3 * File name: ir/ir/irgopt.c
4 * Purpose: Optimizations for a whole ir graph, i.e., a procedure.
5 * Author: Christian Schaefer, Goetz Lindenmaier
6 * Modified by: Sebastian Felis
9 * Copyright: (c) 1998-2003 Universität Karlsruhe
10 * Licence: This file protected by GPL - GNU GENERAL PUBLIC LICENSE.
23 # include "irnode_t.h"
24 # include "irgraph_t.h"
32 # include "pdeq.h" /* Fuer code placement */
35 # include "irbackedge_t.h"
36 # include "irflag_t.h"
38 /* Defined in iropt.c */
39 pset *new_identities (void);
40 void del_identities (pset *value_table);
41 void add_identities (pset *value_table, ir_node *node);
43 /*------------------------------------------------------------------*/
44 /* apply optimizations of iropt to all nodes. */
45 /*------------------------------------------------------------------*/
47 static void init_link (ir_node *n, void *env) {
48 set_irn_link(n, NULL);
51 #if 0 /* Old version. Avoids Ids.
52 This is not necessary: we do a postwalk, and get_irn_n
53 removes ids anyways. So it's much cheaper to call the
54 optimization less often and use the exchange() algorithm. */
56 optimize_in_place_wrapper (ir_node *n, void *env) {
58 ir_node *optimized, *old;
60 irn_arity = intern_get_irn_arity(n);
61 for (i = 0; i < irn_arity; i++) {
62 /* get_irn_n skips Id nodes, so comparison old != optimized does not
63 show all optimizations. Therefore always set new predecessor. */
64 old = intern_get_irn_intra_n(n, i);
65 optimized = optimize_in_place_2(old);
66 set_irn_n(n, i, optimized);
69 if (intern_get_irn_op(n) == op_Block) {
70 optimized = optimize_in_place_2(n);
71 if (optimized != n) exchange (n, optimized);
76 optimize_in_place_wrapper (ir_node *n, void *env) {
77 ir_node *optimized = optimize_in_place_2(n);
78 if (optimized != n) exchange (n, optimized);
85 local_optimize_graph (ir_graph *irg) {
86 ir_graph *rem = current_ir_graph;
87 current_ir_graph = irg;
89 /* Handle graph state */
90 assert(get_irg_phase_state(irg) != phase_building);
91 if (get_opt_global_cse())
92 set_irg_pinned(current_ir_graph, floats);
93 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
94 set_irg_outs_inconsistent(current_ir_graph);
95 if (get_irg_dom_state(current_ir_graph) == dom_consistent)
96 set_irg_dom_inconsistent(current_ir_graph);
98 /* Clean the value_table in irg for the cse. */
99 del_identities(irg->value_table);
100 irg->value_table = new_identities();
102 /* walk over the graph */
103 irg_walk(irg->end, init_link, optimize_in_place_wrapper, NULL);
105 current_ir_graph = rem;
108 /*------------------------------------------------------------------*/
109 /* Routines for dead node elimination / copying garbage collection */
110 /* of the obstack. */
111 /*------------------------------------------------------------------*/
114 * Remember the new node in the old node by using a field all nodes have.
117 set_new_node (ir_node *old, ir_node *new)
123 * Get this new node, before the old node is forgotton.
125 static INLINE ir_node *
126 get_new_node (ir_node * n)
132 * We use the block_visited flag to mark that we have computed the
133 * number of useful predecessors for this block.
134 * Further we encode the new arity in this flag in the old blocks.
135 * Remembering the arity is useful, as it saves a lot of pointer
136 * accesses. This function is called for all Phi and Block nodes
140 compute_new_arity(ir_node *b) {
141 int i, res, irn_arity;
144 irg_v = get_irg_block_visited(current_ir_graph);
145 block_v = get_Block_block_visited(b);
146 if (block_v >= irg_v) {
147 /* we computed the number of preds for this block and saved it in the
149 return block_v - irg_v;
151 /* compute the number of good predecessors */
152 res = irn_arity = intern_get_irn_arity(b);
153 for (i = 0; i < irn_arity; i++)
154 if (intern_get_irn_opcode(intern_get_irn_n(b, i)) == iro_Bad) res--;
155 /* save it in the flag. */
156 set_Block_block_visited(b, irg_v + res);
161 /* TODO: add an ir_op operation */
162 static INLINE void new_backedge_info(ir_node *n) {
163 switch(intern_get_irn_opcode(n)) {
165 n->attr.block.cg_backedge = NULL;
166 n->attr.block.backedge = new_backedge_arr(current_ir_graph->obst, intern_get_irn_arity(n));
169 n->attr.phi_backedge = new_backedge_arr(current_ir_graph->obst, intern_get_irn_arity(n));
172 n->attr.filter.backedge = new_backedge_arr(current_ir_graph->obst, intern_get_irn_arity(n));
179 * Copies the node to the new obstack. The Ins of the new node point to
180 * the predecessors on the old obstack. For block/phi nodes not all
181 * predecessors might be copied. n->link points to the new node.
182 * For Phi and Block nodes the function allocates in-arrays with an arity
183 * only for useful predecessors. The arity is determined by counting
184 * the non-bad predecessors of the block.
187 copy_node (ir_node *n, void *env) {
191 /* The end node looses it's flexible in array. This doesn't matter,
192 as dead node elimination builds End by hand, inlineing doesn't use
194 //assert(n->op == op_End || ((_ARR_DESCR(n->in))->cookie != ARR_F_MAGIC));
196 if (intern_get_irn_opcode(n) == iro_Block) {
198 new_arity = compute_new_arity(n);
199 n->attr.block.graph_arr = NULL;
201 block = get_nodes_Block(n);
202 if (intern_get_irn_opcode(n) == iro_Phi) {
203 new_arity = compute_new_arity(block);
205 new_arity = intern_get_irn_arity(n);
208 nn = new_ir_node(get_irn_dbg_info(n),
211 intern_get_irn_op(n),
212 intern_get_irn_mode(n),
215 /* Copy the attributes. These might point to additional data. If this
216 was allocated on the old obstack the pointers now are dangling. This
217 frees e.g. the memory of the graph_arr allocated in new_immBlock. */
219 new_backedge_info(nn);
222 /* printf("\n old node: "); DDMSG2(n);
223 printf(" new node: "); DDMSG2(nn); */
228 * Copies new predecessors of old node to new node remembered in link.
229 * Spare the Bad predecessors of Phi and Block nodes.
232 copy_preds (ir_node *n, void *env) {
236 nn = get_new_node(n);
238 /* printf("\n old node: "); DDMSG2(n);
239 printf(" new node: "); DDMSG2(nn);
240 printf(" arities: old: %d, new: %d\n", intern_get_irn_arity(n), intern_get_irn_arity(nn)); */
242 if (intern_get_irn_opcode(n) == iro_Block) {
243 /* Don't copy Bad nodes. */
245 irn_arity = intern_get_irn_arity(n);
246 for (i = 0; i < irn_arity; i++)
247 if (intern_get_irn_opcode(intern_get_irn_n(n, i)) != iro_Bad) {
248 set_irn_n (nn, j, get_new_node(intern_get_irn_n(n, i)));
249 /*if (is_backedge(n, i)) set_backedge(nn, j);*/
252 /* repair the block visited flag from above misuse. Repair it in both
253 graphs so that the old one can still be used. */
254 set_Block_block_visited(nn, 0);
255 set_Block_block_visited(n, 0);
256 /* Local optimization could not merge two subsequent blocks if
257 in array contained Bads. Now it's possible.
258 We don't call optimize_in_place as it requires
259 that the fields in ir_graph are set properly. */
260 if ((get_opt_control_flow_straightening()) &&
261 (get_Block_n_cfgpreds(nn) == 1) &&
262 (intern_get_irn_op(get_Block_cfgpred(nn, 0)) == op_Jmp))
263 exchange(nn, get_nodes_Block(get_Block_cfgpred(nn, 0)));
264 } else if (intern_get_irn_opcode(n) == iro_Phi) {
265 /* Don't copy node if corresponding predecessor in block is Bad.
