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 # include "irbackedge_t.h"
33 /* Defined in iropt.c */
34 pset *new_identities (void);
35 void del_identities (pset *value_table);
36 void add_identities (pset *value_table, ir_node *node);
38 /********************************************************************/
39 /* apply optimizations of iropt to all nodes. */
40 /********************************************************************/
42 void init_link (ir_node *n, void *env) {
43 set_irn_link(n, NULL);
47 optimize_in_place_wrapper (ir_node *n, void *env) {
51 for (i = 0; i < get_irn_arity(n); i++) {
52 optimized = optimize_in_place_2(get_irn_n(n, i));
53 set_irn_n(n, i, optimized);
56 if (get_irn_op(n) == op_Block) {
57 optimized = optimize_in_place_2(n);
58 if (optimized != n) exchange (n, optimized);
63 local_optimize_graph (ir_graph *irg) {
64 ir_graph *rem = current_ir_graph;
65 current_ir_graph = irg;
67 /* Handle graph state */
68 assert(get_irg_phase_state(irg) != phase_building);
69 if (get_opt_global_cse())
70 set_irg_pinned(current_ir_graph, floats);
71 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
72 set_irg_outs_inconsistent(current_ir_graph);
73 if (get_irg_dom_state(current_ir_graph) == dom_consistent)
74 set_irg_dom_inconsistent(current_ir_graph);
76 /* Clean the value_table in irg for the cse. */
77 del_identities(irg->value_table);
78 irg->value_table = new_identities();
80 /* walk over the graph */
81 irg_walk(irg->end, init_link, optimize_in_place_wrapper, NULL);
83 current_ir_graph = rem;
86 /********************************************************************/
87 /* Routines for dead node elimination / copying garbage collection */
89 /********************************************************************/
91 /* Remeber the new node in the old node by using a field all nodes have. */
93 set_new_node (ir_node *old, ir_node *new)
98 /* Get this new node, before the old node is forgotton.*/
100 get_new_node (ir_node * n)
105 /* We use the block_visited flag to mark that we have computed the
106 number of useful predecessors for this block.
107 Further we encode the new arity in this flag in the old blocks.
108 Remembering the arity is useful, as it saves a lot of pointer
109 accesses. This function is called for all Phi and Block nodes
112 compute_new_arity(ir_node *b) {
116 irg_v = get_irg_block_visited(current_ir_graph);
117 block_v = get_Block_block_visited(b);
118 if (block_v >= irg_v) {
119 /* we computed the number of preds for this block and saved it in the
121 return block_v - irg_v;
123 /* compute the number of good predecessors */
124 res = get_irn_arity(b);
125 for (i = 0; i < get_irn_arity(b); i++)
126 if (get_irn_opcode(get_irn_n(b, i)) == iro_Bad) res--;
127 /* save it in the flag. */
128 set_Block_block_visited(b, irg_v + res);
133 static INLINE void new_backedge_info(ir_node *n) {
134 switch(get_irn_opcode(n)) {
136 n->attr.block.cg_backedge = NULL;
137 n->attr.block.backedge = new_backedge_arr(current_ir_graph->obst, get_irn_arity(n));
140 n->attr.phi_backedge = new_backedge_arr(current_ir_graph->obst, get_irn_arity(n));
143 n->attr.filter.backedge = new_backedge_arr(current_ir_graph->obst, get_irn_arity(n));
149 /* Copies the node to the new obstack. The Ins of the new node point to
150 the predecessors on the old obstack. For block/phi nodes not all
151 predecessors might be copied. n->link points to the new node.
152 For Phi and Block nodes the function allocates in-arrays with an arity
153 only for useful predecessors. The arity is determined by counting
154 the non-bad predecessors of the block. */
156 copy_node (ir_node *n, void *env) {
160 if (get_irn_opcode(n) == iro_Block) {
162 new_arity = compute_new_arity(n);
163 n->attr.block.graph_arr = NULL;
165 block = get_nodes_Block(n);
166 if (get_irn_opcode(n) == iro_Phi) {
167 new_arity = compute_new_arity(block);
169 new_arity = get_irn_arity(n);
172 nn = new_ir_node(get_irn_dbg_info(n),
179 /* Copy the attributes. These might point to additional data. If this
180 was allocated on the old obstack the pointers now are dangling. This
181 frees e.g. the memory of the graph_arr allocated in new_immBlock. */
183 new_backedge_info(nn);
186 /* printf("\n old node: "); DDMSG2(n);
187 printf(" new node: "); DDMSG2(nn); */
191 /* Copies new predecessors of old node to new node remembered in link.
192 Spare the Bad predecessors of Phi and Block nodes. */
194 copy_preds (ir_node *n, void *env) {
198 nn = get_new_node(n);
200 /* printf("\n old node: "); DDMSG2(n);
201 printf(" new node: "); DDMSG2(nn);
202 printf(" arities: old: %d, new: %d\n", get_irn_arity(n), get_irn_arity(nn)); */
204 if (get_irn_opcode(n) == iro_Block) {
205 /* Don't copy Bad nodes. */
207 for (i = 0; i < get_irn_arity(n); i++)
208 if (get_irn_opcode(get_irn_n(n, i)) != iro_Bad) {
209 set_irn_n (nn, j, get_new_node(get_irn_n(n, i)));
210 //if (is_backedge(n, i)) set_backedge(nn, j);
213 /* repair the block visited flag from above misuse. Repair it in both
214 graphs so that the old one can still be used. */
215 set_Block_block_visited(nn, 0);
216 set_Block_block_visited(n, 0);
217 /* Local optimization could not merge two subsequent blocks if
218 in array contained Bads. Now it's possible.
219 We don't call optimize_in_place as it requires
220 that the fields in ir_graph are set properly. */
221 if ((get_opt_control_flow()) &&
222 (get_Block_n_cfgpreds(nn) == 1) &&
223 (get_irn_op(get_Block_cfgpred(nn, 0)) == op_Jmp))
224 exchange(nn, get_nodes_Block(get_Block_cfgpred(nn, 0)));
225 } else if (get_irn_opcode(n) == iro_Phi) {
226 /* Don't copy node if corresponding predecessor in block is Bad.
