1 /* Coyright (C) 1998 - 2000 by Universitaet Karlsruhe
2 ** All rights reserved.
4 ** Author: Christian Schaefer
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 /* Defined in iropt.c */
32 pset *new_identities (void);
33 void del_identities (pset *value_table);
34 void add_identities (pset *value_table, ir_node *node);
36 /********************************************************************/
37 /* apply optimizations of iropt to all nodes. */
38 /********************************************************************/
40 void init_link (ir_node *n, void *env) {
41 set_irn_link(n, NULL);
45 optimize_in_place_wrapper (ir_node *n, void *env) {
49 for (i = 0; i < get_irn_arity(n); i++) {
50 optimized = optimize_in_place_2(get_irn_n(n, i));
51 set_irn_n(n, i, optimized);
54 if (get_irn_op(n) == op_Block) {
55 optimized = optimize_in_place_2(n);
56 if (optimized != n) exchange (n, optimized);
61 local_optimize_graph (ir_graph *irg) {
62 ir_graph *rem = current_ir_graph;
63 current_ir_graph = irg;
65 /* Handle graph state */
66 assert(get_irg_phase_state(irg) != phase_building);
67 if (get_opt_global_cse())
68 set_irg_pinned(current_ir_graph, floats);
69 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
70 set_irg_outs_inconsistent(current_ir_graph);
71 if (get_irg_dom_state(current_ir_graph) == dom_consistent)
72 set_irg_dom_inconsistent(current_ir_graph);
74 /* Clean the value_table in irg for the cse. */
75 del_identities(irg->value_table);
76 irg->value_table = new_identities();
78 /* walk over the graph */
79 irg_walk(irg->end, init_link, optimize_in_place_wrapper, NULL);
81 current_ir_graph = rem;
84 /********************************************************************/
85 /* Routines for dead node elimination / copying garbage collection */
87 /********************************************************************/
89 /* Remeber the new node in the old node by using a field all nodes have. */
91 set_new_node (ir_node *old, ir_node *new)
96 /* Get this new node, before the old node is forgotton.*/
98 get_new_node (ir_node * n)
103 /* We use the block_visited flag to mark that we have computed the
104 number of useful predecessors for this block.
105 Further we encode the new arity in this flag in the old blocks.
106 Remembering the arity is useful, as it saves a lot of pointer
107 accesses. This function is called for all Phi and Block nodes
110 compute_new_arity(ir_node *b) {
114 irg_v = get_irg_block_visited(current_ir_graph);
115 block_v = get_Block_block_visited(b);
116 if (block_v >= irg_v) {
117 /* we computed the number of preds for this block and saved it in the
119 return block_v - irg_v;
121 /* compute the number of good predecessors */
122 res = get_irn_arity(b);
123 for (i = 0; i < get_irn_arity(b); i++)
124 if (get_irn_opcode(get_irn_n(b, i)) == iro_Bad) res--;
125 /* save it in the flag. */
126 set_Block_block_visited(b, irg_v + res);
131 /* Copies the node to the new obstack. The Ins of the new node point to
132 the predecessors on the old obstack. For block/phi nodes not all
133 predecessors might be copied. n->link points to the new node.
134 For Phi and Block nodes the function allocates in-arrays with an arity
135 only for useful predecessors. The arity is determined by counting
136 the non-bad predecessors of the block. */
138 copy_node (ir_node *n, void *env) {
142 if (get_irn_opcode(n) == iro_Block) {
144 new_arity = compute_new_arity(n);
146 block = get_nodes_Block(n);
147 if (get_irn_opcode(n) == iro_Phi) {
148 new_arity = compute_new_arity(block);
150 new_arity = get_irn_arity(n);
153 nn = new_ir_node(get_irn_dbg_info(n),
160 /* Copy the attributes. These might point to additional data. If this
161 was allocated on the old obstack the pointers now are dangling. This
162 frees e.g. the memory of the graph_arr allocated in new_immBlock. */
166 /* printf("\n old node: "); DDMSG2(n);
167 printf(" new node: "); DDMSG2(nn); */
171 /* Copies new predecessors of old node to new node remembered in link.
172 Spare the Bad predecessors of Phi and Block nodes. */
174 copy_preds (ir_node *n, void *env) {
178 nn = get_new_node(n);
180 /* printf("\n old node: "); DDMSG2(n);
181 printf(" new node: "); DDMSG2(nn);
182 printf(" arities: old: %d, new: %d\n", get_irn_arity(n), get_irn_arity(nn)); */
184 if (get_irn_opcode(n) == iro_Block) {
185 /* Don't copy Bad nodes. */
187 for (i = 0; i < get_irn_arity(n); i++)
188 if (get_irn_opcode(get_irn_n(n, i)) != iro_Bad) {
189 set_irn_n (nn, j, get_new_node(get_irn_n(n, i)));
192 /* repair the block visited flag from above misuse. Repair it in both
193 graphs so that the old one can still be used. */
194 set_Block_block_visited(nn, 0);
195 set_Block_block_visited(n, 0);
196 /* Local optimization could not merge two subsequent blocks if
197 in array contained Bads. Now it's possible.
198 We don't call optimize_in_place as it requires
199 that the fields in ir_graph are set properly. */
200 if (get_Block_n_cfgpreds(nn) == 1
201 && get_irn_op(get_Block_cfgpred(nn, 0)) == op_Jmp)
202 exchange(nn, get_nodes_Block(get_Block_cfgpred(nn, 0)));
203 } else if (get_irn_opcode(n) == iro_Phi) {
204 /* Don't copy node if corresponding predecessor in block is Bad.
