1 /* Coyright (C) 1998 - 2002 by Universitaet Karlsruhe
4 * Author: Christian Schaefer, Goetz Lindenmaier, Sebastian Felis
6 * Optimizations for a whole ir graph, i.e., a procedure.
19 # include "irnode_t.h"
20 # include "irgraph_t.h"
28 # include "pdeq.h" /* Fuer code placement */
31 # include "irbackedge_t.h"
33 /* Defined in iropt.c */
34 pset *new_identities (void);
35 void del_identities (pset *value_table);
36 void add_identities (pset *value_table, ir_node *node);
38 /********************************************************************/
39 /* apply optimizations of iropt to all nodes. */
40 /********************************************************************/
42 static void init_link (ir_node *n, void *env) {
43 set_irn_link(n, NULL);
47 optimize_in_place_wrapper (ir_node *n, void *env) {
51 for (i = 0; i < get_irn_arity(n); i++) {
52 optimized = optimize_in_place_2(get_irn_n(n, i));
53 set_irn_n(n, i, optimized);
56 if (get_irn_op(n) == op_Block) {
57 optimized = optimize_in_place_2(n);
58 if (optimized != n) exchange (n, optimized);
63 local_optimize_graph (ir_graph *irg) {
64 ir_graph *rem = current_ir_graph;
65 current_ir_graph = irg;
67 /* Handle graph state */
68 assert(get_irg_phase_state(irg) != phase_building);
69 if (get_opt_global_cse())
70 set_irg_pinned(current_ir_graph, floats);
71 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
72 set_irg_outs_inconsistent(current_ir_graph);
73 if (get_irg_dom_state(current_ir_graph) == dom_consistent)
74 set_irg_dom_inconsistent(current_ir_graph);
76 /* Clean the value_table in irg for the cse. */
77 del_identities(irg->value_table);
78 irg->value_table = new_identities();
80 /* walk over the graph */
81 irg_walk(irg->end, init_link, optimize_in_place_wrapper, NULL);
83 current_ir_graph = rem;
86 /********************************************************************/
87 /* Routines for dead node elimination / copying garbage collection */
89 /********************************************************************/
91 /* Remeber the new node in the old node by using a field all nodes have. */
93 set_new_node (ir_node *old, ir_node *new)
98 /* Get this new node, before the old node is forgotton.*/
99 static INLINE ir_node *
100 get_new_node (ir_node * n)
105 /* We use the block_visited flag to mark that we have computed the
106 number of useful predecessors for this block.
107 Further we encode the new arity in this flag in the old blocks.
108 Remembering the arity is useful, as it saves a lot of pointer
109 accesses. This function is called for all Phi and Block nodes
112 compute_new_arity(ir_node *b) {
116 irg_v = get_irg_block_visited(current_ir_graph);
117 block_v = get_Block_block_visited(b);
118 if (block_v >= irg_v) {
119 /* we computed the number of preds for this block and saved it in the
121 return block_v - irg_v;
123 /* compute the number of good predecessors */
124 res = get_irn_arity(b);
125 for (i = 0; i < get_irn_arity(b); i++)
126 if (get_irn_opcode(get_irn_n(b, i)) == iro_Bad) res--;
127 /* save it in the flag. */
128 set_Block_block_visited(b, irg_v + res);
133 static INLINE void new_backedge_info(ir_node *n) {
134 switch(get_irn_opcode(n)) {
136 n->attr.block.cg_backedge = NULL;
137 n->attr.block.backedge = new_backedge_arr(current_ir_graph->obst, get_irn_arity(n));
140 n->attr.phi_backedge = new_backedge_arr(current_ir_graph->obst, get_irn_arity(n));
143 n->attr.filter.backedge = new_backedge_arr(current_ir_graph->obst, get_irn_arity(n));
149 /* Copies the node to the new obstack. The Ins of the new node point to
150 the predecessors on the old obstack. For block/phi nodes not all
151 predecessors might be copied. n->link points to the new node.
152 For Phi and Block nodes the function allocates in-arrays with an arity
153 only for useful predecessors. The arity is determined by counting
154 the non-bad predecessors of the block. */
156 copy_node (ir_node *n, void *env) {
160 if (get_irn_opcode(n) == iro_Block) {
162 new_arity = compute_new_arity(n);
163 n->attr.block.graph_arr = NULL;
165 block = get_nodes_Block(n);
166 if (get_irn_opcode(n) == iro_Phi) {
167 new_arity = compute_new_arity(block);
169 new_arity = get_irn_arity(n);
172 nn = new_ir_node(get_irn_dbg_info(n),
179 /* Copy the attributes. These might point to additional data. If this
180 was allocated on the old obstack the pointers now are dangling. This
181 frees e.g. the memory of the graph_arr allocated in new_immBlock. */
183 new_backedge_info(nn);
186 /* printf("\n old node: "); DDMSG2(n);
187 printf(" new node: "); DDMSG2(nn); */
191 /* Copies new predecessors of old node to new node remembered in link.
192 Spare the Bad predecessors of Phi and Block nodes. */
194 copy_preds (ir_node *n, void *env) {
198 nn = get_new_node(n);
200 /* printf("\n old node: "); DDMSG2(n);
201 printf(" new node: "); DDMSG2(nn);
202 printf(" arities: old: %d, new: %d\n", get_irn_arity(n), get_irn_arity(nn)); */
204 if (get_irn_opcode(n) == iro_Block) {
205 /* Don't copy Bad nodes. */
207 for (i = 0; i < get_irn_arity(n); i++)
208 if (get_irn_opcode(get_irn_n(n, i)) != iro_Bad) {
209 set_irn_n (nn, j, get_new_node(get_irn_n(n, i)));
210 //if (is_backedge(n, i)) set_backedge(nn, j);
213 /* repair the block visited flag from above misuse. Repair it in both
214 graphs so that the old one can still be used. */
215 set_Block_block_visited(nn, 0);
216 set_Block_block_visited(n, 0);
217 /* Local optimization could not merge two subsequent blocks if
218 in array contained Bads. Now it's possible.
219 We don't call optimize_in_place as it requires
220 that the fields in ir_graph are set properly. */
221 if ((get_opt_control_flow_straightening()) &&
222 (get_Block_n_cfgpreds(nn) == 1) &&
223 (get_irn_op(get_Block_cfgpred(nn, 0)) == op_Jmp))
224 exchange(nn, get_nodes_Block(get_Block_cfgpred(nn, 0)));
225 } else if (get_irn_opcode(n) == iro_Phi) {
226 /* Don't copy node if corresponding predecessor in block is Bad.