266 The Block itself should not be Bad. */
267 block = get_nodes_Block(n);
268 set_irn_n (nn, -1, get_new_node(block));
270 irn_arity = intern_get_irn_arity(n);
271 for (i = 0; i < irn_arity; i++)
272 if (intern_get_irn_opcode(intern_get_irn_n(block, i)) != iro_Bad) {
273 set_irn_n (nn, j, get_new_node(intern_get_irn_n(n, i)));
274 /*if (is_backedge(n, i)) set_backedge(nn, j);*/
277 /* If the pre walker reached this Phi after the post walker visited the
278 block block_visited is > 0. */
279 set_Block_block_visited(get_nodes_Block(n), 0);
280 /* Compacting the Phi's ins might generate Phis with only one
282 if (intern_get_irn_arity(n) == 1)
283 exchange(n, intern_get_irn_n(n, 0));
285 irn_arity = intern_get_irn_arity(n);
286 for (i = -1; i < irn_arity; i++)
287 set_irn_n (nn, i, get_new_node(intern_get_irn_n(n, i)));
289 /* Now the new node is complete. We can add it to the hash table for cse.
290 @@@ inlinening aborts if we identify End. Why? */
291 if(intern_get_irn_op(nn) != op_End)
292 add_identities (current_ir_graph->value_table, nn);
296 * Copies the graph recursively, compacts the keepalive of the end node.
300 ir_node *oe, *ne; /* old end, new end */
301 ir_node *ka; /* keep alive */
304 oe = get_irg_end(current_ir_graph);
305 /* copy the end node by hand, allocate dynamic in array! */
306 ne = new_ir_node(get_irn_dbg_info(oe),
313 /* Copy the attributes. Well, there might be some in the future... */
315 set_new_node(oe, ne);
317 /* copy the live nodes */
318 irg_walk(get_nodes_Block(oe), copy_node, copy_preds, NULL);
319 /* copy_preds for the end node ... */
320 set_nodes_Block(ne, get_new_node(get_nodes_Block(oe)));
322 /*- ... and now the keep alives. -*/
323 /* First pick the not marked block nodes and walk them. We must pick these
324 first as else we will oversee blocks reachable from Phis. */
325 irn_arity = intern_get_irn_arity(oe);
326 for (i = 0; i < irn_arity; i++) {
327 ka = intern_get_irn_intra_n(oe, i);
328 if ((intern_get_irn_op(ka) == op_Block) &&
329 (get_irn_visited(ka) < get_irg_visited(current_ir_graph))) {
330 /* We must keep the block alive and copy everything reachable */
331 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
332 irg_walk(ka, copy_node, copy_preds, NULL);
333 add_End_keepalive(ne, get_new_node(ka));
337 /* Now pick the Phis. Here we will keep all! */
338 irn_arity = intern_get_irn_arity(oe);
339 for (i = 0; i < irn_arity; i++) {
340 ka = intern_get_irn_intra_n(oe, i);
341 if ((intern_get_irn_op(ka) == op_Phi)) {
342 if (get_irn_visited(ka) < get_irg_visited(current_ir_graph)) {
343 /* We didn't copy the Phi yet. */
344 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
345 irg_walk(ka, copy_node, copy_preds, NULL);
347 add_End_keepalive(ne, get_new_node(ka));
353 * Copies the graph reachable from current_ir_graph->end to the obstack
354 * in current_ir_graph and fixes the environment.
355 * Then fixes the fields in current_ir_graph containing nodes of the
359 copy_graph_env (void) {
361 /* Not all nodes remembered in current_ir_graph might be reachable
362 from the end node. Assure their link is set to NULL, so that
363 we can test whether new nodes have been computed. */
364 set_irn_link(get_irg_frame (current_ir_graph), NULL);
365 set_irn_link(get_irg_globals(current_ir_graph), NULL);
366 set_irn_link(get_irg_args (current_ir_graph), NULL);
368 /* we use the block walk flag for removing Bads from Blocks ins. */
369 inc_irg_block_visited(current_ir_graph);
374 /* fix the fields in current_ir_graph */
375 old_end = get_irg_end(current_ir_graph);
376 set_irg_end (current_ir_graph, get_new_node(old_end));
378 set_irg_end_block (current_ir_graph, get_new_node(get_irg_end_block(current_ir_graph)));
379 if (get_irn_link(get_irg_frame(current_ir_graph)) == NULL) {
380 copy_node (get_irg_frame(current_ir_graph), NULL);
381 copy_preds(get_irg_frame(current_ir_graph), NULL);
383 if (get_irn_link(get_irg_globals(current_ir_graph)) == NULL) {
384 copy_node (get_irg_globals(current_ir_graph), NULL);
385 copy_preds(get_irg_globals(current_ir_graph), NULL);
387 if (get_irn_link(get_irg_args(current_ir_graph)) == NULL) {
388 copy_node (get_irg_args(current_ir_graph), NULL);
389 copy_preds(get_irg_args(current_ir_graph), NULL);
391 set_irg_start (current_ir_graph, get_new_node(get_irg_start(current_ir_graph)));
393 set_irg_start_block(current_ir_graph,
394 get_new_node(get_irg_start_block(current_ir_graph)));
395 set_irg_frame (current_ir_graph, get_new_node(get_irg_frame(current_ir_graph)));
396 set_irg_globals(current_ir_graph, get_new_node(get_irg_globals(current_ir_graph)));
397 set_irg_args (current_ir_graph, get_new_node(get_irg_args(current_ir_graph)));
398 if (get_irn_link(get_irg_bad(current_ir_graph)) == NULL) {
399 copy_node(get_irg_bad(current_ir_graph), NULL);
400 copy_preds(get_irg_bad(current_ir_graph), NULL);
402 set_irg_bad(current_ir_graph, get_new_node(get_irg_bad(current_ir_graph)));
403 /* GL removed: we need unknown with mode for analyses.
404 if (get_irn_link(get_irg_unknown(current_ir_graph)) == NULL) {
405 copy_node(get_irg_unknown(current_ir_graph), NULL);
406 copy_preds(get_irg_unknown(current_ir_graph), NULL);
408 set_irg_unknown(current_ir_graph, get_new_node(get_irg_unknown(current_ir_graph)));
413 * Copies all reachable nodes to a new obstack. Removes bad inputs
414 * from block nodes and the corresponding inputs from Phi nodes.
415 * Merges single exit blocks with single entry blocks and removes
417 * Adds all new nodes to a new hash table for cse. Does not
418 * perform cse, so the hash table might contain common subexpressions.
420 /* Amroq call this emigrate() */
422 dead_node_elimination(ir_graph *irg) {
424 int rem_ipview = interprocedural_view;
425 struct obstack *graveyard_obst = NULL;
426 struct obstack *rebirth_obst = NULL;
428 /* Remember external state of current_ir_graph. */
429 rem = current_ir_graph;
430 current_ir_graph = irg;
431 interprocedural_view = 0;
433 /* Handle graph state */
434 assert(get_irg_phase_state(current_ir_graph) != phase_building);
435 assert(get_irg_callee_info_state(current_ir_graph) == irg_callee_info_none);
436 free_outs(current_ir_graph);
438 /* @@@ so far we loose loops when copying */
439 free_loop_information(current_ir_graph);
441 if (get_opt_optimize() && get_opt_dead_node_elimination()) {
443 /* A quiet place, where the old obstack can rest in peace,
444 until it will be cremated. */
445 graveyard_obst = irg->obst;
447 /* A new obstack, where the reachable nodes will be copied to. */
448 rebirth_obst = (struct obstack *) xmalloc (sizeof (struct obstack));
449 current_ir_graph->obst = rebirth_obst;
450 obstack_init (current_ir_graph->obst);
452 /* We also need a new hash table for cse */
453 del_identities (irg->value_table);
454 irg->value_table = new_identities ();
456 /* Copy the graph from the old to the new obstack */
459 /* Free memory from old unoptimized obstack */
460 obstack_free(graveyard_obst, 0); /* First empty the obstack ... */
461 xfree (graveyard_obst); /* ... then free it. */
464 current_ir_graph = rem;
465 interprocedural_view = rem_ipview;
469 * Relink bad predeseccors of a block and store the old in array to the
470 * link field. This function is called by relink_bad_predecessors().
471 * The array of link field starts with the block operand at position 0.
472 * If block has bad predecessors, create a new in array without bad preds.
473 * Otherwise let in array untouched.
