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(get_type_tpop(get_Call_type(call)) == type_method);
436 if (called_graph == current_ir_graph) return;
439 /** Part the Call node into two nodes. Pre_call collects the parameters of
440 the procedure and later replaces the Start node of the called graph.
441 Post_call is the old Call node and collects the results of the called
442 graph. Both will end up being a tuple. **/
443 post_bl = get_nodes_Block(call);
444 set_irg_current_block(current_ir_graph, post_bl);
445 /* XxMxPxP of Start + parameter of Call */
447 in[1] = get_Call_mem(call);
448 in[2] = get_irg_frame(current_ir_graph);
449 in[3] = get_irg_globals(current_ir_graph);
450 in[4] = new_Tuple (get_Call_n_params(call), get_Call_param_arr(call));
451 pre_call = new_Tuple(5, in);
454 /** Part the block of the Call node into two blocks.
455 The new block gets the ins of the old block, pre_call and all its
456 predecessors and all Phi nodes. **/
457 part_block(pre_call);
459 /** Prepare state for dead node elimination **/
460 /* Visited flags in calling irg must be >= flag in called irg.
461 Else walker and arity computation will not work. */
462 if (get_irg_visited(current_ir_graph) <= get_irg_visited(called_graph))
463 set_irg_visited(current_ir_graph, get_irg_visited(called_graph)+1); /***/
464 if (get_irg_block_visited(current_ir_graph)< get_irg_block_visited(called_graph))
465 set_irg_block_visited(current_ir_graph, get_irg_block_visited(called_graph));
466 /* Set pre_call as new Start node in link field of the start node of
467 calling graph and pre_calls block as new block for the start block
469 Further mark these nodes so that they are not visited by the
471 set_irn_link(get_irg_start(called_graph), pre_call);
472 set_irn_visited(get_irg_start(called_graph),
473 get_irg_visited(current_ir_graph));/***/
474 set_irn_link(get_irg_start_block(called_graph),
475 get_nodes_Block(pre_call));
476 set_irn_visited(get_irg_start_block(called_graph),
477 get_irg_visited(current_ir_graph)); /***/
479 /* Initialize for compaction of in arrays */
480 inc_irg_block_visited(current_ir_graph);
482 /*** Replicate local entities of the called_graph ***/
483 /* copy the entities. */
484 called_frame = get_irg_frame_type(called_graph);
485 for (i = 0; i < get_class_n_member(called_frame); i++) {
486 entity *new_ent, *old_ent;
487 old_ent = get_class_member(called_frame, i);
488 new_ent = copy_entity_own(old_ent, get_cur_frame_type());
489 set_entity_link(old_ent, new_ent);
492 /* visited is > than that of called graph. With this trick visited will
493 remain unchanged so that an outer walker, e.g., searching the call nodes
494 to inline, calling this inline will not visit the inlined nodes. */
495 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
497 /** Performing dead node elimination inlines the graph **/
498 /* Copies the nodes to the obstack of current_ir_graph. Updates links to new
500 /* @@@ endless loops are not copied!! */
501 irg_walk(get_irg_end(called_graph), copy_node_inline, copy_preds,
502 get_irg_frame_type(called_graph));
504 /* Repair called_graph */
505 set_irg_visited(called_graph, get_irg_visited(current_ir_graph));
506 set_irg_block_visited(called_graph, get_irg_block_visited(current_ir_graph));
507 set_Block_block_visited(get_irg_start_block(called_graph), 0);
509 /*** Merge the end of the inlined procedure with the call site ***/
510 /* We will turn the old Call node into a Tuple with the following
513 0: Phi of all Memories of Return statements.
514 1: Jmp from new Block that merges the control flow from all exception
515 predecessors of the old end block.
516 2: Tuple of all arguments.
517 3: Phi of Exception memories.
