1 /*
2 * Copyright (c) 1997, 2014, Oracle and/or its affiliates. All rights reserved.
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4 *
5 * This code is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 only, as
7 * published by the Free Software Foundation.
8 *
9 * This code is distributed in the hope that it will be useful, but WITHOUT
10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
12 * version 2 for more details (a copy is included in the LICENSE file that
13 * accompanied this code).
14 *
15 * You should have received a copy of the GNU General Public License version
16 * 2 along with this work; if not, write to the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
18 *
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20 * or visit www.oracle.com if you need additional information or have any
21 * questions.
22 *
23 */
24
25 #include "precompiled.hpp"
26 #include "libadt/vectset.hpp"
27 #include "memory/allocation.inline.hpp"
28 #include "opto/block.hpp"
29 #include "opto/c2compiler.hpp"
30 #include "opto/callnode.hpp"
31 #include "opto/cfgnode.hpp"
32 #include "opto/machnode.hpp"
33 #include "opto/opcodes.hpp"
34 #include "opto/phaseX.hpp"
35 #include "opto/rootnode.hpp"
36 #include "opto/runtime.hpp"
37 #include "runtime/deoptimization.hpp"
38
39 // Portions of code courtesy of Clifford Click
40
41 // Optimization - Graph Style
42
43 // To avoid float value underflow
44 #define MIN_BLOCK_FREQUENCY 1.e-35f
45
46 //----------------------------schedule_node_into_block-------------------------
47 // Insert node n into block b. Look for projections of n and make sure they
48 // are in b also.
49 void PhaseCFG::schedule_node_into_block( Node *n, Block *b ) {
50 // Set basic block of n, Add n to b,
51 map_node_to_block(n, b);
52 b->add_inst(n);
53
54 // After Matching, nearly any old Node may have projections trailing it.
55 // These are usually machine-dependent flags. In any case, they might
56 // float to another block below this one. Move them up.
57 for (DUIterator_Fast imax, i = n->fast_outs(imax); i < imax; i++) {
58 Node* use = n->fast_out(i);
59 if (use->is_Proj()) {
60 Block* buse = get_block_for_node(use);
61 if (buse != b) { // In wrong block?
62 if (buse != NULL) {
63 buse->find_remove(use); // Remove from wrong block
64 }
65 map_node_to_block(use, b);
66 b->add_inst(use);
67 }
68 }
69 }
70 }
71
72 //----------------------------replace_block_proj_ctrl-------------------------
73 // Nodes that have is_block_proj() nodes as their control need to use
74 // the appropriate Region for their actual block as their control since
75 // the projection will be in a predecessor block.
76 void PhaseCFG::replace_block_proj_ctrl( Node *n ) {
77 const Node *in0 = n->in(0);
78 assert(in0 != NULL, "Only control-dependent");
79 const Node *p = in0->is_block_proj();
80 if (p != NULL && p != n) { // Control from a block projection?
81 assert(!n->pinned() || n->is_MachConstantBase(), "only pinned MachConstantBase node is expected here");
82 // Find trailing Region
83 Block *pb = get_block_for_node(in0); // Block-projection already has basic block
84 uint j = 0;
85 if (pb->_num_succs != 1) { // More then 1 successor?
86 // Search for successor
87 uint max = pb->number_of_nodes();
88 assert( max > 1, "" );
89 uint start = max - pb->_num_succs;
90 // Find which output path belongs to projection
91 for (j = start; j < max; j++) {
92 if( pb->get_node(j) == in0 )
93 break;
94 }
95 assert( j < max, "must find" );
96 // Change control to match head of successor basic block
97 j -= start;
98 }
99 n->set_req(0, pb->_succs[j]->head());
100 }
101 }
102
103 static bool is_dominator(Block* d, Block* n) {
104 return d->dom_lca(n) == d;
105 }
106
107 //------------------------------schedule_pinned_nodes--------------------------
108 // Set the basic block for Nodes pinned into blocks
109 void PhaseCFG::schedule_pinned_nodes(VectorSet &visited) {
110 // Allocate node stack of size C->unique()+8 to avoid frequent realloc
111 GrowableArray <Node *> spstack(C->unique() + 8);
112 spstack.push(_root);
113 while (spstack.is_nonempty()) {
114 Node* node = spstack.pop();
115 if (!visited.test_set(node->_idx)) { // Test node and flag it as visited
116 if (node->pinned() && !has_block(node)) { // Pinned? Nail it down!
