nsz Git - libfirm/blob - ir/opt/opt_inline.c

   1 /*
   2  * Copyright (C) 1995-2008 University of Karlsruhe.  All right reserved.
   3  *
   4  * This file is part of libFirm.
   5  *
   6  * This file may be distributed and/or modified under the terms of the
   7  * GNU General Public License version 2 as published by the Free Software
   8  * Foundation and appearing in the file LICENSE.GPL included in the
   9  * packaging of this file.
  10  *
  11  * Licensees holding valid libFirm Professional Edition licenses may use
  12  * this file in accordance with the libFirm Commercial License.
  13  * Agreement provided with the Software.
  14  *
  15  * This file is provided AS IS with NO WARRANTY OF ANY KIND, INCLUDING THE
  16  * WARRANTY OF DESIGN, MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  17  * PURPOSE.
  18  */
  19
  20 /**
  21  * @file
  22  * @brief    Dead node elimination and Procedure Inlining.
  23  * @author   Michael Beck, Goetz Lindenmaier
  24  * @version  $Id$
  25  */
  26 #include "config.h"
  27
  28 #include <limits.h>
  29 #include <assert.h>
  30
  31 #include "irnode_t.h"
  32 #include "irgraph_t.h"
  33 #include "irprog_t.h"
  34
  35 #include "iroptimize.h"
  36 #include "ircons_t.h"
  37 #include "iropt_t.h"
  38 #include "irgopt.h"
  39 #include "irgmod.h"
  40 #include "irgwalk.h"
  41
  42 #include "array_t.h"
  43 #include "list.h"
  44 #include "pset.h"
  45 #include "pmap.h"
  46 #include "pdeq.h"
  47 #include "xmalloc.h"
  48 #include "pqueue.h"
  49
  50 #include "irouts.h"
  51 #include "irloop_t.h"
  52 #include "irbackedge_t.h"
  53 #include "opt_inline_t.h"
  54 #include "cgana.h"
  55 #include "trouts.h"
  56 #include "error.h"
  57
  58 #include "analyze_irg_args.h"
  59 #include "iredges_t.h"
  60 #include "irflag_t.h"
  61 #include "irhooks.h"
  62 #include "irtools.h"
  63 #include "iropt_dbg.h"
  64
  65 DEBUG_ONLY(static firm_dbg_module_t *dbg;)
  66
  67 /*------------------------------------------------------------------*/
  68 /* Routines for dead node elimination / copying garbage collection  */
  69 /* of the obstack.                                                  */
  70 /*------------------------------------------------------------------*/
  71
  72 /**
  73  * Remember the new node in the old node by using a field all nodes have.
  74  */
  75 #define set_new_node(oldn, newn)  set_irn_link(oldn, newn)
  76
  77 /**
  78  * Get this new node, before the old node is forgotten.
  79  */
  80 #define get_new_node(oldn) get_irn_link(oldn)
  81
  82 /**
  83  * Check if a new node was set.
  84  */
  85 #define has_new_node(n) (get_new_node(n) != NULL)
  86
  87 /**
  88  * We use the block_visited flag to mark that we have computed the
  89  * number of useful predecessors for this block.
  90  * Further we encode the new arity in this flag in the old blocks.
  91  * Remembering the arity is useful, as it saves a lot of pointer
  92  * accesses.  This function is called for all Phi and Block nodes
  93  * in a Block.
  94  */
  95 static INLINE int
  96 compute_new_arity(ir_node *b) {
  97         int i, res, irn_arity;
  98         int irg_v, block_v;
  99
 100         irg_v = get_irg_block_visited(current_ir_graph);
 101         block_v = get_Block_block_visited(b);
 102         if (block_v >= irg_v) {
 103                 /* we computed the number of preds for this block and saved it in the
 104                    block_v flag */
 105                 return block_v - irg_v;
 106         } else {
 107                 /* compute the number of good predecessors */
 108                 res = irn_arity = get_irn_arity(b);
 109                 for (i = 0; i < irn_arity; i++)
 110                         if (is_Bad(get_irn_n(b, i))) res--;
 111                         /* save it in the flag. */
 112                         set_Block_block_visited(b, irg_v + res);
 113                         return res;
 114         }
 115 }
 116
 117 /**
 118  * Copies the node to the new obstack. The Ins of the new node point to
 119  * the predecessors on the old obstack.  For block/phi nodes not all
 120  * predecessors might be copied.  n->link points to the new node.
 121  * For Phi and Block nodes the function allocates in-arrays with an arity
 122  * only for useful predecessors.  The arity is determined by counting
 123  * the non-bad predecessors of the block.
 124  *
 125  * @param n    The node to be copied
 126  * @param env  if non-NULL, the node number attribute will be copied to the new node
 127  *
 128  * Note: Also used for loop unrolling.
 129  */
 130 static void copy_node(ir_node *n, void *env) {
 131         ir_node *nn, *block;
 132         int new_arity;
 133         ir_op *op = get_irn_op(n);
 134         (void) env;
 135
 136         if (op == op_Bad) {
 137                 /* node copied already */
 138                 return;
 139         } else if (op == op_Block) {
 140                 block = NULL;
 141                 new_arity = compute_new_arity(n);
 142                 n->attr.block.graph_arr = NULL;
 143         } else {
 144                 block = get_nodes_block(n);
 145                 if (op == op_Phi) {
 146                         new_arity = compute_new_arity(block);
 147                 } else {
 148                         new_arity = get_irn_arity(n);
 149                 }
 150         }
 151         nn = new_ir_node(get_irn_dbg_info(n),
 152                 current_ir_graph,
 153                 block,
 154                 op,
 155                 get_irn_mode(n),
 156                 new_arity,
 157                 get_irn_in(n) + 1);
 158         /* Copy the attributes.  These might point to additional data.  If this
 159            was allocated on the old obstack the pointers now are dangling.  This
 160            frees e.g. the memory of the graph_arr allocated in new_immBlock. */
 161         if (op == op_Block) {
 162                 /* we cannot allow blocks WITHOUT macroblock input */
 163                 set_Block_MacroBlock(nn, get_Block_MacroBlock(n));
 164         }
 165         copy_node_attr(n, nn);
 166
 167 #ifdef DEBUG_libfirm
 168         {
 169                 int copy_node_nr = env != NULL;
 170                 if (copy_node_nr) {
 171                         /* for easier debugging, we want to copy the node numbers too */
 172                         nn->node_nr = n->node_nr;
 173                 }
 174         }
 175 #endif
 176
 177         set_new_node(n, nn);
 178         hook_dead_node_elim_subst(current_ir_graph, n, nn);
 179 }
 180
 181 /**
 182  * Copies new predecessors of old node to new node remembered in link.
 183  * Spare the Bad predecessors of Phi and Block nodes.
 184  */
 185 static void copy_preds(ir_node *n, void *env) {
 186         ir_node *nn, *block;
 187         int i, j, irn_arity;
 188         (void) env;
 189
 190         nn = get_new_node(n);
 191
 192         if (is_Block(n)) {
 193                 /* copy the macro block header */
 194                 ir_node *mbh = get_Block_MacroBlock(n);
 195
 196                 if (mbh == n) {
 197                         /* this block is a macroblock header */
 198                         set_Block_MacroBlock(nn, nn);
 199                 } else {
 200                         /* get the macro block header */
 201                         ir_node *nmbh = get_new_node(mbh);
 202                         assert(nmbh != NULL);
 203                         set_Block_MacroBlock(nn, nmbh);
 204                 }
 205
 206                 /* Don't copy Bad nodes. */
 207                 j = 0;
 208                 irn_arity = get_irn_arity(n);
 209                 for (i = 0; i < irn_arity; i++) {
 210                         if (! is_Bad(get_irn_n(n, i))) {
 211                                 ir_node *pred = get_irn_n(n, i);
 212                                 set_irn_n(nn, j, get_new_node(pred));
 213                                 j++;
 214                         }
 215                 }
 216                 /* repair the block visited flag from above misuse. Repair it in both
 217                    graphs so that the old one can still be used. */
 218                 set_Block_block_visited(nn, 0);
 219                 set_Block_block_visited(n, 0);
 220                 /* Local optimization could not merge two subsequent blocks if
 221                    in array contained Bads.  Now it's possible.
 222                    We don't call optimize_in_place as it requires
 223                    that the fields in ir_graph are set properly. */
 224                 if ((get_opt_control_flow_straightening()) &&
 225                         (get_Block_n_cfgpreds(nn) == 1) &&
 226                         is_Jmp(get_Block_cfgpred(nn, 0))) {
 227                         ir_node *old = get_nodes_block(get_Block_cfgpred(nn, 0));
 228                         if (nn == old) {
 229                                 /* Jmp jumps into the block it is in -- deal self cycle. */
 230                                 assert(is_Bad(get_new_node(get_irg_bad(current_ir_graph))));
 231                                 exchange(nn, get_new_node(get_irg_bad(current_ir_graph)));
 232                         } else {
 233                                 exchange(nn, old);
 234                         }
 235                 }
 236         } else if (is_Phi(n) && get_irn_arity(n) > 0) {
 237                 /* Don't copy node if corresponding predecessor in block is Bad.
 238                    The Block itself should not be Bad. */
 239                 block = get_nodes_block(n);
 240                 set_nodes_block(nn, get_new_node(block));
 241                 j = 0;
 242                 irn_arity = get_irn_arity(n);
 243                 for (i = 0; i < irn_arity; i++) {
 244                         if (! is_Bad(get_irn_n(block, i))) {
 245                                 ir_node *pred = get_irn_n(n, i);
 246                                 set_irn_n(nn, j, get_new_node(pred));
 247                                 /*if (is_backedge(n, i)) set_backedge(nn, j);*/
 248                                 j++;
 249                         }
 250                 }
 251                 /* If the pre walker reached this Phi after the post walker visited the
 252                    block block_visited is > 0. */
 253                 set_Block_block_visited(get_nodes_block(n), 0);
 254                 /* Compacting the Phi's ins might generate Phis with only one
 255                    predecessor. */
 256                 if (get_irn_arity(nn) == 1)
 257                         exchange(nn, get_irn_n(nn, 0));
 258         } else {
 259                 irn_arity = get_irn_arity(n);
 260                 for (i = -1; i < irn_arity; i++)
 261                         set_irn_n(nn, i, get_new_node(get_irn_n(n, i)));
 262         }
 263         /* Now the new node is complete.  We can add it to the hash table for CSE.
 264            @@@ inlining aborts if we identify End. Why? */
 265         if (!is_End(nn))
 266                 add_identities(current_ir_graph->value_table, nn);
 267 }
 268
 269 /**
 270  * Copies the graph recursively, compacts the keep-alives of the end node.