227 The Block itself should not be Bad. */
228 block = get_nodes_Block(n);
229 set_irn_n (nn, -1, get_new_node(block));
231 for (i = 0; i < get_irn_arity(n); i++)
232 if (get_irn_opcode(get_irn_n(block, i)) != iro_Bad) {
233 set_irn_n (nn, j, get_new_node(get_irn_n(n, i)));
234 //if (is_backedge(n, i)) set_backedge(nn, j);
237 /* If the pre walker reached this Phi after the post walker visited the
238 block block_visited is > 0. */
239 set_Block_block_visited(get_nodes_Block(n), 0);
240 /* Compacting the Phi's ins might generate Phis with only one
242 if (get_irn_arity(n) == 1)
243 exchange(n, get_irn_n(n, 0));
245 for (i = -1; i < get_irn_arity(n); i++)
246 set_irn_n (nn, i, get_new_node(get_irn_n(n, i)));
248 /* Now the new node is complete. We can add it to the hash table for cse.
249 @@@ inlinening aborts if we identify End. Why? */
250 if(get_irn_op(nn) != op_End)
251 add_identities (current_ir_graph->value_table, nn);
254 /* Copies the graph recursively, compacts the keepalive of the end node. */
257 ir_node *oe, *ne; /* old end, new end */
258 ir_node *ka; /* keep alive */
261 oe = get_irg_end(current_ir_graph);
262 /* copy the end node by hand, allocate dynamic in array! */
263 ne = new_ir_node(get_irn_dbg_info(oe),
270 /* Copy the attributes. Well, there might be some in the future... */
272 set_new_node(oe, ne);
274 /* copy the live nodes */
275 irg_walk(get_nodes_Block(oe), copy_node, copy_preds, NULL);
276 /* copy_preds for the end node ... */
277 set_nodes_Block(ne, get_new_node(get_nodes_Block(oe)));
279 /** ... and now the keep alives. **/
280 /* First pick the not marked block nodes and walk them. We must pick these
281 first as else we will oversee blocks reachable from Phis. */
282 for (i = 0; i < get_irn_arity(oe); i++) {
283 ka = get_irn_n(oe, i);
284 if ((get_irn_op(ka) == op_Block) &&
285 (get_irn_visited(ka) < get_irg_visited(current_ir_graph))) {
286 /* We must keep the block alive and copy everything reachable */
287 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
288 irg_walk(ka, copy_node, copy_preds, NULL);
289 add_End_keepalive(ne, get_new_node(ka));
293 /* Now pick the Phis. Here we will keep all! */
294 for (i = 0; i < get_irn_arity(oe); i++) {
295 ka = get_irn_n(oe, i);
296 if ((get_irn_op(ka) == op_Phi)) {
297 if (get_irn_visited(ka) < get_irg_visited(current_ir_graph)) {
298 /* We didn't copy the Phi yet. */
299 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
300 irg_walk(ka, copy_node, copy_preds, NULL);
302 add_End_keepalive(ne, get_new_node(ka));
307 /* Copies the graph reachable from current_ir_graph->end to the obstack
308 in current_ir_graph and fixes the environment.
309 Then fixes the fields in current_ir_graph containing nodes of the
314 /* Not all nodes remembered in current_ir_graph might be reachable
315 from the end node. Assure their link is set to NULL, so that
316 we can test whether new nodes have been computed. */
317 set_irn_link(get_irg_frame (current_ir_graph), NULL);
318 set_irn_link(get_irg_globals(current_ir_graph), NULL);
319 set_irn_link(get_irg_args (current_ir_graph), NULL);
321 /* we use the block walk flag for removing Bads from Blocks ins. */
322 inc_irg_block_visited(current_ir_graph);
327 /* fix the fields in current_ir_graph */
328 old_end = get_irg_end(current_ir_graph);
329 set_irg_end (current_ir_graph, get_new_node(old_end));
331 set_irg_end_block (current_ir_graph, get_new_node(get_irg_end_block(current_ir_graph)));
332 if (get_irn_link(get_irg_frame(current_ir_graph)) == NULL) {
333 copy_node (get_irg_frame(current_ir_graph), NULL);
334 copy_preds(get_irg_frame(current_ir_graph), NULL);
336 if (get_irn_link(get_irg_globals(current_ir_graph)) == NULL) {
337 copy_node (get_irg_globals(current_ir_graph), NULL);
338 copy_preds(get_irg_globals(current_ir_graph), NULL);
340 if (get_irn_link(get_irg_args(current_ir_graph)) == NULL) {
341 copy_node (get_irg_args(current_ir_graph), NULL);
342 copy_preds(get_irg_args(current_ir_graph), NULL);
344 set_irg_start (current_ir_graph, get_new_node(get_irg_start(current_ir_graph)));
346 set_irg_start_block(current_ir_graph,
347 get_new_node(get_irg_start_block(current_ir_graph)));
348 set_irg_frame (current_ir_graph, get_new_node(get_irg_frame(current_ir_graph)));
349 set_irg_globals(current_ir_graph, get_new_node(get_irg_globals(current_ir_graph)));
350 set_irg_args (current_ir_graph, get_new_node(get_irg_args(current_ir_graph)));
351 if (get_irn_link(get_irg_bad(current_ir_graph)) == NULL) {
352 copy_node(get_irg_bad(current_ir_graph), NULL);
353 copy_preds(get_irg_bad(current_ir_graph), NULL);
355 set_irg_bad(current_ir_graph, get_new_node(get_irg_bad(current_ir_graph)));
356 if (get_irn_link(get_irg_unknown(current_ir_graph)) == NULL) {
357 copy_node(get_irg_unknown(current_ir_graph), NULL);
358 copy_preds(get_irg_unknown(current_ir_graph), NULL);
360 set_irg_unknown(current_ir_graph, get_new_node(get_irg_unknown(current_ir_graph)));
363 /* Copies all reachable nodes to a new obstack. Removes bad inputs
364 from block nodes and the corresponding inputs from Phi nodes.