205 The Block itself should not be Bad. */
206 block = get_nodes_Block(n);
207 set_irn_n (nn, -1, get_new_node(block));
209 for (i = 0; i < get_irn_arity(n); i++)
210 if (get_irn_opcode(get_irn_n(block, i)) != iro_Bad) {
211 set_irn_n (nn, j, get_new_node(get_irn_n(n, i)));
214 /* If the pre walker reached this Phi after the post walker visited the
215 block block_visited is > 0. */
216 set_Block_block_visited(get_nodes_Block(n), 0);
217 /* Compacting the Phi's ins might generate Phis with only one
219 if (get_irn_arity(n) == 1)
220 exchange(n, get_irn_n(n, 0));
222 for (i = -1; i < get_irn_arity(n); i++)
223 set_irn_n (nn, i, get_new_node(get_irn_n(n, i)));
225 /* Now the new node is complete. We can add it to the hash table for cse.
226 @@@ inlinening aborts if we identify End. Why? */
227 if(get_irn_op(nn) != op_End)
228 add_identities (current_ir_graph->value_table, nn);
231 /* Copies the graph recursively, compacts the keepalive of the end node. */
234 ir_node *oe, *ne; /* old end, new end */
235 ir_node *ka; /* keep alive */
238 oe = get_irg_end(current_ir_graph);
239 /* copy the end node by hand, allocate dynamic in array! */
240 ne = new_ir_node(get_irn_dbg_info(oe),
247 /* Copy the attributes. Well, there might be some in the future... */
249 set_new_node(oe, ne);
251 /* copy the live nodes */
252 irg_walk(get_nodes_Block(oe), copy_node, copy_preds, NULL);
253 /* copy_preds for the end node ... */
254 set_nodes_Block(ne, get_new_node(get_nodes_Block(oe)));
256 /** ... and now the keep alives. **/
257 /* First pick the not marked block nodes and walk them. We must pick these
258 first as else we will oversee blocks reachable from Phis. */
259 for (i = 0; i < get_irn_arity(oe); i++) {
260 ka = get_irn_n(oe, i);
261 if ((get_irn_op(ka) == op_Block) &&
262 (get_irn_visited(ka) < get_irg_visited(current_ir_graph))) {
263 /* We must keep the block alive and copy everything reachable */
264 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
265 irg_walk(ka, copy_node, copy_preds, NULL);
266 add_End_keepalive(ne, get_new_node(ka));
270 /* Now pick the Phis. Here we will keep all! */
271 for (i = 0; i < get_irn_arity(oe); i++) {
272 ka = get_irn_n(oe, i);
273 if ((get_irn_op(ka) == op_Phi)) {
274 if (get_irn_visited(ka) < get_irg_visited(current_ir_graph)) {
275 /* We didn't copy the Phi yet. */
276 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
277 irg_walk(ka, copy_node, copy_preds, NULL);
279 add_End_keepalive(ne, get_new_node(ka));
284 /* Copies the graph reachable from current_ir_graph->end to the obstack
285 in current_ir_graph and fixes the environment.
286 Then fixes the fields in current_ir_graph containing nodes of the
290 /* Not all nodes remembered in current_ir_graph might be reachable
291 from the end node. Assure their link is set to NULL, so that
292 we can test whether new nodes have been computed. */
293 set_irn_link(get_irg_frame (current_ir_graph), NULL);
294 set_irn_link(get_irg_globals(current_ir_graph), NULL);
295 set_irn_link(get_irg_args (current_ir_graph), NULL);
297 /* we use the block walk flag for removing Bads from Blocks ins. */
298 inc_irg_block_visited(current_ir_graph);
303 /* fix the fields in current_ir_graph */
304 free_End(get_irg_end(current_ir_graph));
305 set_irg_end (current_ir_graph, get_new_node(get_irg_end(current_ir_graph)));
306 set_irg_end_block (current_ir_graph, get_new_node(get_irg_end_block(current_ir_graph)));
307 if (get_irn_link(get_irg_frame(current_ir_graph)) == NULL) {
308 copy_node (get_irg_frame(current_ir_graph), NULL);
309 copy_preds(get_irg_frame(current_ir_graph), NULL);
311 if (get_irn_link(get_irg_globals(current_ir_graph)) == NULL) {
312 copy_node (get_irg_globals(current_ir_graph), NULL);
313 copy_preds(get_irg_globals(current_ir_graph), NULL);
315 if (get_irn_link(get_irg_args(current_ir_graph)) == NULL) {
316 copy_node (get_irg_args(current_ir_graph), NULL);
317 copy_preds(get_irg_args(current_ir_graph), NULL);
319 set_irg_start (current_ir_graph, get_new_node(get_irg_start(current_ir_graph)));
321 set_irg_start_block(current_ir_graph,
322 get_new_node(get_irg_start_block(current_ir_graph)));
323 set_irg_frame (current_ir_graph, get_new_node(get_irg_frame(current_ir_graph)));
324 set_irg_globals(current_ir_graph, get_new_node(get_irg_globals(current_ir_graph)));
325 set_irg_args (current_ir_graph, get_new_node(get_irg_args(current_ir_graph)));
326 if (get_irn_link(get_irg_bad(current_ir_graph)) == NULL) {
327 copy_node(get_irg_bad(current_ir_graph), NULL);
328 copy_preds(get_irg_bad(current_ir_graph), NULL);
330 set_irg_bad(current_ir_graph, get_new_node(get_irg_bad(current_ir_graph)));
331 if (get_irn_link(get_irg_unknown(current_ir_graph)) == NULL) {
332 copy_node(get_irg_unknown(current_ir_graph), NULL);
333 copy_preds(get_irg_unknown(current_ir_graph), NULL);
335 set_irg_unknown(current_ir_graph, get_new_node(get_irg_unknown(current_ir_graph)));
338 /* Copies all reachable nodes to a new obstack. Removes bad inputs
339 from block nodes and the corresponding inputs from Phi nodes.