475 static void relink_bad_block_predecessors(ir_node *n, void *env) {
476 ir_node **new_in, *irn;
477 int i, new_irn_n, old_irn_arity, new_irn_arity = 0;
479 /* if link field of block is NULL, look for bad predecessors otherwise
480 this is allready done */
481 if (intern_get_irn_op(n) == op_Block &&
482 get_irn_link(n) == NULL) {
484 /* save old predecessors in link field (position 0 is the block operand)*/
485 set_irn_link(n, (void *)get_irn_in(n));
487 /* count predecessors without bad nodes */
488 old_irn_arity = intern_get_irn_arity(n);
489 for (i = 0; i < old_irn_arity; i++)
490 if (!is_Bad(intern_get_irn_n(n, i))) new_irn_arity++;
492 /* arity changing: set new predecessors without bad nodes */
493 if (new_irn_arity < old_irn_arity) {
494 /* get new predecessor array without Block predecessor */
495 new_in = NEW_ARR_D (ir_node *, current_ir_graph->obst, (new_irn_arity+1));
497 /* set new predeseccors in array */
500 for (i = 1; i < old_irn_arity; i++) {
501 irn = intern_get_irn_n(n, i);
502 if (!is_Bad(irn)) new_in[new_irn_n++] = irn;
505 } /* ir node has bad predecessors */
507 } /* Block is not relinked */
511 * Relinks Bad predecesors from Bocks and Phis called by walker
512 * remove_bad_predecesors(). If n is a Block, call
513 * relink_bad_block_redecessors(). If n is a Phinode, call also the relinking
514 * function of Phi's Block. If this block has bad predecessors, relink preds
517 static void relink_bad_predecessors(ir_node *n, void *env) {
518 ir_node *block, **old_in;
519 int i, old_irn_arity, new_irn_arity;
521 /* relink bad predeseccors of a block */
522 if (intern_get_irn_op(n) == op_Block)
523 relink_bad_block_predecessors(n, env);
525 /* If Phi node relink its block and its predecessors */
526 if (intern_get_irn_op(n) == op_Phi) {
528 /* Relink predeseccors of phi's block */
529 block = get_nodes_Block(n);
530 if (get_irn_link(block) == NULL)
531 relink_bad_block_predecessors(block, env);
533 old_in = (ir_node **)get_irn_link(block); /* Of Phi's Block */
534 old_irn_arity = ARR_LEN(old_in);
536 /* Relink Phi predeseccors if count of predeseccors changed */
537 if (old_irn_arity != ARR_LEN(get_irn_in(block))) {
538 /* set new predeseccors in array
539 n->in[0] remains the same block */
541 for(i = 1; i < old_irn_arity; i++)
542 if (!is_Bad((ir_node *)old_in[i])) n->in[new_irn_arity++] = n->in[i];
544 ARR_SETLEN(ir_node *, n->in, new_irn_arity);
547 } /* n is a Phi node */
551 * Removes Bad Bad predecesors from Blocks and the corresponding
552 * inputs to Phi nodes as in dead_node_elimination but without
554 * On walking up set the link field to NULL, on walking down call
555 * relink_bad_predecessors() (This function stores the old in array
556 * to the link field and sets a new in array if arity of predecessors
559 void remove_bad_predecessors(ir_graph *irg) {
560 irg_walk_graph(irg, init_link, relink_bad_predecessors, NULL);
564 /*--------------------------------------------------------------------*/
565 /* Funcionality for inlining */
566 /*--------------------------------------------------------------------*/
569 * Copy node for inlineing. Updates attributes that change when
570 * inlineing but not for dead node elimination.
572 * Copies the node by calling copy_node and then updates the entity if
573 * it's a local one. env must be a pointer of the frame type of the
574 * inlined procedure. The new entities must be in the link field of
578 copy_node_inline (ir_node *n, void *env) {
580 type *frame_tp = (type *)env;
583 if (intern_get_irn_op(n) == op_Sel) {
584 new = get_new_node (n);
585 assert(intern_get_irn_op(new) == op_Sel);
586 if (get_entity_owner(get_Sel_entity(n)) == frame_tp) {
587 set_Sel_entity(new, get_entity_link(get_Sel_entity(n)));
589 } else if (intern_get_irn_op(n) == op_Block) {
590 new = get_new_node (n);
591 new->attr.block.irg = current_ir_graph;
596 void inline_method(ir_node *call, ir_graph *called_graph) {
598 ir_node *post_call, *post_bl;
600 ir_node *end, *end_bl;
604 int arity, n_ret, n_exc, n_res, i, j, rem_opt, irn_arity;
608 if ( !(get_irg_inline_property(called_graph) == irg_inline_forced) && (!get_opt_optimize() || !get_opt_inline() ||
609 (get_irg_inline_property(called_graph) == irg_inline_forbidden))) return;
611 /* -- Turn off optimizations, this can cause problems when allocating new nodes. -- */
612 rem_opt = get_opt_optimize();
615 /* Handle graph state */
616 assert(get_irg_phase_state(current_ir_graph) != phase_building);
617 assert(get_irg_pinned(current_ir_graph) == pinned);
618 assert(get_irg_pinned(called_graph) == pinned);
619 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
620 set_irg_outs_inconsistent(current_ir_graph);
622 /* -- Check preconditions -- */
623 assert(get_irn_op(call) == op_Call);
624 /* @@@ does not work for InterfaceIII.java after cgana
625 assert(get_Call_type(call) == get_entity_type(get_irg_ent(called_graph)));
626 assert(smaller_type(get_entity_type(get_irg_ent(called_graph)),
627 get_Call_type(call)));
629 assert(get_type_tpop(get_Call_type(call)) == type_method);
630 if (called_graph == current_ir_graph) {
631 set_optimize(rem_opt);
635 /* -- Decide how to handle exception control flow: Is there a handler
636 for the Call node, or do we branch directly to End on an exception?
637 exc_handling: 0 There is a handler.
639 2 Exception handling not represented in Firm. -- */
641 ir_node *proj, *Mproj = NULL, *Xproj = NULL;
642 for (proj = (ir_node *)get_irn_link(call); proj; proj = (ir_node *)get_irn_link(proj)) {
643 assert(get_irn_op(proj) == op_Proj);
644 if (get_Proj_proj(proj) == pn_Call_X_except) Xproj = proj;
645 if (get_Proj_proj(proj) == pn_Call_M_except) Mproj = proj;
647 if (Mproj) { assert(Xproj); exc_handling = 0; } // Mproj
648 else if (Xproj) { exc_handling = 1; } //!Mproj && Xproj
649 else { exc_handling = 2; } //!Mproj && !Xproj
654 the procedure and later replaces the Start node of the called graph.
655 Post_call is the old Call node and collects the results of the called
656 graph. Both will end up being a tuple. -- */
657 post_bl = get_nodes_Block(call);
658 set_irg_current_block(current_ir_graph, post_bl);
659 /* XxMxPxP of Start + parameter of Call */
661 in[1] = get_Call_mem(call);
662 in[2] = get_irg_frame(current_ir_graph);
663 in[3] = get_irg_globals(current_ir_graph);
664 in[4] = new_Tuple (get_Call_n_params(call), get_Call_param_arr(call));
665 pre_call = new_Tuple(5, in);
669 The new block gets the ins of the old block, pre_call and all its
670 predecessors and all Phi nodes. -- */
671 part_block(pre_call);
673 /* -- Prepare state for dead node elimination -- */
674 /* Visited flags in calling irg must be >= flag in called irg.
675 Else walker and arity computation will not work. */
676 if (get_irg_visited(current_ir_graph) <= get_irg_visited(called_graph))
677 set_irg_visited(current_ir_graph, get_irg_visited(called_graph)+1);
678 if (get_irg_block_visited(current_ir_graph)< get_irg_block_visited(called_graph))
679 set_irg_block_visited(current_ir_graph, get_irg_block_visited(called_graph));
680 /* Set pre_call as new Start node in link field of the start node of
681 calling graph and pre_calls block as new block for the start block
683 Further mark these nodes so that they are not visited by the
685 set_irn_link(get_irg_start(called_graph), pre_call);
686 set_irn_visited(get_irg_start(called_graph),
687 get_irg_visited(current_ir_graph));
688 set_irn_link(get_irg_start_block(called_graph),
689 get_nodes_Block(pre_call));
690 set_irn_visited(get_irg_start_block(called_graph),
691 get_irg_visited(current_ir_graph));
693 /* Initialize for compaction of in arrays */
694 inc_irg_block_visited(current_ir_graph);
696 /* -- Replicate local entities of the called_graph -- */
697 /* copy the entities. */
698 called_frame = get_irg_frame_type(called_graph);
699 for (i = 0; i < get_class_n_members(called_frame); i++) {
700 entity *new_ent, *old_ent;
701 old_ent = get_class_member(called_frame, i);
702 new_ent = copy_entity_own(old_ent, get_cur_frame_type());
703 set_entity_link(old_ent, new_ent);
706 /* visited is > than that of called graph. With this trick visited will
707 remain unchanged so that an outer walker, e.g., searching the call nodes
708 to inline, calling this inline will not visit the inlined nodes. */
709 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
711 /* -- Performing dead node elimination inlines the graph -- */
712 /* Copies the nodes to the obstack of current_ir_graph. Updates links to new
714 /* @@@ endless loops are not copied!! -- they should be, I think... */
715 irg_walk(get_irg_end(called_graph), copy_node_inline, copy_preds,
716 get_irg_frame_type(called_graph));
718 /* Repair called_graph */
719 set_irg_visited(called_graph, get_irg_visited(current_ir_graph));
720 set_irg_block_visited(called_graph, get_irg_block_visited(current_ir_graph));
721 set_Block_block_visited(get_irg_start_block(called_graph), 0);
723 /* -- Merge the end of the inlined procedure with the call site -- */
724 /* We will turn the old Call node into a Tuple with the following
727 0: Phi of all Memories of Return statements.