227 The Block itself should not be Bad. */
228 block = get_nodes_Block(n);
229 set_irn_n (nn, -1, get_new_node(block));
231 for (i = 0; i < get_irn_arity(n); i++)
232 if (get_irn_opcode(get_irn_n(block, i)) != iro_Bad) {
233 set_irn_n (nn, j, get_new_node(get_irn_n(n, i)));
234 //if (is_backedge(n, i)) set_backedge(nn, j);
237 /* If the pre walker reached this Phi after the post walker visited the
238 block block_visited is > 0. */
239 set_Block_block_visited(get_nodes_Block(n), 0);
240 /* Compacting the Phi's ins might generate Phis with only one
242 if (get_irn_arity(n) == 1)
243 exchange(n, get_irn_n(n, 0));
245 for (i = -1; i < get_irn_arity(n); i++)
246 set_irn_n (nn, i, get_new_node(get_irn_n(n, i)));
248 /* Now the new node is complete. We can add it to the hash table for cse.
249 @@@ inlinening aborts if we identify End. Why? */
250 if(get_irn_op(nn) != op_End)
251 add_identities (current_ir_graph->value_table, nn);
254 /* Copies the graph recursively, compacts the keepalive of the end node. */
257 ir_node *oe, *ne; /* old end, new end */
258 ir_node *ka; /* keep alive */
261 oe = get_irg_end(current_ir_graph);
262 /* copy the end node by hand, allocate dynamic in array! */
263 ne = new_ir_node(get_irn_dbg_info(oe),
270 /* Copy the attributes. Well, there might be some in the future... */
272 set_new_node(oe, ne);
274 /* copy the live nodes */
275 irg_walk(get_nodes_Block(oe), copy_node, copy_preds, NULL);
276 /* copy_preds for the end node ... */
277 set_nodes_Block(ne, get_new_node(get_nodes_Block(oe)));
279 /** ... and now the keep alives. **/
280 /* First pick the not marked block nodes and walk them. We must pick these
281 first as else we will oversee blocks reachable from Phis. */
282 for (i = 0; i < get_irn_arity(oe); i++) {
283 ka = get_irn_n(oe, i);
284 if ((get_irn_op(ka) == op_Block) &&
285 (get_irn_visited(ka) < get_irg_visited(current_ir_graph))) {
286 /* We must keep the block alive and copy everything reachable */
287 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
288 irg_walk(ka, copy_node, copy_preds, NULL);
289 add_End_keepalive(ne, get_new_node(ka));
293 /* Now pick the Phis. Here we will keep all! */
294 for (i = 0; i < get_irn_arity(oe); i++) {
295 ka = get_irn_n(oe, i);
296 if ((get_irn_op(ka) == op_Phi)) {
297 if (get_irn_visited(ka) < get_irg_visited(current_ir_graph)) {
298 /* We didn't copy the Phi yet. */
299 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
300 irg_walk(ka, copy_node, copy_preds, NULL);
302 add_End_keepalive(ne, get_new_node(ka));
307 /* Copies the graph reachable from current_ir_graph->end to the obstack
308 in current_ir_graph and fixes the environment.
309 Then fixes the fields in current_ir_graph containing nodes of the
312 copy_graph_env (void) {
314 /* Not all nodes remembered in current_ir_graph might be reachable
315 from the end node. Assure their link is set to NULL, so that
316 we can test whether new nodes have been computed. */
317 set_irn_link(get_irg_frame (current_ir_graph), NULL);
318 set_irn_link(get_irg_globals(current_ir_graph), NULL);
319 set_irn_link(get_irg_args (current_ir_graph), NULL);
321 /* we use the block walk flag for removing Bads from Blocks ins. */
322 inc_irg_block_visited(current_ir_graph);
327 /* fix the fields in current_ir_graph */
328 old_end = get_irg_end(current_ir_graph);
329 set_irg_end (current_ir_graph, get_new_node(old_end));
331 set_irg_end_block (current_ir_graph, get_new_node(get_irg_end_block(current_ir_graph)));
332 if (get_irn_link(get_irg_frame(current_ir_graph)) == NULL) {
333 copy_node (get_irg_frame(current_ir_graph), NULL);
334 copy_preds(get_irg_frame(current_ir_graph), NULL);
336 if (get_irn_link(get_irg_globals(current_ir_graph)) == NULL) {
337 copy_node (get_irg_globals(current_ir_graph), NULL);
338 copy_preds(get_irg_globals(current_ir_graph), NULL);
340 if (get_irn_link(get_irg_args(current_ir_graph)) == NULL) {
341 copy_node (get_irg_args(current_ir_graph), NULL);
342 copy_preds(get_irg_args(current_ir_graph), NULL);
344 set_irg_start (current_ir_graph, get_new_node(get_irg_start(current_ir_graph)));
346 set_irg_start_block(current_ir_graph,
347 get_new_node(get_irg_start_block(current_ir_graph)));
348 set_irg_frame (current_ir_graph, get_new_node(get_irg_frame(current_ir_graph)));
349 set_irg_globals(current_ir_graph, get_new_node(get_irg_globals(current_ir_graph)));
350 set_irg_args (current_ir_graph, get_new_node(get_irg_args(current_ir_graph)));
351 if (get_irn_link(get_irg_bad(current_ir_graph)) == NULL) {
352 copy_node(get_irg_bad(current_ir_graph), NULL);
353 copy_preds(get_irg_bad(current_ir_graph), NULL);
355 set_irg_bad(current_ir_graph, get_new_node(get_irg_bad(current_ir_graph)));
356 if (get_irn_link(get_irg_unknown(current_ir_graph)) == NULL) {
357 copy_node(get_irg_unknown(current_ir_graph), NULL);
358 copy_preds(get_irg_unknown(current_ir_graph), NULL);
360 set_irg_unknown(current_ir_graph, get_new_node(get_irg_unknown(current_ir_graph)));
363 /* Copies all reachable nodes to a new obstack. Removes bad inputs
364 from block nodes and the corresponding inputs from Phi nodes.