340 Merges single exit blocks with single entry blocks and removes
342 Adds all new nodes to a new hash table for cse. Does not
343 perform cse, so the hash table might contain common subexpressions. */
344 /* Amroq call this emigrate() */
346 dead_node_elimination(ir_graph *irg) {
348 struct obstack *graveyard_obst = NULL;
349 struct obstack *rebirth_obst = NULL;
351 /* Remember external state of current_ir_graph. */
352 rem = current_ir_graph;
353 current_ir_graph = irg;
355 /* Handle graph state */
356 assert(get_irg_phase_state(current_ir_graph) != phase_building);
357 free_outs(current_ir_graph);
359 if (get_optimize() && get_opt_dead_node_elimination()) {
361 /* A quiet place, where the old obstack can rest in peace,
362 until it will be cremated. */
363 graveyard_obst = irg->obst;
365 /* A new obstack, where the reachable nodes will be copied to. */
366 rebirth_obst = (struct obstack *) xmalloc (sizeof (struct obstack));
367 current_ir_graph->obst = rebirth_obst;
368 obstack_init (current_ir_graph->obst);
370 /* We also need a new hash table for cse */
371 del_identities (irg->value_table);
372 irg->value_table = new_identities ();
374 /* Copy the graph from the old to the new obstack */
377 /* Free memory from old unoptimized obstack */
378 obstack_free(graveyard_obst, 0); /* First empty the obstack ... */
379 xfree (graveyard_obst); /* ... then free it. */
382 current_ir_graph = rem;
385 /**********************************************************************/
386 /* Funcionality for inlining */
387 /**********************************************************************/
389 /* Copy node for inlineing. Copies the node by calling copy_node and
390 then updates the entity if it's a local one. env must be a pointer
391 to the frame type of the procedure. The new entities must be in
392 the link field of the entities. */
394 copy_node_inline (ir_node *n, void *env) {
396 type *frame_tp = (type *)env;
399 if (get_irn_op(n) == op_Sel) {
400 new = get_new_node (n);
401 assert(get_irn_op(new) == op_Sel);
402 if (get_entity_owner(get_Sel_entity(n)) == frame_tp) {
403 set_Sel_entity(new, get_entity_link(get_Sel_entity(n)));
408 void inline_method(ir_node *call, ir_graph *called_graph) {
410 ir_node *post_call, *post_bl;
412 ir_node *end, *end_bl;
416 ir_node *cf_op = NULL, *bl;
417 int arity, n_ret, n_exc, n_res, i, j, rem_opt;
420 if (!get_optimize() || !get_opt_inline()) return;
421 /** Turn off optimizations, this can cause problems when allocating new nodes. **/
422 rem_opt = get_optimize();
425 /* Handle graph state */
426 assert(get_irg_phase_state(current_ir_graph) != phase_building);
427 assert(get_irg_pinned(current_ir_graph) == pinned);
428 assert(get_irg_pinned(called_graph) == pinned);
429 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
430 set_irg_outs_inconsistent(current_ir_graph);
432 /** Check preconditions **/
433 assert(get_irn_op(call) == op_Call);
434 /* assert(get_Call_type(call) == get_entity_type(get_irg_ent(called_graph))); */
435 assert(smaller_type(get_entity_type(get_irg_ent(called_graph)),
436 get_Call_type(call)));
437 assert(get_type_tpop(get_Call_type(call)) == type_method);
438 if (called_graph == current_ir_graph) return;
441 /** Part the Call node into two nodes. Pre_call collects the parameters of
442 the procedure and later replaces the Start node of the called graph.
443 Post_call is the old Call node and collects the results of the called
444 graph. Both will end up being a tuple. **/
445 post_bl = get_nodes_Block(call);
446 set_irg_current_block(current_ir_graph, post_bl);
447 /* XxMxPxP of Start + parameter of Call */
449 in[1] = get_Call_mem(call);
450 in[2] = get_irg_frame(current_ir_graph);
451 in[3] = get_irg_globals(current_ir_graph);
452 in[4] = new_Tuple (get_Call_n_params(call), get_Call_param_arr(call));
453 pre_call = new_Tuple(5, in);
456 /** Part the block of the Call node into two blocks.
457 The new block gets the ins of the old block, pre_call and all its
458 predecessors and all Phi nodes. **/
459 part_block(pre_call);
461 /** Prepare state for dead node elimination **/
462 /* Visited flags in calling irg must be >= flag in called irg.