365 Merges single exit blocks with single entry blocks and removes
367 Adds all new nodes to a new hash table for cse. Does not
368 perform cse, so the hash table might contain common subexpressions. */
369 /* Amroq call this emigrate() */
371 dead_node_elimination(ir_graph *irg) {
373 struct obstack *graveyard_obst = NULL;
374 struct obstack *rebirth_obst = NULL;
376 /* Remember external state of current_ir_graph. */
377 rem = current_ir_graph;
378 current_ir_graph = irg;
380 /* Handle graph state */
381 assert(get_irg_phase_state(current_ir_graph) != phase_building);
382 free_outs(current_ir_graph);
384 /* @@@ so far we loose loops when copying */
385 set_irg_loop(current_ir_graph, NULL);
387 if (get_optimize() && get_opt_dead_node_elimination()) {
389 /* A quiet place, where the old obstack can rest in peace,
390 until it will be cremated. */
391 graveyard_obst = irg->obst;
393 /* A new obstack, where the reachable nodes will be copied to. */
394 rebirth_obst = (struct obstack *) xmalloc (sizeof (struct obstack));
395 current_ir_graph->obst = rebirth_obst;
396 obstack_init (current_ir_graph->obst);
398 /* We also need a new hash table for cse */
399 del_identities (irg->value_table);
400 irg->value_table = new_identities ();
402 /* Copy the graph from the old to the new obstack */
405 /* Free memory from old unoptimized obstack */
406 obstack_free(graveyard_obst, 0); /* First empty the obstack ... */
407 xfree (graveyard_obst); /* ... then free it. */
410 current_ir_graph = rem;
413 /**********************************************************************/
414 /* Funcionality for inlining */
415 /**********************************************************************/
417 /* Copy node for inlineing. Copies the node by calling copy_node and
418 then updates the entity if it's a local one. env must be a pointer
419 to the frame type of the procedure. The new entities must be in
420 the link field of the entities. */
422 copy_node_inline (ir_node *n, void *env) {
424 type *frame_tp = (type *)env;
427 if (get_irn_op(n) == op_Sel) {
428 new = get_new_node (n);
429 assert(get_irn_op(new) == op_Sel);
430 if (get_entity_owner(get_Sel_entity(n)) == frame_tp) {
431 set_Sel_entity(new, get_entity_link(get_Sel_entity(n)));
436 void inline_method(ir_node *call, ir_graph *called_graph) {
438 ir_node *post_call, *post_bl;
440 ir_node *end, *end_bl;
444 ir_node *cf_op = NULL, *bl;
445 int arity, n_ret, n_exc, n_res, i, j, rem_opt;
448 if (!get_optimize() || !get_opt_inline()) return;
449 /** Turn off optimizations, this can cause problems when allocating new nodes. **/
450 rem_opt = get_optimize();
453 /* Handle graph state */
454 assert(get_irg_phase_state(current_ir_graph) != phase_building);
455 assert(get_irg_pinned(current_ir_graph) == pinned);
456 assert(get_irg_pinned(called_graph) == pinned);
457 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
458 set_irg_outs_inconsistent(current_ir_graph);
460 /** Check preconditions **/
461 assert(get_irn_op(call) == op_Call);
462 /* @@@ TODO does not work for InterfaceIII.java after cgana
463 assert(get_Call_type(call) == get_entity_type(get_irg_ent(called_graph)));
464 assert(smaller_type(get_entity_type(get_irg_ent(called_graph)),
465 get_Call_type(call)));
467 assert(get_type_tpop(get_Call_type(call)) == type_method);
468 if (called_graph == current_ir_graph) return;
471 /** Part the Call node into two nodes. Pre_call collects the parameters of
472 the procedure and later replaces the Start node of the called graph.
473 Post_call is the old Call node and collects the results of the called
474 graph. Both will end up being a tuple. **/
475 post_bl = get_nodes_Block(call);
476 set_irg_current_block(current_ir_graph, post_bl);
477 /* XxMxPxP of Start + parameter of Call */
479 in[1] = get_Call_mem(call);
480 in[2] = get_irg_frame(current_ir_graph);
481 in[3] = get_irg_globals(current_ir_graph);
482 in[4] = new_Tuple (get_Call_n_params(call), get_Call_param_arr(call));
483 pre_call = new_Tuple(5, in);
486 /** Part the block of the Call node into two blocks.
487 The new block gets the ins of the old block, pre_call and all its
488 predecessors and all Phi nodes. **/
489 part_block(pre_call);
491 /** Prepare state for dead node elimination **/
492 /* Visited flags in calling irg must be >= flag in called irg.