365 Merges single exit blocks with single entry blocks and removes
367 Adds all new nodes to a new hash table for cse. Does not
368 perform cse, so the hash table might contain common subexpressions. */
369 /* Amroq call this emigrate() */
371 dead_node_elimination(ir_graph *irg) {
373 struct obstack *graveyard_obst = NULL;
374 struct obstack *rebirth_obst = NULL;
376 /* Remember external state of current_ir_graph. */
377 rem = current_ir_graph;
378 current_ir_graph = irg;
380 /* Handle graph state */
381 assert(get_irg_phase_state(current_ir_graph) != phase_building);
382 free_outs(current_ir_graph);
384 /* @@@ so far we loose loops when copying */
385 set_irg_loop(current_ir_graph, NULL);
387 if (get_optimize() && get_opt_dead_node_elimination()) {
389 /* A quiet place, where the old obstack can rest in peace,
390 until it will be cremated. */
391 graveyard_obst = irg->obst;
393 /* A new obstack, where the reachable nodes will be copied to. */
394 rebirth_obst = (struct obstack *) xmalloc (sizeof (struct obstack));
395 current_ir_graph->obst = rebirth_obst;
396 obstack_init (current_ir_graph->obst);
398 /* We also need a new hash table for cse */
399 del_identities (irg->value_table);
400 irg->value_table = new_identities ();
402 /* Copy the graph from the old to the new obstack */
405 /* Free memory from old unoptimized obstack */
406 obstack_free(graveyard_obst, 0); /* First empty the obstack ... */
407 xfree (graveyard_obst); /* ... then free it. */
410 current_ir_graph = rem;
413 /* Relink bad predeseccors of a block and store the old in array to the
414 link field. This function is called by relink_bad_predecessors().
415 The array of link field starts with the block operand at position 0.
416 If block has bad predecessors, create a new in array without bad preds.
417 Otherwise let in array untouched. */
418 static void relink_bad_block_predecessors(ir_node *n, void *env) {
419 ir_node **new_in, *irn;
420 int i, new_irn_n, old_irn_arity, new_irn_arity = 0;
422 /* if link field of block is NULL, look for bad predecessors otherwise
423 this is allready done */
424 if (get_irn_op(n) == op_Block &&
425 get_irn_link(n) == NULL) {
427 /* save old predecessors in link field (position 0 is the block operand)*/
428 set_irn_link(n, (void *)get_irn_in(n));
430 /* count predecessors without bad nodes */
431 old_irn_arity = get_irn_arity(n);
432 for (i = 0; i < old_irn_arity; i++)
433 if (!is_Bad(get_irn_n(n, i))) new_irn_arity++;
435 /* arity changing: set new predecessors without bad nodes */
436 if (new_irn_arity < old_irn_arity) {
437 /* get new predecessor array without Block predecessor */
438 new_in = NEW_ARR_D (ir_node *, current_ir_graph->obst, (new_irn_arity+1));
440 /* set new predeseccors in array */
443 for (i = 1; i < old_irn_arity; i++) {
444 irn = get_irn_n(n, i);
445 if (!is_Bad(irn)) new_in[new_irn_n++] = irn;
448 } /* ir node has bad predecessors */
450 } /* Block is not relinked */
453 /* Relinks Bad predecesors from Bocks and Phis called by walker
454 remove_bad_predecesors(). If n is a Block, call
455 relink_bad_block_redecessors(). If n is a Phinode, call also the relinking
456 function of Phi's Block. If this block has bad predecessors, relink preds
458 static void relink_bad_predecessors(ir_node *n, void *env) {
459 ir_node *block, **old_in;
460 int i, old_irn_arity, new_irn_arity;
462 /* relink bad predeseccors of a block */
463 if (get_irn_op(n) == op_Block)
464 relink_bad_block_predecessors(n, env);
466 /* If Phi node relink its block and its predecessors */
467 if (get_irn_op(n) == op_Phi) {
469 /* Relink predeseccors of phi's block */
470 block = get_nodes_Block(n);
471 if (get_irn_link(block) == NULL)
472 relink_bad_block_predecessors(block, env);
474 old_in = (ir_node **)get_irn_link(block); /* Of Phi's Block */
475 old_irn_arity = ARR_LEN(old_in);
477 /* Relink Phi predeseccors if count of predeseccors changed */
478 if (old_irn_arity != ARR_LEN(get_irn_in(block))) {
479 /* set new predeseccors in array
480 n->in[0] remains the same block */
482 for(i = 1; i < old_irn_arity; i++)
483 if (!is_Bad((ir_node *)old_in[i])) n->in[new_irn_arity++] = n->in[i];
485 ARR_SETLEN(ir_node *, n->in, new_irn_arity);
488 } /* n is a Phi node */
491 /* Removes Bad Bad predecesors from Blocks and the corresponding
492 inputs to Phi nodes as in dead_node_elimination but without
494 On walking up set the link field to NULL, on walking down call
495 relink_bad_predecessors() (This function stores the old in array
496 to the link field and sets a new in array if arity of predecessors
498 void remove_bad_predecessors(ir_graph *irg) {
499 irg_walk_graph(irg, init_link, relink_bad_predecessors, NULL);
503 /**********************************************************************/
504 /* Funcionality for inlining */
505 /**********************************************************************/
507 /* Copy node for inlineing. Copies the node by calling copy_node and
508 then updates the entity if it's a local one. env must be a pointer
509 to the frame type of the procedure. The new entities must be in
510 the link field of the entities. */
512 copy_node_inline (ir_node *n, void *env) {
514 type *frame_tp = (type *)env;
517 if (get_irn_op(n) == op_Sel) {
518 new = get_new_node (n);
519 assert(get_irn_op(new) == op_Sel);
520 if (get_entity_owner(get_Sel_entity(n)) == frame_tp) {
521 set_Sel_entity(new, get_entity_link(get_Sel_entity(n)));
526 void inline_method(ir_node *call, ir_graph *called_graph) {
528 ir_node *post_call, *post_bl;
530 ir_node *end, *end_bl;
534 ir_node *cf_op = NULL, *bl;
535 int arity, n_ret, n_exc, n_res, i, j, rem_opt;
538 if (!get_optimize() || !get_opt_inline()) return;
539 /* -- Turn off optimizations, this can cause problems when allocating new nodes. -- */
540 rem_opt = get_optimize();
543 /* Handle graph state */
544 assert(get_irg_phase_state(current_ir_graph) != phase_building);
545 assert(get_irg_pinned(current_ir_graph) == pinned);
546 assert(get_irg_pinned(called_graph) == pinned);
547 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
548 set_irg_outs_inconsistent(current_ir_graph);
550 /* -- Check preconditions -- */
551 assert(get_irn_op(call) == op_Call);
552 /* @@@ does not work for InterfaceIII.java after cgana
553 assert(get_Call_type(call) == get_entity_type(get_irg_ent(called_graph)));
554 assert(smaller_type(get_entity_type(get_irg_ent(called_graph)),
555 get_Call_type(call)));
557 assert(get_type_tpop(get_Call_type(call)) == type_method);
558 if (called_graph == current_ir_graph) {
559 set_optimize(rem_opt);
564 the procedure and later replaces the Start node of the called graph.