520 /** Precompute some values **/
521 end_bl = get_new_node(get_irg_end_block(called_graph));
522 end = get_new_node(get_irg_end(called_graph));
523 arity = get_irn_arity(end_bl); /* arity = n_exc + n_ret */
524 n_res = get_method_n_res(get_Call_type(call));
526 res_pred = (ir_node **) malloc (n_res * sizeof (ir_node *));
527 cf_pred = (ir_node **) malloc (arity * sizeof (ir_node *));
529 set_irg_current_block(current_ir_graph, post_bl); /* just to make sure */
531 /** archive keepalives **/
532 for (i = 0; i < get_irn_arity(end); i++)
533 add_End_keepalive(get_irg_end(current_ir_graph), get_irn_n(end, i));
534 /* The new end node will die, but the in array is not on the obstack ... */
537 /** Collect control flow from Return blocks to post_calls block. Replace
538 Return nodes by Jump nodes. **/
540 for (i = 0; i < arity; i++) {
542 ret = get_irn_n(end_bl, i);
543 if (get_irn_op(ret) == op_Return) {
544 cf_pred[n_ret] = new_r_Jmp(current_ir_graph, get_nodes_Block(ret));
548 set_irn_in(post_bl, n_ret, cf_pred);
550 /** Collect results from Return nodes to post_call. Post_call is
551 turned into a tuple. **/
552 turn_into_tuple(post_call, 4);
553 /* First the Memory-Phi */
555 for (i = 0; i < arity; i++) {
556 ret = get_irn_n(end_bl, i);
557 if (get_irn_op(ret) == op_Return) {
558 cf_pred[n_ret] = get_Return_mem(ret);
562 phi = new_Phi(n_ret, cf_pred, mode_M);
563 set_Tuple_pred(call, 0, phi);
564 set_irn_link(phi, get_irn_link(post_bl)); /* Conserve Phi-list for further inlinings */
565 set_irn_link(post_bl, phi);
566 /* Now the real results */
568 for (j = 0; j < n_res; j++) {
570 for (i = 0; i < arity; i++) {
571 ret = get_irn_n(end_bl, i);
572 if (get_irn_op(ret) == op_Return) {
573 cf_pred[n_ret] = get_Return_res(ret, j);
577 phi = new_Phi(n_ret, cf_pred, get_irn_mode(cf_pred[0]));
579 set_irn_link(phi, get_irn_link(post_bl)); /* Conserve Phi-list for further inlinings */
580 set_irn_link(post_bl, phi);
582 set_Tuple_pred(call, 2, new_Tuple(n_res, res_pred));
584 set_Tuple_pred(call, 2, new_Bad());
586 /* Finally the exception control flow. We need to add a Phi node to
587 collect the memory containing the exception objects. Further we need
588 to add another block to get a correct representation of this Phi. To
589 this block we add a Jmp that resolves into the X output of the Call
590 when the Call is turned into a tuple. */
592 for (i = 0; i < arity; i++) {
594 ret = get_irn_n(end_bl, i);
595 if (is_fragile_op(skip_Proj(ret)) || (get_irn_op(skip_Proj(ret)) == op_Raise)) {
596 cf_pred[n_exc] = ret;
601 new_Block(n_exc, cf_pred); /* whatch it: current_block is changed! */
602 set_Tuple_pred(call, 1, new_Jmp());
603 /* The Phi for the memories with the exception objects */
605 for (i = 0; i < arity; i++) {
607 ret = skip_Proj(get_irn_n(end_bl, i));
608 if (get_irn_op(ret) == op_Call) {
609 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 3);
611 } else if (is_fragile_op(ret)) {
612 /* We rely that all cfops have the memory output at the same position. */
613 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 0);
615 } else if (get_irn_op(ret) == op_Raise) {
616 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 1);
620 set_Tuple_pred(call, 3, new_Phi(n_exc, cf_pred, mode_M));
622 set_Tuple_pred(call, 1, new_Bad());
623 set_Tuple_pred(call, 3, new_Bad());
628 /*** Correct the control flow to the end node.
629 If the exception control flow from the Call directly branched to the
630 end block we now have the following control flow predecessor pattern:
631 ProjX -> Tuple -> Jmp.
632 We must remove the Jmp along with it's empty block and add Jmp's
633 predecessors as predecessors of this end block. ***/
634 /* find the problematic predecessor of the end block. */
635 end_bl = get_irg_end_block(current_ir_graph);
636 for (i = 0; i < get_Block_n_cfgpreds(end_bl); i++) {
637 cf_op = get_Block_cfgpred(end_bl, i);
638 if (get_irn_op(cf_op) == op_Proj) {
639 cf_op = get_Proj_pred(cf_op);
640 if (get_irn_op(cf_op) == op_Tuple) {
641 cf_op = get_Tuple_pred(cf_op, 1);
642 assert(get_irn_op(cf_op) == op_Jmp);
648 if (i < get_Block_n_cfgpreds(end_bl)) {
649 bl = get_nodes_Block(cf_op);
650 arity = get_Block_n_cfgpreds(end_bl) + get_Block_n_cfgpreds(bl) - 1;
651 cf_pred = (ir_node **) malloc (arity * sizeof (ir_node *));
652 for (j = 0; j < i; j++)
653 cf_pred[j] = get_Block_cfgpred(end_bl, j);
654 for (j = j; j < i + get_Block_n_cfgpreds(bl); j++)
655 cf_pred[j] = get_Block_cfgpred(bl, j-i);
656 for (j = j; j < arity; j++)
657 cf_pred[j] = get_Block_cfgpred(end_bl, j-get_Block_n_cfgpreds(bl) +1);
658 set_irn_in(end_bl, arity, cf_pred);
662 /** Turn cse back on. **/
663 set_optimize(rem_opt);
666 /********************************************************************/
667 /* Apply inlineing to small methods. */
668 /********************************************************************/
672 /* It makes no sense to inline too many calls in one procedure. Anyways,
673 I didn't get a version with NEW_ARR_F to run. */
674 #define MAX_INLINE 1024
676 static void collect_calls(ir_node *call, void *env) {
677 ir_node **calls = (ir_node **)env;
680 ir_graph *called_irg;
682 if (get_irn_op(call) != op_Call) return;
684 addr = get_Call_ptr(call);
685 if (get_irn_op(addr) == op_Const) {
686 /* Check whether the constant is the pointer to a compiled entity. */
687 tv = get_Const_tarval(addr);
689 called_irg = get_entity_irg(tv->u.p.ent);
690 if (called_irg && pos < MAX_INLINE) {
691 /* The Call node calls a locally defined method. Remember to inline. */
700 /* Inlines all small methods at call sites where the called address comes
701 from a Const node that references the entity representing the called
703 The size argument is a rough measure for the code size of the method:
704 Methods where the obstack containing the firm graph is smaller than
706 void inline_small_irgs(ir_graph *irg, int size) {
708 ir_node *calls[MAX_INLINE];
709 ir_graph *rem = current_ir_graph;
711 if (!(get_optimize() && get_opt_inline())) return;
713 /*DDME(get_irg_ent(current_ir_graph));*/
715 current_ir_graph = irg;
716 /* Handle graph state */
717 assert(get_irg_phase_state(current_ir_graph) != phase_building);
719 /* Find Call nodes to inline.
720 (We can not inline during a walk of the graph, as inlineing the same
721 method several times changes the visited flag of the walked graph:
722 after the first inlineing visited of the callee equals visited of
723 the caller. With the next inlineing both are increased.) */
725 irg_walk(get_irg_end(irg), NULL, collect_calls, (void *) calls);
727 if ((pos > 0) && (pos < MAX_INLINE)) {
728 /* There are calls to inline */
729 collect_phiprojs(irg);
730 for (i = 0; i < pos; i++) {
733 tv = get_Const_tarval(get_Call_ptr(calls[i]));
734 callee = get_entity_irg(tv->u.p.ent);
735 if ((_obstack_memory_used(callee->obst) - obstack_room(callee->obst)) < size) {
736 /*printf(" inlineing "); DDME(tv->u.p.ent);*/
737 inline_method(calls[i], callee);
742 current_ir_graph = rem;
746 /********************************************************************/
747 /* Code Placement. Pinns all floating nodes to a block where they */
748 /* will be executed only if needed. */
749 /********************************************************************/
751 static pdeq *worklist; /* worklist of ir_node*s */
753 /* Find the earliest correct block for N. --- Place N into the
754 same Block as its dominance-deepest Input. */
756 place_floats_early (ir_node *n)
760 /* we must not run into an infinite loop */
761 assert (irn_not_visited(n));
764 /* Place floating nodes. */
765 if (get_op_pinned(get_irn_op(n)) == floats) {
767 ir_node *b = new_Bad(); /* The block to place this node in */
769 assert(get_irn_op(n) != op_Block);
771 if ((get_irn_op(n) == op_Const) ||
