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(current_ir_graph,
159 /* Copy the attributes. These might point to additional data. If this
160 was allocated on the old obstack the pointers now are dangling. This
161 frees e.g. the memory of the graph_arr allocated in new_immBlock. */
165 /* printf("\n old node: "); DDMSG2(n);
166 printf(" new node: "); DDMSG2(nn); */
170 /* Copies new predecessors of old node to new node remembered in link.
171 Spare the Bad predecessors of Phi and Block nodes. */
173 copy_preds (ir_node *n, void *env) {
174 ir_node *nn, *block, *on;
177 nn = get_new_node(n);
179 /* printf("\n old node: "); DDMSG2(n);
180 printf(" new node: "); DDMSG2(nn);
181 printf(" arities: old: %d, new: %d\n", get_irn_arity(n), get_irn_arity(nn)); */
183 if (get_irn_opcode(n) == iro_Block) {
184 /* Don't copy Bad nodes. */
186 for (i = 0; i < get_irn_arity(n); i++)
187 if (get_irn_opcode(get_irn_n(n, i)) != iro_Bad) {
188 set_irn_n (nn, j, get_new_node(get_irn_n(n, i)));
191 /* repair the block visited flag from above misuse. Repair it in both
192 graphs so that the old one can still be used. */
193 set_Block_block_visited(nn, 0);
194 set_Block_block_visited(n, 0);
195 /* Local optimization could not merge two subsequent blocks if
196 in array contained Bads. Now it's possible.
197 We don't call optimize_in_place as it requires
198 that the fields in ir_graph are set properly. */
199 if (get_Block_n_cfgpreds(nn) == 1
200 && get_irn_op(get_Block_cfgpred(nn, 0)) == op_Jmp)
201 exchange(nn, get_nodes_Block(get_Block_cfgpred(nn, 0)));
202 } else if (get_irn_opcode(n) == iro_Phi) {
203 /* Don't copy node if corresponding predecessor in block is Bad.
204 The Block itself should not be Bad. */
205 block = get_nodes_Block(n);
206 set_irn_n (nn, -1, get_new_node(block));
208 for (i = 0; i < get_irn_arity(n); i++)
209 if (get_irn_opcode(get_irn_n(block, i)) != iro_Bad) {
210 set_irn_n (nn, j, get_new_node(get_irn_n(n, i)));
213 /* If the pre walker reached this Phi after the post walker visited the
214 block block_visited is > 0. */
215 set_Block_block_visited(get_nodes_Block(n), 0);
216 /* Compacting the Phi's ins might generate Phis with only one
218 if (get_irn_arity(n) == 1)
219 exchange(n, get_irn_n(n, 0));
221 for (i = -1; i < get_irn_arity(n); i++)
222 set_irn_n (nn, i, get_new_node(get_irn_n(n, i)));
224 /* Now the new node is complete. We can add it to the hash table for cse.
225 @@@ inlinening aborts if we identify End. Why? */
226 if(get_irn_op(nn) != op_End)
227 add_identities (current_ir_graph->value_table, nn);
230 /* Copies the graph recursively, compacts the keepalive of the end node. */
233 ir_node *oe, *ne; /* old end, new end */
234 ir_node *ka; /* keep alive */
237 oe = get_irg_end(current_ir_graph);
238 /* copy the end node by hand, allocate dynamic in array! */
239 ne = new_ir_node(current_ir_graph,
245 /* Copy the attributes. Well, there might be some in the future... */
247 set_new_node(oe, ne);
249 /* copy the live nodes */
250 irg_walk(get_nodes_Block(oe), copy_node, copy_preds, NULL);
251 /* copy_preds for the end node ... */
252 set_nodes_Block(ne, get_new_node(get_nodes_Block(oe)));
254 /** ... and now the keep alives. **/
255 /* First pick the not marked block nodes and walk them. We must pick these
256 first as else we will oversee blocks reachable from Phis. */
257 for (i = 0; i < get_irn_arity(oe); i++) {
258 ka = get_irn_n(oe, i);
259 if ((get_irn_op(ka) == op_Block) &&
260 (get_irn_visited(ka) < get_irg_visited(current_ir_graph))) {
261 /* We must keep the block alive and copy everything reachable */
262 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
263 irg_walk(ka, copy_node, copy_preds, NULL);
264 add_End_keepalive(ne, get_new_node(ka));
268 /* Now pick the Phis. Here we will keep all! */
269 for (i = 0; i < get_irn_arity(oe); i++) {
270 ka = get_irn_n(oe, i);
271 if ((get_irn_op(ka) == op_Phi)) {
272 if (get_irn_visited(ka) < get_irg_visited(current_ir_graph)) {
273 /* We didn't copy the Phi yet. */
274 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
275 irg_walk(ka, copy_node, copy_preds, NULL);
277 add_End_keepalive(ne, get_new_node(ka));
282 /* Copies the graph reachable from current_ir_graph->end to the obstack
283 in current_ir_graph and fixes the environment.