728 1: Jmp from new Block that merges the control flow from all exception
729 predecessors of the old end block.
730 2: Tuple of all arguments.
731 3: Phi of Exception memories.
732 In case the old Call directly branches to End on an exception we don't
733 need the block merging all exceptions nor the Phi of the exception
737 /* -- Precompute some values -- */
738 end_bl = get_new_node(get_irg_end_block(called_graph));
739 end = get_new_node(get_irg_end(called_graph));
740 arity = intern_get_irn_arity(end_bl); /* arity = n_exc + n_ret */
741 n_res = get_method_n_ress(get_Call_type(call));
743 res_pred = (ir_node **) malloc (n_res * sizeof (ir_node *));
744 cf_pred = (ir_node **) malloc (arity * sizeof (ir_node *));
746 set_irg_current_block(current_ir_graph, post_bl); /* just to make sure */
748 /* -- archive keepalives -- */
749 irn_arity = intern_get_irn_arity(end);
750 for (i = 0; i < irn_arity; i++)
751 add_End_keepalive(get_irg_end(current_ir_graph), intern_get_irn_n(end, i));
753 /* The new end node will die. We need not free as the in array is on the obstack:
754 copy_node only generated 'D' arrays. */
756 /* -- Replace Return nodes by Jump nodes. -- */
758 for (i = 0; i < arity; i++) {
760 ret = intern_get_irn_n(end_bl, i);
761 if (intern_get_irn_op(ret) == op_Return) {
762 cf_pred[n_ret] = new_r_Jmp(current_ir_graph, get_nodes_Block(ret));
766 set_irn_in(post_bl, n_ret, cf_pred);
768 /* -- Build a Tuple for all results of the method.
769 Add Phi node if there was more than one Return. -- */
770 turn_into_tuple(post_call, 4);
771 /* First the Memory-Phi */
773 for (i = 0; i < arity; i++) {
774 ret = intern_get_irn_n(end_bl, i);
775 if (intern_get_irn_op(ret) == op_Return) {
776 cf_pred[n_ret] = get_Return_mem(ret);
780 phi = new_Phi(n_ret, cf_pred, mode_M);
781 set_Tuple_pred(call, 0, phi);
782 /* Conserve Phi-list for further inlinings -- but might be optimized */
783 if (get_nodes_Block(phi) == post_bl) {
784 set_irn_link(phi, get_irn_link(post_bl));
785 set_irn_link(post_bl, phi);
787 /* Now the real results */
789 for (j = 0; j < n_res; j++) {
791 for (i = 0; i < arity; i++) {
792 ret = intern_get_irn_n(end_bl, i);
793 if (intern_get_irn_op(ret) == op_Return) {
794 cf_pred[n_ret] = get_Return_res(ret, j);
798 phi = new_Phi(n_ret, cf_pred, intern_get_irn_mode(cf_pred[0]));
800 /* Conserve Phi-list for further inlinings -- but might be optimized */
801 if (get_nodes_Block(phi) == post_bl) {
802 set_irn_link(phi, get_irn_link(post_bl));
803 set_irn_link(post_bl, phi);
806 set_Tuple_pred(call, 2, new_Tuple(n_res, res_pred));
808 set_Tuple_pred(call, 2, new_Bad());
810 /* Finally the exception control flow.
811 We have two (three) possible situations:
812 First if the Call branches to an exception handler: We need to add a Phi node to
813 collect the memory containing the exception objects. Further we need
814 to add another block to get a correct representation of this Phi. To
815 this block we add a Jmp that resolves into the X output of the Call
816 when the Call is turned into a tuple.
817 Second the Call branches to End, the exception is not handled. Just
818 add all inlined exception branches to the End node.
819 Third: there is no Exception edge at all. Handle as case two. */
820 if (exc_handling == 0) {
822 for (i = 0; i < arity; i++) {
824 ret = intern_get_irn_n(end_bl, i);
825 if (is_fragile_op(skip_Proj(ret)) || (intern_get_irn_op(skip_Proj(ret)) == op_Raise)) {
826 cf_pred[n_exc] = ret;
831 new_Block(n_exc, cf_pred); /* watch it: current_block is changed! */
832 set_Tuple_pred(call, 1, new_Jmp());
833 /* The Phi for the memories with the exception objects */
835 for (i = 0; i < arity; i++) {
837 ret = skip_Proj(intern_get_irn_n(end_bl, i));
838 if (intern_get_irn_op(ret) == op_Call) {
839 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 3);
841 } else if (is_fragile_op(ret)) {
842 /* We rely that all cfops have the memory output at the same position. */
843 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 0);
845 } else if (intern_get_irn_op(ret) == op_Raise) {
846 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 1);
850 set_Tuple_pred(call, 3, new_Phi(n_exc, cf_pred, mode_M));
852 set_Tuple_pred(call, 1, new_Bad());
853 set_Tuple_pred(call, 3, new_Bad());
856 ir_node *main_end_bl;
857 int main_end_bl_arity;
860 /* assert(exc_handling == 1 || no exceptions. ) */
862 for (i = 0; i < arity; i++) {
863 ir_node *ret = intern_get_irn_n(end_bl, i);
865 if (is_fragile_op(skip_Proj(ret)) || (intern_get_irn_op(skip_Proj(ret)) == op_Raise)) {
866 cf_pred[n_exc] = ret;
870 main_end_bl = get_irg_end_block(current_ir_graph);
871 main_end_bl_arity = intern_get_irn_arity(main_end_bl);
872 end_preds = (ir_node **) malloc ((n_exc + main_end_bl_arity) * sizeof (ir_node *));
874 for (i = 0; i < main_end_bl_arity; ++i)
875 end_preds[i] = intern_get_irn_n(main_end_bl, i);
876 for (i = 0; i < n_exc; ++i)
877 end_preds[main_end_bl_arity + i] = cf_pred[i];
878 set_irn_in(main_end_bl, n_exc + main_end_bl_arity, end_preds);
879 set_Tuple_pred(call, 1, new_Bad());
880 set_Tuple_pred(call, 3, new_Bad());
886 #if 0 /* old. now better, correcter, faster implementation. */
888 /* -- If the exception control flow from the inlined Call directly
889 branched to the end block we now have the following control
890 flow predecessor pattern: ProjX -> Tuple -> Jmp. We must
891 remove the Jmp along with it's empty block and add Jmp's
892 predecessors as predecessors of this end block. No problem if
893 there is no exception, because then branches Bad to End which
895 @@@ can't we know this beforehand: by getting the Proj(1) from
896 the Call link list and checking whether it goes to Proj. */
897 /* find the problematic predecessor of the end block. */
898 end_bl = get_irg_end_block(current_ir_graph);
899 for (i = 0; i < get_Block_n_cfgpreds(end_bl); i++) {
900 cf_op = get_Block_cfgpred(end_bl, i);
901 if (intern_get_irn_op(cf_op) == op_Proj) {
902 cf_op = get_Proj_pred(cf_op);
903 if ((intern_get_irn_op(cf_op) == op_Tuple) && (cf_op == call)) {
904 // There are unoptimized tuples from inlineing before when no exc
905 assert(get_Proj_proj(get_Block_cfgpred(end_bl, i)) == pn_Call_X_except);
906 cf_op = get_Tuple_pred(cf_op, pn_Call_X_except);
907 assert(intern_get_irn_op(cf_op) == op_Jmp);
913 if (i < get_Block_n_cfgpreds(end_bl)) {
914 bl = get_nodes_Block(cf_op);
915 arity = get_Block_n_cfgpreds(end_bl) + get_Block_n_cfgpreds(bl) - 1;
916 cf_pred = (ir_node **) malloc (arity * sizeof (ir_node *));
917 for (j = 0; j < i; j++)
918 cf_pred[j] = get_Block_cfgpred(end_bl, j);
919 for (j = j; j < i + get_Block_n_cfgpreds(bl); j++)
920 cf_pred[j] = get_Block_cfgpred(bl, j-i);
921 for (j = j; j < arity; j++)
922 cf_pred[j] = get_Block_cfgpred(end_bl, j-get_Block_n_cfgpreds(bl) +1);
923 set_irn_in(end_bl, arity, cf_pred);
925 // Remove the exception pred from post-call Tuple.