493 Else walker and arity computation will not work. */
494 if (get_irg_visited(current_ir_graph) <= get_irg_visited(called_graph))
495 set_irg_visited(current_ir_graph, get_irg_visited(called_graph)+1);
496 if (get_irg_block_visited(current_ir_graph)< get_irg_block_visited(called_graph))
497 set_irg_block_visited(current_ir_graph, get_irg_block_visited(called_graph));
498 /* Set pre_call as new Start node in link field of the start node of
499 calling graph and pre_calls block as new block for the start block
501 Further mark these nodes so that they are not visited by the
503 set_irn_link(get_irg_start(called_graph), pre_call);
504 set_irn_visited(get_irg_start(called_graph),
505 get_irg_visited(current_ir_graph));
506 set_irn_link(get_irg_start_block(called_graph),
507 get_nodes_Block(pre_call));
508 set_irn_visited(get_irg_start_block(called_graph),
509 get_irg_visited(current_ir_graph));
511 /* Initialize for compaction of in arrays */
512 inc_irg_block_visited(current_ir_graph);
514 /*** Replicate local entities of the called_graph ***/
515 /* copy the entities. */
516 called_frame = get_irg_frame_type(called_graph);
517 for (i = 0; i < get_class_n_members(called_frame); i++) {
518 entity *new_ent, *old_ent;
519 old_ent = get_class_member(called_frame, i);
520 new_ent = copy_entity_own(old_ent, get_cur_frame_type());
521 set_entity_link(old_ent, new_ent);
524 /* visited is > than that of called graph. With this trick visited will
525 remain unchanged so that an outer walker, e.g., searching the call nodes
526 to inline, calling this inline will not visit the inlined nodes. */
527 set_irg_visited(current_ir_graph, get_irg_visited(current_ir_graph)-1);
529 /** Performing dead node elimination inlines the graph **/
530 /* Copies the nodes to the obstack of current_ir_graph. Updates links to new
532 /* @@@ endless loops are not copied!! -- they should be, I think... */
533 irg_walk(get_irg_end(called_graph), copy_node_inline, copy_preds,
534 get_irg_frame_type(called_graph));
536 /* Repair called_graph */
537 set_irg_visited(called_graph, get_irg_visited(current_ir_graph));
538 set_irg_block_visited(called_graph, get_irg_block_visited(current_ir_graph));
539 set_Block_block_visited(get_irg_start_block(called_graph), 0);
541 /*** Merge the end of the inlined procedure with the call site ***/
542 /* We will turn the old Call node into a Tuple with the following
545 0: Phi of all Memories of Return statements.
546 1: Jmp from new Block that merges the control flow from all exception
547 predecessors of the old end block.
548 2: Tuple of all arguments.
549 3: Phi of Exception memories.
552 /** Precompute some values **/
553 end_bl = get_new_node(get_irg_end_block(called_graph));
554 end = get_new_node(get_irg_end(called_graph));
555 arity = get_irn_arity(end_bl); /* arity = n_exc + n_ret */
556 n_res = get_method_n_ress(get_Call_type(call));
558 res_pred = (ir_node **) malloc (n_res * sizeof (ir_node *));
559 cf_pred = (ir_node **) malloc (arity * sizeof (ir_node *));
561 set_irg_current_block(current_ir_graph, post_bl); /* just to make sure */
563 /** archive keepalives **/
564 for (i = 0; i < get_irn_arity(end); i++)
565 add_End_keepalive(get_irg_end(current_ir_graph), get_irn_n(end, i));
566 /* The new end node will die, but the in array is not on the obstack ... */
569 /** Collect control flow from Return blocks to post_calls block. Replace
570 Return nodes by Jump nodes. **/
572 for (i = 0; i < arity; i++) {
574 ret = get_irn_n(end_bl, i);
575 if (get_irn_op(ret) == op_Return) {
576 cf_pred[n_ret] = new_r_Jmp(current_ir_graph, get_nodes_Block(ret));
580 set_irn_in(post_bl, n_ret, cf_pred);
582 /** Collect results from Return nodes to post_call. Post_call is
583 turned into a tuple. **/
584 turn_into_tuple(post_call, 4);
585 /* First the Memory-Phi */
587 for (i = 0; i < arity; i++) {
588 ret = get_irn_n(end_bl, i);
589 if (get_irn_op(ret) == op_Return) {
590 cf_pred[n_ret] = get_Return_mem(ret);
594 phi = new_Phi(n_ret, cf_pred, mode_M);
595 set_Tuple_pred(call, 0, phi);
596 /* Conserve Phi-list for further inlinings -- but might be optimized */
597 if (get_nodes_Block(phi) == post_bl) {
598 set_irn_link(phi, get_irn_link(post_bl));
599 set_irn_link(post_bl, phi);
601 /* Now the real results */
603 for (j = 0; j < n_res; j++) {
605 for (i = 0; i < arity; i++) {
606 ret = get_irn_n(end_bl, i);
607 if (get_irn_op(ret) == op_Return) {
608 cf_pred[n_ret] = get_Return_res(ret, j);
612 phi = new_Phi(n_ret, cf_pred, get_irn_mode(cf_pred[0]));
614 /* Conserve Phi-list for further inlinings -- but might be optimized */
615 if (get_nodes_Block(phi) == post_bl) {
616 set_irn_link(phi, get_irn_link(post_bl));
617 set_irn_link(post_bl, phi);
620 set_Tuple_pred(call, 2, new_Tuple(n_res, res_pred));
622 set_Tuple_pred(call, 2, new_Bad());
624 /* Finally the exception control flow. We need to add a Phi node to
625 collect the memory containing the exception objects. Further we need
626 to add another block to get a correct representation of this Phi. To
627 this block we add a Jmp that resolves into the X output of the Call
628 when the Call is turned into a tuple. */
630 for (i = 0; i < arity; i++) {
632 ret = get_irn_n(end_bl, i);
633 if (is_fragile_op(skip_Proj(ret)) || (get_irn_op(skip_Proj(ret)) == op_Raise)) {
634 cf_pred[n_exc] = ret;
639 new_Block(n_exc, cf_pred); /* watch it: current_block is changed! */
640 set_Tuple_pred(call, 1, new_Jmp());
641 /* The Phi for the memories with the exception objects */
643 for (i = 0; i < arity; i++) {
645 ret = skip_Proj(get_irn_n(end_bl, i));
646 if (get_irn_op(ret) == op_Call) {
647 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 3);
649 } else if (is_fragile_op(ret)) {
650 /* We rely that all cfops have the memory output at the same position. */
651 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 0);
653 } else if (get_irn_op(ret) == op_Raise) {
654 cf_pred[n_exc] = new_r_Proj(current_ir_graph, get_nodes_Block(ret), ret, mode_M, 1);
658 set_Tuple_pred(call, 3, new_Phi(n_exc, cf_pred, mode_M));
660 set_Tuple_pred(call, 1, new_Bad());
661 set_Tuple_pred(call, 3, new_Bad());
666 /*** Correct the control flow to the end node.