284 Then fixes the fields in current_ir_graph containing nodes of the
288 /* Not all nodes remembered in current_ir_graph might be reachable
289 from the end node. Assure their link is set to NULL, so that
290 we can test whether new nodes have been computed. */
291 set_irn_link(get_irg_frame (current_ir_graph), NULL);
292 set_irn_link(get_irg_globals(current_ir_graph), NULL);
293 set_irn_link(get_irg_args (current_ir_graph), NULL);
295 /* we use the block walk flag for removing Bads from Blocks ins. */
296 inc_irg_block_visited(current_ir_graph);
301 /* fix the fields in current_ir_graph */
302 free_End(get_irg_end(current_ir_graph));
303 set_irg_end (current_ir_graph, get_new_node(get_irg_end(current_ir_graph)));
304 set_irg_end_block (current_ir_graph, get_new_node(get_irg_end_block(current_ir_graph)));
305 if (get_irn_link(get_irg_frame(current_ir_graph)) == NULL) {
306 copy_node (get_irg_frame(current_ir_graph), NULL);
307 copy_preds(get_irg_frame(current_ir_graph), NULL);
309 if (get_irn_link(get_irg_globals(current_ir_graph)) == NULL) {
310 copy_node (get_irg_globals(current_ir_graph), NULL);
311 copy_preds(get_irg_globals(current_ir_graph), NULL);
313 if (get_irn_link(get_irg_args(current_ir_graph)) == NULL) {
314 copy_node (get_irg_args(current_ir_graph), NULL);
315 copy_preds(get_irg_args(current_ir_graph), NULL);
317 set_irg_start (current_ir_graph, get_new_node(get_irg_start(current_ir_graph)));
319 set_irg_start_block(current_ir_graph,
320 get_new_node(get_irg_start_block(current_ir_graph)));
321 set_irg_frame (current_ir_graph, get_new_node(get_irg_frame(current_ir_graph)));
322 set_irg_globals(current_ir_graph, get_new_node(get_irg_globals(current_ir_graph)));
323 set_irg_args (current_ir_graph, get_new_node(get_irg_args(current_ir_graph)));
324 if (get_irn_link(get_irg_bad(current_ir_graph)) == NULL) {
325 copy_node(get_irg_bad(current_ir_graph), NULL);
326 copy_preds(get_irg_bad(current_ir_graph), NULL);
328 set_irg_bad(current_ir_graph, get_new_node(get_irg_bad(current_ir_graph)));
331 /* Copies all reachable nodes to a new obstack. Removes bad inputs
332 from block nodes and the corresponding inputs from Phi nodes.
333 Merges single exit blocks with single entry blocks and removes
335 Adds all new nodes to a new hash table for cse. Does not
336 perform cse, so the hash table might contain common subexpressions. */
337 /* Amroq call this emigrate() */
339 dead_node_elimination(ir_graph *irg) {
341 struct obstack *graveyard_obst = NULL;
342 struct obstack *rebirth_obst = NULL;
344 /* Remember external state of current_ir_graph. */
345 rem = current_ir_graph;
346 current_ir_graph = irg;
348 /* Handle graph state */
349 assert(get_irg_phase_state(current_ir_graph) != phase_building);
350 free_outs(current_ir_graph);
352 if (get_optimize() && get_opt_dead_node_elimination()) {
354 /* A quiet place, where the old obstack can rest in peace,
355 until it will be cremated. */
356 graveyard_obst = irg->obst;
358 /* A new obstack, where the reachable nodes will be copied to. */
359 rebirth_obst = (struct obstack *) xmalloc (sizeof (struct obstack));
360 current_ir_graph->obst = rebirth_obst;
361 obstack_init (current_ir_graph->obst);
363 /* We also need a new hash table for cse */
364 del_identities (irg->value_table);
365 irg->value_table = new_identities ();
367 /* Copy the graph from the old to the new obstack */
370 /* Free memory from old unoptimized obstack */
371 obstack_free(graveyard_obst, 0); /* First empty the obstack ... */
372 xfree (graveyard_obst); /* ... then free it. */
375 current_ir_graph = rem;
378 /**********************************************************************/
379 /* Funcionality for inlining */
380 /**********************************************************************/
382 /* Copy node for inlineing. Copies the node by calling copy_node and
383 then updates the entity if it's a local one. env must be a pointer
384 to the frame type of the procedure. The new entities must be in
385 the link field of the entities. */
387 copy_node_inline (ir_node *n, void *env) {
389 type *frame_tp = (type *)env;
392 if (get_irn_op(n) == op_Sel) {
393 new = get_new_node (n);
394 assert(get_irn_op(new) == op_Sel);
395 if (get_entity_owner(get_Sel_entity(n)) == frame_tp) {
396 set_Sel_entity(new, get_entity_link(get_Sel_entity(n)));
401 void inline_method(ir_node *call, ir_graph *called_graph) {
403 ir_node *post_call, *post_bl;
405 ir_node *end, *end_bl;
410 int arity, n_ret, n_exc, n_res, i, j, rem_opt;
411 type *called_frame, *caller_frame;
413 if (!get_optimize() || !get_opt_inline()) return;
414 /** Turn off optimizations, this can cause problems when allocating new nodes. **/
415 rem_opt = get_optimize();
418 /* Handle graph state */
419 assert(get_irg_phase_state(current_ir_graph) != phase_building);
420 assert(get_irg_pinned(current_ir_graph) == pinned);
421 assert(get_irg_pinned(called_graph) == pinned);
422 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
423 set_irg_outs_inconsistent(current_ir_graph);
425 /** Check preconditions **/
426 assert(get_irn_op(call) == op_Call);
427 assert(get_Call_type(call) == get_entity_type(get_irg_ent(called_graph)));
428 assert(get_type_tpop(get_Call_type(call)) == type_method);
429 if (called_graph == current_ir_graph) return;
432 /** Part the Call node into two nodes. Pre_call collects the parameters of
433 the procedure and later replaces the Start node of the called graph.