 271  *
 272  * @param irg           the graph to be copied
 273  * @param copy_node_nr  If non-zero, the node number will be copied
 274  */
 275 static void copy_graph(ir_graph *irg, int copy_node_nr) {
 276         ir_node *oe, *ne, *ob, *nb, *om, *nm; /* old end, new end, old bad, new bad, old NoMem, new NoMem */
 277         ir_node *ka;      /* keep alive */
 278         int i, irn_arity;
 279         unsigned long vfl;
 280
 281         /* Some nodes must be copied by hand, sigh */
 282         vfl = get_irg_visited(irg);
 283         set_irg_visited(irg, vfl + 1);
 284
 285         oe = get_irg_end(irg);
 286         mark_irn_visited(oe);
 287         /* copy the end node by hand, allocate dynamic in array! */
 288         ne = new_ir_node(get_irn_dbg_info(oe),
 289                 irg,
 290                 NULL,
 291                 op_End,
 292                 mode_X,
 293                 -1,
 294                 NULL);
 295         /* Copy the attributes.  Well, there might be some in the future... */
 296         copy_node_attr(oe, ne);
 297         set_new_node(oe, ne);
 298
 299         /* copy the Bad node */
 300         ob = get_irg_bad(irg);
 301         mark_irn_visited(ob);
 302         nb = new_ir_node(get_irn_dbg_info(ob),
 303                 irg,
 304                 NULL,
 305                 op_Bad,
 306                 mode_T,
 307                 0,
 308                 NULL);
 309         copy_node_attr(ob, nb);
 310         set_new_node(ob, nb);
 311
 312         /* copy the NoMem node */
 313         om = get_irg_no_mem(irg);
 314         mark_irn_visited(om);
 315         nm = new_ir_node(get_irn_dbg_info(om),
 316                 irg,
 317                 NULL,
 318                 op_NoMem,
 319                 mode_M,
 320                 0,
 321                 NULL);
 322         copy_node_attr(om, nm);
 323         set_new_node(om, nm);
 324
 325         /* copy the live nodes */
 326         set_irg_visited(irg, vfl);
 327         irg_walk(get_nodes_block(oe), copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
 328
 329         /* Note: from yet, the visited flag of the graph is equal to vfl + 1 */
 330
 331         /* visit the anchors as well */
 332         for (i = get_irg_n_anchors(irg) - 1; i >= 0; --i) {
 333                 ir_node *n = get_irg_anchor(irg, i);
 334
 335                 if (n && (get_irn_visited(n) <= vfl)) {
 336                         set_irg_visited(irg, vfl);
 337                         irg_walk(n, copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
 338                 }
 339         }
 340
 341         /* copy_preds for the end node ... */
 342         set_nodes_block(ne, get_new_node(get_nodes_block(oe)));
 343
 344         /*- ... and now the keep alives. -*/
 345         /* First pick the not marked block nodes and walk them.  We must pick these
 346            first as else we will oversee blocks reachable from Phis. */
 347         irn_arity = get_End_n_keepalives(oe);
 348         for (i = 0; i < irn_arity; i++) {
 349                 ka = get_End_keepalive(oe, i);
 350                 if (is_Block(ka)) {
 351                         if (get_irn_visited(ka) <= vfl) {
 352                                 /* We must keep the block alive and copy everything reachable */
 353                                 set_irg_visited(irg, vfl);
 354                                 irg_walk(ka, copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
 355                         }
 356                         add_End_keepalive(ne, get_new_node(ka));
 357                 }
 358         }
 359
 360         /* Now pick other nodes.  Here we will keep all! */
 361         irn_arity = get_End_n_keepalives(oe);
 362         for (i = 0; i < irn_arity; i++) {
 363                 ka = get_End_keepalive(oe, i);
 364                 if (!is_Block(ka)) {
 365                         if (get_irn_visited(ka) <= vfl) {
 366                                 /* We didn't copy the node yet.  */
 367                                 set_irg_visited(irg, vfl);
 368                                 irg_walk(ka, copy_node, copy_preds, INT_TO_PTR(copy_node_nr));
 369                         }
 370                         add_End_keepalive(ne, get_new_node(ka));
 371                 }
 372         }
 373
 374         /* start block sometimes only reached after keep alives */
 375         set_nodes_block(nb, get_new_node(get_nodes_block(ob)));
 376         set_nodes_block(nm, get_new_node(get_nodes_block(om)));
 377 }
 378
 379 /**
 380  * Copies the graph reachable from current_ir_graph->end to the obstack
 381  * in current_ir_graph and fixes the environment.
 382  * Then fixes the fields in current_ir_graph containing nodes of the
 383  * graph.
 384  *
 385  * @param copy_node_nr  If non-zero, the node number will be copied
 386  */
 387 static void
 388 copy_graph_env(int copy_node_nr) {
 389         ir_graph *irg = current_ir_graph;
 390         ir_node *old_end, *new_anchor;
 391         int i;
 392
 393         /* remove end_except and end_reg nodes */
 394         old_end = get_irg_end(irg);
 395         set_irg_end_except (irg, old_end);
 396         set_irg_end_reg    (irg, old_end);
 397
 398         /* Not all nodes remembered in irg might be reachable
 399            from the end node.  Assure their link is set to NULL, so that
 400            we can test whether new nodes have been computed. */
 401         for (i = get_irg_n_anchors(irg) - 1; i >= 0; --i) {
 402                 ir_node *n = get_irg_anchor(irg, i);
 403                 if (n != NULL)
 404                         set_new_node(n, NULL);
 405         }
 406         /* we use the block walk flag for removing Bads from Blocks ins. */
 407         inc_irg_block_visited(irg);
 408
 409         /* copy the graph */
 410         copy_graph(irg, copy_node_nr);
 411
 412         /* fix the anchor */
 413         old_end    = get_irg_end(irg);
 414         new_anchor = new_Anchor(irg);
 415
 416         for (i = get_irg_n_anchors(irg) - 1; i >= 0; --i) {
 417                 ir_node *n = get_irg_anchor(irg, i);
 418                 if (n)
 419                         set_irn_n(new_anchor, i, get_new_node(n));
 420         }
 421         free_End(old_end);
 422         irg->anchor = new_anchor;
 423
 424         /* ensure the new anchor is placed in the endblock */
 425         set_nodes_block(new_anchor, get_irg_end_block(irg));
 426 }
 427
 428 /**
 429  * Copies all reachable nodes to a new obstack.  Removes bad inputs
 430  * from block nodes and the corresponding inputs from Phi nodes.
 431  * Merges single exit blocks with single entry blocks and removes
 432  * 1-input Phis.
 433  * Adds all new nodes to a new hash table for CSE.  Does not
 434  * perform CSE, so the hash table might contain common subexpressions.
 435  */
 436 void dead_node_elimination(ir_graph *irg) {
 437         ir_graph *rem;
 438 #ifdef INTERPROCEDURAL_VIEW
 439         int rem_ipview = get_interprocedural_view();
 440 #endif
 441         struct obstack *graveyard_obst = NULL;
 442         struct obstack *rebirth_obst   = NULL;
 443
 444         edges_deactivate(irg);
 445
 446         /* inform statistics that we started a dead-node elimination run */
 447         hook_dead_node_elim(irg, 1);
 448
 449         /* Remember external state of current_ir_graph. */
 450         rem = current_ir_graph;
 451         current_ir_graph = irg;
 452 #ifdef INTERPROCEDURAL_VIEW
 453         set_interprocedural_view(0);
 454 #endif
 455
 456         assert(get_irg_phase_state(irg) != phase_building);
 457
 458         /* Handle graph state */
 459         free_callee_info(irg);
 460         free_irg_outs(irg);
 461         free_trouts();
 462
 463         /* @@@ so far we loose loops when copying */
 464         free_loop_information(irg);
 465
 466         set_irg_doms_inconsistent(irg);
 467
 468         /* A quiet place, where the old obstack can rest in peace,
 469            until it will be cremated. */
 470         graveyard_obst = irg->obst;
 471
 472         /* A new obstack, where the reachable nodes will be copied to. */
 473         rebirth_obst = XMALLOC(struct obstack);
 474         irg->obst = rebirth_obst;
 475         obstack_init(irg->obst);
 476         irg->last_node_idx = 0;
 477
 478         /* We also need a new value table for CSE */
 479         del_identities(irg->value_table);
 480         irg->value_table = new_identities();
 481
 482         /* Copy the graph from the old to the new obstack */
 483         copy_graph_env(/*copy_node_nr=*/1);
 484
 485         /* Free memory from old unoptimized obstack */
 486         obstack_free(graveyard_obst, 0);  /* First empty the obstack ... */
 487         xfree(graveyard_obst);            /* ... then free it.           */
 488
 489         /* inform statistics that the run is over */
 490         hook_dead_node_elim(irg, 0);
 491
 492         current_ir_graph = rem;
 493 #ifdef INTERPROCEDURAL_VIEW
 494         set_interprocedural_view(rem_ipview);
 495 #endif
 496 }
 497
 498 /**
 499  * Relink bad predecessors of a block and store the old in array to the
 500  * link field. This function is called by relink_bad_predecessors().
 501  * The array of link field starts with the block operand at position 0.
 502  * If block has bad predecessors, create a new in array without bad preds.
 503  * Otherwise let in array untouched.
 504  */
 505 static void relink_bad_block_predecessors(ir_node *n, void *env) {
 506         ir_node **new_in, *irn;
 507         int i, new_irn_n, old_irn_arity, new_irn_arity = 0;
 508         (void) env;
 509
 510         /* if link field of block is NULL, look for bad predecessors otherwise
 511            this is already done */
 512         if (is_Block(n) && get_irn_link(n) == NULL) {
 513                 /* save old predecessors in link field (position 0 is the block operand)*/
 514                 set_irn_link(n, get_irn_in(n));
 515
 516                 /* count predecessors without bad nodes */
 517                 old_irn_arity = get_irn_arity(n);
 518                 for (i = 0; i < old_irn_arity; i++)
 519                         if (!is_Bad(get_irn_n(n, i)))
 520                                 ++new_irn_arity;
 521
 522                 /* arity changing: set new predecessors without bad nodes */
 523                 if (new_irn_arity < old_irn_arity) {
 524                         /* Get new predecessor array. We do not resize the array, as we must
 525                            keep the old one to update Phis. */
 526                         new_in = NEW_ARR_D(ir_node *, current_ir_graph->obst, (new_irn_arity+1));
 527
 528                         /* set new predecessors in array */
 529                         new_in[0] = NULL;
 530                         new_irn_n = 1;
 531                         for (i = 0; i < old_irn_arity; i++) {
 532                                 irn = get_irn_n(n, i);
 533                                 if (!is_Bad(irn)) {
 534                                         new_in[new_irn_n] = irn;
 535                                         is_backedge(n, i) ? set_backedge(n, new_irn_n-1) : set_not_backedge(n, new_irn_n-1);
 536                                         ++new_irn_n;
 537                                 }
 538                         }
 539                         /* ARR_SETLEN(int, n->attr.block.backedge, new_irn_arity); */
 540                         ARR_SHRINKLEN(n->attr.block.backedge, new_irn_arity);
 541                         n->in = new_in;
 542                 } /* ir node has bad predecessors */
 543         } /* Block is not relinked */
 544 }
 545
 546 /**
 547  * Relinks Bad predecessors from Blocks and Phis called by walker
 548  * remove_bad_predecesors(). If n is a Block, call
 549  * relink_bad_block_redecessors(). If n is a Phi-node, call also the relinking
 550  * function of Phi's Block. If this block has bad predecessors, relink preds
 551  * of the Phi-node.
 552  */
 553 static void relink_bad_predecessors(ir_node *n, void *env) {
 554         ir_node *block, **old_in;
 555         int i, old_irn_arity, new_irn_arity;
 556
 557         /* relink bad predecessors of a block */
 558         if (is_Block(n))
 559                 relink_bad_block_predecessors(n, env);
 560
 561         /* If Phi node relink its block and its predecessors */
 562         if (is_Phi(n)) {
 563                 /* Relink predecessors of phi's block */
 564                 block = get_nodes_block(n);
 565                 if (get_irn_link(block) == NULL)
 566                         relink_bad_block_predecessors(block, env);
 567
 568                 old_in = (ir_node **)get_irn_link(block); /* Of Phi's Block */
 569                 old_irn_arity = ARR_LEN(old_in);
 570
 571                 /* Relink Phi predecessors if count of predecessors changed */
 572                 if (old_irn_arity != ARR_LEN(get_irn_in(block))) {
 573                         /* set new predecessors in array
 574                            n->in[0] remains the same block */
 575                         new_irn_arity = 1;
 576                         for(i = 1; i < old_irn_arity; i++)
 577                                 if (!is_Bad(old_in[i])) {
 578                                         n->in[new_irn_arity] = n->in[i];
 579                                         is_backedge(n, i) ? set_backedge(n, new_irn_arity) : set_not_backedge(n, new_irn_arity);
 580                                         ++new_irn_arity;
 581                                 }
 582
 583                                 ARR_SETLEN(ir_node *, n->in, new_irn_arity);
 584                                 ARR_SETLEN(int, n->attr.phi.u.backedge, new_irn_arity);
 585                 }
 586         } /* n is a Phi node */
 587 }
 588
 589 /*
 590  * Removes Bad Bad predecessors from Blocks and the corresponding
 591  * inputs to Phi nodes as in dead_node_elimination but without
 592  * copying the graph.