772 (get_irn_op(n) == op_SymConst) ||
774 /* These nodes will not be placed by the loop below. */
775 b = get_irg_start_block(current_ir_graph);
779 /* find the block for this node. */
780 for (i = 0; i < get_irn_arity(n); i++) {
781 ir_node *dep = get_irn_n(n, i);
783 if ((irn_not_visited(dep)) &&
784 (get_op_pinned(get_irn_op(dep)) == floats)) {
785 place_floats_early (dep);
787 /* Because all loops contain at least one pinned node, now all
788 our inputs are either pinned or place_early has already
789 been finished on them. We do not have any unfinished inputs! */
790 dep_block = get_nodes_Block(dep);
791 if ((!is_Bad(dep_block)) &&
792 (get_Block_dom_depth(dep_block) > depth)) {
794 depth = get_Block_dom_depth(dep_block);
796 /* Avoid that the node is placed in the Start block */
797 if ((depth == 1) && (get_Block_dom_depth(get_nodes_Block(n)) > 1)) {
798 b = get_Block_cfg_out(get_irg_start_block(current_ir_graph), 0);
799 assert(b != get_irg_start_block(current_ir_graph));
803 set_nodes_Block(n, b);
806 /* Add predecessors of non floating nodes on worklist. */
807 start = (get_irn_op(n) == op_Block) ? 0 : -1;
808 for (i = start; i < get_irn_arity(n); i++) {
809 ir_node *pred = get_irn_n(n, i);
810 if (irn_not_visited(pred)) {
811 pdeq_putr (worklist, pred);
816 /* Floating nodes form subgraphs that begin at nodes as Const, Load,
817 Start, Call and end at pinned nodes as Store, Call. Place_early
818 places all floating nodes reachable from its argument through floating
819 nodes and adds all beginnings at pinned nodes to the worklist. */
820 INLINE void place_early () {
822 inc_irg_visited(current_ir_graph);
824 /* this inits the worklist */
825 place_floats_early (get_irg_end(current_ir_graph));
827 /* Work the content of the worklist. */
828 while (!pdeq_empty (worklist)) {
829 ir_node *n = pdeq_getl (worklist);
830 if (irn_not_visited(n)) place_floats_early (n);
833 set_irg_outs_inconsistent(current_ir_graph);
834 current_ir_graph->pinned = pinned;
838 /* deepest common dominance ancestor of DCA and CONSUMER of PRODUCER */
840 consumer_dom_dca (ir_node *dca, ir_node *consumer, ir_node *producer)
842 ir_node *block = NULL;
844 /* Compute the latest block into which we can place a node so that it is
846 if (get_irn_op(consumer) == op_Phi) {
847 /* our comsumer is a Phi-node, the effective use is in all those
848 blocks through which the Phi-node reaches producer */
850 ir_node *phi_block = get_nodes_Block(consumer);
851 for (i = 0; i < get_irn_arity(consumer); i++) {
852 if (get_irn_n(consumer, i) == producer) {
853 block = get_nodes_Block(get_Block_cfgpred(phi_block, i));
857 assert(is_no_Block(consumer));
858 block = get_nodes_Block(consumer);
861 /* Compute the deepest common ancestor of block and dca. */
863 if (!dca) return block;
864 while (get_Block_dom_depth(block) > get_Block_dom_depth(dca))
865 block = get_Block_idom(block);
866 while (get_Block_dom_depth(dca) > get_Block_dom_depth(block))
867 dca = get_Block_idom(dca);
869 { block = get_Block_idom(block); dca = get_Block_idom(dca); }
875 /* @@@ Needs loop informations. Will implement later interprocedural. */
877 move_out_of_loops (ir_node *n, ir_node *dca)
881 /* Find the region deepest in the dominator tree dominating
882 dca with the least loop nesting depth, but still dominated
883 by our early placement. */
885 while (dca != get_nodes_Block(n)) {
886 dca = get_Block_idom(dca);
887 if (!dca) break; /* should we put assert(dca)? */
888 if (get_Block_loop_depth(dca) < get_Block_loop_depth(best)) {
892 if (get_Block_dom_depth(best) >= get_Block_dom_depth(get_nodes_Block(n)))
893 set_nodes_Block(n, best);
897 /* Find the latest legal block for N and place N into the
898 `optimal' Block between the latest and earliest legal block.
899 The `optimal' block is the dominance-deepest block of those
900 with the least loop-nesting-depth. This places N out of as many
901 loops as possible and then makes it as controldependant as
904 place_floats_late (ir_node *n)
908 assert (irn_not_visited(n)); /* no multiple placement */
910 /* no need to place block nodes, control nodes are already placed. */
911 if ((get_irn_op(n) != op_Block) && (!is_cfop(n)) && (get_irn_mode(n) != mode_X)) {
912 /* Assure that our users are all placed, except the Phi-nodes.