434 Post_call is the old Call node and collects the results of the called
435 graph. Both will end up being a tuple. **/
436 post_bl = get_nodes_Block(call);
437 set_irg_current_block(current_ir_graph, post_bl);
438 /* XxMxPxP of Start + parameter of Call */
440 in[1] = get_Call_mem(call);
441 in[2] = get_irg_frame(current_ir_graph);
442 in[3] = get_irg_globals(current_ir_graph);
443 in[4] = new_Tuple (get_Call_n_params(call), get_Call_param_arr(call));
444 pre_call = new_Tuple(5, in);
447 /** Part the block of the Call node into two blocks.
448 The new block gets the ins of the old block, pre_call and all its
449 predecessors and all Phi nodes. **/
450 part_block(pre_call);
452 /** Prepare state for dead node elimination **/
453 /* Visited flags in calling irg must be >= flag in called irg.
454 Else walker and arity computation will not work. */
455 if (get_irg_visited(current_ir_graph) <= get_irg_visited(called_graph))
456 set_irg_visited(current_ir_graph, get_irg_visited(called_graph)+1); /***/
457 if (get_irg_block_visited(current_ir_graph)< get_irg_block_visited(called_graph))
458 set_irg_block_visited(current_ir_graph, get_irg_block_visited(called_graph));
459 /* Set pre_call as new Start node in link field of the start node of
460 calling graph and pre_calls block as new block for the start block
462 Further mark these nodes so that they are not visited by the
464 set_irn_link(get_irg_start(called_graph), pre_call);
465 set_irn_visited(get_irg_start(called_graph),
466 get_irg_visited(current_ir_graph));/***/
467 set_irn_link(get_irg_start_block(called_graph),
468 get_nodes_Block(pre_call));
469 set_irn_visited(get_irg_start_block(called_graph),
470 get_irg_visited(current_ir_graph)); /***/
472 /* Initialize for compaction of in arrays */
473 inc_irg_block_visited(current_ir_graph);
475 set_Block_block_visited(get_irg_start_block(called_graph),
476 get_irg_block_visited(current_ir_graph) +1 +1); /* count for self edge */
478 /*** Replicate local entities of the called_graph ***/
479 /* copy the entities. */
480 called_frame = get_irg_frame_type(called_graph);
481 for (i = 0; i < get_class_n_member(called_frame); i++) {
482 entity *new_ent, *old_ent;
483 old_ent = get_class_member(called_frame, i);
484 new_ent = copy_entity_own(old_ent, get_cur_frame_type());
485 set_entity_link(old_ent, new_ent);
488 /* visited is > than that of called graph. With this trick visited will
489 remain unchanged so that an outer walker, e.g., searching the call nodes
490 to inline, calling this inline will not visit the inlined nodes. */
491 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
493 /** Performing dead node elimination inlines the graph **/
494 /* Copies the nodes to the obstack of current_ir_graph. Updates links to new
496 /* @@@ endless loops are not copied!! */
497 irg_walk(get_irg_end(called_graph), copy_node_inline, copy_preds,
498 get_irg_frame_type(called_graph));
500 /* Repair called_graph */
501 set_irg_visited(called_graph, get_irg_visited(current_ir_graph));
502 set_irg_block_visited(called_graph, get_irg_block_visited(current_ir_graph));
503 set_Block_block_visited(get_irg_start_block(called_graph), 0);
505 /*** Merge the end of the inlined procedure with the call site ***/
506 /* We will turn the old Call node into a Tuple with the following
509 0: Phi of all Memories of Return statements.
510 1: Jmp from new Block that merges the control flow from all exception
511 predecessors of the old end block.
512 2: Tuple of all arguments.
513 3: Phi of Exception memories.