117 assert(node->in(0), "pinned Node must have Control");
118 // Before setting block replace block_proj control edge
119 replace_block_proj_ctrl(node);
120 Node* input = node->in(0);
121 while (!input->is_block_start()) {
122 input = input->in(0);
123 }
124 Block* block = get_block_for_node(input); // Basic block of controlling input
125 schedule_node_into_block(node, block);
126 }
127
128 // If the node has precedence edges (added when CastPP nodes are
129 // removed in final_graph_reshaping), fix the control of the
130 // node to cover the precedence edges and remove the
131 // dependencies.
132 Node* n = NULL;
133 for (uint i = node->len()-1; i >= node->req(); i--) {
134 Node* m = node->in(i);
135 if (m == NULL) continue;
136 if (m->is_block_proj()) {
137 node->rm_prec(i);
138 if (n == NULL) {
139 n = m;
140 } else {
141 Block* bn = get_block_for_node(n);
142 Block* bm = get_block_for_node(m);
143 n = is_dominator(bn, bm) ? m : n;
144 }
145 }
146 }
147 if (n != NULL) {
148 assert(node->in(0), "control should have been set");
149 Block* bn = get_block_for_node(n);
150 Block* bnode = get_block_for_node(node->in(0));
151 if (!is_dominator(bn, bnode)) {
152 node->set_req(0, n);
153 }
154 }
155
156 // process all inputs that are non NULL
157 for (int i = node->req() - 1; i >= 0; --i) {
158 if (node->in(i) != NULL) {
159 spstack.push(node->in(i));
160 }
161 }
162 }
163 }
164 }
165
166 #ifdef ASSERT
167 // Assert that new input b2 is dominated by all previous inputs.
168 // Check this by by seeing that it is dominated by b1, the deepest
169 // input observed until b2.
170 static void assert_dom(Block* b1, Block* b2, Node* n, const PhaseCFG* cfg) {
171 if (b1 == NULL) return;
172 assert(b1->_dom_depth < b2->_dom_depth, "sanity");
173 Block* tmp = b2;
174 while (tmp != b1 && tmp != NULL) {
175 tmp = tmp->_idom;
176 }
177 if (tmp != b1) {
178 // Detected an unschedulable graph. Print some nice stuff and die.
179 tty->print_cr("!!! Unschedulable graph !!!");
180 for (uint j=0; j<n->len(); j++) { // For all inputs
181 Node* inn = n->in(j); // Get input
182 if (inn == NULL) continue; // Ignore NULL, missing inputs
183 Block* inb = cfg->get_block_for_node(inn);
184 tty->print("B%d idom=B%d depth=%2d ",inb->_pre_order,
185 inb->_idom ? inb->_idom->_pre_order : 0, inb->_dom_depth);
186 inn->dump();
187 }
188 tty->print("Failing node: ");
189 n->dump();
190 assert(false, "unscheduable graph");
191 }
192 }
193 #endif
194
195 static Block* find_deepest_input(Node* n, const PhaseCFG* cfg) {
196 // Find the last input dominated by all other inputs.
197 Block* deepb = NULL; // Deepest block so far
198 int deepb_dom_depth = 0;
199 for (uint k = 0; k < n->len(); k++) { // For all inputs
200 Node* inn = n->in(k); // Get input
201 if (inn == NULL) continue; // Ignore NULL, missing inputs
202 Block* inb = cfg->get_block_for_node(inn);
203 assert(inb != NULL, "must already have scheduled this input");
204 if (deepb_dom_depth < (int) inb->_dom_depth) {
205 // The new inb must be dominated by the previous deepb.
206 // The various inputs must be linearly ordered in the dom
207 // tree, or else there will not be a unique deepest block.
208 DEBUG_ONLY(assert_dom(deepb, inb, n, cfg));
209 deepb = inb; // Save deepest block
210 deepb_dom_depth = deepb->_dom_depth;
211 }
212 }
213 assert(deepb != NULL, "must be at least one input to n");
214 return deepb;
215 }
216
217
218 //------------------------------schedule_early---------------------------------
219 // Find the earliest Block any instruction can be placed in. Some instructions
220 // are pinned into Blocks. Unpinned instructions can appear in last block in
221 // which all their inputs occur.