 593  * On walking up set the link field to NULL, on walking down call
 594  * relink_bad_predecessors() (This function stores the old in array
 595  * to the link field and sets a new in array if arity of predecessors
 596  * changes).
 597  */
 598 void remove_bad_predecessors(ir_graph *irg) {
 599         panic("Fix backedge handling first");
 600         irg_walk_graph(irg, firm_clear_link, relink_bad_predecessors, NULL);
 601 }
 602
 603
 604 /*
 605    __                      _  __ __
 606   (_     __    o     _    | \/  |_
 607   __)|_| | \_/ | \_/(/_   |_/\__|__
 608
 609   The following stuff implements a facility that automatically patches
 610   registered ir_node pointers to the new node when a dead node elimination occurs.
 611 */
 612
 613 struct _survive_dce_t {
 614         struct obstack obst;
 615         pmap *places;
 616         pmap *new_places;
 617         hook_entry_t dead_node_elim;
 618         hook_entry_t dead_node_elim_subst;
 619 };
 620
 621 typedef struct _survive_dce_list_t {
 622         struct _survive_dce_list_t *next;
 623         ir_node **place;
 624 } survive_dce_list_t;
 625
 626 static void dead_node_hook(void *context, ir_graph *irg, int start) {
 627         survive_dce_t *sd = context;
 628         (void) irg;
 629
 630         /* Create a new map before the dead node elimination is performed. */
 631         if (start) {
 632                 sd->new_places = pmap_create_ex(pmap_count(sd->places));
 633         } else {
 634                 /* Patch back all nodes if dead node elimination is over and something is to be done. */
 635                 pmap_destroy(sd->places);
 636                 sd->places     = sd->new_places;
 637                 sd->new_places = NULL;
 638         }
 639 }
 640
 641 /**
 642  * Hook called when dead node elimination replaces old by nw.
 643  */
 644 static void dead_node_subst_hook(void *context, ir_graph *irg, ir_node *old, ir_node *nw) {
 645         survive_dce_t *sd = context;
 646         survive_dce_list_t *list = pmap_get(sd->places, old);
 647         (void) irg;
 648
 649         /* If the node is to be patched back, write the new address to all registered locations. */
 650         if (list) {
 651                 survive_dce_list_t *p;
 652
 653                 for (p = list; p; p = p->next)
 654                         *(p->place) = nw;
 655
 656                 pmap_insert(sd->new_places, nw, list);
 657         }
 658 }
 659
 660 /**
 661  * Make a new Survive DCE environment.
 662  */
 663 survive_dce_t *new_survive_dce(void) {
 664         survive_dce_t *res = XMALLOC(survive_dce_t);
 665         obstack_init(&res->obst);
 666         res->places     = pmap_create();
 667         res->new_places = NULL;
 668
 669         res->dead_node_elim.hook._hook_dead_node_elim = dead_node_hook;
 670         res->dead_node_elim.context                   = res;
 671         res->dead_node_elim.next                      = NULL;
 672
 673         res->dead_node_elim_subst.hook._hook_dead_node_elim_subst = dead_node_subst_hook;
 674         res->dead_node_elim_subst.context = res;
 675         res->dead_node_elim_subst.next    = NULL;
 676
 677         register_hook(hook_dead_node_elim, &res->dead_node_elim);
 678         register_hook(hook_dead_node_elim_subst, &res->dead_node_elim_subst);
 679         return res;
 680 }
 681
 682 /**
 683  * Free a Survive DCE environment.
 684  */
 685 void free_survive_dce(survive_dce_t *sd) {
 686         obstack_free(&sd->obst, NULL);
 687         pmap_destroy(sd->places);
 688         unregister_hook(hook_dead_node_elim, &sd->dead_node_elim);
 689         unregister_hook(hook_dead_node_elim_subst, &sd->dead_node_elim_subst);
 690         xfree(sd);
 691 }
 692
 693 /**
 694  * Register a node pointer to be patched upon DCE.
 695  * When DCE occurs, the node pointer specified by @p place will be
 696  * patched to the new address of the node it is pointing to.
 697  *
 698  * @param sd    The Survive DCE environment.
 699  * @param place The address of the node pointer.
 700  */
 701 void survive_dce_register_irn(survive_dce_t *sd, ir_node **place) {
 702         if (*place != NULL) {
 703                 ir_node *irn      = *place;
 704                 survive_dce_list_t *curr = pmap_get(sd->places, irn);
 705                 survive_dce_list_t *nw   = obstack_alloc(&sd->obst, sizeof(nw[0]));
 706
 707                 nw->next  = curr;
 708                 nw->place = place;
 709
 710                 pmap_insert(sd->places, irn, nw);
 711         }
 712 }
 713
 714 /*--------------------------------------------------------------------*/
 715 /*  Functionality for inlining                                         */
 716 /*--------------------------------------------------------------------*/
 717
 718 /**
 719  * Copy node for inlineing.  Updates attributes that change when
 720  * inlineing but not for dead node elimination.
 721  *
 722  * Copies the node by calling copy_node() and then updates the entity if
 723  * it's a local one.  env must be a pointer of the frame type of the
 724  * inlined procedure. The new entities must be in the link field of
 725  * the entities.
 726  */
 727 static void copy_node_inline(ir_node *n, void *env) {
 728         ir_node *nn;
 729         ir_type *frame_tp = (ir_type *)env;
 730
 731         copy_node(n, NULL);
 732         if (is_Sel(n)) {
 733                 nn = get_new_node (n);
 734                 assert(is_Sel(nn));
 735                 if (get_entity_owner(get_Sel_entity(n)) == frame_tp) {
 736                         set_Sel_entity(nn, get_entity_link(get_Sel_entity(n)));
 737                 }
 738         } else if (is_Block(n)) {
 739                 nn = get_new_node (n);
 740                 nn->attr.block.irg = current_ir_graph;
 741         }
 742 }
 743
 744 /**
 745  * Copies new predecessors of old node and move constants to
 746  * the Start Block.
 747  */
 748 static void copy_preds_inline(ir_node *n, void *env) {
 749         ir_node *nn;
 750
 751         copy_preds(n, env);
 752         nn = skip_Id(get_new_node(n));
 753         if (is_irn_constlike(nn)) {
 754                 /* move Constants into the start block */
 755                 set_nodes_block(nn, get_irg_start_block(current_ir_graph));
 756
 757                 n = identify_remember(current_ir_graph->value_table, nn);
 758                 if (nn != n) {
 759                         DBG_OPT_CSE(nn, n);
 760                         exchange(nn, n);
 761                 }
 762         }
 763 }
 764
 765 /**
 766  * Walker: checks if P_value_arg_base is used.
 767  */
 768 static void find_addr(ir_node *node, void *env) {
 769         int *allow_inline = env;
 770         if (is_Proj(node) &&
 771                         is_Start(get_Proj_pred(node)) &&
 772                         get_Proj_proj(node) == pn_Start_P_value_arg_base) {
 773                 *allow_inline = 0;
 774         } else if (is_Alloc(node) && get_Alloc_where(node) == stack_alloc) {
 775                 /* From GCC:
 776                  * Refuse to inline alloca call unless user explicitly forced so as this
 777                  * may change program's memory overhead drastically when the function
 778                  * using alloca is called in loop.  In GCC present in SPEC2000 inlining
 779                  * into schedule_block cause it to require 2GB of ram instead of 256MB.
 780                  *
 781                  * Sorrily this is true with our implementation also.
 782                  * Moreover, we cannot differentiate between alloca() and VLA yet, so this
 783                  * disables inlining of functions using VLA (with are completely save).
 784                  *
 785                  * 2 Solutions:
 786                  * - add a flag to the Alloc node for "real" alloca() calls
 787                  * - add a new Stack-Restore node at the end of a function using alloca()
 788                  */
 789                 *allow_inline = 0;
 790         }
 791 }
 792
 793 /**
 794  * Check if we can inline a given call.
 795  * Currently, we cannot inline two cases:
 796  * - call with compound arguments
 797  * - graphs that take the address of a parameter
 798  *
 799  * check these conditions here
 800  */
 801 static int can_inline(ir_node *call, ir_graph *called_graph) {
 802         ir_type *call_type = get_Call_type(call);
 803         int params, ress, i, res;
 804         assert(is_Method_type(call_type));
 805
 806         params = get_method_n_params(call_type);
 807         ress   = get_method_n_ress(call_type);
 808
 809         /* check parameters for compound arguments */
 810         for (i = 0; i < params; ++i) {
 811                 ir_type *p_type = get_method_param_type(call_type, i);
 812
 813                 if (is_compound_type(p_type))
 814                         return 0;
 815         }
 816
 817         /* check results for compound arguments */
 818         for (i = 0; i < ress; ++i) {
 819                 ir_type *r_type = get_method_res_type(call_type, i);
 820
 821                 if (is_compound_type(r_type))
 822                         return 0;
 823         }
 824
 825         res = 1;
 826         irg_walk_graph(called_graph, find_addr, NULL, &res);
 827
 828         return res;
 829 }
 830
 831 enum exc_mode {
 832         exc_handler    = 0, /**< There is a handler. */
 833         exc_to_end     = 1, /**< Branches to End. */
 834         exc_no_handler = 2  /**< Exception handling not represented. */
 835 };
 836
 837 /* Inlines a method at the given call site. */
 838 int inline_method(ir_node *call, ir_graph *called_graph) {
 839         ir_node             *pre_call;
 840         ir_node             *post_call, *post_bl;
 841         ir_node             *in[pn_Start_max];
 842         ir_node             *end, *end_bl, *block;
 843         ir_node             **res_pred;
 844         ir_node             **cf_pred;
 845         ir_node             **args_in;
 846         ir_node             *ret, *phi;
 847         int                 arity, n_ret, n_exc, n_res, i, n, j, rem_opt, irn_arity, n_params;
 848         enum exc_mode       exc_handling;
 849         ir_type             *called_frame, *curr_frame, *mtp, *ctp;
 850         ir_entity           *ent;
 851         ir_graph            *rem, *irg;
 852         irg_inline_property prop = get_irg_inline_property(called_graph);
 853         unsigned long       visited;
 854
 855         if (prop == irg_inline_forbidden)
 856                 return 0;
 857
 858         ent = get_irg_entity(called_graph);
 859
 860         mtp = get_entity_type(ent);
 861         ctp = get_Call_type(call);
 862         if (get_method_n_params(mtp) > get_method_n_params(ctp)) {
 863                 /* this is a bad feature of C: without a prototype, we can can call a function with less
 864                 parameters than needed. Currently we don't support this, although it would be
 865                 to use Unknown than. */
 866                 return 0;
 867         }
 868
 869         /* Argh, compiling C has some bad consequences:
 870            the call type AND the method type might be different.