926 set_Tuple_pred(call, pn_Call_X_except, new_Bad());
931 /* -- Turn cse back on. -- */
932 set_optimize(rem_opt);
935 /********************************************************************/
936 /* Apply inlineing to small methods. */
937 /********************************************************************/
941 /* It makes no sense to inline too many calls in one procedure. Anyways,
942 I didn't get a version with NEW_ARR_F to run. */
943 #define MAX_INLINE 1024
946 * Returns the irg called from a Call node. If the irg is not
947 * known, NULL is returned.
949 static ir_graph *get_call_called_irg(ir_node *call) {
952 ir_graph *called_irg = NULL;
954 assert(get_irn_op(call) == op_Call);
956 addr = get_Call_ptr(call);
957 if (intern_get_irn_op(addr) == op_Const) {
958 /* Check whether the constant is the pointer to a compiled entity. */
959 tv = get_Const_tarval(addr);
960 if (tarval_to_entity(tv))
961 called_irg = get_entity_irg(tarval_to_entity(tv));
966 static void collect_calls(ir_node *call, void *env) {
968 ir_node **calls = (ir_node **)env;
971 ir_graph *called_irg;
973 if (intern_get_irn_op(call) != op_Call) return;
975 addr = get_Call_ptr(call);
976 if (intern_get_irn_op(addr) == op_Const) {
977 /* Check whether the constant is the pointer to a compiled entity. */
978 tv = get_Const_tarval(addr);
979 if (tarval_to_entity(tv)) {
980 called_irg = get_entity_irg(tarval_to_entity(tv));
981 if (called_irg && pos < MAX_INLINE) {
982 /* The Call node calls a locally defined method. Remember to inline. */
991 * Inlines all small methods at call sites where the called address comes
992 * from a Const node that references the entity representing the called
994 * The size argument is a rough measure for the code size of the method:
995 * Methods where the obstack containing the firm graph is smaller than
998 void inline_small_irgs(ir_graph *irg, int size) {
1000 ir_node *calls[MAX_INLINE];
1001 ir_graph *rem = current_ir_graph;
1003 if (!(get_opt_optimize() && get_opt_inline())) return;
1005 current_ir_graph = irg;
1006 /* Handle graph state */
1007 assert(get_irg_phase_state(current_ir_graph) != phase_building);
1008 assert(get_irg_callee_info_state(current_ir_graph) == irg_callee_info_none);
1010 /* Find Call nodes to inline.
1011 (We can not inline during a walk of the graph, as inlineing the same
1012 method several times changes the visited flag of the walked graph:
1013 after the first inlineing visited of the callee equals visited of
1014 the caller. With the next inlineing both are increased.) */
1016 irg_walk(get_irg_end(irg), NULL, collect_calls, (void *) calls);
1018 if ((pos > 0) && (pos < MAX_INLINE)) {
1019 /* There are calls to inline */
1020 collect_phiprojs(irg);
1021 for (i = 0; i < pos; i++) {
1024 tv = get_Const_tarval(get_Call_ptr(calls[i]));
1025 callee = get_entity_irg(tarval_to_entity(tv));
1026 if (((_obstack_memory_used(callee->obst) - obstack_room(callee->obst)) < size) ||
1027 (get_irg_inline_property(callee) == irg_inline_forced)) {
1028 inline_method(calls[i], callee);
1033 current_ir_graph = rem;
1037 * Environment for inlining irgs.
1040 int n_nodes; /**< Nodes in graph except Id, Tuple, Proj, Start, End */
1041 int n_nodes_orig; /**< for statistics */
1042 eset *call_nodes; /**< All call nodes in this graph */
1044 int n_call_nodes_orig; /**< for statistics */
1045 int n_callers; /**< Number of known graphs that call this graphs. */
1046 int n_callers_orig; /**< for statistics */
1049 static inline_irg_env *new_inline_irg_env(void) {
1050 inline_irg_env *env = malloc(sizeof(inline_irg_env));
1051 env->n_nodes = -2; /* uncount Start, End */
1052 env->n_nodes_orig = -2; /* uncount Start, End */
1053 env->call_nodes = eset_create();
1054 env->n_call_nodes = 0;
1055 env->n_call_nodes_orig = 0;
1057 env->n_callers_orig = 0;
1061 static void free_inline_irg_env(inline_irg_env *env) {
1062 eset_destroy(env->call_nodes);
1066 static void collect_calls2(ir_node *call, void *env) {
1067 inline_irg_env *x = (inline_irg_env *)env;
1068 ir_op *op = intern_get_irn_op(call);
1071 /* count nodes in irg */
1072 if (op != op_Proj && op != op_Tuple && op != op_Sync) {
1077 if (op != op_Call) return;
1079 /* collect all call nodes */
1080 eset_insert(x->call_nodes, (void *)call);
1082 x->n_call_nodes_orig++;
1084 /* count all static callers */
1085 callee = get_call_called_irg(call);
1087 ((inline_irg_env *)get_irg_link(callee))->n_callers++;
1088 ((inline_irg_env *)get_irg_link(callee))->n_callers_orig++;
1092 INLINE static int is_leave(ir_graph *irg) {
1093 return (((inline_irg_env *)get_irg_link(irg))->n_call_nodes == 0);
1096 INLINE static int is_smaller(ir_graph *callee, int size) {
1097 return (((inline_irg_env *)get_irg_link(callee))->n_nodes < size);
1102 * Inlines small leave methods at call sites where the called address comes
1103 * from a Const node that references the entity representing the called
1105 * The size argument is a rough measure for the code size of the method:
1106 * Methods where the obstack containing the firm graph is smaller than
1109 void inline_leave_functions(int maxsize, int leavesize, int size) {
1110 inline_irg_env *env;
1111 int i, n_irgs = get_irp_n_irgs();
1112 ir_graph *rem = current_ir_graph;
1115 if (!(get_opt_optimize() && get_opt_inline())) return;
1117 /* extend all irgs by a temporary data structure for inlineing. */
1118 for (i = 0; i < n_irgs; ++i)
1119 set_irg_link(get_irp_irg(i), new_inline_irg_env());
1121 /* Precompute information in temporary data structure. */
1122 for (i = 0; i < n_irgs; ++i) {
1123 current_ir_graph = get_irp_irg(i);
1124 assert(get_irg_phase_state(current_ir_graph) != phase_building);
1125 assert(get_irg_callee_info_state(current_ir_graph) == irg_callee_info_none);
1127 irg_walk(get_irg_end(current_ir_graph), NULL, collect_calls2,
1128 get_irg_link(current_ir_graph));
1129 env = (inline_irg_env *)get_irg_link(current_ir_graph);
1133 Inline leaves recursively -- we might construct new leaves. */
1135 while (did_inline) {
1136 //printf("iteration %d\n", itercnt++);
1138 for (i = 0; i < n_irgs; ++i) {
1141 int phiproj_computed = 0;
1143 current_ir_graph = get_irp_irg(i);
1144 env = (inline_irg_env *)get_irg_link(current_ir_graph);
1146 /* we can not walk and change a set, nor remove from it.
1148 walkset = env->call_nodes;
1149 env->call_nodes = eset_create();
1150 for (call = eset_first(walkset); call; call = eset_next(walkset)) {
1151 inline_irg_env *callee_env;
1152 ir_graph *callee = get_call_called_irg(call);
1154 if (env->n_nodes > maxsize) break;
1156 ((is_leave(callee) && is_smaller(callee, leavesize)) ||
1157 (get_irg_inline_property(callee) == irg_inline_forced))) {
1158 if (!phiproj_computed) {
1159 phiproj_computed = 1;
1160 collect_phiprojs(current_ir_graph);
1162 callee_env = (inline_irg_env *)get_irg_link(callee);
1163 // printf(" %s: Inlineing %s.\n", get_entity_name(get_irg_entity(current_ir_graph)),
1164 // get_entity_name(get_irg_entity(callee)));
1165 inline_method(call, callee);
1167 env->n_call_nodes--;
1168 eset_insert_all(env->call_nodes, callee_env->call_nodes);
1169 env->n_call_nodes += callee_env->n_call_nodes;
1170 env->n_nodes += callee_env->n_nodes;
1171 callee_env->n_callers--;
1173 eset_insert(env->call_nodes, call);
1176 eset_destroy(walkset);
1180 //printf("Non leaves\n");
1181 /* inline other small functions. */
1182 for (i = 0; i < n_irgs; ++i) {
1185 int phiproj_computed = 0;
1187 current_ir_graph = get_irp_irg(i);
1188 env = (inline_irg_env *)get_irg_link(current_ir_graph);
1190 /* we can not walk and change a set, nor remove from it.