516 /** Precompute some values **/
517 end_bl = get_new_node(get_irg_end_block(called_graph));
518 end = get_new_node(get_irg_end(called_graph));
519 arity = get_irn_arity(end_bl); /* arity = n_exc + n_ret */
520 n_res = get_method_n_res(get_Call_type(call));
522 res_pred = (ir_node **) malloc (n_res * sizeof (ir_node *));
523 cf_pred = (ir_node **) malloc (arity * sizeof (ir_node *));
525 set_irg_current_block(current_ir_graph, post_bl); /* just to make sure */
527 /** archive keepalives **/
528 for (i = 0; i < get_irn_arity(end); i++)
529 add_End_keepalive(get_irg_end(current_ir_graph), get_irn_n(end, i));
530 /* The new end node will die, but the in array is not on the obstack ... */
533 /** Collect control flow from Return blocks to post_calls block. Replace
534 Return nodes by Jump nodes. **/
536 for (i = 0; i < arity; i++) {
538 ret = get_irn_n(end_bl, i);
539 if (get_irn_op(ret) == op_Return) {
540 cf_pred[n_ret] = new_r_Jmp(current_ir_graph, get_nodes_Block(ret));
544 set_irn_in(post_bl, n_ret, cf_pred);
546 /** Collect results from Return nodes to post_call. Post_call is
547 turned into a tuple. **/
548 turn_into_tuple(post_call, 4);
549 /* First the Memory-Phi */
551 for (i = 0; i < arity; i++) {
552 ret = get_irn_n(end_bl, i);
553 if (get_irn_op(ret) == op_Return) {
554 cf_pred[n_ret] = get_Return_mem(ret);
558 phi = new_Phi(n_ret, cf_pred, mode_M);
559 set_Tuple_pred(call, 0, phi);
560 set_irn_link(phi, get_irn_link(post_bl)); /* Conserve Phi-list for further inlinings */
561 set_irn_link(post_bl, phi);
562 /* Now the real results */
564 for (j = 0; j < n_res; j++) {
566 for (i = 0; i < arity; i++) {
567 ret = get_irn_n(end_bl, i);
568 if (get_irn_op(ret) == op_Return) {
569 cf_pred[n_ret] = get_Return_res(ret, j);
573 phi = new_Phi(n_ret, cf_pred, get_irn_mode(cf_pred[0]));
575 set_irn_link(phi, get_irn_link(post_bl)); /* Conserve Phi-list for further inlinings */
576 set_irn_link(post_bl, phi);
578 set_Tuple_pred(call, 2, new_Tuple(n_res, res_pred));
580 set_Tuple_pred(call, 2, new_Bad());
582 /* Finally the exception control flow. We need to add a Phi node to
583 collect the memory containing the exception objects. Further we need
584 to add another block to get a correct representation of this Phi. To
585 this block we add a Jmp that resolves into the X output of the Call
586 when the Call is turned into a tuple. */
588 for (i = 0; i < arity; i++) {
590 ret = get_irn_n(end_bl, i);
591 if (is_fragile_op(skip_Proj(ret)) || (get_irn_op(skip_Proj(ret)) == op_Raise)) {
592 cf_pred[n_exc] = ret;
597 new_Block(n_exc, cf_pred); /* whatch it: current_block is changed! */
598 set_Tuple_pred(call, 1, new_Jmp());
599 /* The Phi for the memories with the exception objects */
601 for (i = 0; i < arity; i++) {
603 ret = skip_Proj(get_irn_n(end_bl, i));
604 if (get_irn_op(ret) == op_Call) {
605 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 3);
607 } else if (is_fragile_op(ret)) {
608 /* We rely that all cfops have the memory output at the same position. */
609 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 0);
611 } else if (get_irn_op(ret) == op_Raise) {
612 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 1);
616 set_Tuple_pred(call, 3, new_Phi(n_exc, cf_pred, mode_M));
618 set_Tuple_pred(call, 1, new_Bad());
619 set_Tuple_pred(call, 3, new_Bad());
624 /*** Correct the control flow to the end node.
625 If the exception control flow from the Call directly branched to the
626 end block we now have the following control flow predecessor pattern:
627 ProjX -> Tuple -> Jmp.
628 We must remove the Jmp along with it's empty block and add Jmp's
629 predecessors as predecessors of this end block. ***/
630 /* find the problematic predecessor of the end block. */
631 end_bl = get_irg_end_block(current_ir_graph);
632 for (i = 0; i < get_Block_n_cfgpreds(end_bl); i++) {
633 cf_op = get_Block_cfgpred(end_bl, i);
634 if (get_irn_op(cf_op) == op_Proj) {
635 cf_op = get_Proj_pred(cf_op);
636 if (get_irn_op(cf_op) == op_Tuple) {
637 cf_op = get_Tuple_pred(cf_op, 1);
638 assert(get_irn_op(cf_op) == op_Jmp);
644 if (i < get_Block_n_cfgpreds(end_bl)) {
645 bl = get_nodes_Block(cf_op);
646 arity = get_Block_n_cfgpreds(end_bl) + get_Block_n_cfgpreds(bl) - 1;
647 cf_pred = (ir_node **) malloc (arity * sizeof (ir_node *));
648 for (j = 0; j < i; j++)
649 cf_pred[j] = get_Block_cfgpred(end_bl, j);
650 for (j = j; j < i + get_Block_n_cfgpreds(bl); j++)
651 cf_pred[j] = get_Block_cfgpred(bl, j-i);
652 for (j = j; j < arity; j++)
653 cf_pred[j] = get_Block_cfgpred(end_bl, j-get_Block_n_cfgpreds(bl) +1);
654 set_irn_in(end_bl, arity, cf_pred);
658 /** Turn cse back on. **/
659 set_optimize(rem_opt);
662 /********************************************************************/
663 /* Apply inlineing to small methods. */
664 /********************************************************************/
668 /* It makes no sense to inline too many calls in one procedure. Anyways,
669 I didn't get a version with NEW_ARR_F to run. */
670 #define MAX_INLINE 1024
672 static void collect_calls(ir_node *call, void *env) {
673 ir_node **calls = (ir_node **)env;
676 ir_graph *called_irg;
678 if (get_irn_op(call) != op_Call) return;
680 addr = get_Call_ptr(call);
681 if (get_irn_op(addr) == op_Const) {
682 /* Check whether the constant is the pointer to a compiled entity. */
683 tv = get_Const_tarval(addr);
685 called_irg = get_entity_irg(tv->u.p.ent);
686 if (called_irg && pos < MAX_INLINE) {
687 /* The Call node calls a locally defined method. Remember to inline. */
696 /* Inlines all small methods at call sites where the called address comes
697 from a Const node that references the entity representing the called
699 The size argument is a rough measure for the code size of the method:
700 Methods where the obstack containing the firm graph is smaller than
702 void inline_small_irgs(ir_graph *irg, int size) {
704 ir_node *calls[MAX_INLINE];
705 ir_graph *rem = current_ir_graph;
707 if (!(get_optimize() && get_opt_inline())) return;
709 /*DDME(get_irg_ent(current_ir_graph));*/
711 current_ir_graph = irg;
712 /* Handle graph state */
713 assert(get_irg_phase_state(current_ir_graph) != phase_building);
715 /* Find Call nodes to inline.
716 (We can not inline during a walk of the graph, as inlineing the same
717 method several times changes the visited flag of the walked graph:
718 after the first inlineing visited of the callee equals visited of
719 the caller. With the next inlineing both are increased.) */
721 irg_walk(get_irg_end(irg), NULL, collect_calls, (void *) calls);
723 if ((pos > 0) && (pos < MAX_INLINE)) {
724 /* There are calls to inline */
725 collect_phiprojs(irg);
726 for (i = 0; i < pos; i++) {
729 tv = get_Const_tarval(get_Call_ptr(calls[i]));
730 callee = get_entity_irg(tv->u.p.ent);
731 if ((_obstack_memory_used(callee->obst) - obstack_room(callee->obst)) < size) {
732 /*printf(" inlineing "); DDME(tv->u.p.ent);*/
733 inline_method(calls[i], callee);
738 current_ir_graph = rem;
742 /********************************************************************/
743 /* Code Placement. Pinns all floating nodes to a block where they */
744 /* will be executed only if needed. */
745 /********************************************************************/
747 static pdeq *worklist; /* worklist of ir_node*s */
749 /* Find the earliest correct block for N. --- Place N into the
750 same Block as its dominance-deepest Input. */
752 place_floats_early (ir_node *n)
756 /* we must not run into an infinite loop */
757 assert (irn_not_visited(n));
760 /* Place floating nodes. */
761 if (get_op_pinned(get_irn_op(n)) == floats) {
763 ir_node *b = new_Bad(); /* The block to place this node in */
765 assert(get_irn_op(n) != op_Block);
767 if ((get_irn_op(n) == op_Const) ||
768 (get_irn_op(n) == op_SymConst) ||
770 /* These nodes will not be placed by the loop below. */
771 b = get_irg_start_block(current_ir_graph);
775 /* find the block for this node. */
