_entity(copy, NULL);
2300 free_ir_graph(copy);
2302 pmap_destroy(copied_graphs);
2306 obstack_free(&temp_obst, NULL);
2307 current_ir_graph = rem;
2310 void firm_init_inline(void) {
2311 FIRM_DBG_REGISTER(dbg, "firm.opt.inline");