1192 walkset = env->call_nodes;
1193 env->call_nodes = eset_create();
1194 for (call = eset_first(walkset); call; call = eset_next(walkset)) {
1195 inline_irg_env *callee_env;
1196 ir_graph *callee = get_call_called_irg(call);
1198 if (env->n_nodes > maxsize) break;
1199 if (callee && is_smaller(callee, size)) {
1200 if (!phiproj_computed) {
1201 phiproj_computed = 1;
1202 collect_phiprojs(current_ir_graph);
1204 callee_env = (inline_irg_env *)get_irg_link(callee);
1205 // printf(" %s: Inlineing %s.\n", get_entity_name(get_irg_entity(current_ir_graph)),
1206 // get_entity_name(get_irg_entity(callee)));
1207 inline_method(call, callee);
1209 env->n_call_nodes--;
1210 eset_insert_all(env->call_nodes, callee_env->call_nodes);
1211 env->n_call_nodes += callee_env->n_call_nodes;
1212 env->n_nodes += callee_env->n_nodes;
1213 callee_env->n_callers--;
1215 eset_insert(env->call_nodes, call);
1218 eset_destroy(walkset);
1221 for (i = 0; i < n_irgs; ++i) {
1222 current_ir_graph = get_irp_irg(i);
1224 env = (inline_irg_env *)get_irg_link(current_ir_graph);
1225 if ((env->n_call_nodes_orig != env->n_call_nodes) ||
1226 (env->n_callers_orig != env->n_callers))
1227 printf("Nodes:%3d ->%3d, calls:%3d ->%3d, callers:%3d ->%3d, -- %s\n",
1228 env->n_nodes_orig, env->n_nodes, env->n_call_nodes_orig, env->n_call_nodes,
1229 env->n_callers_orig, env->n_callers,
1230 get_entity_name(get_irg_entity(current_ir_graph)));
1232 free_inline_irg_env((inline_irg_env *)get_irg_link(current_ir_graph));
1235 current_ir_graph = rem;
1238 /*-----------------------------------------------------------------*/
1239 /* Code Placement. Pins all floating nodes to a block where they */
1240 /* will be executed only if needed. */
1241 /*-----------------------------------------------------------------*/
1244 * Find the earliest correct block for N. --- Place N into the
1245 * same Block as its dominance-deepest Input.
1248 place_floats_early(ir_node *n, pdeq *worklist)
1250 int i, start, irn_arity;
1252 /* we must not run into an infinite loop */
1253 assert (irn_not_visited(n));
1254 mark_irn_visited(n);
1256 /* Place floating nodes. */
1257 if (get_op_pinned(intern_get_irn_op(n)) == floats) {
1259 ir_node *b = new_Bad(); /* The block to place this node in */
1261 assert(intern_get_irn_op(n) != op_Block);
1263 if ((intern_get_irn_op(n) == op_Const) ||
1264 (intern_get_irn_op(n) == op_SymConst) ||
1266 (intern_get_irn_op(n) == op_Unknown)) {
1267 /* These nodes will not be placed by the loop below. */
1268 b = get_irg_start_block(current_ir_graph);
1272 /* find the block for this node. */
1273 irn_arity = intern_get_irn_arity(n);
1274 for (i = 0; i < irn_arity; i++) {
1275 ir_node *dep = intern_get_irn_n(n, i);
1277 if ((irn_not_visited(dep)) &&
1278 (get_op_pinned(intern_get_irn_op(dep)) == floats)) {
1279 place_floats_early(dep, worklist);
1281 /* Because all loops contain at least one pinned node, now all
1282 our inputs are either pinned or place_early has already
1283 been finished on them. We do not have any unfinished inputs! */
1284 dep_block = get_nodes_Block(dep);
1285 if ((!is_Bad(dep_block)) &&
1286 (get_Block_dom_depth(dep_block) > depth)) {
1288 depth = get_Block_dom_depth(dep_block);
1290 /* Avoid that the node is placed in the Start block */
1291 if ((depth == 1) && (get_Block_dom_depth(get_nodes_Block(n)) > 1)) {
1292 b = get_Block_cfg_out(get_irg_start_block(current_ir_graph), 0);
1293 assert(b != get_irg_start_block(current_ir_graph));
1297 set_nodes_Block(n, b);
1300 /* Add predecessors of non floating nodes on worklist. */
1301 start = (intern_get_irn_op(n) == op_Block) ? 0 : -1;
1302 irn_arity = intern_get_irn_arity(n);
1303 for (i = start; i < irn_arity; i++) {
1304 ir_node *pred = intern_get_irn_n(n, i);
1305 if (irn_not_visited(pred)) {
1306 pdeq_putr (worklist, pred);
1312 * Floating nodes form subgraphs that begin at nodes as Const, Load,
1313 * Start, Call and end at pinned nodes as Store, Call. Place_early
1314 * places all floating nodes reachable from its argument through floating
1315 * nodes and adds all beginnings at pinned nodes to the worklist.
1317 static INLINE void place_early(pdeq* worklist) {
1319 inc_irg_visited(current_ir_graph);
1321 /* this inits the worklist */
1322 place_floats_early(get_irg_end(current_ir_graph), worklist);
1324 /* Work the content of the worklist. */
1325 while (!pdeq_empty (worklist)) {
1326 ir_node *n = pdeq_getl (worklist);
1327 if (irn_not_visited(n)) place_floats_early(n, worklist);
1330 set_irg_outs_inconsistent(current_ir_graph);
1331 current_ir_graph->pinned = pinned;
1335 /** deepest common dominance ancestor of DCA and CONSUMER of PRODUCER. */
1337 consumer_dom_dca (ir_node *dca, ir_node *consumer, ir_node *producer)
1339 ir_node *block = NULL;
1341 /* Compute the latest block into which we can place a node so that it is
1343 if (intern_get_irn_op(consumer) == op_Phi) {
1344 /* our consumer is a Phi-node, the effective use is in all those
1345 blocks through which the Phi-node reaches producer */
1347 ir_node *phi_block = get_nodes_Block(consumer);
1348 irn_arity = intern_get_irn_arity(consumer);
1349 for (i = 0; i < irn_arity; i++) {
1350 if (intern_get_irn_n(consumer, i) == producer) {
1351 block = get_nodes_Block(get_Block_cfgpred(phi_block, i));
1355 assert(is_no_Block(consumer));
1356 block = get_nodes_Block(consumer);
1359 /* Compute the deepest common ancestor of block and dca. */
1361 if (!dca) return block;
1362 while (get_Block_dom_depth(block) > get_Block_dom_depth(dca))
1363 block = get_Block_idom(block);
1364 while (get_Block_dom_depth(dca) > get_Block_dom_depth(block))
1365 dca = get_Block_idom(dca);
1366 while (block != dca)
1367 { block = get_Block_idom(block); dca = get_Block_idom(dca); }
1372 static INLINE int get_irn_loop_depth(ir_node *n) {
1373 return get_loop_depth(get_irn_loop(n));
1377 * Move n to a block with less loop depth than it's current block. The
1378 * new block must be dominated by early.
1381 move_out_of_loops (ir_node *n, ir_node *early)
1383 ir_node *best, *dca;
1387 /* Find the region deepest in the dominator tree dominating
1388 dca with the least loop nesting depth, but still dominated
1389 by our early placement. */
1390 dca = get_nodes_Block(n);
1392 while (dca != early) {
1393 dca = get_Block_idom(dca);
1394 if (!dca) break; /* should we put assert(dca)? */
1395 if (get_irn_loop_depth(dca) < get_irn_loop_depth(best)) {
1399 if (best != get_nodes_Block(n)) {
1401 printf("Moving out of loop: "); DDMN(n);
1402 printf(" Outermost block: "); DDMN(early);
1403 printf(" Best block: "); DDMN(best);
1404 printf(" Innermost block: "); DDMN(get_nodes_Block(n));
1406 set_nodes_Block(n, best);
1411 * Find the latest legal block for N and place N into the
1412 * `optimal' Block between the latest and earliest legal block.