776 for (i = 0; i < get_irn_arity(n); i++) {
777 ir_node *dep = get_irn_n(n, i);
779 if ((irn_not_visited(dep)) &&
780 (get_op_pinned(get_irn_op(dep)) == floats)) {
781 place_floats_early (dep);
783 /* Because all loops contain at least one pinned node, now all
784 our inputs are either pinned or place_early has already
785 been finished on them. We do not have any unfinished inputs! */
786 dep_block = get_nodes_Block(dep);
787 if ((!is_Bad(dep_block)) &&
788 (get_Block_dom_depth(dep_block) > depth)) {
790 depth = get_Block_dom_depth(dep_block);
792 /* Avoid that the node is placed in the Start block */
793 if ((depth == 1) && (get_Block_dom_depth(get_nodes_Block(n)) > 1)) {
794 b = get_Block_cfg_out(get_irg_start_block(current_ir_graph), 0);
795 assert(b != get_irg_start_block(current_ir_graph));
799 set_nodes_Block(n, b);
802 /* Add predecessors of non floating nodes on worklist. */
803 start = (get_irn_op(n) == op_Block) ? 0 : -1;
804 for (i = start; i < get_irn_arity(n); i++) {
805 ir_node *pred = get_irn_n(n, i);
806 if (irn_not_visited(pred)) {
807 pdeq_putr (worklist, pred);
812 /* Floating nodes form subgraphs that begin at nodes as Const, Load,
813 Start, Call and end at pinned nodes as Store, Call. Place_early
814 places all floating nodes reachable from its argument through floating
815 nodes and adds all beginnings at pinned nodes to the worklist. */
816 inline void place_early () {
821 inc_irg_visited(current_ir_graph);
823 /* this inits the worklist */
824 place_floats_early (get_irg_end(current_ir_graph));
826 /* Work the content of the worklist. */
827 while (!pdeq_empty (worklist)) {
828 ir_node *n = pdeq_getl (worklist);
829 if (irn_not_visited(n)) place_floats_early (n);
832 set_irg_outs_inconsistent(current_ir_graph);
833 current_ir_graph->pinned = pinned;
837 /* deepest common dominance ancestor of DCA and CONSUMER of PRODUCER */
839 consumer_dom_dca (ir_node *dca, ir_node *consumer, ir_node *producer)
843 /* Compute the latest block into which we can place a node so that it is
845 if (get_irn_op(consumer) == op_Phi) {
846 /* our comsumer is a Phi-node, the effective use is in all those
847 blocks through which the Phi-node reaches producer */
849 ir_node *phi_block = get_nodes_Block(consumer);
850 for (i = 0; i < get_irn_arity(consumer); i++) {
851 if (get_irn_n(consumer, i) == producer) {
852 block = get_nodes_Block(get_Block_cfgpred(phi_block, i));
856 assert(is_no_Block(consumer));
857 block = get_nodes_Block(consumer);
860 /* Compute the deepest common ancestor of block and dca. */
862 if (!dca) return block;
863 while (get_Block_dom_depth(block) > get_Block_dom_depth(dca))
864 block = get_Block_idom(block);
865 while (get_Block_dom_depth(dca) > get_Block_dom_depth(block))
866 dca = get_Block_idom(dca);
868 { block = get_Block_idom(block); dca = get_Block_idom(dca); }
874 /* @@@ Needs loop informations. Will implement later interprocedural. */
876 move_out_of_loops (ir_node *n, ir_node *dca)
880 /* Find the region deepest in the dominator tree dominating
881 dca with the least loop nesting depth, but still dominated
882 by our early placement. */
884 while (dca != get_nodes_Block(n)) {
885 dca = get_Block_idom(dca);
886 if (!dca) break; /* should we put assert(dca)? */
887 if (get_Block_loop_depth(dca) < get_Block_loop_depth(best)) {
891 if (get_Block_dom_depth(best) >= get_Block_dom_depth(get_nodes_Block(n)))
892 set_nodes_Block(n, best);
896 /* Find the latest legal block for N and place N into the
897 `optimal' Block between the latest and earliest legal block.
898 The `optimal' block is the dominance-deepest block of those
899 with the least loop-nesting-depth. This places N out of as many
900 loops as possible and then makes it as controldependant as
903 place_floats_late (ir_node *n)
907 assert (irn_not_visited(n)); /* no multiple placement */
909 /* no need to place block nodes, control nodes are already placed. */
910 if ((get_irn_op(n) != op_Block) && (!is_cfop(n)) && (get_irn_mode(n) != mode_X)) {
911 /* Assure that our users are all placed, except the Phi-nodes.
912 --- Each dataflow cycle contains at least one Phi-node. We
913 have to break the `user has to be placed before the
914 producer' dependance cycle and the Phi-nodes are the
915 place to do so, because we need to base our placement on the
916 final region of our users, which is OK with Phi-nodes, as they
917 are pinned, and they never have to be placed after a