222 bool PhaseCFG::schedule_early(VectorSet &visited, Node_List &roots) {
223 // Allocate stack with enough space to avoid frequent realloc
224 Node_Stack nstack(roots.Size() + 8);
225 // _root will be processed among C->top() inputs
226 roots.push(C->top());
227 visited.set(C->top()->_idx);
228
229 while (roots.size() != 0) {
230 // Use local variables nstack_top_n & nstack_top_i to cache values
231 // on stack's top.
232 Node* parent_node = roots.pop();
233 uint input_index = 0;
234
235 while (true) {
236 if (input_index == 0) {
237 // Fixup some control. Constants without control get attached
238 // to root and nodes that use is_block_proj() nodes should be attached
239 // to the region that starts their block.
240 const Node* control_input = parent_node->in(0);
241 if (control_input != NULL) {
242 replace_block_proj_ctrl(parent_node);
243 } else {
244 // Is a constant with NO inputs?
245 if (parent_node->req() == 1) {
246 parent_node->set_req(0, _root);
247 }
248 }
249 }
250
251 // First, visit all inputs and force them to get a block. If an
252 // input is already in a block we quit following inputs (to avoid
253 // cycles). Instead we put that Node on a worklist to be handled
254 // later (since IT'S inputs may not have a block yet).
255
256 // Assume all n's inputs will be processed
257 bool done = true;
258
259 while (input_index < parent_node->len()) {
260 Node* in = parent_node->in(input_index++);
261 if (in == NULL) {
262 continue;
263 }
264
265 int is_visited = visited.test_set(in->_idx);
266 if (!has_block(in)) {
267 if (is_visited) {
268 return false;
269 }
270 // Save parent node and next input's index.
271 nstack.push(parent_node, input_index);
272 // Process current input now.
273 parent_node = in;
274 input_index = 0;
275 // Not all n's inputs processed.
276 done = false;
277 break;
278 } else if (!is_visited) {
279 // Visit this guy later, using worklist
280 roots.push(in);
281 }
282 }
283
284 if (done) {
285 // All of n's inputs have been processed, complete post-processing.
286
287 // Some instructions are pinned into a block. These include Region,
288 // Phi, Start, Return, and other control-dependent instructions and
289 // any projections which depend on them.
290 if (!parent_node->pinned()) {
291 // Set earliest legal block.
292 Block* earliest_block = find_deepest_input(parent_node, this);
293 map_node_to_block(parent_node, earliest_block);
294 } else {
295 assert(get_block_for_node(parent_node) == get_block_for_node(parent_node->in(0)), "Pinned Node should be at the same block as its control edge");
296 }
297
298 if (nstack.is_empty()) {
299 // Finished all nodes on stack.
300 // Process next node on the worklist 'roots'.
301 break;
302 }
303 // Get saved parent node and next input's index.
304 parent_node = nstack.node();
305 input_index = nstack.index();
306 nstack.pop();
307 }
308 }
309 }
310 return true;
311 }
312
313 //------------------------------dom_lca----------------------------------------
314 // Find least common ancestor in dominator tree
315 // LCA is a current notion of LCA, to be raised above 'this'.
316 // As a convenient boundary condition, return 'this' if LCA is NULL.
317 // Find the LCA of those two nodes.
318 Block* Block::dom_lca(Block* LCA) {
319 if (LCA == NULL || LCA == this) return this;
320
321 Block* anc = this;
322 while (anc->_dom_depth > LCA->_dom_depth)
323 anc = anc->_idom; // Walk up till anc is as high as LCA
324
325 while (LCA->_dom_depth > anc->_dom_depth)
326 LCA = LCA->_idom; // Walk up till LCA is as high as anc
327
328 while (LCA != anc) { // Walk both up till they are the same
329 LCA = LCA->_idom;
330 anc = anc->_idom;
331 }
332
333 return LCA;
334 }
335
336 //--------------------------raise_LCA_above_use--------------------------------
337 // We are placing a definition, and have been given a def->use edge.
338 // The definition must dominate the use, so move the LCA upward in the
339 // dominator tree to dominate the use. If the use is a phi, adjust
340 // the LCA only with the phi input paths which actually use this def.