463 Else walker and arity computation will not work. */
464 if (get_irg_visited(current_ir_graph) <= get_irg_visited(called_graph))
465 set_irg_visited(current_ir_graph, get_irg_visited(called_graph)+1); /***/
466 if (get_irg_block_visited(current_ir_graph)< get_irg_block_visited(called_graph))
467 set_irg_block_visited(current_ir_graph, get_irg_block_visited(called_graph));
468 /* Set pre_call as new Start node in link field of the start node of
469 calling graph and pre_calls block as new block for the start block
471 Further mark these nodes so that they are not visited by the
473 set_irn_link(get_irg_start(called_graph), pre_call);
474 set_irn_visited(get_irg_start(called_graph),
475 get_irg_visited(current_ir_graph));/***/
476 set_irn_link(get_irg_start_block(called_graph),
477 get_nodes_Block(pre_call));
478 set_irn_visited(get_irg_start_block(called_graph),
479 get_irg_visited(current_ir_graph)); /***/
481 /* Initialize for compaction of in arrays */
482 inc_irg_block_visited(current_ir_graph);
484 /*** Replicate local entities of the called_graph ***/
485 /* copy the entities. */
486 called_frame = get_irg_frame_type(called_graph);
487 for (i = 0; i < get_class_n_members(called_frame); i++) {
488 entity *new_ent, *old_ent;
489 old_ent = get_class_member(called_frame, i);
490 new_ent = copy_entity_own(old_ent, get_cur_frame_type());
491 set_entity_link(old_ent, new_ent);
494 /* visited is > than that of called graph. With this trick visited will
495 remain unchanged so that an outer walker, e.g., searching the call nodes
496 to inline, calling this inline will not visit the inlined nodes. */
497 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
499 /** Performing dead node elimination inlines the graph **/
500 /* Copies the nodes to the obstack of current_ir_graph. Updates links to new
502 /* @@@ endless loops are not copied!! */
503 irg_walk(get_irg_end(called_graph), copy_node_inline, copy_preds,
504 get_irg_frame_type(called_graph));
506 /* Repair called_graph */
507 set_irg_visited(called_graph, get_irg_visited(current_ir_graph));
508 set_irg_block_visited(called_graph, get_irg_block_visited(current_ir_graph));
509 set_Block_block_visited(get_irg_start_block(called_graph), 0);
511 /*** Merge the end of the inlined procedure with the call site ***/
512 /* We will turn the old Call node into a Tuple with the following
515 0: Phi of all Memories of Return statements.
516 1: Jmp from new Block that merges the control flow from all exception
517 predecessors of the old end block.
518 2: Tuple of all arguments.
519 3: Phi of Exception memories.
522 /** Precompute some values **/
523 end_bl = get_new_node(get_irg_end_block(called_graph));
524 end = get_new_node(get_irg_end(called_graph));
525 arity = get_irn_arity(end_bl); /* arity = n_exc + n_ret */
526 n_res = get_method_n_ress(get_Call_type(call));
528 res_pred = (ir_node **) malloc (n_res * sizeof (ir_node *));
529 cf_pred = (ir_node **) malloc (arity * sizeof (ir_node *));
531 set_irg_current_block(current_ir_graph, post_bl); /* just to make sure */
533 /** archive keepalives **/
534 for (i = 0; i < get_irn_arity(end); i++)
535 add_End_keepalive(get_irg_end(current_ir_graph), get_irn_n(end, i));
536 /* The new end node will die, but the in array is not on the obstack ... */
539 /** Collect control flow from Return blocks to post_calls block. Replace
540 Return nodes by Jump nodes. **/
542 for (i = 0; i < arity; i++) {
544 ret = get_irn_n(end_bl, i);
545 if (get_irn_op(ret) == op_Return) {
546 cf_pred[n_ret] = new_r_Jmp(current_ir_graph, get_nodes_Block(ret));
550 set_irn_in(post_bl, n_ret, cf_pred);
552 /** Collect results from Return nodes to post_call. Post_call is
553 turned into a tuple. **/
554 turn_into_tuple(post_call, 4);
555 /* First the Memory-Phi */
557 for (i = 0; i < arity; i++) {
558 ret = get_irn_n(end_bl, i);
559 if (get_irn_op(ret) == op_Return) {
560 cf_pred[n_ret] = get_Return_mem(ret);
564 phi = new_Phi(n_ret, cf_pred, mode_M);
565 set_Tuple_pred(call, 0, phi);
566 set_irn_link(phi, get_irn_link(post_bl)); /* Conserve Phi-list for further inlinings */
567 set_irn_link(post_bl, phi);
568 /* Now the real results */
570 for (j = 0; j < n_res; j++) {
572 for (i = 0; i < arity; i++) {
573 ret = get_irn_n(end_bl, i);
574 if (get_irn_op(ret) == op_Return) {
575 cf_pred[n_ret] = get_Return_res(ret, j);
579 phi = new_Phi(n_ret, cf_pred, get_irn_mode(cf_pred[0]));
581 set_irn_link(phi, get_irn_link(post_bl)); /* Conserve Phi-list for further inlinings */
582 set_irn_link(post_bl, phi);
584 set_Tuple_pred(call, 2, new_Tuple(n_res, res_pred));
586 set_Tuple_pred(call, 2, new_Bad());
588 /* Finally the exception control flow. We need to add a Phi node to
589 collect the memory containing the exception objects. Further we need
590 to add another block to get a correct representation of this Phi. To
591 this block we add a Jmp that resolves into the X output of the Call
592 when the Call is turned into a tuple. */
594 for (i = 0; i < arity; i++) {
596 ret = get_irn_n(end_bl, i);
597 if (is_fragile_op(skip_Proj(ret)) || (get_irn_op(skip_Proj(ret)) == op_Raise)) {
598 cf_pred[n_exc] = ret;
603 new_Block(n_exc, cf_pred); /* whatch it: current_block is changed! */
604 set_Tuple_pred(call, 1, new_Jmp());
605 /* The Phi for the memories with the exception objects */
607 for (i = 0; i < arity; i++) {
609 ret = skip_Proj(get_irn_n(end_bl, i));
610 if (get_irn_op(ret) == op_Call) {
611 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 3);
613 } else if (is_fragile_op(ret)) {
614 /* We rely that all cfops have the memory output at the same position. */
615 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 0);
617 } else if (get_irn_op(ret) == op_Raise) {
618 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 1);
622 set_Tuple_pred(call, 3, new_Phi(n_exc, cf_pred, mode_M));
624 set_Tuple_pred(call, 1, new_Bad());
625 set_Tuple_pred(call, 3, new_Bad());
630 /*** Correct the control flow to the end node.