667 If the exception control flow from the Call directly branched to the
668 end block we now have the following control flow predecessor pattern:
669 ProjX -> Tuple -> Jmp.
670 We must remove the Jmp along with it's empty block and add Jmp's
671 predecessors as predecessors of this end block. ***/
672 /* find the problematic predecessor of the end block. */
673 end_bl = get_irg_end_block(current_ir_graph);
674 for (i = 0; i < get_Block_n_cfgpreds(end_bl); i++) {
675 cf_op = get_Block_cfgpred(end_bl, i);
676 if (get_irn_op(cf_op) == op_Proj) {
677 cf_op = get_Proj_pred(cf_op);
678 if (get_irn_op(cf_op) == op_Tuple) {
679 cf_op = get_Tuple_pred(cf_op, 1);
680 assert(get_irn_op(cf_op) == op_Jmp);
686 if (i < get_Block_n_cfgpreds(end_bl)) {
687 bl = get_nodes_Block(cf_op);
688 arity = get_Block_n_cfgpreds(end_bl) + get_Block_n_cfgpreds(bl) - 1;
689 cf_pred = (ir_node **) malloc (arity * sizeof (ir_node *));
690 for (j = 0; j < i; j++)
691 cf_pred[j] = get_Block_cfgpred(end_bl, j);
692 for (j = j; j < i + get_Block_n_cfgpreds(bl); j++)
693 cf_pred[j] = get_Block_cfgpred(bl, j-i);
694 for (j = j; j < arity; j++)
695 cf_pred[j] = get_Block_cfgpred(end_bl, j-get_Block_n_cfgpreds(bl) +1);
696 set_irn_in(end_bl, arity, cf_pred);
700 /** Turn cse back on. **/
701 set_optimize(rem_opt);
704 /********************************************************************/
705 /* Apply inlineing to small methods. */
706 /********************************************************************/
710 /* It makes no sense to inline too many calls in one procedure. Anyways,
711 I didn't get a version with NEW_ARR_F to run. */
712 #define MAX_INLINE 1024
714 static void collect_calls(ir_node *call, void *env) {
715 ir_node **calls = (ir_node **)env;
718 ir_graph *called_irg;
720 if (get_irn_op(call) != op_Call) return;
722 addr = get_Call_ptr(call);
723 if (get_irn_op(addr) == op_Const) {
724 /* Check whether the constant is the pointer to a compiled entity. */
725 tv = get_Const_tarval(addr);
727 called_irg = get_entity_irg(tv->u.p.ent);
728 if (called_irg && pos < MAX_INLINE) {
729 /* The Call node calls a locally defined method. Remember to inline. */
737 /* Inlines all small methods at call sites where the called address comes
738 from a Const node that references the entity representing the called
740 The size argument is a rough measure for the code size of the method:
741 Methods where the obstack containing the firm graph is smaller than
743 void inline_small_irgs(ir_graph *irg, int size) {
745 ir_node *calls[MAX_INLINE];
746 ir_graph *rem = current_ir_graph;
748 if (!(get_optimize() && get_opt_inline())) return;
750 current_ir_graph = irg;
751 /* Handle graph state */
752 assert(get_irg_phase_state(current_ir_graph) != phase_building);
754 /* Find Call nodes to inline.
755 (We can not inline during a walk of the graph, as inlineing the same
756 method several times changes the visited flag of the walked graph:
757 after the first inlineing visited of the callee equals visited of
758 the caller. With the next inlineing both are increased.) */
760 irg_walk(get_irg_end(irg), NULL, collect_calls, (void *) calls);
762 if ((pos > 0) && (pos < MAX_INLINE)) {
763 /* There are calls to inline */
764 collect_phiprojs(irg);
765 for (i = 0; i < pos; i++) {
769 tv = get_Const_tarval(get_Call_ptr(calls[i]));
770 callee = get_entity_irg(tv->u.p.ent);
771 if ((_obstack_memory_used(callee->obst) - obstack_room(callee->obst)) < size) {
772 inline_method(calls[i], callee);
777 current_ir_graph = rem;
781 /********************************************************************/
782 /* Code Placement. Pinns all floating nodes to a block where they */
783 /* will be executed only if needed. */
784 /********************************************************************/
786 static pdeq *worklist; /* worklist of ir_node*s */
788 /* Find the earliest correct block for N. --- Place N into the
789 same Block as its dominance-deepest Input. */
791 place_floats_early (ir_node *n)
795 /* we must not run into an infinite loop */
796 assert (irn_not_visited(n));
799 /* Place floating nodes. */
800 if (get_op_pinned(get_irn_op(n)) == floats) {
802 ir_node *b = new_Bad(); /* The block to place this node in */
804 assert(get_irn_op(n) != op_Block);
806 if ((get_irn_op(n) == op_Const) ||
807 (get_irn_op(n) == op_SymConst) ||
809 /* These nodes will not be placed by the loop below. */
810 b = get_irg_start_block(current_ir_graph);
814 /* find the block for this node. */
815 for (i = 0; i < get_irn_arity(n); i++) {