565 Post_call is the old Call node and collects the results of the called
566 graph. Both will end up being a tuple. -- */
567 post_bl = get_nodes_Block(call);
568 set_irg_current_block(current_ir_graph, post_bl);
569 /* XxMxPxP of Start + parameter of Call */
571 in[1] = get_Call_mem(call);
572 in[2] = get_irg_frame(current_ir_graph);
573 in[3] = get_irg_globals(current_ir_graph);
574 in[4] = new_Tuple (get_Call_n_params(call), get_Call_param_arr(call));
575 pre_call = new_Tuple(5, in);
579 The new block gets the ins of the old block, pre_call and all its
580 predecessors and all Phi nodes. -- */
581 part_block(pre_call);
583 /* -- Prepare state for dead node elimination -- */
584 /* Visited flags in calling irg must be >= flag in called irg.
585 Else walker and arity computation will not work. */
586 if (get_irg_visited(current_ir_graph) <= get_irg_visited(called_graph))
587 set_irg_visited(current_ir_graph, get_irg_visited(called_graph)+1);
588 if (get_irg_block_visited(current_ir_graph)< get_irg_block_visited(called_graph))
589 set_irg_block_visited(current_ir_graph, get_irg_block_visited(called_graph));
590 /* Set pre_call as new Start node in link field of the start node of
591 calling graph and pre_calls block as new block for the start block
593 Further mark these nodes so that they are not visited by the
595 set_irn_link(get_irg_start(called_graph), pre_call);
596 set_irn_visited(get_irg_start(called_graph),
597 get_irg_visited(current_ir_graph));
598 set_irn_link(get_irg_start_block(called_graph),
599 get_nodes_Block(pre_call));
600 set_irn_visited(get_irg_start_block(called_graph),
601 get_irg_visited(current_ir_graph));
603 /* Initialize for compaction of in arrays */
604 inc_irg_block_visited(current_ir_graph);
606 /* -- Replicate local entities of the called_graph -- */
607 /* copy the entities. */
608 called_frame = get_irg_frame_type(called_graph);
609 for (i = 0; i < get_class_n_members(called_frame); i++) {
610 entity *new_ent, *old_ent;
611 old_ent = get_class_member(called_frame, i);
612 new_ent = copy_entity_own(old_ent, get_cur_frame_type());
613 set_entity_link(old_ent, new_ent);
616 /* visited is > than that of called graph. With this trick visited will
617 remain unchanged so that an outer walker, e.g., searching the call nodes
618 to inline, calling this inline will not visit the inlined nodes. */
619 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
621 /* -- Performing dead node elimination inlines the graph -- */
622 /* Copies the nodes to the obstack of current_ir_graph. Updates links to new
624 /* @@@ endless loops are not copied!! -- they should be, I think... */
625 irg_walk(get_irg_end(called_graph), copy_node_inline, copy_preds,
626 get_irg_frame_type(called_graph));
628 /* Repair called_graph */
629 set_irg_visited(called_graph, get_irg_visited(current_ir_graph));
630 set_irg_block_visited(called_graph, get_irg_block_visited(current_ir_graph));
631 set_Block_block_visited(get_irg_start_block(called_graph), 0);
633 /* -- Merge the end of the inlined procedure with the call site -- */
634 /* We will turn the old Call node into a Tuple with the following
637 0: Phi of all Memories of Return statements.
638 1: Jmp from new Block that merges the control flow from all exception
639 predecessors of the old end block.
640 2: Tuple of all arguments.
641 3: Phi of Exception memories.
644 /* -- Precompute some values -- */
645 end_bl = get_new_node(get_irg_end_block(called_graph));
646 end = get_new_node(get_irg_end(called_graph));
647 arity = get_irn_arity(end_bl); /* arity = n_exc + n_ret */
648 n_res = get_method_n_ress(get_Call_type(call));
650 res_pred = (ir_node **) malloc (n_res * sizeof (ir_node *));
651 cf_pred = (ir_node **) malloc (arity * sizeof (ir_node *));
653 set_irg_current_block(current_ir_graph, post_bl); /* just to make sure */
655 /* -- archive keepalives -- */
656 for (i = 0; i < get_irn_arity(end); i++)
657 add_End_keepalive(get_irg_end(current_ir_graph), get_irn_n(end, i));
658 /* The new end node will die, but the in array is not on the obstack ... */
662 Return nodes by Jump nodes. -- */
664 for (i = 0; i < arity; i++) {
666 ret = get_irn_n(end_bl, i);
667 if (get_irn_op(ret) == op_Return) {
668 cf_pred[n_ret] = new_r_Jmp(current_ir_graph, get_nodes_Block(ret));
672 set_irn_in(post_bl, n_ret, cf_pred);
675 turned into a tuple. -- */
676 turn_into_tuple(post_call, 4);
677 /* First the Memory-Phi */
679 for (i = 0; i < arity; i++) {
680 ret = get_irn_n(end_bl, i);
681 if (get_irn_op(ret) == op_Return) {
682 cf_pred[n_ret] = get_Return_mem(ret);
686 phi = new_Phi(n_ret, cf_pred, mode_M);
687 set_Tuple_pred(call, 0, phi);
688 /* Conserve Phi-list for further inlinings -- but might be optimized */
689 if (get_nodes_Block(phi) == post_bl) {
690 set_irn_link(phi, get_irn_link(post_bl));
691 set_irn_link(post_bl, phi);
693 /* Now the real results */
695 for (j = 0; j < n_res; j++) {
697 for (i = 0; i < arity; i++) {
698 ret = get_irn_n(end_bl, i);
699 if (get_irn_op(ret) == op_Return) {
700 cf_pred[n_ret] = get_Return_res(ret, j);
704 phi = new_Phi(n_ret, cf_pred, get_irn_mode(cf_pred[0]));
706 /* Conserve Phi-list for further inlinings -- but might be optimized */
707 if (get_nodes_Block(phi) == post_bl) {
708 set_irn_link(phi, get_irn_link(post_bl));
709 set_irn_link(post_bl, phi);
712 set_Tuple_pred(call, 2, new_Tuple(n_res, res_pred));
714 set_Tuple_pred(call, 2, new_Bad());
716 /* Finally the exception control flow. We need to add a Phi node to
717 collect the memory containing the exception objects. Further we need
718 to add another block to get a correct representation of this Phi. To
719 this block we add a Jmp that resolves into the X output of the Call
720 when the Call is turned into a tuple. */
722 for (i = 0; i < arity; i++) {
724 ret = get_irn_n(end_bl, i);
725 if (is_fragile_op(skip_Proj(ret)) || (get_irn_op(skip_Proj(ret)) == op_Raise)) {
726 cf_pred[n_exc] = ret;
731 new_Block(n_exc, cf_pred); /* watch it: current_block is changed! */
732 set_Tuple_pred(call, 1, new_Jmp());
733 /* The Phi for the memories with the exception objects */
735 for (i = 0; i < arity; i++) {
737 ret = skip_Proj(get_irn_n(end_bl, i));
738 if (get_irn_op(ret) == op_Call) {
739 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 3);
741 } else if (is_fragile_op(ret)) {
742 /* We rely that all cfops have the memory output at the same position. */
743 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 0);
745 } else if (get_irn_op(ret) == op_Raise) {
746 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 1);
750 set_Tuple_pred(call, 3, new_Phi(n_exc, cf_pred, mode_M));
752 set_Tuple_pred(call, 1, new_Bad());
753 set_Tuple_pred(call, 3, new_Bad());
759 If the exception control flow from the Call directly branched to the
760 end block we now have the following control flow predecessor pattern:
761 ProjX -> Tuple -> Jmp.
762 We must remove the Jmp along with it's empty block and add Jmp's
763 predecessors as predecessors of this end block. -- */
764 /* find the problematic predecessor of the end block. */
765 end_bl = get_irg_end_block(current_ir_graph);
766 for (i = 0; i < get_Block_n_cfgpreds(end_bl); i++) {
767 cf_op = get_Block_cfgpred(end_bl, i);
768 if (get_irn_op(cf_op) == op_Proj) {
769 cf_op = get_Proj_pred(cf_op);
770 if (get_irn_op(cf_op) == op_Tuple) {
771 cf_op = get_Tuple_pred(cf_op, 1);
772 assert(get_irn_op(cf_op) == op_Jmp);
778 if (i < get_Block_n_cfgpreds(end_bl)) {
779 bl = get_nodes_Block(cf_op);
780 arity = get_Block_n_cfgpreds(end_bl) + get_Block_n_cfgpreds(bl) - 1;
781 cf_pred = (ir_node **) malloc (arity * sizeof (ir_node *));
782 for (j = 0; j < i; j++)
783 cf_pred[j] = get_Block_cfgpred(end_bl, j);
784 for (j = j; j < i + get_Block_n_cfgpreds(bl); j++)
785 cf_pred[j] = get_Block_cfgpred(bl, j-i);
786 for (j = j; j < arity; j++)
787 cf_pred[j] = get_Block_cfgpred(end_bl, j-get_Block_n_cfgpreds(bl) +1);
788 set_irn_in(end_bl, arity, cf_pred);
792 /* -- Turn cse back on. -- */
793 set_optimize(rem_opt);
796 /********************************************************************/
797 /* Apply inlineing to small methods. */
798 /********************************************************************/
802 /* It makes no sense to inline too many calls in one procedure. Anyways,
803 I didn't get a version with NEW_ARR_F to run. */
804 #define MAX_INLINE 1024
806 static void collect_calls(ir_node *call, void *env) {
807 ir_node **calls = (ir_node **)env;
810 ir_graph *called_irg;
812 if (get_irn_op(call) != op_Call) return;
814 addr = get_Call_ptr(call);
815 if (get_irn_op(addr) == op_Const) {
816 /* Check whether the constant is the pointer to a compiled entity. */
817 tv = get_Const_tarval(addr);
818 if (tarval_to_entity(tv)) {
819 called_irg = get_entity_irg(tarval_to_entity(tv));
820 if (called_irg && pos < MAX_INLINE) {
821 /* The Call node calls a locally defined method. Remember to inline. */
829 /* Inlines all small methods at call sites where the called address comes
830 from a Const node that references the entity representing the called
832 The size argument is a rough measure for the code size of the method:
833 Methods where the obstack containing the firm graph is smaller than
835 void inline_small_irgs(ir_graph *irg, int size) {
837 ir_node *calls[MAX_INLINE];
838 ir_graph *rem = current_ir_graph;
840 if (!(get_optimize() && get_opt_inline())) return;
842 current_ir_graph = irg;
843 /* Handle graph state */
844 assert(get_irg_phase_state(current_ir_graph) != phase_building);
846 /* Find Call nodes to inline.
847 (We can not inline during a walk of the graph, as inlineing the same
848 method several times changes the visited flag of the walked graph:
849 after the first inlineing visited of the callee equals visited of
850 the caller. With the next inlineing both are increased.) */
852 irg_walk(get_irg_end(irg), NULL, collect_calls, (void *) calls);
854 if ((pos > 0) && (pos < MAX_INLINE)) {
855 /* There are calls to inline */
856 collect_phiprojs(irg);
857 for (i = 0; i < pos; i++) {
860 tv = get_Const_tarval(get_Call_ptr(calls[i]));
861 callee = get_entity_irg(tarval_to_entity(tv));
862 if ((_obstack_memory_used(callee->obst) - obstack_room(callee->obst)) < size) {
863 inline_method(calls[i], callee);
868 current_ir_graph = rem;
872 /********************************************************************/
873 /* Code Placement. Pinns all floating nodes to a block where they */
874 /* will be executed only if needed. */
875 /********************************************************************/
877 static pdeq *worklist; /* worklist of ir_node*s */
879 /* Find the earliest correct block for N. --- Place N into the
880 same Block as its dominance-deepest Input. */
882 place_floats_early (ir_node *n)
886 /* we must not run into an infinite loop */
887 assert (irn_not_visited(n));
890 /* Place floating nodes. */
891 if (get_op_pinned(get_irn_op(n)) == floats) {
893 ir_node *b = new_Bad(); /* The block to place this node in */
895 assert(get_irn_op(n) != op_Block);
897 if ((get_irn_op(n) == op_Const) ||
898 (get_irn_op(n) == op_SymConst) ||
900 /* These nodes will not be placed by the loop below. */
901 b = get_irg_start_block(current_ir_graph);
905 /* find the block for this node. */
906 for (i = 0; i < get_irn_arity(n); i++) {
907 ir_node *dep = get_irn_n(n, i);
909 if ((irn_not_visited(dep)) &&
910 (get_op_pinned(get_irn_op(dep)) == floats)) {
911 place_floats_early (dep);
913 /* Because all loops contain at least one pinned node, now all
914 our inputs are either pinned or place_early has already
915 been finished on them. We do not have any unfinished inputs! */
916 dep_block = get_nodes_Block(dep);
917 if ((!is_Bad(dep_block)) &&
918 (get_Block_dom_depth(dep_block) > depth)) {
920 depth = get_Block_dom_depth(dep_block);
922 /* Avoid that the node is placed in the Start block */
923 if ((depth == 1) && (get_Block_dom_depth(get_nodes_Block(n)) > 1)) {
924 b = get_Block_cfg_out(get_irg_start_block(current_ir_graph), 0);
925 assert(b != get_irg_start_block(current_ir_graph));
929 set_nodes_Block(n, b);
932 /* Add predecessors of non floating nodes on worklist. */
933 start = (get_irn_op(n) == op_Block) ? 0 : -1;
934 for (i = start; i < get_irn_arity(n); i++) {
935 ir_node *pred = get_irn_n(n, i);
936 if (irn_not_visited(pred)) {
937 pdeq_putr (worklist, pred);
942 /* Floating nodes form subgraphs that begin at nodes as Const, Load,
943 Start, Call and end at pinned nodes as Store, Call. Place_early
944 places all floating nodes reachable from its argument through floating
945 nodes and adds all beginnings at pinned nodes to the worklist. */
946 static INLINE void place_early (void) {
948 inc_irg_visited(current_ir_graph);
950 /* this inits the worklist */
951 place_floats_early (get_irg_end(current_ir_graph));
953 /* Work the content of the worklist. */
954 while (!pdeq_empty (worklist)) {