631 If the exception control flow from the Call directly branched to the
632 end block we now have the following control flow predecessor pattern:
633 ProjX -> Tuple -> Jmp.
634 We must remove the Jmp along with it's empty block and add Jmp's
635 predecessors as predecessors of this end block. ***/
636 /* find the problematic predecessor of the end block. */
637 end_bl = get_irg_end_block(current_ir_graph);
638 for (i = 0; i < get_Block_n_cfgpreds(end_bl); i++) {
639 cf_op = get_Block_cfgpred(end_bl, i);
640 if (get_irn_op(cf_op) == op_Proj) {
641 cf_op = get_Proj_pred(cf_op);
642 if (get_irn_op(cf_op) == op_Tuple) {
643 cf_op = get_Tuple_pred(cf_op, 1);
644 assert(get_irn_op(cf_op) == op_Jmp);
650 if (i < get_Block_n_cfgpreds(end_bl)) {
651 bl = get_nodes_Block(cf_op);
652 arity = get_Block_n_cfgpreds(end_bl) + get_Block_n_cfgpreds(bl) - 1;
653 cf_pred = (ir_node **) malloc (arity * sizeof (ir_node *));
654 for (j = 0; j < i; j++)
655 cf_pred[j] = get_Block_cfgpred(end_bl, j);
656 for (j = j; j < i + get_Block_n_cfgpreds(bl); j++)
657 cf_pred[j] = get_Block_cfgpred(bl, j-i);
658 for (j = j; j < arity; j++)
659 cf_pred[j] = get_Block_cfgpred(end_bl, j-get_Block_n_cfgpreds(bl) +1);
660 set_irn_in(end_bl, arity, cf_pred);
664 /** Turn cse back on. **/
665 set_optimize(rem_opt);
668 /********************************************************************/
669 /* Apply inlineing to small methods. */
670 /********************************************************************/
674 /* It makes no sense to inline too many calls in one procedure. Anyways,
675 I didn't get a version with NEW_ARR_F to run. */
676 #define MAX_INLINE 1024
678 static void collect_calls(ir_node *call, void *env) {
679 ir_node **calls = (ir_node **)env;
682 ir_graph *called_irg;
684 if (get_irn_op(call) != op_Call) return;
686 addr = get_Call_ptr(call);
687 if (get_irn_op(addr) == op_Const) {
688 /* Check whether the constant is the pointer to a compiled entity. */
689 tv = get_Const_tarval(addr);
691 called_irg = get_entity_irg(tv->u.p.ent);
692 if (called_irg && pos < MAX_INLINE) {
693 /* The Call node calls a locally defined method. Remember to inline. */
702 /* Inlines all small methods at call sites where the called address comes
703 from a Const node that references the entity representing the called
705 The size argument is a rough measure for the code size of the method:
706 Methods where the obstack containing the firm graph is smaller than
708 void inline_small_irgs(ir_graph *irg, int size) {
710 ir_node *calls[MAX_INLINE];
711 ir_graph *rem = current_ir_graph;
713 if (!(get_optimize() && get_opt_inline())) return;
715 /*DDME(get_irg_ent(current_ir_graph));*/
717 current_ir_graph = irg;
718 /* Handle graph state */
719 assert(get_irg_phase_state(current_ir_graph) != phase_building);
721 /* Find Call nodes to inline.
722 (We can not inline during a walk of the graph, as inlineing the same
723 method several times changes the visited flag of the walked graph:
724 after the first inlineing visited of the callee equals visited of
725 the caller. With the next inlineing both are increased.) */
727 irg_walk(get_irg_end(irg), NULL, collect_calls, (void *) calls);
729 if ((pos > 0) && (pos < MAX_INLINE)) {
730 /* There are calls to inline */
731 collect_phiprojs(irg);
732 for (i = 0; i < pos; i++) {
735 tv = get_Const_tarval(get_Call_ptr(calls[i]));
736 callee = get_entity_irg(tv->u.p.ent);
737 if ((_obstack_memory_used(callee->obst) - obstack_room(callee->obst)) < size) {
738 /*printf(" inlineing "); DDME(tv->u.p.ent);*/
739 inline_method(calls[i], callee);
744 current_ir_graph = rem;
748 /********************************************************************/
749 /* Code Placement. Pinns all floating nodes to a block where they */
750 /* will be executed only if needed. */
751 /********************************************************************/
753 static pdeq *worklist; /* worklist of ir_node*s */
755 /* Find the earliest correct block for N. --- Place N into the
756 same Block as its dominance-deepest Input. */
758 place_floats_early (ir_node *n)
762 /* we must not run into an infinite loop */
763 assert (irn_not_visited(n));
766 /* Place floating nodes. */
767 if (get_op_pinned(get_irn_op(n)) == floats) {
769 ir_node *b = new_Bad(); /* The block to place this node in */
771 assert(get_irn_op(n) != op_Block);
773 if ((get_irn_op(n) == op_Const) ||
774 (get_irn_op(n) == op_SymConst) ||
776 /* These nodes will not be placed by the loop below. */
777 b = get_irg_start_block(current_ir_graph);
781 /* find the block for this node. */
782 for (i = 0; i < get_irn_arity(n); i++) {
783 ir_node *dep = get_irn_n(n, i);
785 if ((irn_not_visited(dep)) &&
786 (get_op_pinned(get_irn_op(dep)) == floats)) {
787 place_floats_early (dep);
789 /* Because all loops contain at least one pinned node, now all
790 our inputs are either pinned or place_early has already
791 been finished on them. We do not have any unfinished inputs! */
792 dep_block = get_nodes_Block(dep);
793 if ((!is_Bad(dep_block)) &&
794 (get_Block_dom_depth(dep_block) > depth)) {
796 depth = get_Block_dom_depth(dep_block);
798 /* Avoid that the node is placed in the Start block */
799 if ((depth == 1) && (get_Block_dom_depth(get_nodes_Block(n)) > 1)) {
800 b = get_Block_cfg_out(get_irg_start_block(current_ir_graph), 0);
801 assert(b != get_irg_start_block(current_ir_graph));
805 set_nodes_Block(n, b);