 871            It is implementation defendant what happens in that case.
 872            We support inlining, if the bitsize of the types matches AND
 873            the same arithmetic is used. */
 874         n_params = get_method_n_params(mtp);
 875         for (i = n_params - 1; i >= 0; --i) {
 876                 ir_type *param_tp = get_method_param_type(mtp, i);
 877                 ir_type *arg_tp   = get_method_param_type(ctp, i);
 878
 879                 if (param_tp != arg_tp) {
 880                         ir_mode *pmode = get_type_mode(param_tp);
 881                         ir_mode *amode = get_type_mode(arg_tp);
 882
 883                         if (pmode == NULL || amode == NULL)
 884                                 return 0;
 885                         if (get_mode_size_bits(pmode) != get_mode_size_bits(amode))
 886                                 return 0;
 887                         if (get_mode_arithmetic(pmode) != get_mode_arithmetic(amode))
 888                                 return 0;
 889                         /* otherwise we can simply "reinterpret" the bits */
 890                 }
 891         }
 892
 893         irg = get_irn_irg(call);
 894
 895         /*
 896          * We cannot inline a recursive call. The graph must be copied before
 897          * the call the inline_method() using create_irg_copy().
 898          */
 899         if (called_graph == irg)
 900                 return 0;
 901
 902         /*
 903          * currently, we cannot inline two cases:
 904          * - call with compound arguments
 905          * - graphs that take the address of a parameter
 906          */
 907         if (! can_inline(call, called_graph))
 908                 return 0;
 909
 910         rem = current_ir_graph;
 911         current_ir_graph = irg;
 912
 913         DB((dbg, LEVEL_1, "Inlining %+F(%+F) into %+F\n", call, called_graph, irg));
 914
 915         /* --  Turn off optimizations, this can cause problems when allocating new nodes. -- */
 916         rem_opt = get_opt_optimize();
 917         set_optimize(0);
 918
 919         /* Handle graph state */
 920         assert(get_irg_phase_state(irg) != phase_building);
 921         assert(get_irg_pinned(irg) == op_pin_state_pinned);
 922         assert(get_irg_pinned(called_graph) == op_pin_state_pinned);
 923         set_irg_outs_inconsistent(irg);
 924         set_irg_extblk_inconsistent(irg);
 925         set_irg_doms_inconsistent(irg);
 926         set_irg_loopinfo_inconsistent(irg);
 927         set_irg_callee_info_state(irg, irg_callee_info_inconsistent);
 928
 929         /* -- Check preconditions -- */
 930         assert(is_Call(call));
 931
 932         /* here we know we WILL inline, so inform the statistics */
 933         hook_inline(call, called_graph);
 934
 935         /* -- Decide how to handle exception control flow: Is there a handler
 936            for the Call node, or do we branch directly to End on an exception?
 937            exc_handling:
 938            0 There is a handler.
 939            1 Branches to End.
 940            2 Exception handling not represented in Firm. -- */
 941         {
 942                 ir_node *proj, *Mproj = NULL, *Xproj = NULL;
 943                 for (proj = get_irn_link(call); proj; proj = get_irn_link(proj)) {
 944                         long proj_nr = get_Proj_proj(proj);
 945                         if (proj_nr == pn_Call_X_except) Xproj = proj;
 946                         if (proj_nr == pn_Call_M_except) Mproj = proj;
 947                 }
 948                 if      (Mproj) { assert(Xproj); exc_handling = exc_handler; } /*  Mproj           */
 949                 else if (Xproj) {                exc_handling = exc_to_end; } /* !Mproj &&  Xproj   */
 950                 else            {                exc_handling = exc_no_handler; } /* !Mproj && !Xproj   */
 951         }
 952
 953         /* create the argument tuple */
 954         NEW_ARR_A(ir_type *, args_in, n_params);
 955
 956         block = get_nodes_block(call);
 957         for (i = n_params - 1; i >= 0; --i) {
 958                 ir_node *arg      = get_Call_param(call, i);
 959                 ir_type *param_tp = get_method_param_type(mtp, i);
 960                 ir_mode *mode     = get_type_mode(param_tp);
 961
 962                 if (mode != get_irn_mode(arg)) {
 963                         arg = new_r_Conv(irg, block, arg, mode);
 964                 }
 965                 args_in[i] = arg;
 966         }
 967
 968         /* --
 969            the procedure and later replaces the Start node of the called graph.
 970            Post_call is the old Call node and collects the results of the called
 971            graph. Both will end up being a tuple.  -- */
 972         post_bl = get_nodes_block(call);
 973         set_irg_current_block(irg, post_bl);
 974         /* XxMxPxPxPxT of Start + parameter of Call */
 975         in[pn_Start_X_initial_exec]   = new_Jmp();
 976         in[pn_Start_M]                = get_Call_mem(call);
 977         in[pn_Start_P_frame_base]     = get_irg_frame(irg);
 978         in[pn_Start_P_tls]            = get_irg_tls(irg);
 979         in[pn_Start_T_args]           = new_Tuple(n_params, args_in);
 980         /* in[pn_Start_P_value_arg_base] = ??? */
 981         assert(pn_Start_P_value_arg_base == pn_Start_max - 1 && "pn_Start_P_value_arg_base not supported, fix");
 982         pre_call = new_Tuple(pn_Start_max - 1, in);
 983         post_call = call;
 984
 985         /* --
 986            The new block gets the ins of the old block, pre_call and all its
 987            predecessors and all Phi nodes. -- */
 988         part_block(pre_call);
 989
 990         /* -- Prepare state for dead node elimination -- */
 991         /* Visited flags in calling irg must be >= flag in called irg.
 992            Else walker and arity computation will not work. */
 993         if (get_irg_visited(irg) <= get_irg_visited(called_graph))
 994                 set_irg_visited(irg, get_irg_visited(called_graph) + 1);
 995         if (get_irg_block_visited(irg) < get_irg_block_visited(called_graph))
 996                 set_irg_block_visited(irg, get_irg_block_visited(called_graph));
 997         visited = get_irg_visited(irg);
 998
 999         /* Set pre_call as new Start node in link field of the start node of
1000            calling graph and pre_calls block as new block for the start block
1001            of calling graph.
1002            Further mark these nodes so that they are not visited by the
1003            copying. */
1004         set_irn_link(get_irg_start(called_graph), pre_call);
1005         set_irn_visited(get_irg_start(called_graph), visited);
1006         set_irn_link(get_irg_start_block(called_graph), get_nodes_block(pre_call));
1007         set_irn_visited(get_irg_start_block(called_graph), visited);
1008
1009         set_irn_link(get_irg_bad(called_graph), get_irg_bad(current_ir_graph));
1010         set_irn_visited(get_irg_bad(called_graph), visited);
1011
1012         set_irn_link(get_irg_no_mem(called_graph), get_irg_no_mem(current_ir_graph));
1013         set_irn_visited(get_irg_no_mem(called_graph), visited);
1014
1015         /* Initialize for compaction of in arrays */
1016         inc_irg_block_visited(irg);
1017
1018         /* -- Replicate local entities of the called_graph -- */
1019         /* copy the entities. */
1020         called_frame = get_irg_frame_type(called_graph);
1021         curr_frame   = get_irg_frame_type(irg);
1022         for (i = 0, n = get_class_n_members(called_frame); i < n; ++i) {
1023                 ir_entity *new_ent, *old_ent;
1024                 old_ent = get_class_member(called_frame, i);
1025                 new_ent = copy_entity_own(old_ent, curr_frame);
1026                 set_entity_link(old_ent, new_ent);
1027         }
1028
1029         /* visited is > than that of called graph.  With this trick visited will
1030            remain unchanged so that an outer walker, e.g., searching the call nodes
1031             to inline, calling this inline will not visit the inlined nodes. */
1032         set_irg_visited(irg, get_irg_visited(irg)-1);
1033
1034         /* -- Performing dead node elimination inlines the graph -- */
1035         /* Copies the nodes to the obstack of current_ir_graph. Updates links to new
1036            entities. */
1037         irg_walk(get_irg_end(called_graph), copy_node_inline, copy_preds_inline,
1038                  get_irg_frame_type(called_graph));
1039
1040         /* Repair called_graph */
1041         set_irg_visited(called_graph, get_irg_visited(irg));
1042         set_irg_block_visited(called_graph, get_irg_block_visited(irg));
1043         set_Block_block_visited(get_irg_start_block(called_graph), 0);
1044
1045         /* -- Merge the end of the inlined procedure with the call site -- */
1046         /* We will turn the old Call node into a Tuple with the following
1047            predecessors:
1048            -1:  Block of Tuple.
1049            0: Phi of all Memories of Return statements.
1050            1: Jmp from new Block that merges the control flow from all exception
1051            predecessors of the old end block.
1052            2: Tuple of all arguments.
1053            3: Phi of Exception memories.
1054            In case the old Call directly branches to End on an exception we don't
1055            need the block merging all exceptions nor the Phi of the exception
1056            memories.
1057         */
1058
1059         /* -- Precompute some values -- */
1060         end_bl = get_new_node(get_irg_end_block(called_graph));
1061         end = get_new_node(get_irg_end(called_graph));
1062         arity = get_Block_n_cfgpreds(end_bl);    /* arity = n_exc + n_ret  */
1063         n_res = get_method_n_ress(get_Call_type(call));
1064
1065         res_pred = XMALLOCN(ir_node*, n_res);
1066         cf_pred  = XMALLOCN(ir_node*, arity);
1067
1068         set_irg_current_block(irg, post_bl); /* just to make sure */
1069
1070         /* -- archive keepalives -- */
1071         irn_arity = get_irn_arity(end);
1072         for (i = 0; i < irn_arity; i++) {
1073                 ir_node *ka = get_End_keepalive(end, i);
1074                 if (! is_Bad(ka))
1075                         add_End_keepalive(get_irg_end(irg), ka);
1076         }
1077
1078         /* The new end node will die.  We need not free as the in array is on the obstack:
1079            copy_node() only generated 'D' arrays. */
1080
1081         /* -- Replace Return nodes by Jump nodes. -- */
1082         n_ret = 0;
1083         for (i = 0; i < arity; i++) {
1084                 ir_node *ret;
1085                 ret = get_Block_cfgpred(end_bl, i);
1086                 if (is_Return(ret)) {
1087                         cf_pred[n_ret] = new_r_Jmp(irg, get_nodes_block(ret));
1088                         n_ret++;
1089                 }
1090         }
1091         set_irn_in(post_bl, n_ret, cf_pred);
1092
1093         /* -- Build a Tuple for all results of the method.
1094            Add Phi node if there was more than one Return.  -- */
1095         turn_into_tuple(post_call, pn_Call_max);
1096         /* First the Memory-Phi */
1097         n_ret = 0;
1098         for (i = 0; i < arity; i++) {
1099                 ret = get_Block_cfgpred(end_bl, i);
1100                 if (is_Return(ret)) {
1101                         cf_pred[n_ret] = get_Return_mem(ret);
1102                         n_ret++;
1103                 }
1104         }
1105         phi = new_Phi(n_ret, cf_pred, mode_M);
1106         set_Tuple_pred(call, pn_Call_M_regular, phi);
1107         /* Conserve Phi-list for further inlinings -- but might be optimized */
1108         if (get_nodes_block(phi) == post_bl) {
1109                 set_irn_link(phi, get_irn_link(post_bl));
1110                 set_irn_link(post_bl, phi);
1111         }
1112         /* Now the real results */
1113         if (n_res > 0) {
1114                 for (j = 0; j < n_res; j++) {
1115                         n_ret = 0;
1116                         for (i = 0; i < arity; i++) {
1117                                 ret = get_Block_cfgpred(end_bl, i);
1118                                 if (is_Return(ret)) {
1119                                         cf_pred[n_ret] = get_Return_res(ret, j);
1120                                         n_ret++;
1121                                 }
1122                         }
1123                         if (n_ret > 0)
1124                                 phi = new_Phi(n_ret, cf_pred, get_irn_mode(cf_pred[0]));
1125                         else
1126                                 phi = new_Bad();
1127                         res_pred[j] = phi;
1128                         /* Conserve Phi-list for further inlinings -- but might be optimized */
1129                         if (get_nodes_block(phi) == post_bl) {
1130                                 set_Phi_next(phi, get_Block_phis(post_bl));
1131                                 set_Block_phis(post_bl, phi);
1132                         }
1133                 }
1134                 set_Tuple_pred(call, pn_Call_T_result, new_Tuple(n_res, res_pred));
1135         } else {
1136                 set_Tuple_pred(call, pn_Call_T_result, new_Bad());
1137         }
1138         /* handle the regular call */
1139         set_Tuple_pred(call, pn_Call_X_regular, new_Jmp());
1140
1141         /* For now, we cannot inline calls with value_base */
1142         set_Tuple_pred(call, pn_Call_P_value_res_base, new_Bad());
1143
1144         /* Finally the exception control flow.