816 ir_node *dep = get_irn_n(n, i);
818 if ((irn_not_visited(dep)) &&
819 (get_op_pinned(get_irn_op(dep)) == floats)) {
820 place_floats_early (dep);
822 /* Because all loops contain at least one pinned node, now all
823 our inputs are either pinned or place_early has already
824 been finished on them. We do not have any unfinished inputs! */
825 dep_block = get_nodes_Block(dep);
826 if ((!is_Bad(dep_block)) &&
827 (get_Block_dom_depth(dep_block) > depth)) {
829 depth = get_Block_dom_depth(dep_block);
831 /* Avoid that the node is placed in the Start block */
832 if ((depth == 1) && (get_Block_dom_depth(get_nodes_Block(n)) > 1)) {
833 b = get_Block_cfg_out(get_irg_start_block(current_ir_graph), 0);
834 assert(b != get_irg_start_block(current_ir_graph));
838 set_nodes_Block(n, b);
841 /* Add predecessors of non floating nodes on worklist. */
842 start = (get_irn_op(n) == op_Block) ? 0 : -1;
843 for (i = start; i < get_irn_arity(n); i++) {
844 ir_node *pred = get_irn_n(n, i);
845 if (irn_not_visited(pred)) {
846 pdeq_putr (worklist, pred);
851 /* Floating nodes form subgraphs that begin at nodes as Const, Load,
852 Start, Call and end at pinned nodes as Store, Call. Place_early
853 places all floating nodes reachable from its argument through floating
854 nodes and adds all beginnings at pinned nodes to the worklist. */
855 INLINE void place_early () {
857 inc_irg_visited(current_ir_graph);
859 /* this inits the worklist */
860 place_floats_early (get_irg_end(current_ir_graph));
862 /* Work the content of the worklist. */
863 while (!pdeq_empty (worklist)) {
864 ir_node *n = pdeq_getl (worklist);
865 if (irn_not_visited(n)) place_floats_early (n);
868 set_irg_outs_inconsistent(current_ir_graph);
869 current_ir_graph->pinned = pinned;
873 /* deepest common dominance ancestor of DCA and CONSUMER of PRODUCER */
875 consumer_dom_dca (ir_node *dca, ir_node *consumer, ir_node *producer)
877 ir_node *block = NULL;
879 /* Compute the latest block into which we can place a node so that it is
881 if (get_irn_op(consumer) == op_Phi) {
882 /* our comsumer is a Phi-node, the effective use is in all those
883 blocks through which the Phi-node reaches producer */
885 ir_node *phi_block = get_nodes_Block(consumer);
886 for (i = 0; i < get_irn_arity(consumer); i++) {
887 if (get_irn_n(consumer, i) == producer) {
888 block = get_nodes_Block(get_Block_cfgpred(phi_block, i));
892 assert(is_no_Block(consumer));
893 block = get_nodes_Block(consumer);
896 /* Compute the deepest common ancestor of block and dca. */
898 if (!dca) return block;
899 while (get_Block_dom_depth(block) > get_Block_dom_depth(dca))
900 block = get_Block_idom(block);
901 while (get_Block_dom_depth(dca) > get_Block_dom_depth(block))
902 dca = get_Block_idom(dca);
904 { block = get_Block_idom(block); dca = get_Block_idom(dca); }
909 INLINE int get_irn_loop_depth(ir_node *n) {
910 return get_loop_depth(get_irn_loop(n));
913 /* Move n to a block with less loop depth than it's current block. The
914 new block must be dominated by early. */
916 move_out_of_loops (ir_node *n, ir_node *early)
922 /* Find the region deepest in the dominator tree dominating
923 dca with the least loop nesting depth, but still dominated
924 by our early placement. */
925 dca = get_nodes_Block(n);
927 while (dca != early) {
928 dca = get_Block_idom(dca);
929 if (!dca) break; /* should we put assert(dca)? */
930 if (get_irn_loop_depth(dca) < get_irn_loop_depth(best)) {
934 if (best != get_nodes_Block(n)) {
936 printf("Moving out of loop: "); DDMN(n);
937 printf(" Outermost block: "); DDMN(early);
938 printf(" Best block: "); DDMN(best);
939 printf(" Innermost block: "); DDMN(get_nodes_Block(n));
941 set_nodes_Block(n, best);
945 /* Find the latest legal block for N and place N into the
946 `optimal' Block between the latest and earliest legal block.
947 The `optimal' block is the dominance-deepest block of those
948 with the least loop-nesting-depth. This places N out of as many
949 loops as possible and then makes it as controldependant as
952 place_floats_late (ir_node *n)
957 assert (irn_not_visited(n)); /* no multiple placement */
959 /* no need to place block nodes, control nodes are already placed. */
960 if ((get_irn_op(n) != op_Block) &&
962 (get_irn_mode(n) != mode_X)) {
963 /* Remember the early palacement of this block to move it
964 out of loop no further than the early placement. */
965 early = get_nodes_Block(n);
966 /* Assure that our users are all placed, except the Phi-nodes.
