2 * Copyright (C) 1995-2008 University of Karlsruhe. All right reserved.
4 * This file is part of libFirm.
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.
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.
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
22 * @brief Move nodes to a block where they will be executed the least
24 * @author Christian Schaefer, Goetz Lindenmaier, Sebastian Felis,
28 * The idea here is to push nodes as deep into the dominance tree as their
29 * dependencies allow. After pushing them back up out of as many loops as
36 #include "iroptimize.h"
43 static bool is_block_reachable(ir_node *block)
45 return get_Block_dom_depth(block) >= 0;
49 * Find the earliest correct block for node n. --- Place n into the
50 * same Block as its dominance-deepest Input.
52 * move_out_of_loops() expects that place_floats_early() have placed
53 * all "living" nodes into a living block. That's why we must
54 * move nodes in dead block with "live" successors into a valid
56 * We move them just into the same block as its successor (or
57 * in case of a Phi into the effective use block). For Phi successors,
58 * this may still be a dead block, but then there is no real use, as
59 * the control flow will be dead later.
61 static void place_floats_early(ir_node *n, waitq *worklist)
71 /* we must not run into an infinite loop */
72 if (irn_visited_else_mark(n))
75 /* The algorithm relies on the fact that all predecessors of a block are
76 * moved up after a call to place_float_early of the predecessors
77 * (see the loop below).
78 * However we have to break cycles somewhere. Relying on the visited flag
79 * will result in nodes not being moved up despite their place_floats_early
81 * Instead we break cycles at pinned nodes which won't move anyway:
82 * This works because in firm each cycle contains a Phi or Block node
85 if (get_irn_pinned(n) != op_pin_state_floats) {
86 /* we can't move pinned nodes */
87 arity = get_irn_arity(n);
88 for (i = 0; i < arity; ++i) {
89 ir_node *pred = get_irn_n(n, i);
90 pdeq_putr(worklist, pred);
93 pdeq_putr(worklist, get_nodes_block(n));
97 block = get_nodes_block(n);
98 /* do not move unreachable code (or its predecessors around) since dominance
100 if (!is_block_reachable(block))
103 /* first move predecessors up */
104 arity = get_irn_arity(n);
105 place_floats_early(block, worklist);
106 for (i = 0; i < arity; ++i) {
107 ir_node *pred = get_irn_n(n, i);
108 place_floats_early(pred, worklist);
111 /* determine earliest point */
114 for (i = 0; i < arity; ++i) {
115 ir_node *pred = get_irn_n(n, i);
116 ir_node *pred_block = get_nodes_block(pred);
117 int pred_depth = get_Block_dom_depth(pred_block);
118 if (pred_depth > new_depth) {
119 new_depth = pred_depth;
120 new_block = pred_block;
124 /* avoid moving nodes into the start block if we are not in the backend */
125 irg = get_irn_irg(n);
126 start_block = get_irg_start_block(irg);
127 if (new_block == start_block && block != start_block &&
128 get_irg_phase_state(irg) != phase_backend) {
129 assert(get_Block_n_cfg_outs(start_block) == 1);
130 new_block = get_Block_cfg_out(start_block, 0);
133 /* Set the new block */
134 if (new_block != NULL)
135 set_nodes_block(n, new_block);
139 * Floating nodes form subgraphs that begin at nodes as Const, Load,
140 * Start, Call and that end at op_pin_state_pinned nodes as Store, Call.
141 * Place_early places all floating nodes reachable from its argument through
142 * floating nodes and adds all beginnings at op_pin_state_pinned nodes to the
145 * @param worklist a worklist, used for the algorithm, empty on in/output
147 static void place_early(ir_graph *irg, waitq *worklist)
150 inc_irg_visited(irg);
152 /* this inits the worklist */
153 place_floats_early(get_irg_end(irg), worklist);
155 /* Work the content of the worklist. */
156 while (!waitq_empty(worklist)) {
157 ir_node *n = (ir_node*)waitq_get(worklist);
159 place_floats_early(n, worklist);
161 set_irg_pinned(irg, op_pin_state_pinned);
165 * Compute the deepest common dominator tree ancestor of block and dca.