1145            We have two (three) possible situations:
1146            First if the Call branches to an exception handler: We need to add a Phi node to
1147            collect the memory containing the exception objects.  Further we need
1148            to add another block to get a correct representation of this Phi.  To
1149            this block we add a Jmp that resolves into the X output of the Call
1150            when the Call is turned into a tuple.
1151            Second the Call branches to End, the exception is not handled.  Just
1152            add all inlined exception branches to the End node.
1153            Third: there is no Exception edge at all. Handle as case two. */
1154         if (exc_handling == exc_handler) {
1155                 n_exc = 0;
1156                 for (i = 0; i < arity; i++) {
1157                         ir_node *ret, *irn;
1158                         ret = get_Block_cfgpred(end_bl, i);
1159                         irn = skip_Proj(ret);
1160                         if (is_fragile_op(irn) || is_Raise(irn)) {
1161                                 cf_pred[n_exc] = ret;
1162                                 ++n_exc;
1163                         }
1164                 }
1165                 if (n_exc > 0) {
1166                         new_Block(n_exc, cf_pred);      /* watch it: current_block is changed! */
1167                         set_Tuple_pred(call, pn_Call_X_except, new_Jmp());
1168                         /* The Phi for the memories with the exception objects */
1169                         n_exc = 0;
1170                         for (i = 0; i < arity; i++) {
1171                                 ir_node *ret;
1172                                 ret = skip_Proj(get_Block_cfgpred(end_bl, i));
1173                                 if (is_Call(ret)) {
1174                                         cf_pred[n_exc] = new_r_Proj(irg, get_nodes_block(ret), ret, mode_M, 3);
1175                                         n_exc++;
1176                                 } else if (is_fragile_op(ret)) {
1177                                         /* We rely that all cfops have the memory output at the same position. */
1178                                         cf_pred[n_exc] = new_r_Proj(irg, get_nodes_block(ret), ret, mode_M, 0);
1179                                         n_exc++;
1180                                 } else if (is_Raise(ret)) {
1181                                         cf_pred[n_exc] = new_r_Proj(irg, get_nodes_block(ret), ret, mode_M, 1);
1182                                         n_exc++;
1183                                 }
1184                         }
1185                         set_Tuple_pred(call, pn_Call_M_except, new_Phi(n_exc, cf_pred, mode_M));
1186                 } else {
1187                         set_Tuple_pred(call, pn_Call_X_except, new_Bad());
1188                         set_Tuple_pred(call, pn_Call_M_except, new_Bad());
1189                 }
1190         } else {
1191                 ir_node *main_end_bl;
1192                 int main_end_bl_arity;
1193                 ir_node **end_preds;
1194
1195                 /* assert(exc_handling == 1 || no exceptions. ) */
1196                 n_exc = 0;
1197                 for (i = 0; i < arity; i++) {
1198                         ir_node *ret = get_Block_cfgpred(end_bl, i);
1199                         ir_node *irn = skip_Proj(ret);
1200
1201                         if (is_fragile_op(irn) || is_Raise(irn)) {
1202                                 cf_pred[n_exc] = ret;
1203                                 n_exc++;
1204                         }
1205                 }
1206                 main_end_bl       = get_irg_end_block(irg);
1207                 main_end_bl_arity = get_irn_arity(main_end_bl);
1208                 end_preds         = XMALLOCN(ir_node*, n_exc + main_end_bl_arity);
1209
1210                 for (i = 0; i < main_end_bl_arity; ++i)
1211                         end_preds[i] = get_irn_n(main_end_bl, i);
1212                 for (i = 0; i < n_exc; ++i)
1213                         end_preds[main_end_bl_arity + i] = cf_pred[i];
1214                 set_irn_in(main_end_bl, n_exc + main_end_bl_arity, end_preds);
1215                 set_Tuple_pred(call, pn_Call_X_except,  new_Bad());
1216                 set_Tuple_pred(call, pn_Call_M_except,  new_Bad());
1217                 free(end_preds);
1218         }
1219         free(res_pred);
1220         free(cf_pred);
1221
1222         /* --  Turn CSE back on. -- */
1223         set_optimize(rem_opt);
1224         current_ir_graph = rem;
1225
1226         return 1;
1227 }
1228
1229 /********************************************************************/
1230 /* Apply inlining to small methods.                                 */
1231 /********************************************************************/
1232
1233 static struct obstack  temp_obst;
1234
1235 /** Represents a possible inlinable call in a graph. */
1236 typedef struct _call_entry {
1237         ir_node    *call;       /**< The Call node. */
1238         ir_graph   *callee;     /**< The callee IR-graph. */
1239         list_head  list;        /**< List head for linking the next one. */
1240         int        loop_depth;  /**< The loop depth of this call. */
1241         int        benefice;    /**< The calculated benefice of this call. */
1242         unsigned   local_adr:1; /**< Set if this call gets an address of a local variable. */
1243         unsigned   all_const:1; /**< Set if this call has only constant parameters. */
1244 } call_entry;
1245
1246 /**
1247  * environment for inlining small irgs
1248  */
1249 typedef struct _inline_env_t {
1250         struct obstack obst;  /**< An obstack where call_entries are allocated on. */
1251         list_head      calls; /**< The call entry list. */
1252 } inline_env_t;
1253
1254 /**
1255  * Returns the irg called from a Call node. If the irg is not
1256  * known, NULL is returned.
1257  *
1258  * @param call  the call node
1259  */
1260 static ir_graph *get_call_called_irg(ir_node *call) {
1261         ir_node *addr;
1262
1263         addr = get_Call_ptr(call);
1264         if (is_Global(addr)) {
1265                 ir_entity *ent = get_Global_entity(addr);
1266                 return get_entity_irg(ent);
1267         }
1268
1269         return NULL;
1270 }
1271
1272 /**
1273  * Walker: Collect all calls to known graphs inside a graph.
1274  */
1275 static void collect_calls(ir_node *call, void *env) {
1276         (void) env;
1277         if (is_Call(call)) {
1278                 ir_graph *called_irg = get_call_called_irg(call);
1279
1280                 if (called_irg != NULL) {
1281                         /* The Call node calls a locally defined method.  Remember to inline. */
1282                         inline_env_t *ienv  = env;
1283                         call_entry   *entry = obstack_alloc(&ienv->obst, sizeof(*entry));
1284                         entry->call       = call;
1285                         entry->callee     = called_irg;
1286                         entry->loop_depth = 0;
1287                         entry->benefice   = 0;
1288                         entry->local_adr  = 0;
1289                         entry->all_const  = 0;
1290
1291                         list_add_tail(&entry->list, &ienv->calls);
1292                 }
1293         }
1294 }
1295
1296 /**
1297  * Inlines all small methods at call sites where the called address comes
1298  * from a Const node that references the entity representing the called
1299  * method.
1300  * The size argument is a rough measure for the code size of the method:
1301  * Methods where the obstack containing the firm graph is smaller than
1302  * size are inlined.
1303  */
1304 void inline_small_irgs(ir_graph *irg, int size) {
1305         ir_graph *rem = current_ir_graph;
1306         inline_env_t env;
1307         call_entry *entry;
1308
1309         current_ir_graph = irg;
1310         /* Handle graph state */
1311         assert(get_irg_phase_state(irg) != phase_building);
1312         free_callee_info(irg);
1313
1314         /* Find Call nodes to inline.
1315            (We can not inline during a walk of the graph, as inlining the same
1316            method several times changes the visited flag of the walked graph:
1317            after the first inlining visited of the callee equals visited of
1318            the caller.  With the next inlining both are increased.) */
1319         obstack_init(&env.obst);
1320         INIT_LIST_HEAD(&env.calls);
1321         irg_walk_graph(irg, NULL, collect_calls, &env);
1322
1323         if (! list_empty(&env.calls)) {
1324                 /* There are calls to inline */
1325                 collect_phiprojs(irg);
1326
1327                 list_for_each_entry(call_entry, entry, &env.calls, list) {
1328                         ir_graph            *callee = entry->callee;
1329                         irg_inline_property prop    = get_irg_inline_property(callee);
1330
1331                         if (prop == irg_inline_forbidden || get_irg_additional_properties(callee) & mtp_property_weak) {
1332                                 /* do not inline forbidden / weak graphs */
1333                                 continue;
1334                         }
1335
1336                         if (prop >= irg_inline_forced ||
1337                             _obstack_memory_used(callee->obst) - (int)obstack_room(callee->obst) < size) {
1338                                 inline_method(entry->call, callee);
1339                         }
1340                 }
1341         }
1342         obstack_free(&env.obst, NULL);
1343         current_ir_graph = rem;
1344 }
1345
1346 /**
1347  * Environment for inlining irgs.
1348  */
1349 typedef struct {
1350         list_head calls;             /**< List of of all call nodes in this graph. */
1351         unsigned  *local_weights;    /**< Once allocated, the beneficial weight for transmitting local addresses. */
1352         unsigned  n_nodes;           /**< Number of nodes in graph except Id, Tuple, Proj, Start, End. */
1353         unsigned  n_blocks;          /**< Number of Blocks in graph without Start and End block. */
1354         unsigned  n_nodes_orig;      /**< for statistics */
1355         unsigned  n_call_nodes;      /**< Number of Call nodes in the graph. */
1356         unsigned  n_call_nodes_orig; /**< for statistics */
1357         unsigned  n_callers;         /**< Number of known graphs that call this graphs. */
1358         unsigned  n_callers_orig;    /**< for statistics */
1359         unsigned  got_inline:1;      /**< Set, if at least one call inside this graph was inlined. */
1360         unsigned  local_vars:1;      /**< Set, if an inlined function got the address of a local variable. */
1361         unsigned  recursive:1;       /**< Set, if this function is self recursive. */
1362 } inline_irg_env;
1363
1364 /**
1365  * Allocate a new environment for inlining.
1366  */
1367 static inline_irg_env *alloc_inline_irg_env(void) {
1368         inline_irg_env *env    = obstack_alloc(&temp_obst, sizeof(*env));
1369         INIT_LIST_HEAD(&env->calls);
1370         env->local_weights     = NULL;
1371         env->n_nodes           = -2; /* do not count count Start, End */
1372         env->n_blocks          = -2; /* do not count count Start, End Block */
1373         env->n_nodes_orig      = -2; /* do not count Start, End */
1374         env->n_call_nodes      = 0;
1375         env->n_call_nodes_orig = 0;
1376         env->n_callers         = 0;
1377         env->n_callers_orig    = 0;
1378         env->got_inline        = 0;
1379         env->local_vars        = 0;
1380         env->recursive         = 0;
1381         return env;
1382 }
1383
1384 typedef struct walker_env {
1385         inline_irg_env *x;     /**< the inline environment */
1386         char ignore_runtime;   /**< the ignore runtime flag */
1387         char ignore_callers;   /**< if set, do change callers data */
1388 } wenv_t;
1389
1390 /**
1391  * post-walker: collect all calls in the inline-environment
1392  * of a graph and sum some statistics.