808 /* Add predecessors of non floating nodes on worklist. */
809 start = (get_irn_op(n) == op_Block) ? 0 : -1;
810 for (i = start; i < get_irn_arity(n); i++) {
811 ir_node *pred = get_irn_n(n, i);
812 if (irn_not_visited(pred)) {
813 pdeq_putr (worklist, pred);
818 /* Floating nodes form subgraphs that begin at nodes as Const, Load,
819 Start, Call and end at pinned nodes as Store, Call. Place_early
820 places all floating nodes reachable from its argument through floating
821 nodes and adds all beginnings at pinned nodes to the worklist. */
822 INLINE void place_early () {
824 inc_irg_visited(current_ir_graph);
826 /* this inits the worklist */
827 place_floats_early (get_irg_end(current_ir_graph));
829 /* Work the content of the worklist. */
830 while (!pdeq_empty (worklist)) {
831 ir_node *n = pdeq_getl (worklist);
832 if (irn_not_visited(n)) place_floats_early (n);
835 set_irg_outs_inconsistent(current_ir_graph);
836 current_ir_graph->pinned = pinned;
840 /* deepest common dominance ancestor of DCA and CONSUMER of PRODUCER */
842 consumer_dom_dca (ir_node *dca, ir_node *consumer, ir_node *producer)
844 ir_node *block = NULL;
846 /* Compute the latest block into which we can place a node so that it is
848 if (get_irn_op(consumer) == op_Phi) {
849 /* our comsumer is a Phi-node, the effective use is in all those
850 blocks through which the Phi-node reaches producer */
852 ir_node *phi_block = get_nodes_Block(consumer);
853 for (i = 0; i < get_irn_arity(consumer); i++) {
854 if (get_irn_n(consumer, i) == producer) {
855 block = get_nodes_Block(get_Block_cfgpred(phi_block, i));
859 assert(is_no_Block(consumer));
860 block = get_nodes_Block(consumer);
863 /* Compute the deepest common ancestor of block and dca. */
865 if (!dca) return block;
866 while (get_Block_dom_depth(block) > get_Block_dom_depth(dca))
867 block = get_Block_idom(block);
868 while (get_Block_dom_depth(dca) > get_Block_dom_depth(block))
869 dca = get_Block_idom(dca);
871 { block = get_Block_idom(block); dca = get_Block_idom(dca); }
877 /* @@@ Needs loop informations. Will implement later interprocedural. */
879 move_out_of_loops (ir_node *n, ir_node *dca)
883 /* Find the region deepest in the dominator tree dominating
884 dca with the least loop nesting depth, but still dominated
885 by our early placement. */
887 while (dca != get_nodes_Block(n)) {
888 dca = get_Block_idom(dca);
889 if (!dca) break; /* should we put assert(dca)? */
890 if (get_Block_loop_depth(dca) < get_Block_loop_depth(best)) {
894 if (get_Block_dom_depth(best) >= get_Block_dom_depth(get_nodes_Block(n)))
895 set_nodes_Block(n, best);
899 /* Find the latest legal block for N and place N into the
900 `optimal' Block between the latest and earliest legal block.
901 The `optimal' block is the dominance-deepest block of those
902 with the least loop-nesting-depth. This places N out of as many
903 loops as possible and then makes it as controldependant as
906 place_floats_late (ir_node *n)
910 assert (irn_not_visited(n)); /* no multiple placement */
912 /* no need to place block nodes, control nodes are already placed. */
913 if ((get_irn_op(n) != op_Block) && (!is_cfop(n)) && (get_irn_mode(n) != mode_X)) {
914 /* Assure that our users are all placed, except the Phi-nodes.
915 --- Each dataflow cycle contains at least one Phi-node. We
916 have to break the `user has to be placed before the
917 producer' dependance cycle and the Phi-nodes are the
918 place to do so, because we need to base our placement on the
919 final region of our users, which is OK with Phi-nodes, as they
920 are pinned, and they never have to be placed after a
921 producer of one of their inputs in the same block anyway. */
922 for (i = 0; i < get_irn_n_outs(n); i++) {
923 ir_node *succ = get_irn_out(n, i);
924 if (irn_not_visited(succ) && (get_irn_op(succ) != op_Phi))
925 place_floats_late (succ);
928 /* We have to determine the final block of this node... except for constants. */
929 if ((get_op_pinned(get_irn_op(n)) == floats) &&
930 (get_irn_op(n) != op_Const) &&
931 (get_irn_op(n) != op_SymConst)) {
932 ir_node *dca = NULL; /* deepest common ancestor in the
933 dominator tree of all nodes'
934 blocks depending on us; our final
935 placement has to dominate DCA. */
936 for (i = 0; i < get_irn_n_outs(n); i++) {
937 dca = consumer_dom_dca (dca, get_irn_out(n, i), n);
939 set_nodes_Block(n, dca);
941 move_out_of_loops (n, dca);
948 /* Add predecessors of all non-floating nodes on list. (Those of floating
949 nodes are placeded already and therefore are marked.) */
950 for (i = 0; i < get_irn_n_outs(n); i++) {
951 if (irn_not_visited(get_irn_out(n, i))) {
952 pdeq_putr (worklist, get_irn_out(n, i));
957 INLINE void place_late() {
959 inc_irg_visited(current_ir_graph);
961 /* This fills the worklist initially. */
962 place_floats_late(get_irg_start_block(current_ir_graph));
963 /* And now empty the worklist again... */
964 while (!pdeq_empty (worklist)) {
965 ir_node *n = pdeq_getl (worklist);
966 if (irn_not_visited(n)) place_floats_late(n);
970 void place_code(ir_graph *irg) {
971 ir_graph *rem = current_ir_graph;
972 current_ir_graph = irg;
974 if (!(get_optimize() && get_opt_global_cse())) return;
976 /* Handle graph state */
977 assert(get_irg_phase_state(irg) != phase_building);
978 if (get_irg_dom_state(irg) != dom_consistent)
981 /* Place all floating nodes as early as possible. This guarantees
982 a legal code placement. */
983 worklist = new_pdeq ();