167 * @param dca the deepest common dominator tree ancestor so far,
169 * @param block a block
171 * @return the deepest common dominator tree ancestor of block and dca
173 static ir_node *calc_dom_dca(ir_node *dca, ir_node *block)
175 assert(block != NULL);
177 /* We found a first legal placement. */
181 /* Find a placement that is dominates both, dca and block. */
182 while (get_Block_dom_depth(block) > get_Block_dom_depth(dca))
183 block = get_Block_idom(block);
185 while (get_Block_dom_depth(dca) > get_Block_dom_depth(block)) {
186 dca = get_Block_idom(dca);
189 while (block != dca) {
190 block = get_Block_idom(block); dca = get_Block_idom(dca);
196 * Deepest common dominance ancestor of DCA and CONSUMER of PRODUCER.
197 * I.e., DCA is the block where we might place PRODUCER.
198 * A data flow edge points from producer to consumer.
200 static ir_node *consumer_dom_dca(ir_node *dca, ir_node *consumer,
203 /* Compute the last block into which we can place a node so that it is
205 if (is_Phi(consumer)) {
206 /* our consumer is a Phi-node, the effective use is in all those
207 blocks through which the Phi-node reaches producer */
208 ir_node *phi_block = get_nodes_block(consumer);
209 int arity = get_irn_arity(consumer);
212 for (i = 0; i < arity; i++) {
213 if (get_Phi_pred(consumer, i) == producer) {
214 ir_node *new_block = get_Block_cfgpred_block(phi_block, i);
215 if (is_Bad(new_block))
218 if (is_block_reachable(new_block))
219 dca = calc_dom_dca(dca, new_block);
223 dca = calc_dom_dca(dca, get_nodes_block(consumer));
228 static inline int get_block_loop_depth(ir_node *block)
230 return get_loop_depth(get_irn_loop(block));
234 * Move n to a block with less loop depth than its current block. The
235 * new block must be dominated by early.
237 * @param n the node that should be moved
238 * @param early the earliest block we can n move to
240 static void move_out_of_loops(ir_node *n, ir_node *early)
242 ir_node *block = get_nodes_block(n);
243 ir_node *best = block;
244 int best_depth = get_block_loop_depth(best);
246 /* Find the region deepest in the dominator tree dominating
247 dca with the least loop nesting depth, but still dominated
248 by our early placement. */
249 while (block != early) {
250 ir_node *idom = get_Block_idom(block);
251 int idom_depth = get_block_loop_depth(idom);
252 if (idom_depth < best_depth) {
254 best_depth = idom_depth;
258 if (best != get_nodes_block(n))
259 set_nodes_block(n, best);
263 * Calculate the deepest common ancestor in the dominator tree of all nodes'
264 * blocks depending on node; our final placement has to dominate DCA.
266 * @param node the definition node
267 * @param dca the deepest common ancestor block so far, initially
270 * @return the deepest common dominator ancestor of all blocks of node's users
272 static ir_node *get_deepest_common_dom_ancestor(ir_node *node, ir_node *dca)
276 for (i = get_irn_n_outs(node) - 1; i >= 0; --i) {
277 ir_node *succ = get_irn_out(node, i);
279 /* keepalive edges are special and don't respect the dominance */
284 /* Proj nodes are in the same block as node, so
285 * the users of Proj are our users. */
286 dca = get_deepest_common_dom_ancestor(succ, dca);
288 /* ignore successors in unreachable code */
289 ir_node *succ_blk = get_nodes_block(succ);
290 if (!is_block_reachable(succ_blk))
292 dca = consumer_dom_dca(dca, succ, node);
299 * Put all the Proj nodes of a node into a given block.