955 ir_node *n = pdeq_getl (worklist);
956 if (irn_not_visited(n)) place_floats_early (n);
959 set_irg_outs_inconsistent(current_ir_graph);
960 current_ir_graph->pinned = pinned;
964 /* deepest common dominance ancestor of DCA and CONSUMER of PRODUCER */
966 consumer_dom_dca (ir_node *dca, ir_node *consumer, ir_node *producer)
968 ir_node *block = NULL;
970 /* Compute the latest block into which we can place a node so that it is
972 if (get_irn_op(consumer) == op_Phi) {
973 /* our comsumer is a Phi-node, the effective use is in all those
974 blocks through which the Phi-node reaches producer */
976 ir_node *phi_block = get_nodes_Block(consumer);
977 for (i = 0; i < get_irn_arity(consumer); i++) {
978 if (get_irn_n(consumer, i) == producer) {
979 block = get_nodes_Block(get_Block_cfgpred(phi_block, i));
983 assert(is_no_Block(consumer));
984 block = get_nodes_Block(consumer);
987 /* Compute the deepest common ancestor of block and dca. */
989 if (!dca) return block;
990 while (get_Block_dom_depth(block) > get_Block_dom_depth(dca))
991 block = get_Block_idom(block);
992 while (get_Block_dom_depth(dca) > get_Block_dom_depth(block))
993 dca = get_Block_idom(dca);
995 { block = get_Block_idom(block); dca = get_Block_idom(dca); }
1000 static INLINE int get_irn_loop_depth(ir_node *n) {
1001 return get_loop_depth(get_irn_loop(n));
1004 /* Move n to a block with less loop depth than it's current block. The
1005 new block must be dominated by early. */
1007 move_out_of_loops (ir_node *n, ir_node *early)
1009 ir_node *best, *dca;
1013 /* Find the region deepest in the dominator tree dominating
1014 dca with the least loop nesting depth, but still dominated
1015 by our early placement. */
1016 dca = get_nodes_Block(n);
1018 while (dca != early) {
1019 dca = get_Block_idom(dca);
1020 if (!dca) break; /* should we put assert(dca)? */
1021 if (get_irn_loop_depth(dca) < get_irn_loop_depth(best)) {
1025 if (best != get_nodes_Block(n)) {
1027 printf("Moving out of loop: "); DDMN(n);
1028 printf(" Outermost block: "); DDMN(early);
1029 printf(" Best block: "); DDMN(best);
1030 printf(" Innermost block: "); DDMN(get_nodes_Block(n));
1032 set_nodes_Block(n, best);
1036 /* Find the latest legal block for N and place N into the
1037 `optimal' Block between the latest and earliest legal block.
1038 The `optimal' block is the dominance-deepest block of those
1039 with the least loop-nesting-depth. This places N out of as many
1040 loops as possible and then makes it as controldependant as
1043 place_floats_late (ir_node *n)
1048 assert (irn_not_visited(n)); /* no multiple placement */
1050 /* no need to place block nodes, control nodes are already placed. */
1051 if ((get_irn_op(n) != op_Block) &&
1053 (get_irn_mode(n) != mode_X)) {
1054 /* Remember the early palacement of this block to move it
1055 out of loop no further than the early placement. */
1056 early = get_nodes_Block(n);
1057 /* Assure that our users are all placed, except the Phi-nodes.
1058 --- Each dataflow cycle contains at least one Phi-node. We
1059 have to break the `user has to be placed before the
1060 producer' dependance cycle and the Phi-nodes are the
1061 place to do so, because we need to base our placement on the
1062 final region of our users, which is OK with Phi-nodes, as they
1063 are pinned, and they never have to be placed after a
1064 producer of one of their inputs in the same block anyway. */
1065 for (i = 0; i < get_irn_n_outs(n); i++) {
1066 ir_node *succ = get_irn_out(n, i);
1067 if (irn_not_visited(succ) && (get_irn_op(succ) != op_Phi))
1068 place_floats_late (succ);
1071 /* We have to determine the final block of this node... except for
1073 if ((get_op_pinned(get_irn_op(n)) == floats) &&
1074 (get_irn_op(n) != op_Const) &&
1075 (get_irn_op(n) != op_SymConst)) {
1076 ir_node *dca = NULL; /* deepest common ancestor in the
1077 dominator tree of all nodes'
1078 blocks depending on us; our final
1079 placement has to dominate DCA. */
1080 for (i = 0; i < get_irn_n_outs(n); i++) {
1081 dca = consumer_dom_dca (dca, get_irn_out(n, i), n);
1083 set_nodes_Block(n, dca);
1085 move_out_of_loops (n, early);
1089 mark_irn_visited(n);
1091 /* Add predecessors of all non-floating nodes on list. (Those of floating
1092 nodes are placeded already and therefore are marked.) */
1093 for (i = 0; i < get_irn_n_outs(n); i++) {
1094 if (irn_not_visited(get_irn_out(n, i))) {
1095 pdeq_putr (worklist, get_irn_out(n, i));
1100 static INLINE void place_late(void) {
1102 inc_irg_visited(current_ir_graph);
1104 /* This fills the worklist initially. */
1105 place_floats_late(get_irg_start_block(current_ir_graph));
1106 /* And now empty the worklist again... */
1107 while (!pdeq_empty (worklist)) {
1108 ir_node *n = pdeq_getl (worklist);
1109 if (irn_not_visited(n)) place_floats_late(n);
1113 void place_code(ir_graph *irg) {
1114 ir_graph *rem = current_ir_graph;
1115 current_ir_graph = irg;
1117 if (!(get_optimize() && get_opt_global_cse())) return;
1119 /* Handle graph state */
1120 assert(get_irg_phase_state(irg) != phase_building);
1121 if (get_irg_dom_state(irg) != dom_consistent)
1124 construct_backedges(irg);
1126 /* Place all floating nodes as early as possible. This guarantees
1127 a legal code placement. */
1128 worklist = new_pdeq ();
1131 /* place_early invalidates the outs, place_late needs them. */
1133 /* Now move the nodes down in the dominator tree. This reduces the
1134 unnecessary executions of the node. */
1137 set_irg_outs_inconsistent(current_ir_graph);
1138 del_pdeq (worklist);
1139 current_ir_graph = rem;
1144 /********************************************************************/
1145 /* Control flow optimization. */
1146 /* Removes Bad control flow predecessors and empty blocks. A block */
1147 /* is empty if it contains only a Jmp node. */
1148 /* Blocks can only be removed if they are not needed for the */
1149 /* semantics of Phi nodes. */
1150 /********************************************************************/
1152 /* Removes Tuples from Block control flow predecessors.