986 /* place_early invalidates the outs, place_late needs them. */
988 /* Now move the nodes down in the dominator tree. This reduces the
989 unnecessary executions of the node. */
992 set_irg_outs_inconsistent(current_ir_graph);
994 current_ir_graph = rem;
999 /********************************************************************/
1000 /* Control flow optimization. */
1001 /* Removes Bad control flow predecessors and empty blocks. A block */
1002 /* is empty if it contains only a Jmp node. */
1004 /********************************************************************/
1007 static void merge_blocks(ir_node *n, void *env) {
1009 set_irn_link(n, NULL);
1011 if (get_irn_op(n) == op_Block) {
1013 for (i = 0; i < get_Block_n_cfgpreds(n); i++)
1014 set_Block_cfgpred(n, i, skip_Tuple(get_Block_cfgpred(n, i)));
1015 } else if (get_irn_mode(n) == mode_X) {
1016 /* We will soon visit a block. Optimize it before visiting! */
1017 ir_node *b = get_nodes_Block(n);
1018 ir_node *new = equivalent_node(b);
1019 while (irn_not_visited(b) && (!is_Bad(new)) && (new != b)) {
1020 /* We would have to run gigo if new is bad. */
1021 if (get_optimize() && get_opt_control_flow()) exchange (b, new);
1023 new = equivalent_node(b);
1025 if (is_Bad(new)) exchange (n, new_Bad());
1029 static void collect_nodes(ir_node *n, void *env) {
1030 if (is_no_Block(n)) {
1031 ir_node *b = get_nodes_Block(n);
1033 if ((get_irn_op(n) == op_Phi)) {
1034 /* Collect Phi nodes to compact ins along with block's ins. */
1035 set_irn_link(n, get_irn_link(b));
1037 } else if (get_irn_op(n) != op_Jmp) { /* Check for non empty block. */
1038 mark_Block_block_visited(b);
1043 /* Returns true if pred is pred of block */
1044 int is_pred_of(ir_node *pred, ir_node *b) {
1046 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1047 ir_node *b_pred = get_nodes_Block(get_Block_cfgpred(b, i));
1048 if (b_pred == pred) return 1;
1053 int test_whether_dispensable(ir_node *b, int pos) {
1054 int i, j, n_preds = 1;
1055 int dispensable = 1;
1056 ir_node *cfop = get_Block_cfgpred(b, pos);
1057 ir_node *pred = get_nodes_Block(cfop);
1059 if (get_Block_block_visited(pred) + 1
1060 < get_irg_block_visited(current_ir_graph)) {
1061 if (!get_optimize() || !get_opt_control_flow()) {
1062 /* Mark block so that is will not be removed. */
1063 set_Block_block_visited(pred, get_irg_block_visited(current_ir_graph)-1);
1066 /* Seems to be empty. */
1067 if (!get_irn_link(b)) {
1068 /* There are no Phi nodes ==> dispensable. */
1069 n_preds = get_Block_n_cfgpreds(pred);
1071 /* b's pred blocks and pred's pred blocks must be pairwise disjunct.
1072 Work preds < pos as if they were already removed. */
1073 for (i = 0; i < pos; i++) {
1074 ir_node *b_pred = get_nodes_Block(get_Block_cfgpred(b, i));
1075 if (get_Block_block_visited(b_pred) + 1
1076 < get_irg_block_visited(current_ir_graph)) {
1077 for (j = 0; j < get_Block_n_cfgpreds(b_pred); j++) {
1078 ir_node *b_pred_pred = get_nodes_Block(get_Block_cfgpred(b_pred, j));
1079 if (is_pred_of(b_pred_pred, pred)) dispensable = 0;
1082 if (is_pred_of(b_pred, pred)) dispensable = 0;
1085 for (i = pos +1; i < get_Block_n_cfgpreds(b); i++) {
1086 ir_node *b_pred = get_nodes_Block(get_Block_cfgpred(b, i));
1087 if (is_pred_of(b_pred, pred)) dispensable = 0;
1090 set_Block_block_visited(pred, get_irg_block_visited(current_ir_graph)-1);
1093 n_preds = get_Block_n_cfgpreds(pred);
1101 void optimize_blocks(ir_node *b, void *env) {
1102 int i, j, k, max_preds, n_preds;
1103 ir_node *pred, *phi;
1107 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1108 pred = get_Block_cfgpred(b, i);
1109 max_preds += test_whether_dispensable(b, i);
1111 in = (ir_node **) malloc(max_preds * sizeof(ir_node *));
1115 printf(" working on "); DDMN(b);
1116 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1117 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1118 if (is_Bad(get_Block_cfgpred(b, i))) {
1119 printf(" removing Bad %i\n ", i);
1120 } else if (get_Block_block_visited(pred) +1
1121 < get_irg_block_visited(current_ir_graph)) {
1122 printf(" removing pred %i ", i); DDMN(pred);
1123 } else { printf(" Nothing to do for "); DDMN(pred); }
1125 ** end Debug output **/
1127 /** Fix the Phi nodes **/
1128 phi = get_irn_link(b);
1130 assert(get_irn_op(phi) == op_Phi);
1131 /* Find the new predecessors for the Phi */
1133 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1134 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1135 if (is_Bad(get_Block_cfgpred(b, i))) {
1137 } else if (get_Block_block_visited(pred) +1
1138 < get_irg_block_visited(current_ir_graph)) {
1139 /* It's an empty block and not yet visited. */
1140 ir_node *phi_pred = get_Phi_pred(phi, i);
1141 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1142 if (get_nodes_Block(phi_pred) == pred) {
1143 assert(get_irn_op(phi_pred) == op_Phi); /* Block is empty!! */
1144 in[n_preds] = get_Phi_pred(phi_pred, j);
1146 in[n_preds] = phi_pred;
1151 /* @@@ hier brauche ich schleifeninformation!!! Wenn keine Rueckwaertskante
1152 dann darfs auch keine Verwendung geben. */
1153 if (get_nodes_Block(phi_pred) == pred) {
1154 /* remove the Phi as it might be kept alive. Further there
1155 might be other users. */
1156 exchange(phi_pred, phi); /* geht, is aber doch semantisch falsch! */
1160 in[n_preds] = get_Phi_pred(phi, i);
1165 set_irn_in(phi, n_preds, in);
1166 phi = get_irn_link(phi);
1169 /** Move Phi nodes from removed blocks to this one.