913 --- Each dataflow cycle contains at least one Phi-node. We
914 have to break the `user has to be placed before the
915 producer' dependance cycle and the Phi-nodes are the
916 place to do so, because we need to base our placement on the
917 final region of our users, which is OK with Phi-nodes, as they
918 are pinned, and they never have to be placed after a
919 producer of one of their inputs in the same block anyway. */
920 for (i = 0; i < get_irn_n_outs(n); i++) {
921 ir_node *succ = get_irn_out(n, i);
922 if (irn_not_visited(succ) && (get_irn_op(succ) != op_Phi))
923 place_floats_late (succ);
926 /* We have to determine the final block of this node... except for constants. */
927 if ((get_op_pinned(get_irn_op(n)) == floats) &&
928 (get_irn_op(n) != op_Const) &&
929 (get_irn_op(n) != op_SymConst)) {
930 ir_node *dca = NULL; /* deepest common ancestor in the
931 dominator tree of all nodes'
932 blocks depending on us; our final
933 placement has to dominate DCA. */
934 for (i = 0; i < get_irn_n_outs(n); i++) {
935 dca = consumer_dom_dca (dca, get_irn_out(n, i), n);
937 set_nodes_Block(n, dca);
939 move_out_of_loops (n, dca);
946 /* Add predecessors of all non-floating nodes on list. (Those of floating
947 nodes are placeded already and therefore are marked.) */
948 for (i = 0; i < get_irn_n_outs(n); i++) {
949 if (irn_not_visited(get_irn_out(n, i))) {
950 pdeq_putr (worklist, get_irn_out(n, i));
955 INLINE void place_late() {
957 inc_irg_visited(current_ir_graph);
959 /* This fills the worklist initially. */
960 place_floats_late(get_irg_start_block(current_ir_graph));
961 /* And now empty the worklist again... */
962 while (!pdeq_empty (worklist)) {
963 ir_node *n = pdeq_getl (worklist);
964 if (irn_not_visited(n)) place_floats_late(n);
968 void place_code(ir_graph *irg) {
969 ir_graph *rem = current_ir_graph;
970 current_ir_graph = irg;
972 if (!(get_optimize() && get_opt_global_cse())) return;
974 /* Handle graph state */
975 assert(get_irg_phase_state(irg) != phase_building);
976 if (get_irg_dom_state(irg) != dom_consistent)
979 /* Place all floating nodes as early as possible. This guarantees
980 a legal code placement. */
981 worklist = new_pdeq ();
984 /* place_early invalidates the outs, place_late needs them. */
986 /* Now move the nodes down in the dominator tree. This reduces the
987 unnecessary executions of the node. */
990 set_irg_outs_inconsistent(current_ir_graph);
992 current_ir_graph = rem;
997 /********************************************************************/
998 /* Control flow optimization. */
999 /* Removes Bad control flow predecessors and empty blocks. A block */
1000 /* is empty if it contains only a Jmp node. */
1002 /********************************************************************/
1005 static void merge_blocks(ir_node *n, void *env) {
1007 set_irn_link(n, NULL);
1009 if (get_irn_op(n) == op_Block) {
1011 for (i = 0; i < get_Block_n_cfgpreds(n); i++)
1012 set_Block_cfgpred(n, i, skip_Tuple(get_Block_cfgpred(n, i)));
1013 } else if (get_irn_mode(n) == mode_X) {
1014 /* We will soon visit a block. Optimize it before visiting! */
1015 ir_node *b = get_nodes_Block(n);
1016 ir_node *new = equivalent_node(b);
1017 while (irn_not_visited(b) && (!is_Bad(new)) && (new != b)) {
1018 /* We would have to run gigo if new is bad. */
1019 if (get_optimize() && get_opt_control_flow()) exchange (b, new);
1021 new = equivalent_node(b);
1023 if (is_Bad(new)) exchange (n, new_Bad());
1027 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 /* Use block visited flag to mark non-empty blocks. */
1267 inc_irg_block_visited(irg);
1268 irg_walk(end, merge_blocks, collect_nodes, NULL);
1270 /* Optimize the standard code. */
1271 irg_block_walk(get_irg_end_block(irg), optimize_blocks, NULL, NULL);
1273 /* Walk all keep alives, optimize them if block, add to new in-array
1274 for end if useful. */
1275 in = NEW_ARR_F (ir_node *, 1);
1276 in[0] = get_nodes_Block(end);
1278 inc_irg_visited(current_ir_graph);
1279 for(i = 0; i < get_End_n_keepalives(end); i++) {
1280 ir_node *ka = get_End_keepalive(end, i);
1281 if (irn_not_visited(ka)) {
1282 if ((get_irn_op(ka) == op_Block) && Block_not_block_visited(ka)) {
1283 set_irg_block_visited(current_ir_graph, /* Don't walk all the way to Start. */
1284 get_irg_block_visited(current_ir_graph)-1);
1285 irg_block_walk(ka, optimize_blocks, NULL, NULL);
1286 mark_irn_visited(ka);
1287 ARR_APP1 (ir_node *, in, ka);
1288 } else if (get_irn_op(ka) == op_Phi) {
1289 mark_irn_visited(ka);
1290 ARR_APP1 (ir_node *, in, ka);
1294 /* DEL_ARR_F(end->in); GL @@@ tut nicht ! */
1297 current_ir_graph = rem;