1153 Optimizes blocks with equivalent_node().
1154 Replaces n by Bad if n is unreachable control flow. */
1155 static void merge_blocks(ir_node *n, void *env) {
1157 set_irn_link(n, NULL);
1159 if (get_irn_op(n) == op_Block) {
1161 for (i = 0; i < get_Block_n_cfgpreds(n); i++)
1162 /* GL @@@ : is this possible? if (get_opt_normalize()) -- added, all tests go throug.
1163 A different order of optimizations might cause problems. */
1164 if (get_opt_normalize())
1165 set_Block_cfgpred(n, i, skip_Tuple(get_Block_cfgpred(n, i)));
1166 } else if (get_optimize() && (get_irn_mode(n) == mode_X)) {
1167 /* We will soon visit a block. Optimize it before visiting! */
1168 ir_node *b = get_nodes_Block(n);
1169 ir_node *new = equivalent_node(b);
1170 while (irn_not_visited(b) && (!is_Bad(new)) && (new != b)) {
1171 /* We would have to run gigo if new is bad, so we
1172 promote it directly below. */
1173 assert(((b == new) || get_opt_control_flow_straightening() || get_opt_control_flow_weak_simplification()) &&
1174 ("strange flag setting"));
1177 new = equivalent_node(b);
1179 /* GL @@@ get_opt_normalize hinzugefuegt, 5.5.2003 */
1180 if (is_Bad(new) && get_opt_normalize()) exchange (n, new_Bad());
1184 /* Collects all Phi nodes in link list of Block.
1185 Marks all blocks "block_visited" if they contain a node other
1187 static void collect_nodes(ir_node *n, void *env) {
1188 if (is_no_Block(n)) {
1189 ir_node *b = get_nodes_Block(n);
1191 if ((get_irn_op(n) == op_Phi)) {
1192 /* Collect Phi nodes to compact ins along with block's ins. */
1193 set_irn_link(n, get_irn_link(b));
1195 } else if (get_irn_op(n) != op_Jmp) { /* Check for non empty block. */
1196 mark_Block_block_visited(b);
1201 /* Returns true if pred is pred of block */
1202 static int is_pred_of(ir_node *pred, ir_node *b) {
1204 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1205 ir_node *b_pred = get_nodes_Block(get_Block_cfgpred(b, i));
1206 if (b_pred == pred) return 1;
1211 static int test_whether_dispensable(ir_node *b, int pos) {
1212 int i, j, n_preds = 1;
1213 int dispensable = 1;
1214 ir_node *cfop = get_Block_cfgpred(b, pos);
1215 ir_node *pred = get_nodes_Block(cfop);
1217 if (get_Block_block_visited(pred) + 1
1218 < get_irg_block_visited(current_ir_graph)) {
1219 if (!get_optimize() || !get_opt_control_flow_strong_simplification()) {
1220 /* Mark block so that is will not be removed. */
1221 set_Block_block_visited(pred, get_irg_block_visited(current_ir_graph)-1);
1224 /* Seems to be empty. */
1225 if (!get_irn_link(b)) {
1226 /* There are no Phi nodes ==> dispensable. */
1227 n_preds = get_Block_n_cfgpreds(pred);
1229 /* b's pred blocks and pred's pred blocks must be pairwise disjunct.
1230 Work preds < pos as if they were already removed. */
1231 for (i = 0; i < pos; i++) {
1232 ir_node *b_pred = get_nodes_Block(get_Block_cfgpred(b, i));
1233 if (get_Block_block_visited(b_pred) + 1
1234 < get_irg_block_visited(current_ir_graph)) {
1235 for (j = 0; j < get_Block_n_cfgpreds(b_pred); j++) {
1236 ir_node *b_pred_pred = get_nodes_Block(get_Block_cfgpred(b_pred, j));
1237 if (is_pred_of(b_pred_pred, pred)) dispensable = 0;
1240 if (is_pred_of(b_pred, pred)) dispensable = 0;
1243 for (i = pos +1; i < get_Block_n_cfgpreds(b); i++) {
1244 ir_node *b_pred = get_nodes_Block(get_Block_cfgpred(b, i));
1245 if (is_pred_of(b_pred, pred)) dispensable = 0;
1248 set_Block_block_visited(pred, get_irg_block_visited(current_ir_graph)-1);
1251 n_preds = get_Block_n_cfgpreds(pred);
1259 static void optimize_blocks(ir_node *b, void *env) {
1260 int i, j, k, max_preds, n_preds;
1261 ir_node *pred, *phi;
1264 /* Count the number of predecessor if this block is merged with pred blocks
1267 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1268 max_preds += test_whether_dispensable(b, i);
1270 in = (ir_node **) malloc(max_preds * sizeof(ir_node *));
1273 printf(" working on "); DDMN(b);
1274 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1275 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1276 if (is_Bad(get_Block_cfgpred(b, i))) {
1277 printf(" removing Bad %i\n ", i);
1278 } else if (get_Block_block_visited(pred) +1
1279 < get_irg_block_visited(current_ir_graph)) {
1280 printf(" removing pred %i ", i); DDMN(pred);
1281 } else { printf(" Nothing to do for "); DDMN(pred); }
1283 * end Debug output **/
1285 /** Fix the Phi nodes **/
1286 phi = get_irn_link(b);
1288 assert(get_irn_op(phi) == op_Phi);
1289 /* Find the new predecessors for the Phi */
1291 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1292 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1293 if (is_Bad(get_Block_cfgpred(b, i))) {
1295 } else if (get_Block_block_visited(pred) +1
1296 < get_irg_block_visited(current_ir_graph)) {
1297 /* It's an empty block and not yet visited. */
1298 ir_node *phi_pred = get_Phi_pred(phi, i);
1299 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1300 if (get_nodes_Block(phi_pred) == pred) {
1301 assert(get_irn_op(phi_pred) == op_Phi); /* Block is empty!! */
1302 in[n_preds] = get_Phi_pred(phi_pred, j);
1304 in[n_preds] = phi_pred;
1308 /* The Phi_pred node is replaced now if it is a Phi.