1170 This happens only if merge between loop backedge and single loop entry. **/
1171 for (k = 0; k < get_Block_n_cfgpreds(b); k++) {
1172 pred = get_nodes_Block(get_Block_cfgpred(b, k));
1173 if (get_Block_block_visited(pred) +1
1174 < get_irg_block_visited(current_ir_graph)) {
1175 phi = get_irn_link(pred);
1177 if (get_irn_op(phi) == op_Phi) {
1178 set_nodes_Block(phi, b);
1181 for (i = 0; i < k; i++) {
1182 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1183 if (is_Bad(get_Block_cfgpred(b, i))) {
1185 } else if (get_Block_block_visited(pred) +1
1186 < get_irg_block_visited(current_ir_graph)) {
1187 /* It's an empty block and not yet visited. */
1188 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1189 /* @@@ Hier brauche ich schleifeninformation!!! Kontrllflusskante
1190 muss Rueckwaertskante sein! (An allen vier in[n_preds] = phi
1191 Anweisungen.) Trotzdem tuts bisher!! */
1200 for (i = 0; i < get_Phi_n_preds(phi); i++) {
1201 in[n_preds] = get_Phi_pred(phi, i);
1204 for (i = k+1; i < get_Block_n_cfgpreds(b); i++) {
1205 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1206 if (is_Bad(get_Block_cfgpred(b, i))) {
1208 } else if (get_Block_block_visited(pred) +1
1209 < get_irg_block_visited(current_ir_graph)) {
1210 /* It's an empty block and not yet visited. */
1211 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1220 set_irn_in(phi, n_preds, in);
1222 phi = get_irn_link(phi);
1227 /** Fix the block **/
1229 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1230 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1231 if (is_Bad(get_Block_cfgpred(b, i))) {
1233 } else if (get_Block_block_visited(pred) +1
1234 < get_irg_block_visited(current_ir_graph)) {
1235 /* It's an empty block and not yet visited. */
1236 assert(get_Block_n_cfgpreds(b) > 1);
1237 /* Else it should be optimized by equivalent_node. */
1238 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1239 in[n_preds] = get_Block_cfgpred(pred, j);
1242 /* Remove block as it might be kept alive. */
1243 exchange(pred, b/*new_Bad()*/);
1245 in[n_preds] = get_Block_cfgpred(b, i);
1249 set_irn_in(b, n_preds, in);
1253 void optimize_cf(ir_graph *irg) {
1256 ir_node *end = get_irg_end(irg);
1257 ir_graph *rem = current_ir_graph;
1258 current_ir_graph = irg;
1261 /* Handle graph state */
1262 assert(get_irg_phase_state(irg) != phase_building);
1263 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
1264 set_irg_outs_inconsistent(current_ir_graph);
1265 if (get_irg_dom_state(current_ir_graph) == dom_consistent)
1266 set_irg_dom_inconsistent(current_ir_graph);
1268 /* Use block visited flag to mark non-empty blocks. */
1269 inc_irg_block_visited(irg);
1270 irg_walk(end, merge_blocks, collect_nodes, NULL);
1272 /* Optimize the standard code. */
1273 irg_block_walk(get_irg_end_block(irg), optimize_blocks, NULL, NULL);
1275 /* Walk all keep alives, optimize them if block, add to new in-array
1276 for end if useful. */
1277 in = NEW_ARR_F (ir_node *, 1);
1278 in[0] = get_nodes_Block(end);
1280 inc_irg_visited(current_ir_graph);
1281 for(i = 0; i < get_End_n_keepalives(end); i++) {
1282 ir_node *ka = get_End_keepalive(end, i);
1283 if (irn_not_visited(ka)) {
1284 if ((get_irn_op(ka) == op_Block) && Block_not_block_visited(ka)) {
1285 set_irg_block_visited(current_ir_graph, /* Don't walk all the way to Start. */
1286 get_irg_block_visited(current_ir_graph)-1);
1287 irg_block_walk(ka, optimize_blocks, NULL, NULL);
1288 mark_irn_visited(ka);
1289 ARR_APP1 (ir_node *, in, ka);
1290 } else if (get_irn_op(ka) == op_Phi) {
1291 mark_irn_visited(ka);
1292 ARR_APP1 (ir_node *, in, ka);
1296 /* DEL_ARR_F(end->in); GL @@@ tut nicht ! */
1299 current_ir_graph = rem;