918 producer of one of their inputs in the same block anyway. */
919 for (i = 0; i < get_irn_n_outs(n); i++) {
920 ir_node *succ = get_irn_out(n, i);
921 if (irn_not_visited(succ) && (get_irn_op(succ) != op_Phi))
922 place_floats_late (succ);
925 /* We have to determine the final block of this node... except for constants. */
926 if ((get_op_pinned(get_irn_op(n)) == floats) &&
927 (get_irn_op(n) != op_Const) &&
928 (get_irn_op(n) != op_SymConst)) {
929 ir_node *dca = NULL; /* deepest common ancestor in the
930 dominator tree of all nodes'
931 blocks depending on us; our final
932 placement has to dominate DCA. */
933 for (i = 0; i < get_irn_n_outs(n); i++) {
934 dca = consumer_dom_dca (dca, get_irn_out(n, i), n);
936 set_nodes_Block(n, dca);
938 move_out_of_loops (n, dca);
945 /* Add predecessors of all non-floating nodes on list. (Those of floating
946 nodes are placeded already and therefore are marked.) */
947 for (i = 0; i < get_irn_n_outs(n); i++) {
948 if (irn_not_visited(get_irn_out(n, i))) {
949 pdeq_putr (worklist, get_irn_out(n, i));
954 inline void place_late() {
956 inc_irg_visited(current_ir_graph);
958 /* This fills the worklist initially. */
959 place_floats_late(get_irg_start_block(current_ir_graph));
960 /* And now empty the worklist again... */
961 while (!pdeq_empty (worklist)) {
962 ir_node *n = pdeq_getl (worklist);
963 if (irn_not_visited(n)) place_floats_late(n);
967 void place_code(ir_graph *irg) {
968 ir_graph *rem = current_ir_graph;
969 current_ir_graph = irg;
971 if (!(get_optimize() && get_opt_global_cse())) return;
973 /* Handle graph state */
974 assert(get_irg_phase_state(irg) != phase_building);
975 if (get_irg_dom_state(irg) != dom_consistent)
978 /* Place all floating nodes as early as possible. This guarantees
979 a legal code placement. */
980 worklist = new_pdeq ();
983 /* place_early invalidates the outs, place_late needs them. */
985 /* Now move the nodes down in the dominator tree. This reduces the
986 unnecessary executions of the node. */
989 set_irg_outs_inconsistent(current_ir_graph);
991 current_ir_graph = rem;
996 /********************************************************************/
997 /* Control flow optimization. */
998 /* Removes Bad control flow predecessors and empty blocks. A block */
999 /* is empty if it contains only a Jmp node. */
1001 /********************************************************************/
1004 static void merge_blocks(ir_node *n, void *env) {
1006 set_irn_link(n, NULL);
1008 if (get_irn_op(n) == op_Block) {
1010 for (i = 0; i < get_Block_n_cfgpreds(n); i++)
1011 set_Block_cfgpred(n, i, skip_Tuple(get_Block_cfgpred(n, i)));
1012 } else if (get_irn_mode(n) == mode_X) {
1013 /* We will soon visit a block. Optimize it before visiting! */
1014 ir_node *b = get_nodes_Block(n);
1015 ir_node *new = equivalent_node(b);
1016 while (irn_not_visited(b) && (!is_Bad(new)) && (new != b)) {
1017 /* We would have to run gigo if new is bad. */
1018 if (get_optimize() && get_opt_control_flow()) exchange (b, new);
1020 new = equivalent_node(b);
1022 if (is_Bad(new)) exchange (n, new_Bad());
1026 static void collect_nodes(ir_node *n, void *env) {
1028 if (is_no_Block(n)) {
1029 ir_node *b = get_nodes_Block(n);
1031 if ((get_irn_op(n) == op_Phi)) {
1032 /* Collect Phi nodes to compact ins along with block's ins. */
1033 set_irn_link(n, get_irn_link(b));
1035 } else if (get_irn_op(n) != op_Jmp) { /* Check for non empty block. */
1036 mark_Block_block_visited(b);
1041 /* Returns true if pred is pred of block */
1042 int is_pred_of(ir_node *pred, ir_node *b) {
1044 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1045 ir_node *b_pred = get_nodes_Block(get_Block_cfgpred(b, i));
1046 if (b_pred == pred) return 1;
1051 int test_whether_dispensable(ir_node *b, int pos) {
1052 int i, j, n_preds = 1;
1053 int dispensable = 1;
1054 ir_node *cfop = get_Block_cfgpred(b, pos);
1055 ir_node *pred = get_nodes_Block(cfop);
1057 if (get_Block_block_visited(pred) + 1
1058 < get_irg_block_visited(current_ir_graph)) {
1059 if (!get_optimize() || !get_opt_control_flow()) {
1060 /* Mark block so that is will not be removed. */
1061 set_Block_block_visited(pred, get_irg_block_visited(current_ir_graph)-1);
1064 /* Seems to be empty. */
1065 if (!get_irn_link(b)) {
1066 /* There are no Phi nodes ==> dispensable. */
1067 n_preds = get_Block_n_cfgpreds(pred);
1069 /* b's pred blocks and pred's pred blocks must be pairwise disjunct.