967 --- Each dataflow cycle contains at least one Phi-node. We
968 have to break the `user has to be placed before the
969 producer' dependance cycle and the Phi-nodes are the
970 place to do so, because we need to base our placement on the
971 final region of our users, which is OK with Phi-nodes, as they
972 are pinned, and they never have to be placed after a
973 producer of one of their inputs in the same block anyway. */
974 for (i = 0; i < get_irn_n_outs(n); i++) {
975 ir_node *succ = get_irn_out(n, i);
976 if (irn_not_visited(succ) && (get_irn_op(succ) != op_Phi))
977 place_floats_late (succ);
980 /* We have to determine the final block of this node... except for
982 if ((get_op_pinned(get_irn_op(n)) == floats) &&
983 (get_irn_op(n) != op_Const) &&
984 (get_irn_op(n) != op_SymConst)) {
985 ir_node *dca = NULL; /* deepest common ancestor in the
986 dominator tree of all nodes'
987 blocks depending on us; our final
988 placement has to dominate DCA. */
989 for (i = 0; i < get_irn_n_outs(n); i++) {
990 dca = consumer_dom_dca (dca, get_irn_out(n, i), n);
992 set_nodes_Block(n, dca);
994 move_out_of_loops (n, early);
1000 /* Add predecessors of all non-floating nodes on list. (Those of floating
1001 nodes are placeded already and therefore are marked.) */
1002 for (i = 0; i < get_irn_n_outs(n); i++) {
1003 if (irn_not_visited(get_irn_out(n, i))) {
1004 pdeq_putr (worklist, get_irn_out(n, i));
1009 INLINE void place_late() {
1011 inc_irg_visited(current_ir_graph);
1013 /* This fills the worklist initially. */
1014 place_floats_late(get_irg_start_block(current_ir_graph));
1015 /* And now empty the worklist again... */
1016 while (!pdeq_empty (worklist)) {
1017 ir_node *n = pdeq_getl (worklist);
1018 if (irn_not_visited(n)) place_floats_late(n);
1022 void place_code(ir_graph *irg) {
1023 ir_graph *rem = current_ir_graph;
1024 current_ir_graph = irg;
1026 if (!(get_optimize() && get_opt_global_cse())) return;
1028 /* Handle graph state */
1029 assert(get_irg_phase_state(irg) != phase_building);
1030 if (get_irg_dom_state(irg) != dom_consistent)
1033 construct_backedges(irg);
1035 /* Place all floating nodes as early as possible. This guarantees
1036 a legal code placement. */
1037 worklist = new_pdeq ();
1040 /* place_early invalidates the outs, place_late needs them. */
1042 /* Now move the nodes down in the dominator tree. This reduces the
1043 unnecessary executions of the node. */
1046 set_irg_outs_inconsistent(current_ir_graph);
1047 del_pdeq (worklist);
1048 current_ir_graph = rem;
1053 /********************************************************************/
1054 /* Control flow optimization. */
1055 /* Removes Bad control flow predecessors and empty blocks. A block */
1056 /* is empty if it contains only a Jmp node. */
1058 /********************************************************************/
1061 static void merge_blocks(ir_node *n, void *env) {
1063 set_irn_link(n, NULL);
1065 if (get_irn_op(n) == op_Block) {
1067 for (i = 0; i < get_Block_n_cfgpreds(n); i++)
1068 set_Block_cfgpred(n, i, skip_Tuple(get_Block_cfgpred(n, i)));
1069 } else if (get_irn_mode(n) == mode_X) {
1070 /* We will soon visit a block. Optimize it before visiting! */
1071 ir_node *b = get_nodes_Block(n);
1072 ir_node *new = equivalent_node(b);
1073 while (irn_not_visited(b) && (!is_Bad(new)) && (new != b)) {
1074 /* We would have to run gigo if new is bad. */
1075 if (get_optimize() && get_opt_control_flow()) exchange (b, new);
1077 new = equivalent_node(b);
1079 if (is_Bad(new)) exchange (n, new_Bad());
1083 static void collect_nodes(ir_node *n, void *env) {
1084 if (is_no_Block(n)) {
1085 ir_node *b = get_nodes_Block(n);
1087 if ((get_irn_op(n) == op_Phi)) {
1088 /* Collect Phi nodes to compact ins along with block's ins. */
1089 set_irn_link(n, get_irn_link(b));
1091 } else if (get_irn_op(n) != op_Jmp) { /* Check for non empty block. */
1092 mark_Block_block_visited(b);
1097 /* Returns true if pred is pred of block */
1098 int is_pred_of(ir_node *pred, ir_node *b) {
1100 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1101 ir_node *b_pred = get_nodes_Block(get_Block_cfgpred(b, i));
1102 if (b_pred == pred) return 1;
1107 int test_whether_dispensable(ir_node *b, int pos) {
1108 int i, j, n_preds = 1;
1109 int dispensable = 1;
1110 ir_node *cfop = get_Block_cfgpred(b, pos);
1111 ir_node *pred = get_nodes_Block(cfop);
1113 if (get_Block_block_visited(pred) + 1
1114 < get_irg_block_visited(current_ir_graph)) {
1115 if (!get_optimize() || !get_opt_control_flow()) {
1116 /* Mark block so that is will not be removed. */
1117 set_Block_block_visited(pred, get_irg_block_visited(current_ir_graph)-1);
1120 /* Seems to be empty. */
1121 if (!get_irn_link(b)) {
1122 /* There are no Phi nodes ==> dispensable. */
1123 n_preds = get_Block_n_cfgpreds(pred);
1125 /* b's pred blocks and pred's pred blocks must be pairwise disjunct.
1126 Work preds < pos as if they were already removed. */
1127 for (i = 0; i < pos; i++) {
1128 ir_node *b_pred = get_nodes_Block(get_Block_cfgpred(b, i));
1129 if (get_Block_block_visited(b_pred) + 1
1130 < get_irg_block_visited(current_ir_graph)) {
1131 for (j = 0; j < get_Block_n_cfgpreds(b_pred); j++) {
1132 ir_node *b_pred_pred = get_nodes_Block(get_Block_cfgpred(b_pred, j));
1133 if (is_pred_of(b_pred_pred, pred)) dispensable = 0;
1136 if (is_pred_of(b_pred, pred)) dispensable = 0;
1139 for (i = pos +1; i < get_Block_n_cfgpreds(b); i++) {
1140 ir_node *b_pred = get_nodes_Block(get_Block_cfgpred(b, i));
1141 if (is_pred_of(b_pred, pred)) dispensable = 0;
1144 set_Block_block_visited(pred, get_irg_block_visited(current_ir_graph)-1);
1147 n_preds = get_Block_n_cfgpreds(pred);
1155 void optimize_blocks(ir_node *b, void *env) {
1156 int i, j, k, max_preds, n_preds;
1157 ir_node *pred, *phi;
1160 /* Count the number of predecessor if this block is merged with pred blocks
1163 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1164 max_preds += test_whether_dispensable(b, i);
1166 in = (ir_node **) malloc(max_preds * sizeof(ir_node *));
1169 printf(" working on "); DDMN(b);
1170 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1171 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1172 if (is_Bad(get_Block_cfgpred(b, i))) {
1173 printf(" removing Bad %i\n ", i);
1174 } else if (get_Block_block_visited(pred) +1
1175 < get_irg_block_visited(current_ir_graph)) {
1176 printf(" removing pred %i ", i); DDMN(pred);
1177 } else { printf(" Nothing to do for "); DDMN(pred); }
1179 ** end Debug output **/
1181 /** Fix the Phi nodes **/
1182 phi = get_irn_link(b);
1184 assert(get_irn_op(phi) == op_Phi);
1185 /* Find the new predecessors for the Phi */
1187 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1188 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1189 if (is_Bad(get_Block_cfgpred(b, i))) {
1191 } else if (get_Block_block_visited(pred) +1
1192 < get_irg_block_visited(current_ir_graph)) {
1193 /* It's an empty block and not yet visited. */
1194 ir_node *phi_pred = get_Phi_pred(phi, i);
1195 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1196 if (get_nodes_Block(phi_pred) == pred) {
1197 assert(get_irn_op(phi_pred) == op_Phi); /* Block is empty!! */
1198 in[n_preds] = get_Phi_pred(phi_pred, j);
1200 in[n_preds] = phi_pred;
1205 /* @@@ hier brauche ich schleifeninformation!!! Wenn keine Rueckwaertskante
1206 dann darfs auch keine Verwendung geben. */
1207 if (get_nodes_Block(phi_pred) == pred) {
1208 /* remove the Phi as it might be kept alive. Further there
1209 might be other users. */
1210 exchange(phi_pred, phi); /* geht, is aber doch semantisch falsch! */
1214 in[n_preds] = get_Phi_pred(phi, i);
1219 set_irn_in(phi, n_preds, in);
1220 phi = get_irn_link(phi);
1223 /** Move Phi nodes from removed blocks to this one.