1393  */
1394 static void collect_calls2(ir_node *call, void *ctx) {
1395         wenv_t         *env = ctx;
1396         inline_irg_env *x = env->x;
1397         ir_opcode      code = get_irn_opcode(call);
1398         ir_graph       *callee;
1399         call_entry     *entry;
1400
1401         /* count meaningful nodes in irg */
1402         if (code != iro_Proj && code != iro_Tuple && code != iro_Sync) {
1403                 if (code != iro_Block) {
1404                         ++x->n_nodes;
1405                         ++x->n_nodes_orig;
1406                 } else {
1407                         ++x->n_blocks;
1408                 }
1409         }
1410
1411         if (code != iro_Call) return;
1412
1413         /* check, if it's a runtime call */
1414         if (env->ignore_runtime) {
1415                 ir_node *symc = get_Call_ptr(call);
1416
1417                 if (is_Global(symc)) {
1418                         ir_entity *ent = get_Global_entity(symc);
1419
1420                         if (get_entity_additional_properties(ent) & mtp_property_runtime)
1421                                 return;
1422                 }
1423         }
1424
1425         /* collect all call nodes */
1426         ++x->n_call_nodes;
1427         ++x->n_call_nodes_orig;
1428
1429         callee = get_call_called_irg(call);
1430         if (callee != NULL) {
1431                 if (! env->ignore_callers) {
1432                         inline_irg_env *callee_env = get_irg_link(callee);
1433                         /* count all static callers */
1434                         ++callee_env->n_callers;
1435                         ++callee_env->n_callers_orig;
1436                 }
1437                 if (callee == current_ir_graph)
1438                         x->recursive = 1;
1439
1440                 /* link it in the list of possible inlinable entries */
1441                 entry = obstack_alloc(&temp_obst, sizeof(*entry));
1442                 entry->call       = call;
1443                 entry->callee     = callee;
1444                 entry->loop_depth = get_irn_loop(get_nodes_block(call))->depth;
1445                 entry->benefice   = 0;
1446                 entry->local_adr  = 0;
1447                 entry->all_const  = 0;
1448
1449                 list_add_tail(&entry->list, &x->calls);
1450         }
1451 }
1452
1453 /**
1454  * Returns TRUE if the number of callers is 0 in the irg's environment,
1455  * hence this irg is a leave.
1456  */
1457 INLINE static int is_leave(ir_graph *irg) {
1458         inline_irg_env *env = get_irg_link(irg);
1459         return env->n_call_nodes == 0;
1460 }
1461
1462 /**
1463  * Returns TRUE if the number of nodes in the callee is
1464  * smaller then size in the irg's environment.
1465  */
1466 INLINE static int is_smaller(ir_graph *callee, unsigned size) {
1467         inline_irg_env *env = get_irg_link(callee);
1468         return env->n_nodes < size;
1469 }
1470
1471 /**
1472  * Duplicate a call entry.
1473  *
1474  * @param entry     the original entry to duplicate
1475  * @param new_call  the new call node
1476  * @param loop_depth_delta
1477  *                  delta value for the loop depth
1478  */
1479 static call_entry *duplicate_call_entry(const call_entry *entry,
1480                                         ir_node *new_call, int loop_depth_delta) {
1481         call_entry *nentry = obstack_alloc(&temp_obst, sizeof(*nentry));
1482         nentry->call       = new_call;
1483         nentry->callee     = entry->callee;
1484         nentry->benefice   = entry->benefice;
1485         nentry->loop_depth = entry->loop_depth + loop_depth_delta;
1486         nentry->local_adr  = entry->local_adr;
1487         nentry->all_const  = entry->all_const;
1488
1489         return nentry;
1490 }
1491
1492 /**
1493  * Append all call nodes of the source environment to the nodes of in the destination
1494  * environment.
1495  *
1496  * @param dst         destination environment
1497  * @param src         source environment
1498  * @param loop_depth  the loop depth of the call that is replaced by the src list
1499  */
1500 static void append_call_list(inline_irg_env *dst, inline_irg_env *src, int loop_depth) {
1501         call_entry *entry, *nentry;
1502
1503         /* Note that the src list points to Call nodes in the inlined graph, but
1504            we need Call nodes in our graph. Luckily the inliner leaves this information
1505            in the link field. */
1506         list_for_each_entry(call_entry, entry, &src->calls, list) {
1507                 nentry = duplicate_call_entry(entry, get_irn_link(entry->call), loop_depth);
1508                 list_add_tail(&nentry->list, &dst->calls);
1509         }
1510         dst->n_call_nodes += src->n_call_nodes;
1511         dst->n_nodes      += src->n_nodes;
1512 }
1513
1514 /*
1515  * Inlines small leave methods at call sites where the called address comes
1516  * from a Const node that references the entity representing the called
1517  * method.
1518  * The size argument is a rough measure for the code size of the method:
1519  * Methods where the obstack containing the firm graph is smaller than
1520  * size are inlined.
1521  */
1522 void inline_leave_functions(unsigned maxsize, unsigned leavesize,
1523                             unsigned size, int ignore_runtime)
1524 {
1525         inline_irg_env   *env;
1526         ir_graph         *irg;
1527         int              i, n_irgs;
1528         ir_graph         *rem;
1529         int              did_inline;
1530         wenv_t           wenv;
1531         call_entry       *entry, *next;
1532         const call_entry *centry;
1533         pmap             *copied_graphs;
1534         pmap_entry       *pm_entry;
1535
1536         rem = current_ir_graph;
1537         obstack_init(&temp_obst);
1538
1539         /* a map for the copied graphs, used to inline recursive calls */
1540         copied_graphs = pmap_create();
1541
1542         /* extend all irgs by a temporary data structure for inlining. */
1543         n_irgs = get_irp_n_irgs();
1544         for (i = 0; i < n_irgs; ++i)
1545                 set_irg_link(get_irp_irg(i), alloc_inline_irg_env());
1546
1547         /* Pre-compute information in temporary data structure. */
1548         wenv.ignore_runtime = ignore_runtime;
1549         wenv.ignore_callers = 0;
1550         for (i = 0; i < n_irgs; ++i) {
1551                 ir_graph *irg = get_irp_irg(i);
1552
1553                 assert(get_irg_phase_state(irg) != phase_building);
1554                 free_callee_info(irg);
1555
1556                 assure_cf_loop(irg);
1557                 wenv.x = get_irg_link(irg);
1558                 irg_walk_graph(irg, NULL, collect_calls2, &wenv);
1559         }
1560
1561         /* -- and now inline. -- */
1562
1563         /* Inline leaves recursively -- we might construct new leaves. */
1564         do {
1565                 did_inline = 0;
1566
1567                 for (i = 0; i < n_irgs; ++i) {
1568                         ir_node *call;
1569                         int phiproj_computed = 0;
1570
1571                         current_ir_graph = get_irp_irg(i);
1572                         env              = get_irg_link(current_ir_graph);
1573
1574                         list_for_each_entry_safe(call_entry, entry, next, &env->calls, list) {
1575                                 ir_graph            *callee;
1576                                 irg_inline_property  prop;
1577
1578                                 if (env->n_nodes > maxsize)
1579                                         break;
1580
1581                                 call   = entry->call;
1582                                 callee = entry->callee;
1583
1584                                 prop = get_irg_inline_property(callee);
1585                                 if (prop == irg_inline_forbidden || get_irg_additional_properties(callee) & mtp_property_weak) {
1586                                         /* do not inline forbidden / weak graphs */
1587                                         continue;
1588                                 }
1589
1590                                 if (is_leave(callee) && (
1591                                     is_smaller(callee, leavesize) || prop >= irg_inline_forced)) {
1592                                         if (!phiproj_computed) {
1593                                                 phiproj_computed = 1;
1594                                                 collect_phiprojs(current_ir_graph);
1595                                         }
1596                                         did_inline = inline_method(call, callee);
1597
1598                                         if (did_inline) {
1599                                                 inline_irg_env *callee_env = get_irg_link(callee);
1600
1601                                                 /* call was inlined, Phi/Projs for current graph must be recomputed */
1602                                                 phiproj_computed = 0;
1603
1604                                                 /* Do some statistics */
1605                                                 env->got_inline = 1;
1606                                                 --env->n_call_nodes;
1607                                                 env->n_nodes += callee_env->n_nodes;
1608                                                 --callee_env->n_callers;
1609
1610                                                 /* remove this call from the list */
1611                                                 list_del(&entry->list);
1612                                                 continue;
1613                                         }
1614                                 }
1615                         }
1616                 }
1617         } while (did_inline);
1618
1619         /* inline other small functions. */
1620         for (i = 0; i < n_irgs; ++i) {
1621                 ir_node *call;
1622                 int phiproj_computed = 0;
1623
1624                 current_ir_graph = get_irp_irg(i);
1625                 env              = get_irg_link(current_ir_graph);
1626
1627                 /* note that the list of possible calls is updated during the process */
1628                 list_for_each_entry_safe(call_entry, entry, next, &env->calls, list) {
1629                         irg_inline_property prop;
1630                         ir_graph            *callee;
1631                         pmap_entry          *e;
1632
1633                         call   = entry->call;
1634                         callee = entry->callee;
1635
1636                         prop = get_irg_inline_property(callee);
1637                         if (prop == irg_inline_forbidden || get_irg_additional_properties(callee) & mtp_property_weak) {
1638                                 /* do not inline forbidden / weak graphs */
1639                                 continue;
1640                         }
1641
1642                         e = pmap_find(copied_graphs, callee);
1643                         if (e != NULL) {
1644                                 /*
1645                                  * Remap callee if we have a copy.
1646                                  * FIXME: Should we do this only for recursive Calls ?
1647                                  */
1648                                 callee = e->value;
1649                         }
1650
1651                         if (prop >= irg_inline_forced ||
1652                             (is_smaller(callee, size) && env->n_nodes < maxsize) /* small function */) {
1653                                 if (current_ir_graph == callee) {
1654                                         /*
1655                                          * Recursive call: we cannot directly inline because we cannot walk
1656                                          * the graph and change it. So we have to make a copy of the graph
1657                                          * first.
1658                                          */
1659
1660                                         inline_irg_env *callee_env;
1661                                         ir_graph       *copy;
1662
1663                                         /*
1664                                          * No copy yet, create one.
1665                                          * Note that recursive methods are never leaves, so it is sufficient
1666                                          * to test this condition here.
1667                                          */
1668                                         copy = create_irg_copy(callee);
1669
1670                                         /* create_irg_copy() destroys the Proj links, recompute them */
1671                                         phiproj_computed = 0;
1672
1673                                         /* allocate new environment */
1674                                         callee_env = alloc_inline_irg_env();
1675                                         set_irg_link(copy, callee_env);
1676
1677                                         assure_cf_loop(copy);
1678                                         wenv.x              = callee_env;
1679                                         wenv.ignore_callers = 1;
1680                                         irg_walk_graph(copy, NULL, collect_calls2, &wenv);
1681
1682                                         /*
1683                                          * Enter the entity of the original graph. This is needed
1684                                          * for inline_method(). However, note that ent->irg still points
1685                                          * to callee, NOT to copy.