1413 * The `optimal' block is the dominance-deepest block of those
1414 * with the least loop-nesting-depth. This places N out of as many
1415 * loops as possible and then makes it as control dependant as
1419 place_floats_late(ir_node *n, pdeq *worklist)
1424 assert (irn_not_visited(n)); /* no multiple placement */
1426 /* no need to place block nodes, control nodes are already placed. */
1427 if ((intern_get_irn_op(n) != op_Block) &&
1429 (intern_get_irn_mode(n) != mode_X)) {
1430 /* Remember the early placement of this block to move it
1431 out of loop no further than the early placement. */
1432 early = get_nodes_Block(n);
1433 /* Assure that our users are all placed, except the Phi-nodes.
1434 --- Each data flow cycle contains at least one Phi-node. We
1435 have to break the `user has to be placed before the
1436 producer' dependence cycle and the Phi-nodes are the
1437 place to do so, because we need to base our placement on the
1438 final region of our users, which is OK with Phi-nodes, as they
1439 are pinned, and they never have to be placed after a
1440 producer of one of their inputs in the same block anyway. */
1441 for (i = 0; i < get_irn_n_outs(n); i++) {
1442 ir_node *succ = get_irn_out(n, i);
1443 if (irn_not_visited(succ) && (intern_get_irn_op(succ) != op_Phi))
1444 place_floats_late(succ, worklist);
1447 /* We have to determine the final block of this node... except for
1449 if ((get_op_pinned(intern_get_irn_op(n)) == floats) &&
1450 (intern_get_irn_op(n) != op_Const) &&
1451 (intern_get_irn_op(n) != op_SymConst)) {
1452 ir_node *dca = NULL; /* deepest common ancestor in the
1453 dominator tree of all nodes'
1454 blocks depending on us; our final
1455 placement has to dominate DCA. */
1456 for (i = 0; i < get_irn_n_outs(n); i++) {
1457 dca = consumer_dom_dca (dca, get_irn_out(n, i), n);
1459 set_nodes_Block(n, dca);
1461 move_out_of_loops (n, early);
1465 mark_irn_visited(n);
1467 /* Add predecessors of all non-floating nodes on list. (Those of floating
1468 nodes are placeded already and therefore are marked.) */
1469 for (i = 0; i < get_irn_n_outs(n); i++) {
1470 if (irn_not_visited(get_irn_out(n, i))) {
1471 pdeq_putr (worklist, get_irn_out(n, i));
1476 static INLINE void place_late(pdeq* worklist) {
1478 inc_irg_visited(current_ir_graph);
1480 /* This fills the worklist initially. */
1481 place_floats_late(get_irg_start_block(current_ir_graph), worklist);
1482 /* And now empty the worklist again... */
1483 while (!pdeq_empty (worklist)) {
1484 ir_node *n = pdeq_getl (worklist);
1485 if (irn_not_visited(n)) place_floats_late(n, worklist);
1489 void place_code(ir_graph *irg) {
1491 ir_graph *rem = current_ir_graph;
1493 current_ir_graph = irg;
1495 if (!(get_opt_optimize() && get_opt_global_cse())) return;
1497 /* Handle graph state */
1498 assert(get_irg_phase_state(irg) != phase_building);
1499 if (get_irg_dom_state(irg) != dom_consistent)
1502 if (get_irg_loopinfo_state(irg) != loopinfo_consistent) {
1503 free_loop_information(irg);
1504 construct_backedges(irg);
1507 /* Place all floating nodes as early as possible. This guarantees
1508 a legal code placement. */
1509 worklist = new_pdeq();
1510 place_early(worklist);
1512 /* place_early invalidates the outs, place_late needs them. */
1514 /* Now move the nodes down in the dominator tree. This reduces the
1515 unnecessary executions of the node. */
1516 place_late(worklist);
1518 set_irg_outs_inconsistent(current_ir_graph);
1519 set_irg_loopinfo_inconsistent(current_ir_graph);
1521 current_ir_graph = rem;
1526 /********************************************************************/
1527 /* Control flow optimization. */
1528 /* Removes Bad control flow predecessors and empty blocks. A block */
1529 /* is empty if it contains only a Jmp node. */
1530 /* Blocks can only be removed if they are not needed for the */
1531 /* semantics of Phi nodes. */
1532 /********************************************************************/
1535 * Removes Tuples from Block control flow predecessors.
1536 * Optimizes blocks with equivalent_node().
1537 * Replaces n by Bad if n is unreachable control flow.
1539 static void merge_blocks(ir_node *n, void *env) {
1541 set_irn_link(n, NULL);
1543 if (intern_get_irn_op(n) == op_Block) {
1545 for (i = 0; i < get_Block_n_cfgpreds(n); i++)
1546 /* GL @@@ : is this possible? if (get_opt_normalize()) -- added, all tests go through.
1547 A different order of optimizations might cause problems. */
1548 if (get_opt_normalize())
1549 set_Block_cfgpred(n, i, skip_Tuple(get_Block_cfgpred(n, i)));
1550 } else if (get_opt_optimize() && (intern_get_irn_mode(n) == mode_X)) {
1551 /* We will soon visit a block. Optimize it before visiting! */
1552 ir_node *b = get_nodes_Block(n);
1553 ir_node *new_node = equivalent_node(b);
1554 while (irn_not_visited(b) && (!is_Bad(new_node)) && (new_node != b)) {
1555 /* We would have to run gigo if new is bad, so we
1556 promote it directly below. */
1557 assert(((b == new_node) ||
1558 get_opt_control_flow_straightening() ||
1559 get_opt_control_flow_weak_simplification()) &&
1560 ("strange flag setting"));
1561 exchange (b, new_node);
1563 new_node = equivalent_node(b);
1565 /* GL @@@ get_opt_normalize hinzugefuegt, 5.5.2003 */
1566 if (is_Bad(new_node) && get_opt_normalize()) exchange(n, new_Bad());
1571 * Collects all Phi nodes in link list of Block.
1572 * Marks all blocks "block_visited" if they contain a node other
1575 static void collect_nodes(ir_node *n, void *env) {
1576 if (is_no_Block(n)) {
1577 ir_node *b = get_nodes_Block(n);
1579 if ((intern_get_irn_op(n) == op_Phi)) {
1580 /* Collect Phi nodes to compact ins along with block's ins. */
1581 set_irn_link(n, get_irn_link(b));
1583 } else if (intern_get_irn_op(n) != op_Jmp) { /* Check for non empty block. */
1584 mark_Block_block_visited(b);
1589 /** Returns true if pred is predecessor of block. */
1590 static int is_pred_of(ir_node *pred, ir_node *b) {
1592 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1593 ir_node *b_pred = get_nodes_Block(get_Block_cfgpred(b, i));
1594 if (b_pred == pred) return 1;
1599 static int test_whether_dispensable(ir_node *b, int pos) {
1600 int i, j, n_preds = 1;
1601 int dispensable = 1;
1602 ir_node *cfop = get_Block_cfgpred(b, pos);
1603 ir_node *pred = get_nodes_Block(cfop);
1605 if (get_Block_block_visited(pred) + 1
1606 < get_irg_block_visited(current_ir_graph)) {
1607 if (!get_opt_optimize() || !get_opt_control_flow_strong_simplification()) {
1608 /* Mark block so that is will not be removed. */
1609 set_Block_block_visited(pred, get_irg_block_visited(current_ir_graph)-1);
1612 /* Seems to be empty. */
1613 if (!get_irn_link(b)) {
1614 /* There are no Phi nodes ==> dispensable. */
1615 n_preds = get_Block_n_cfgpreds(pred);
1617 /* b's pred blocks and pred's pred blocks must be pairwise disjunct.