1309 In Schleifen kann offenbar der entfernte Phi Knoten legal verwendet werden.
1310 Daher muss der Phiknoten durch den neuen ersetzt werden.
1311 Weiter muss der alte Phiknoten entfernt werden (durch ersetzen oder
1312 durch einen Bad) damit er aus den keep_alive verschwinden kann.
1313 Man sollte also, falls keine Schleife vorliegt, exchange mit new_Bad
1315 if (get_nodes_Block(phi_pred) == pred) {
1316 /* remove the Phi as it might be kept alive. Further there
1317 might be other users. */
1318 exchange(phi_pred, phi); /* geht, ist aber doch semantisch falsch! Warum?? */
1321 in[n_preds] = get_Phi_pred(phi, i);
1326 set_irn_in(phi, n_preds, in);
1328 phi = get_irn_link(phi);
1332 This happens only if merge between loop backedge and single loop entry. **/
1333 for (k = 0; k < get_Block_n_cfgpreds(b); k++) {
1334 pred = get_nodes_Block(get_Block_cfgpred(b, k));
1335 if (get_Block_block_visited(pred) +1
1336 < get_irg_block_visited(current_ir_graph)) {
1337 phi = get_irn_link(pred);
1339 if (get_irn_op(phi) == op_Phi) {
1340 set_nodes_Block(phi, b);
1343 for (i = 0; i < k; i++) {
1344 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1345 if (is_Bad(get_Block_cfgpred(b, i))) {
1347 } else if (get_Block_block_visited(pred) +1
1348 < get_irg_block_visited(current_ir_graph)) {
1349 /* It's an empty block and not yet visited. */
1350 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1351 /* @@@ Hier brauche ich Schleifeninformation!!! Kontrollflusskante
1352 muss Rueckwaertskante sein! (An allen vier in[n_preds] = phi
1353 Anweisungen.) Trotzdem tuts bisher!! */
1362 for (i = 0; i < get_Phi_n_preds(phi); i++) {
1363 in[n_preds] = get_Phi_pred(phi, i);
1366 for (i = k+1; i < get_Block_n_cfgpreds(b); i++) {
1367 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1368 if (is_Bad(get_Block_cfgpred(b, i))) {
1370 } else if (get_Block_block_visited(pred) +1
1371 < get_irg_block_visited(current_ir_graph)) {
1372 /* It's an empty block and not yet visited. */
1373 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1382 set_irn_in(phi, n_preds, in);
1384 phi = get_irn_link(phi);
1389 /** Fix the block **/
1391 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1392 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1393 if (is_Bad(get_Block_cfgpred(b, i))) {
1395 } else if (get_Block_block_visited(pred) +1
1396 < get_irg_block_visited(current_ir_graph)) {
1397 /* It's an empty block and not yet visited. */
1398 assert(get_Block_n_cfgpreds(b) > 1);
1399 /* Else it should be optimized by equivalent_node. */
1400 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1401 in[n_preds] = get_Block_cfgpred(pred, j);
1404 /* Remove block as it might be kept alive. */
1405 exchange(pred, b/*new_Bad()*/);
1407 in[n_preds] = get_Block_cfgpred(b, i);
1411 set_irn_in(b, n_preds, in);
1415 void optimize_cf(ir_graph *irg) {
1418 ir_node *end = get_irg_end(irg);
1419 ir_graph *rem = current_ir_graph;
1420 current_ir_graph = irg;
1422 /* Handle graph state */
1423 assert(get_irg_phase_state(irg) != phase_building);
1424 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
1425 set_irg_outs_inconsistent(current_ir_graph);
1426 if (get_irg_dom_state(current_ir_graph) == dom_consistent)
1427 set_irg_dom_inconsistent(current_ir_graph);
1429 /* Use block visited flag to mark non-empty blocks. */
1430 inc_irg_block_visited(irg);
1431 irg_walk(end, merge_blocks, collect_nodes, NULL);
1433 /* Optimize the standard code. */
1434 irg_block_walk(get_irg_end_block(irg), optimize_blocks, NULL, NULL);
1436 /* Walk all keep alives, optimize them if block, add to new in-array
1437 for end if useful. */
1438 in = NEW_ARR_F (ir_node *, 1);
1439 in[0] = get_nodes_Block(end);
1440 inc_irg_visited(current_ir_graph);
1441 for(i = 0; i < get_End_n_keepalives(end); i++) {
1442 ir_node *ka = get_End_keepalive(end, i);
1443 if (irn_not_visited(ka)) {
1444 if ((get_irn_op(ka) == op_Block) && Block_not_block_visited(ka)) {
1445 set_irg_block_visited(current_ir_graph, /* Don't walk all the way to Start. */
1446 get_irg_block_visited(current_ir_graph)-1);
1447 irg_block_walk(ka, optimize_blocks, NULL, NULL);
1448 mark_irn_visited(ka);
1449 ARR_APP1 (ir_node *, in, ka);
1450 } else if (get_irn_op(ka) == op_Phi) {
1451 mark_irn_visited(ka);
1452 ARR_APP1 (ir_node *, in, ka);
1456 /* DEL_ARR_F(end->in); GL @@@ tut nicht ! */
1459 current_ir_graph = rem;
1464 * Called by walker of remove_critical_cf_edges.
1466 * Place an empty block to an edge between a blocks of multiple
1467 * predecessors and a block of multiple sucessors.
1470 * @param env Envirnment of walker. This field is unused and has
1473 static void walk_critical_cf_edges(ir_node *n, void *env) {
1475 ir_node *pre, *block, **in, *jmp;
1477 /* Block has multiple predecessors */
1478 if ((op_Block == get_irn_op(n)) &&
1479 (get_irn_arity(n) > 1)) {
1480 arity = get_irn_arity(n);
1482 for (i=0; i<arity; i++) {
1483 pre = get_irn_n(n, i);
1484 /* Predecessor has multiple sucessors. Insert new flow edge */
1485 if ((NULL != pre) && (op_Proj == get_irn_op(pre))) {
1487 /* set predeseccor array for new block */
1488 in = NEW_ARR_D (ir_node *, current_ir_graph->obst, 1);
1489 /* set predecessor of new block */
1491 block = new_Block(1, in);
1492 /* insert new jmp node to new block */
1493 switch_block(block);
1496 /* set sucessor of new block */
1497 set_irn_n(n, i, jmp);
1499 } /* predecessor has multiple sucessors */
1500 } /* for all predecessors */
1501 } /* n is a block */
1504 void remove_critical_cf_edges(ir_graph *irg) {
1505 if (get_opt_critical_edges())
1506 irg_walk_graph(irg, NULL, walk_critical_cf_edges, NULL);