1686                                          */
1687                                         set_irg_entity(copy, get_irg_entity(callee));
1688
1689                                         pmap_insert(copied_graphs, callee, copy);
1690                                         callee = copy;
1691
1692                                         /* we have only one caller: the original graph */
1693                                         callee_env->n_callers      = 1;
1694                                         callee_env->n_callers_orig = 1;
1695                                 }
1696                                 if (! phiproj_computed) {
1697                                         phiproj_computed = 1;
1698                                         collect_phiprojs(current_ir_graph);
1699                                 }
1700                                 did_inline = inline_method(call, callee);
1701                                 if (did_inline) {
1702                                         inline_irg_env *callee_env = (inline_irg_env *)get_irg_link(callee);
1703
1704                                         /* call was inlined, Phi/Projs for current graph must be recomputed */
1705                                         phiproj_computed = 0;
1706
1707                                         /* callee was inline. Append it's call list. */
1708                                         env->got_inline = 1;
1709                                         --env->n_call_nodes;
1710                                         append_call_list(env, callee_env, entry->loop_depth);
1711                                         --callee_env->n_callers;
1712
1713                                         /* after we have inlined callee, all called methods inside callee
1714                                            are now called once more */
1715                                         list_for_each_entry(call_entry, centry, &callee_env->calls, list) {
1716                                                 inline_irg_env *penv = get_irg_link(centry->callee);
1717                                                 ++penv->n_callers;
1718                                         }
1719
1720                                         /* remove this call from the list */
1721                                         list_del(&entry->list);
1722                                         continue;
1723                                 }
1724                         }
1725                 }
1726         }
1727
1728         for (i = 0; i < n_irgs; ++i) {
1729                 irg = get_irp_irg(i);
1730                 env = get_irg_link(irg);
1731
1732                 if (env->got_inline) {
1733                         optimize_graph_df(irg);
1734                         optimize_cf(irg);
1735                 }
1736                 if (env->got_inline || (env->n_callers_orig != env->n_callers)) {
1737                         DB((dbg, LEVEL_1, "Nodes:%3d ->%3d, calls:%3d ->%3d, callers:%3d ->%3d, -- %s\n",
1738                         env->n_nodes_orig, env->n_nodes, env->n_call_nodes_orig, env->n_call_nodes,
1739                         env->n_callers_orig, env->n_callers,
1740                         get_entity_name(get_irg_entity(irg))));
1741                 }
1742         }
1743
1744         /* kill the copied graphs: we don't need them anymore */
1745         foreach_pmap(copied_graphs, pm_entry) {
1746                 ir_graph *copy = pm_entry->value;
1747
1748                 /* reset the entity, otherwise it will be deleted in the next step ... */
1749                 set_irg_entity(copy, NULL);
1750                 free_ir_graph(copy);
1751         }
1752         pmap_destroy(copied_graphs);
1753
1754         obstack_free(&temp_obst, NULL);
1755         current_ir_graph = rem;
1756 }
1757
1758 /**
1759  * Calculate the parameter weights for transmitting the address of a local variable.
1760  */
1761 static unsigned calc_method_local_weight(ir_node *arg) {
1762         int      i, j, k;
1763         unsigned v, weight = 0;
1764
1765         for (i = get_irn_n_outs(arg) - 1; i >= 0; --i) {
1766                 ir_node *succ = get_irn_out(arg, i);
1767
1768                 switch (get_irn_opcode(succ)) {
1769                 case iro_Load:
1770                 case iro_Store:
1771                         /* Loads and Store can be removed */
1772                         weight += 3;
1773                         break;
1774                 case iro_Sel:
1775                         /* check if all args are constant */
1776                         for (j = get_Sel_n_indexs(succ) - 1; j >= 0; --j) {
1777                                 ir_node *idx = get_Sel_index(succ, j);
1778                                 if (! is_Const(idx))
1779                                         return 0;
1780                         }
1781                         /* Check users on this Sel. Note: if a 0 is returned here, there was
1782                            some unsupported node. */
1783                         v = calc_method_local_weight(succ);
1784                         if (v == 0)
1785                                 return 0;
1786                         /* we can kill one Sel with constant indexes, this is cheap */
1787                         weight += v + 1;
1788                         break;
1789                 case iro_Id:
1790                         /* when looking backward we might find Id nodes */
1791                         weight += calc_method_local_weight(succ);
1792                         break;
1793                 case iro_Tuple:
1794                         /* unoptimized tuple */
1795                         for (j = get_Tuple_n_preds(succ) - 1; j >= 0; --j) {
1796                                 ir_node *pred = get_Tuple_pred(succ, j);
1797                                 if (pred == arg) {
1798                                         /* look for Proj(j) */
1799                                         for (k = get_irn_n_outs(succ) - 1; k >= 0; --k) {
1800                                                 ir_node *succ_succ = get_irn_out(succ, k);
1801                                                 if (is_Proj(succ_succ)) {
1802                                                         if (get_Proj_proj(succ_succ) == j) {
1803                                                                 /* found */
1804                                                                 weight += calc_method_local_weight(succ_succ);
1805                                                         }
1806                                                 } else {
1807                                                         /* this should NOT happen */
1808                                                         return 0;
1809                                                 }
1810                                         }
1811                                 }
1812                         }
1813                         break;
1814                 default:
1815                         /* any other node: unsupported yet or bad. */
1816                         return 0;
1817                 }
1818         }
1819         return weight;
1820 }
1821
1822 /**
1823  * Calculate the parameter weights for transmitting the address of a local variable.
1824  */
1825 static void analyze_irg_local_weights(inline_irg_env *env, ir_graph *irg) {
1826         ir_entity *ent = get_irg_entity(irg);
1827         ir_type  *mtp;
1828         int      nparams, i, proj_nr;
1829         ir_node  *irg_args, *arg;
1830
1831         mtp      = get_entity_type(ent);
1832         nparams  = get_method_n_params(mtp);
1833
1834         /* allocate a new array. currently used as 'analysed' flag */
1835         env->local_weights = NEW_ARR_D(unsigned, &temp_obst, nparams);
1836
1837         /* If the method haven't parameters we have nothing to do. */
1838         if (nparams <= 0)
1839                 return;
1840
1841         assure_irg_outs(irg);
1842         irg_args = get_irg_args(irg);
1843         for (i = get_irn_n_outs(irg_args) - 1; i >= 0; --i) {
1844                 arg     = get_irn_out(irg_args, i);
1845                 proj_nr = get_Proj_proj(arg);
1846                 env->local_weights[proj_nr] = calc_method_local_weight(arg);
1847         }
1848 }
1849
1850 /**
1851  * Calculate the benefice for transmitting an local variable address.
1852  * After inlining, the local variable might be transformed into a
1853  * SSA variable by scalar_replacement().
1854  */
1855 static unsigned get_method_local_adress_weight(ir_graph *callee, int pos) {
1856         inline_irg_env *env = get_irg_link(callee);
1857
1858         if (env->local_weights != NULL) {
1859                 if (pos < ARR_LEN(env->local_weights))
1860                         return env->local_weights[pos];
1861                 return 0;
1862         }
1863
1864         analyze_irg_local_weights(env, callee);
1865
1866         if (pos < ARR_LEN(env->local_weights))
1867                 return env->local_weights[pos];
1868         return 0;
1869 }
1870
1871 /**
1872  * Calculate a benefice value for inlining the given call.
1873  *
1874  * @param call       the call node we have to inspect
1875  * @param callee     the called graph
1876  */
1877 static int calc_inline_benefice(call_entry *entry, ir_graph *callee)
1878 {
1879         ir_node   *call = entry->call;
1880         ir_entity *ent  = get_irg_entity(callee);
1881         ir_node   *frame_ptr;
1882         ir_type   *mtp;
1883         int       weight = 0;
1884         int       i, n_params, all_const;
1885         unsigned  cc, v;
1886         irg_inline_property prop;
1887
1888         inline_irg_env *callee_env;
1889
1890         prop = get_irg_inline_property(callee);
1891         if (prop == irg_inline_forbidden) {
1892                 DB((dbg, LEVEL_2, "In %+F Call to %+F: inlining forbidden\n",
1893                     call, callee));
1894                 return entry->benefice = INT_MIN;
1895         }
1896
1897         if (get_irg_additional_properties(callee) & (mtp_property_noreturn | mtp_property_weak)) {
1898                 DB((dbg, LEVEL_2, "In %+F Call to %+F: not inlining noreturn or weak\n",
1899                     call, callee));
1900                 return entry->benefice = INT_MIN;
1901         }
1902
1903         /* costs for every passed parameter */
1904         n_params = get_Call_n_params(call);
1905         mtp      = get_entity_type(ent);
1906         cc       = get_method_calling_convention(mtp);
1907         if (cc & cc_reg_param) {
1908                 /* register parameter, smaller costs for register parameters */
1909                 int max_regs = cc & ~cc_bits;
1910
1911                 if (max_regs < n_params)
1912                         weight += max_regs * 2 + (n_params - max_regs) * 5;
1913                 else
1914                         weight += n_params * 2;
1915         } else {
1916                 /* parameters are passed an stack */
1917                 weight += 5 * n_params;
1918         }
1919
1920         /* constant parameters improve the benefice */
1921         frame_ptr = get_irg_frame(current_ir_graph);
1922         all_const = 1;
1923         for (i = 0; i < n_params; ++i) {
1924                 ir_node *param = get_Call_param(call, i);
1925
1926                 if (is_Const(param)) {
1927                         weight += get_method_param_weight(ent, i);
1928                 } else {
1929                         all_const = 0;
1930                         if (is_SymConst(param))
1931                                 weight += get_method_param_weight(ent, i);
1932                         else if (is_Sel(param) && get_Sel_ptr(param) == frame_ptr) {
1933                                 /*
1934                                  * An address of a local variable is transmitted. After
1935                                  * inlining, scalar_replacement might be able to remove the
1936                                  * local variable, so honor this.
1937                                  */
1938                                 v = get_method_local_adress_weight(callee, i);
1939                                 weight += v;
1940                                 if (v > 0)
1941                                         entry->local_adr = 1;
1942                         }
1943                 }
1944         }
1945         entry->all_const = all_const;
1946
1947         callee_env = get_irg_link(callee);
1948         if (callee_env->n_callers == 1 &&
1949             callee != current_ir_graph &&
1950                 get_entity_visibility(ent) == visibility_local) {
1951                 weight += 700;
1952         }
1953
1954         /* give a bonus for functions with one block */
1955         if (callee_env->n_blocks == 1)
1956                 weight = weight * 3 / 2;
1957
1958         /* and one for small non-recursive functions: we want them to be inlined in mostly every case */
1959         if (callee_env->n_nodes < 30 && !callee_env->recursive)
1960                 weight += 2000;
1961
1962         /* and finally for leaves: they do not increase the register pressure
1963            because of callee safe registers */
1964         if (callee_env->n_call_nodes == 0)
1965                 weight += 400;
1966
1967         /** it's important to inline inner loops first */
1968         if (entry->loop_depth > 30)
1969                 weight += 30 * 1024;
1970         else
1971                 weight += entry->loop_depth * 1024;
1972
1973         /*
1974          * All arguments constant is probably a good sign, give an extra bonus
1975          */
1976         if (all_const)
1977                 weight += 1024;
1978
1979         return entry->benefice = weight;
1980 }
1981
1982 static ir_graph **irgs;
1983 static int      last_irg;
1984
1985 /**
1986  * Callgraph walker, collect all visited graphs.
1987  */
1988 static void callgraph_walker(ir_graph *irg, void *data) {
1989         (void) data;
1990         irgs[last_irg++] = irg;
1991 }
1992
1993 /**
1994  * Creates an inline order for all graphs.
1995  *
1996  * @return the list of graphs.