1618 Work preds < pos as if they were already removed. */
1619 for (i = 0; i < pos; i++) {
1620 ir_node *b_pred = get_nodes_Block(get_Block_cfgpred(b, i));
1621 if (get_Block_block_visited(b_pred) + 1
1622 < get_irg_block_visited(current_ir_graph)) {
1623 for (j = 0; j < get_Block_n_cfgpreds(b_pred); j++) {
1624 ir_node *b_pred_pred = get_nodes_Block(get_Block_cfgpred(b_pred, j));
1625 if (is_pred_of(b_pred_pred, pred)) dispensable = 0;
1628 if (is_pred_of(b_pred, pred)) dispensable = 0;
1631 for (i = pos +1; i < get_Block_n_cfgpreds(b); i++) {
1632 ir_node *b_pred = get_nodes_Block(get_Block_cfgpred(b, i));
1633 if (is_pred_of(b_pred, pred)) dispensable = 0;
1636 set_Block_block_visited(pred, get_irg_block_visited(current_ir_graph)-1);
1639 n_preds = get_Block_n_cfgpreds(pred);
1647 static void optimize_blocks(ir_node *b, void *env) {
1648 int i, j, k, max_preds, n_preds;
1649 ir_node *pred, *phi;
1652 /* Count the number of predecessor if this block is merged with pred blocks
1655 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1656 max_preds += test_whether_dispensable(b, i);
1658 in = (ir_node **) malloc(max_preds * sizeof(ir_node *));
1661 printf(" working on "); DDMN(b);
1662 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1663 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1664 if (is_Bad(get_Block_cfgpred(b, i))) {
1665 printf(" removing Bad %i\n ", i);
1666 } else if (get_Block_block_visited(pred) +1
1667 < get_irg_block_visited(current_ir_graph)) {
1668 printf(" removing pred %i ", i); DDMN(pred);
1669 } else { printf(" Nothing to do for "); DDMN(pred); }
1671 * end Debug output -*/
1673 /*- Fix the Phi nodes -*/
1674 phi = get_irn_link(b);
1676 assert(intern_get_irn_op(phi) == op_Phi);
1677 /* Find the new predecessors for the Phi */
1679 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1680 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1681 if (is_Bad(get_Block_cfgpred(b, i))) {
1683 } else if (get_Block_block_visited(pred) +1
1684 < get_irg_block_visited(current_ir_graph)) {
1685 /* It's an empty block and not yet visited. */
1686 ir_node *phi_pred = get_Phi_pred(phi, i);
1687 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1688 if (get_nodes_Block(phi_pred) == pred) {
1689 assert(intern_get_irn_op(phi_pred) == op_Phi); /* Block is empty!! */
1690 in[n_preds] = get_Phi_pred(phi_pred, j);
1692 in[n_preds] = phi_pred;
1696 /* The Phi_pred node is replaced now if it is a Phi.
1697 In Schleifen kann offenbar der entfernte Phi Knoten legal verwendet werden.
1698 Daher muss der Phiknoten durch den neuen ersetzt werden.
1699 Weiter muss der alte Phiknoten entfernt werden (durch ersetzen oder
1700 durch einen Bad) damit er aus den keep_alive verschwinden kann.
1701 Man sollte also, falls keine Schleife vorliegt, exchange mit new_Bad
1703 if (get_nodes_Block(phi_pred) == pred) {
1704 /* remove the Phi as it might be kept alive. Further there
1705 might be other users. */
1706 exchange(phi_pred, phi); /* geht, ist aber doch semantisch falsch! Warum?? */
1709 in[n_preds] = get_Phi_pred(phi, i);
1714 set_irn_in(phi, n_preds, in);
1716 phi = get_irn_link(phi);
1720 This happens only if merge between loop backedge and single loop entry. -*/
1721 for (k = 0; k < get_Block_n_cfgpreds(b); k++) {
1722 pred = get_nodes_Block(get_Block_cfgpred(b, k));
1723 if (get_Block_block_visited(pred) +1
1724 < get_irg_block_visited(current_ir_graph)) {
1725 phi = get_irn_link(pred);
1727 if (intern_get_irn_op(phi) == op_Phi) {
1728 set_nodes_Block(phi, b);
1731 for (i = 0; i < k; i++) {
1732 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1733 if (is_Bad(get_Block_cfgpred(b, i))) {
1735 } else if (get_Block_block_visited(pred) +1
1736 < get_irg_block_visited(current_ir_graph)) {
1737 /* It's an empty block and not yet visited. */
1738 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1739 /* @@@ Hier brauche ich Schleifeninformation!!! Kontrollflusskante
1740 muss Rueckwaertskante sein! (An allen vier in[n_preds] = phi
1741 Anweisungen.) Trotzdem tuts bisher!! */
1750 for (i = 0; i < get_Phi_n_preds(phi); i++) {
1751 in[n_preds] = get_Phi_pred(phi, i);
1754 for (i = k+1; i < get_Block_n_cfgpreds(b); i++) {
1755 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1756 if (is_Bad(get_Block_cfgpred(b, i))) {
1758 } else if (get_Block_block_visited(pred) +1
1759 < get_irg_block_visited(current_ir_graph)) {
1760 /* It's an empty block and not yet visited. */
1761 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1770 set_irn_in(phi, n_preds, in);
1772 phi = get_irn_link(phi);
1777 /*- Fix the block -*/
1779 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1780 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1781 if (is_Bad(get_Block_cfgpred(b, i))) {
1783 } else if (get_Block_block_visited(pred) +1
1784 < get_irg_block_visited(current_ir_graph)) {
1785 /* It's an empty block and not yet visited. */
1786 assert(get_Block_n_cfgpreds(b) > 1);
1787 /* Else it should be optimized by equivalent_node. */
1788 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1789 in[n_preds] = get_Block_cfgpred(pred, j);
1792 /* Remove block as it might be kept alive. */
1793 exchange(pred, b/*new_Bad()*/);
1795 in[n_preds] = get_Block_cfgpred(b, i);
1799 set_irn_in(b, n_preds, in);
1803 void optimize_cf(ir_graph *irg) {
1806 ir_node *end = get_irg_end(irg);
1807 ir_graph *rem = current_ir_graph;
1808 current_ir_graph = irg;
1810 /* Handle graph state */
1811 assert(get_irg_phase_state(irg) != phase_building);
1812 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
1813 set_irg_outs_inconsistent(current_ir_graph);
1814 if (get_irg_dom_state(current_ir_graph) == dom_consistent)
1815 set_irg_dom_inconsistent(current_ir_graph);
1817 /* Use block visited flag to mark non-empty blocks. */
1818 inc_irg_block_visited(irg);
1819 irg_walk(end, merge_blocks, collect_nodes, NULL);
1821 /* Optimize the standard code. */
1822 irg_block_walk(get_irg_end_block(irg), optimize_blocks, NULL, NULL);
1824 /* Walk all keep alives, optimize them if block, add to new in-array
1825 for end if useful. */
1826 in = NEW_ARR_F (ir_node *, 1);
1827 in[0] = get_nodes_Block(end);
1828 inc_irg_visited(current_ir_graph);
1829 for(i = 0; i < get_End_n_keepalives(end); i++) {
1830 ir_node *ka = get_End_keepalive(end, i);
1831 if (irn_not_visited(ka)) {
1832 if ((intern_get_irn_op(ka) == op_Block) && Block_not_block_visited(ka)) {
1833 set_irg_block_visited(current_ir_graph, /* Don't walk all the way to Start. */
1834 get_irg_block_visited(current_ir_graph)-1);
1835 irg_block_walk(ka, optimize_blocks, NULL, NULL);
1836 mark_irn_visited(ka);
1837 ARR_APP1 (ir_node *, in, ka);
1838 } else if (intern_get_irn_op(ka) == op_Phi) {
1839 mark_irn_visited(ka);
1840 ARR_APP1 (ir_node *, in, ka);
1844 /* DEL_ARR_F(end->in); GL @@@ tut nicht ! */
1847 current_ir_graph = rem;
1852 * Called by walker of remove_critical_cf_edges.
1854 * Place an empty block to an edge between a blocks of multiple
1855 * predecessors and a block of multiple successors.
1858 * @param env Environment of walker. This field is unused and has
1861 static void walk_critical_cf_edges(ir_node *n, void *env) {
1863 ir_node *pre, *block, **in, *jmp;
1865 /* Block has multiple predecessors */
1866 if ((op_Block == intern_get_irn_op(n)) &&
1867 (intern_get_irn_arity(n) > 1)) {
1868 arity = intern_get_irn_arity(n);
1870 if (n == get_irg_end_block(current_ir_graph))
1871 return; // No use to add a block here.
1873 for (i=0; i<arity; i++) {
1874 pre = intern_get_irn_n(n, i);
1875 /* Predecessor has multiple successors. Insert new flow edge */
1876 if ((NULL != pre) &&
1877 (op_Proj == intern_get_irn_op(pre)) &&
1878 op_Raise != intern_get_irn_op(skip_Proj(pre))) {
1880 /* set predecessor array for new block */
1881 in = NEW_ARR_D (ir_node *, current_ir_graph->obst, 1);
1882 /* set predecessor of new block */
1884 block = new_Block(1, in);
1885 /* insert new jmp node to new block */
1886 switch_block(block);
1889 /* set successor of new block */
1890 set_irn_n(n, i, jmp);
1892 } /* predecessor has multiple successors */
1893 } /* for all predecessors */
1894 } /* n is a block */
1897 void remove_critical_cf_edges(ir_graph *irg) {
1898 if (get_opt_critical_edges())
1899 irg_walk_graph(irg, NULL, walk_critical_cf_edges, NULL);