1224 This happens only if merge between loop backedge and single loop entry. **/
1225 for (k = 0; k < get_Block_n_cfgpreds(b); k++) {
1226 pred = get_nodes_Block(get_Block_cfgpred(b, k));
1227 if (get_Block_block_visited(pred) +1
1228 < get_irg_block_visited(current_ir_graph)) {
1229 phi = get_irn_link(pred);
1231 if (get_irn_op(phi) == op_Phi) {
1232 set_nodes_Block(phi, b);
1235 for (i = 0; i < k; i++) {
1236 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1237 if (is_Bad(get_Block_cfgpred(b, i))) {
1239 } else if (get_Block_block_visited(pred) +1
1240 < get_irg_block_visited(current_ir_graph)) {
1241 /* It's an empty block and not yet visited. */
1242 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1243 /* @@@ Hier brauche ich schleifeninformation!!! Kontrllflusskante
1244 muss Rueckwaertskante sein! (An allen vier in[n_preds] = phi
1245 Anweisungen.) Trotzdem tuts bisher!! */
1254 for (i = 0; i < get_Phi_n_preds(phi); i++) {
1255 in[n_preds] = get_Phi_pred(phi, i);
1258 for (i = k+1; i < get_Block_n_cfgpreds(b); i++) {
1259 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1260 if (is_Bad(get_Block_cfgpred(b, i))) {
1262 } else if (get_Block_block_visited(pred) +1
1263 < get_irg_block_visited(current_ir_graph)) {
1264 /* It's an empty block and not yet visited. */
1265 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1274 set_irn_in(phi, n_preds, in);
1276 phi = get_irn_link(phi);
1281 /** Fix the block **/
1283 for (i = 0; i < get_Block_n_cfgpreds(b); i++) {
1284 pred = get_nodes_Block(get_Block_cfgpred(b, i));
1285 if (is_Bad(get_Block_cfgpred(b, i))) {
1287 } else if (get_Block_block_visited(pred) +1
1288 < get_irg_block_visited(current_ir_graph)) {
1289 /* It's an empty block and not yet visited. */
1290 assert(get_Block_n_cfgpreds(b) > 1);
1291 /* Else it should be optimized by equivalent_node. */
1292 for (j = 0; j < get_Block_n_cfgpreds(pred); j++) {
1293 in[n_preds] = get_Block_cfgpred(pred, j);
1296 /* Remove block as it might be kept alive. */
1297 exchange(pred, b/*new_Bad()*/);
1299 in[n_preds] = get_Block_cfgpred(b, i);
1303 set_irn_in(b, n_preds, in);
1307 void optimize_cf(ir_graph *irg) {
1310 ir_node *end = get_irg_end(irg);
1311 ir_graph *rem = current_ir_graph;
1312 current_ir_graph = irg;
1315 /* Handle graph state */
1316 assert(get_irg_phase_state(irg) != phase_building);
1317 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
1318 set_irg_outs_inconsistent(current_ir_graph);
1319 if (get_irg_dom_state(current_ir_graph) == dom_consistent)
1320 set_irg_dom_inconsistent(current_ir_graph);
1322 /* Use block visited flag to mark non-empty blocks. */
1323 inc_irg_block_visited(irg);
1324 irg_walk(end, merge_blocks, collect_nodes, NULL);
1326 /* Optimize the standard code. */
1327 irg_block_walk(get_irg_end_block(irg), optimize_blocks, NULL, NULL);
1329 /* Walk all keep alives, optimize them if block, add to new in-array
1330 for end if useful. */
1331 in = NEW_ARR_F (ir_node *, 1);
1332 in[0] = get_nodes_Block(end);
1334 inc_irg_visited(current_ir_graph);
1335 for(i = 0; i < get_End_n_keepalives(end); i++) {
1336 ir_node *ka = get_End_keepalive(end, i);
1337 if (irn_not_visited(ka)) {
1338 if ((get_irn_op(ka) == op_Block) && Block_not_block_visited(ka)) {
1339 set_irg_block_visited(current_ir_graph, /* Don't walk all the way to Start. */
1340 get_irg_block_visited(current_ir_graph)-1);
1341 irg_block_walk(ka, optimize_blocks, NULL, NULL);
1342 mark_irn_visited(ka);
1343 ARR_APP1 (ir_node *, in, ka);
1344 } else if (get_irn_op(ka) == op_Phi) {
1345 mark_irn_visited(ka);
1346 ARR_APP1 (ir_node *, in, ka);
1350 /* DEL_ARR_F(end->in); GL @@@ tut nicht ! */
1353 current_ir_graph = rem;