1070 Work preds < pos as if they were already removed. */
1071 for (i = 0; i < pos; i++) {
1072 ir_node *b_pred = get_nodes_Block(get_Block_cfgpred(b, i));
1073 if (get_Block_block_visited(b_pred) + 1
1074 < get_irg_block_visited(current_ir_graph)) {
1075 for (j = 0; j < get_Block_n_cfgpreds(b_pred); j++) {
1076 ir_node *b_pred_pred = get_nodes_Block(get_Block_cfgpred(b_pred, j));
1077 if (is_pred_of(b_pred_pred, pred)) dispensable = 0;
1080 if (is_pred_of(b_pred, pred)) dispensable = 0;
1083 for (i = pos +1; i < get_Block_n_cfgpreds(b); i++) {
1084 ir_node *b_pred = get_nodes_Block(get_Block_cfgpred(b, i));
1085 if (is_pred_of(b_pred, pred)) dispensable = 0;
1088 set_Block_block_visited(pred, get_irg_block_visited(current_ir_graph)-1);
1091 n_preds = get_Block_n_cfgpreds(pred);
1099 void optimize_blocks(ir_node *b, void *env) {
1100 int i, j, k, max_preds, n_preds;
1101 ir_node *pred, *phi;
1105 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1106 pred = get_Block_cfgpred(b, i);
1107 max_preds += test_whether_dispensable(b, i);
1109 in = (ir_node **) malloc(max_preds * sizeof(ir_node *));
1113 printf(" working on "); DDMN(b);
1114 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1115 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1116 if (is_Bad(get_Block_cfgpred(b, i))) {
1117 printf(" removing Bad %i\n ", i);
1118 } else if (get_Block_block_visited(pred) +1
1119 < get_irg_block_visited(current_ir_graph)) {
1120 printf(" removing pred %i ", i); DDMN(pred);
1121 } else { printf(" Nothing to do for "); DDMN(pred); }
1123 ** end Debug output **/
1125 /** Fix the Phi nodes **/
1126 phi = get_irn_link(b);
1128 assert(get_irn_op(phi) == op_Phi);
1129 /* Find the new predecessors for the Phi */
1131 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1132 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1133 if (is_Bad(get_Block_cfgpred(b, i))) {
1135 } else if (get_Block_block_visited(pred) +1
1136 < get_irg_block_visited(current_ir_graph)) {
1137 /* It's an empty block and not yet visited. */
1138 ir_node *phi_pred = get_Phi_pred(phi, i);
1139 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1140 if (get_nodes_Block(phi_pred) == pred) {
1141 assert(get_irn_op(phi_pred) == op_Phi); /* Block is empty!! */
1142 in[n_preds] = get_Phi_pred(phi_pred, j);
1144 in[n_preds] = phi_pred;
1149 /* @@@ hier brauche ich schleifeninformation!!! Wenn keine Rueckwaertskante
1150 dann darfs auch keine Verwendung geben. */
1151 if (get_nodes_Block(phi_pred) == pred) {
1152 /* remove the Phi as it might be kept alive. Further there
1153 might be other users. */
1154 exchange(phi_pred, phi); /* geht, is aber doch semantisch falsch! */
1158 in[n_preds] = get_Phi_pred(phi, i);
1163 set_irn_in(phi, n_preds, in);
1164 phi = get_irn_link(phi);
1167 /** Move Phi nodes from removed blocks to this one.
1168 This happens only if merge between loop backedge and single loop entry. **/
1169 for (k = 0; k < get_Block_n_cfgpreds(b); k++) {
1170 pred = get_nodes_Block(get_Block_cfgpred(b, k));
1171 if (get_Block_block_visited(pred) +1
1172 < get_irg_block_visited(current_ir_graph)) {
1173 phi = get_irn_link(pred);
1175 if (get_irn_op(phi) == op_Phi) {
1176 set_nodes_Block(phi, b);
1179 for (i = 0; i < k; i++) {
1180 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1181 if (is_Bad(get_Block_cfgpred(b, i))) {
1183 } else if (get_Block_block_visited(pred) +1
1184 < get_irg_block_visited(current_ir_graph)) {
1185 /* It's an empty block and not yet visited. */
1186 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1187 /* @@@ Hier brauche ich schleifeninformation!!! Kontrllflusskante
1188 muss Rueckwaertskante sein! (An allen vier in[n_preds] = phi
1189 Anweisungen.) Trotzdem tuts bisher!! */
1198 for (i = 0; i < get_Phi_n_preds(phi); i++) {
1199 in[n_preds] = get_Phi_pred(phi, i);
1202 for (i = k+1; i < get_Block_n_cfgpreds(b); i++) {
1203 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1204 if (is_Bad(get_Block_cfgpred(b, i))) {
1206 } else if (get_Block_block_visited(pred) +1
1207 < get_irg_block_visited(current_ir_graph)) {
1208 /* It's an empty block and not yet visited. */
1209 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1218 set_irn_in(phi, n_preds, in);
1220 phi = get_irn_link(phi);
1225 /** Fix the block **/
1227 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1228 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1229 if (is_Bad(get_Block_cfgpred(b, i))) {
1231 } else if (get_Block_block_visited(pred) +1
1232 < get_irg_block_visited(current_ir_graph)) {
1233 /* It's an empty block and not yet visited. */
1234 assert(get_Block_n_cfgpreds(b) > 1);
1235 /* Else it should be optimized by equivalent_node. */
1236 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1237 in[n_preds] = get_Block_cfgpred(pred, j);
1240 /* Remove block as it might be kept alive. */
1241 exchange(pred, b/*new_Bad()*/);
1243 in[n_preds] = get_Block_cfgpred(b, i);
1247 set_irn_in(b, n_preds, in);
1251 void optimize_cf(ir_graph *irg) {
1254 ir_node *end = get_irg_end(irg);
1255 ir_graph *rem = current_ir_graph;
1256 current_ir_graph = irg;
1259 /* Handle graph state */
1260 assert(get_irg_phase_state(irg) != phase_building);
1261 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
1262 set_irg_outs_inconsistent(current_ir_graph);
1263 if (get_irg_dom_state(current_ir_graph) == dom_consistent)
1264 set_irg_dom_inconsistent(current_ir_graph);
1266 //DDME(get_irg_ent(irg));
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);
1295 /* DEL_ARR_F(end->in); GL @@@ tut nicht ! */
1298 current_ir_graph = rem;