1997  */
1998 static ir_graph **create_irg_list(void) {
1999         ir_entity **free_methods;
2000         int       arr_len;
2001         int       n_irgs = get_irp_n_irgs();
2002
2003         cgana(&arr_len, &free_methods);
2004         xfree(free_methods);
2005
2006         compute_callgraph();
2007
2008         last_irg = 0;
2009         irgs     = XMALLOCNZ(ir_graph*, n_irgs);
2010
2011         callgraph_walk(NULL, callgraph_walker, NULL);
2012         assert(n_irgs == last_irg);
2013
2014         return irgs;
2015 }
2016
2017 /**
2018  * Push a call onto the priority list if its benefice is big enough.
2019  *
2020  * @param pqueue   the priority queue of calls
2021  * @param call     the call entry
2022  * @param inlien_threshold
2023  *                 the threshold value
2024  */
2025 static void maybe_push_call(pqueue_t *pqueue, call_entry *call,
2026                             int inline_threshold)
2027 {
2028         ir_graph            *callee  = call->callee;
2029         irg_inline_property prop     = get_irg_inline_property(callee);
2030         int                 benefice = calc_inline_benefice(call, callee);
2031
2032         DB((dbg, LEVEL_2, "In %+F Call %+F to %+F has benefice %d\n",
2033             get_irn_irg(call->call), call->call, callee, benefice));
2034
2035         if (prop < irg_inline_forced && benefice < inline_threshold) {
2036                 return;
2037         }
2038
2039         pqueue_put(pqueue, call, benefice);
2040 }
2041
2042 /**
2043  * Try to inline calls into a graph.
2044  *
2045  * @param irg      the graph into which we inline
2046  * @param maxsize  do NOT inline if the size of irg gets
2047  *                 bigger than this amount
2048  * @param inline_threshold
2049  *                 threshold value for inline decision
2050  * @param copied_graphs
2051  *                 map containing copied of recursive graphs
2052  */
2053 static void inline_into(ir_graph *irg, unsigned maxsize,
2054                         int inline_threshold, pmap *copied_graphs)
2055 {
2056         int            phiproj_computed = 0;
2057         inline_irg_env *env = get_irg_link(irg);
2058         call_entry     *curr_call;
2059         wenv_t         wenv;
2060         pqueue_t       *pqueue;
2061
2062         if (env->n_call_nodes == 0)
2063                 return;
2064
2065         if (env->n_nodes > maxsize) {
2066                 DB((dbg, LEVEL_2, "%+F: too big (%d)\n", irg, env->n_nodes));
2067                 return;
2068         }
2069
2070         current_ir_graph = irg;
2071
2072         /* put irgs into the pqueue */
2073         pqueue = new_pqueue();
2074
2075         list_for_each_entry(call_entry, curr_call, &env->calls, list) {
2076                 assert(is_Call(curr_call->call));
2077                 maybe_push_call(pqueue, curr_call, inline_threshold);
2078         }
2079
2080         /* note that the list of possible calls is updated during the process */
2081         while (!pqueue_empty(pqueue)) {
2082                 int                 did_inline;
2083                 call_entry          *curr_call  = pqueue_pop_front(pqueue);
2084                 ir_graph            *callee     = curr_call->callee;
2085                 ir_node             *call_node  = curr_call->call;
2086                 inline_irg_env      *callee_env = get_irg_link(callee);
2087                 irg_inline_property prop        = get_irg_inline_property(callee);
2088                 int                 loop_depth;
2089                 const call_entry    *centry;
2090                 pmap_entry          *e;
2091
2092                 if ((prop < irg_inline_forced) && env->n_nodes + callee_env->n_nodes > maxsize) {
2093                         DB((dbg, LEVEL_2, "%+F: too big (%d) + %+F (%d)\n", irg,
2094                                                 env->n_nodes, callee, callee_env->n_nodes));
2095                         continue;
2096                 }
2097
2098                 e = pmap_find(copied_graphs, callee);
2099                 if (e != NULL) {
2100                         int benefice = curr_call->benefice;
2101                         /*
2102                          * Reduce the weight for recursive function IFF not all arguments are const.
2103                          * inlining recursive functions is rarely good.
2104                          */
2105                         if (!curr_call->all_const)
2106                                 benefice -= 2000;
2107                         if (benefice < inline_threshold)
2108                                 continue;
2109
2110                         /*
2111                          * Remap callee if we have a copy.
2112                          */
2113                         callee     = e->value;
2114                         callee_env = get_irg_link(callee);
2115                 }
2116
2117                 if (current_ir_graph == callee) {
2118                         /*
2119                          * Recursive call: we cannot directly inline because we cannot
2120                          * walk the graph and change it. So we have to make a copy of
2121                          * the graph first.
2122                          */
2123                         int benefice = curr_call->benefice;
2124                         ir_graph *copy;
2125
2126                         /*
2127                          * Reduce the weight for recursive function IFF not all arguments are const.
2128                          * inlining recursive functions is rarely good.
2129                          */
2130                         if (!curr_call->all_const)
2131                                 benefice -= 2000;
2132                         if (benefice < inline_threshold)
2133                                 continue;
2134
2135                         /*
2136                          * No copy yet, create one.
2137                          * Note that recursive methods are never leaves, so it is
2138                          * sufficient to test this condition here.
2139                          */
2140                         copy = create_irg_copy(callee);
2141
2142                         /* create_irg_copy() destroys the Proj links, recompute them */
2143                         phiproj_computed = 0;
2144
2145                         /* allocate a new environment */
2146                         callee_env = alloc_inline_irg_env();
2147                         set_irg_link(copy, callee_env);
2148
2149                         assure_cf_loop(copy);
2150                         wenv.x              = callee_env;
2151                         wenv.ignore_callers = 1;
2152                         irg_walk_graph(copy, NULL, collect_calls2, &wenv);
2153
2154                         /*
2155                          * Enter the entity of the original graph. This is needed
2156                          * for inline_method(). However, note that ent->irg still points
2157                          * to callee, NOT to copy.
2158                          */
2159                         set_irg_entity(copy, get_irg_entity(callee));
2160
2161                         pmap_insert(copied_graphs, callee, copy);
2162                         callee = copy;
2163
2164                         /* we have only one caller: the original graph */
2165                         callee_env->n_callers      = 1;
2166                         callee_env->n_callers_orig = 1;
2167                 }
2168                 if (! phiproj_computed) {
2169                         phiproj_computed = 1;
2170                         collect_phiprojs(current_ir_graph);
2171                 }
2172                 did_inline = inline_method(call_node, callee);
2173                 if (!did_inline)
2174                         continue;
2175
2176                 /* call was inlined, Phi/Projs for current graph must be recomputed */
2177                 phiproj_computed = 0;
2178
2179                 /* remove it from the caller list */
2180                 list_del(&curr_call->list);
2181
2182                 /* callee was inline. Append it's call list. */
2183                 env->got_inline = 1;
2184                 if (curr_call->local_adr)
2185                         env->local_vars = 1;
2186                 --env->n_call_nodes;
2187
2188                 /* we just generate a bunch of new calls */
2189                 loop_depth = curr_call->loop_depth;
2190                 list_for_each_entry(call_entry, centry, &callee_env->calls, list) {
2191                         inline_irg_env *penv = get_irg_link(centry->callee);
2192                         ir_node        *new_call;
2193                         call_entry     *new_entry;
2194
2195                         /* after we have inlined callee, all called methods inside
2196                          * callee are now called once more */
2197                         ++penv->n_callers;
2198
2199                         /* Note that the src list points to Call nodes in the inlined graph,
2200                          * but we need Call nodes in our graph. Luckily the inliner leaves
2201                          * this information in the link field. */
2202                         new_call = get_irn_link(centry->call);
2203                         assert(is_Call(new_call));
2204
2205                         new_entry = duplicate_call_entry(centry, new_call, loop_depth);
2206                         list_add_tail(&new_entry->list, &env->calls);
2207                         maybe_push_call(pqueue, new_entry, inline_threshold);
2208                 }
2209
2210                 env->n_call_nodes += callee_env->n_call_nodes;
2211                 env->n_nodes += callee_env->n_nodes;
2212                 --callee_env->n_callers;
2213         }
2214
2215         del_pqueue(pqueue);
2216 }
2217
2218 /*
2219  * Heuristic inliner. Calculates a benefice value for every call and inlines
2220  * those calls with a value higher than the threshold.
2221  */
2222 void inline_functions(unsigned maxsize, int inline_threshold) {
2223         inline_irg_env   *env;
2224         int              i, n_irgs;
2225         ir_graph         *rem;
2226         wenv_t           wenv;
2227         pmap             *copied_graphs;
2228         pmap_entry       *pm_entry;
2229         ir_graph         **irgs;
2230
2231         rem = current_ir_graph;
2232         obstack_init(&temp_obst);
2233
2234         irgs = create_irg_list();
2235
2236         /* a map for the copied graphs, used to inline recursive calls */
2237         copied_graphs = pmap_create();
2238
2239         /* extend all irgs by a temporary data structure for inlining. */
2240         n_irgs = get_irp_n_irgs();
2241         for (i = 0; i < n_irgs; ++i)
2242                 set_irg_link(irgs[i], alloc_inline_irg_env());
2243
2244         /* Pre-compute information in temporary data structure. */
2245         wenv.ignore_runtime = 0;
2246         wenv.ignore_callers = 0;
2247         for (i = 0; i < n_irgs; ++i) {
2248                 ir_graph *irg = irgs[i];
2249
2250                 free_callee_info(irg);
2251
2252                 wenv.x = get_irg_link(irg);
2253                 assure_cf_loop(irg);
2254                 irg_walk_graph(irg, NULL, collect_calls2, &wenv);
2255         }
2256
2257         /* -- and now inline. -- */
2258         for (i = 0; i < n_irgs; ++i) {
2259                 ir_graph *irg = irgs[i];
2260
2261                 inline_into(irg, maxsize, inline_threshold, copied_graphs);
2262         }
2263
2264         for (i = 0; i < n_irgs; ++i) {
2265                 ir_graph *irg = irgs[i];
2266
2267                 env = get_irg_link(irg);
2268                 if (env->got_inline) {
2269                         /* this irg got calls inlined: optimize it */
2270                         if (get_opt_combo()) {
2271                                 if (env->local_vars) {
2272                                         scalar_replacement_opt(irg);
2273                                 }
2274                                 combo(irg);
2275                         } else {
2276                                 if (env->local_vars) {
2277                                         if (scalar_replacement_opt(irg)) {
2278                                                 optimize_graph_df(irg);
2279                                         }
2280                                 }
2281                                 optimize_cf(irg);
2282                         }
2283                 }
2284                 if (env->got_inline || (env->n_callers_orig != env->n_callers)) {
2285                         DB((dbg, LEVEL_1, "Nodes:%3d ->%3d, calls:%3d ->%3d, callers:%3d ->%3d, -- %s\n",
2286                         env->n_nodes_orig, env->n_nodes, env->n_call_nodes_orig, env->n_call_nodes,
2287                         env->n_callers_orig, env->n_callers,
2288                         get_entity_name(get_irg_entity(irg))));
2289                 }
2290         }
2291
2292         /* kill the copied graphs: we don't need them anymore */
2293         foreach_pmap(copied_graphs, pm_entry) {
2294                 ir_graph *copy = pm_entry->value;
2295
2296                 /* reset the entity, otherwise it will be deleted in the next step ... */
2297                 set_irg_entity(copy, NULL);
2298                 free_ir_graph(copy);
2299         }
2300         pmap_destroy(copied_graphs);
2301
2302         xfree(irgs);
2303
2304         obstack_free(&temp_obst, NULL);
2305         current_ir_graph = rem;
2306 }
2307
2308 void firm_init_inline(void) {
2309         FIRM_DBG_REGISTER(dbg, "firm.opt.inline");
2310 }