301 * @param node the mode_T node
302 * @param block the block to put the Proj nodes to
304 static void set_projs_block(ir_node *node, ir_node *block)
308 for (i = get_irn_n_outs(node) - 1; i >= 0; --i) {
309 ir_node *succ = get_irn_out(node, i);
311 assert(is_Proj(succ));
313 if (get_irn_mode(succ) == mode_T) {
314 set_projs_block(succ, block);
316 set_nodes_block(succ, block);
321 * Find the latest legal block for N and place N into the
322 * `optimal' Block between the latest and earliest legal block.
323 * The `optimal' block is the dominance-deepest block of those
324 * with the least loop-nesting-depth. This places N out of as many
325 * loops as possible and then makes it as control dependent as
328 static void place_floats_late(ir_node *n, pdeq *worklist)
335 if (irn_visited_else_mark(n))
338 n_outs = get_irn_n_outs(n);
339 /* break cycles at pinned nodes (see place place_floats_early) as to why */
340 if (get_irn_pinned(n) != op_pin_state_floats) {
341 for (i = 0; i < n_outs; ++i) {
342 ir_node *succ = get_irn_out(n, i);
343 pdeq_putr(worklist, succ);
348 /* place our users */
349 for (i = 0; i < n_outs; ++i) {
350 ir_node *succ = get_irn_out(n, i);
351 place_floats_late(succ, worklist);
354 /* no point in moving Projs around, they are moved with their predecessor */
357 /* some nodes should simply stay in the startblock */
358 if (is_irn_start_block_placed(n)) {
360 ir_graph *irg = get_irn_irg(n);
361 ir_node *start_block = get_irg_start_block(irg);
362 assert(get_nodes_block(n) == start_block);
367 /* don't move unreachable code around */
368 block = get_nodes_block(n);
369 if (!is_block_reachable(block))
372 /* deepest common ancestor in the dominator tree of all nodes'
373 blocks depending on us; our final placement has to dominate
375 dca = get_deepest_common_dom_ancestor(n, NULL);
377 set_nodes_block(n, dca);
378 move_out_of_loops(n, block);
379 if (get_irn_mode(n) == mode_T) {
380 set_projs_block(n, get_nodes_block(n));
386 * Place floating nodes on the given worklist as late as possible using
387 * the dominance tree.
389 * @param worklist the worklist containing the nodes to place
391 static void place_late(ir_graph *irg, waitq *worklist)
394 inc_irg_visited(irg);
396 /* This fills the worklist initially. */
397 place_floats_late(get_irg_start_block(irg), worklist);
399 /* And now empty the worklist again... */
400 while (!waitq_empty(worklist)) {
401 ir_node *n = (ir_node*)waitq_get(worklist);
403 place_floats_late(n, worklist);
407 /* Code Placement. */
408 void place_code(ir_graph *irg)
412 remove_critical_cf_edges(irg);
414 /* Handle graph state */
415 assert(get_irg_phase_state(irg) != phase_building);
416 assure_irg_outs(irg);
420 /* Place all floating nodes as early as possible. This guarantees
421 a legal code placement. */
422 worklist = new_waitq();
423 place_early(irg, worklist);
425 /* Note: place_early changes only blocks, no data edges. So, the
426 * data out edges are still valid, no need to recalculate them here. */
428 /* Now move the nodes down in the dominator tree. This reduces the
429 unnecessary executions of the node. */
430 place_late(irg, worklist);
432 set_irg_outs_inconsistent(irg);
433 set_irg_loopinfo_inconsistent(irg);
438 * Wrapper for place_code() inside the place_code pass.
440 static void place_code_wrapper(ir_graph *irg)
442 set_opt_global_cse(1);
443 optimize_graph_df(irg);
445 set_opt_global_cse(0);
448 ir_graph_pass_t *place_code_pass(const char *name)
450 return def_graph_pass(name ? name : "place", place_code_wrapper);