341 static Block* raise_LCA_above_use(Block* LCA, Node* use, Node* def, const PhaseCFG* cfg) {
342 Block* buse = cfg->get_block_for_node(use);
343 if (buse == NULL) return LCA; // Unused killing Projs have no use block
344 if (!use->is_Phi()) return buse->dom_lca(LCA);
345 uint pmax = use->req(); // Number of Phi inputs
346 // Why does not this loop just break after finding the matching input to
347 // the Phi? Well...it's like this. I do not have true def-use/use-def
348 // chains. Means I cannot distinguish, from the def-use direction, which
349 // of many use-defs lead from the same use to the same def. That is, this
350 // Phi might have several uses of the same def. Each use appears in a
351 // different predecessor block. But when I enter here, I cannot distinguish
352 // which use-def edge I should find the predecessor block for. So I find
353 // them all. Means I do a little extra work if a Phi uses the same value
354 // more than once.
355 for (uint j=1; j<pmax; j++) { // For all inputs
356 if (use->in(j) == def) { // Found matching input?
357 Block* pred = cfg->get_block_for_node(buse->pred(j));
358 LCA = pred->dom_lca(LCA);
359 }
360 }
361 return LCA;
362 }
363
364 //----------------------------raise_LCA_above_marks----------------------------
365 // Return a new LCA that dominates LCA and any of its marked predecessors.
366 // Search all my parents up to 'early' (exclusive), looking for predecessors
367 // which are marked with the given index. Return the LCA (in the dom tree)
368 // of all marked blocks. If there are none marked, return the original
369 // LCA.
370 static Block* raise_LCA_above_marks(Block* LCA, node_idx_t mark, Block* early, const PhaseCFG* cfg) {
371 Block_List worklist;
372 worklist.push(LCA);
373 while (worklist.size() > 0) {
374 Block* mid = worklist.pop();
375 if (mid == early) continue; // stop searching here
376
377 // Test and set the visited bit.
378 if (mid->raise_LCA_visited() == mark) continue; // already visited
379
380 // Don't process the current LCA, otherwise the search may terminate early
381 if (mid != LCA && mid->raise_LCA_mark() == mark) {
382 // Raise the LCA.
383 LCA = mid->dom_lca(LCA);
384 if (LCA == early) break; // stop searching everywhere
385 assert(early->dominates(LCA), "early is high enough");
386 // Resume searching at that point, skipping intermediate levels.
387 worklist.push(LCA);
388 if (LCA == mid)
389 continue; // Don't mark as visited to avoid early termination.
390 } else {
391 // Keep searching through this block's predecessors.
392 for (uint j = 1, jmax = mid->num_preds(); j < jmax; j++) {
393 Block* mid_parent = cfg->get_block_for_node(mid->pred(j));
394 worklist.push(mid_parent);
395 }
396 }
397 mid->set_raise_LCA_visited(mark);
398 }
399 return LCA;
400 }
401
402 //--------------------------memory_early_block--------------------------------
403 // This is a variation of find_deepest_input, the heart of schedule_early.
404 // Find the "early" block for a load, if we considered only memory and
405 // address inputs, that is, if other data inputs were ignored.
406 //
407 // Because a subset of edges are considered, the resulting block will
408 // be earlier (at a shallower dom_depth) than the true schedule_early
409 // point of the node. We compute this earlier block as a more permissive
410 // site for anti-dependency insertion, but only if subsume_loads is enabled.
411 static Block* memory_early_block(Node* load, Block* early, const PhaseCFG* cfg) {
412 Node* base;
413 Node* index;
414 Node* store = load->in(MemNode::Memory);
415 load->as_Mach()->memory_inputs(base, index);
416
417 assert(base != NodeSentinel && index != NodeSentinel,
418 "unexpected base/index inputs");
419
420 Node* mem_inputs[4];
421 int mem_inputs_length = 0;
422 if (base != NULL) mem_inputs[mem_inputs_length++] = base;
423 if (index != NULL) mem_inputs[mem_inputs_length++] = index;
424 if (store != NULL) mem_inputs[mem_inputs_length++] = store;
425
426 // In the comparision below, add one to account for the control input,
427 // which may be null, but always takes up a spot in the in array.
428 if (mem_inputs_length + 1 < (int) load->req()) {
429 // This "load" has more inputs than just the memory, base and index inputs.
430 // For purposes of checking anti-dependences, we need to start
431 // from the early block of only the address portion of the instruction,
432 // and ignore other blocks that may have factored into the wider
433 // schedule_early calculation.
434 if (load->in(0) != NULL) mem_inputs[mem_inputs_length++] = load->in(0);
435
436 Block* deepb = NULL; // Deepest block so far
437 int deepb_dom_depth = 0;
438 for (int i = 0; i < mem_inputs_length; i++) {
439 Block* inb = cfg->get_block_for_node(mem_inputs[i]);
440 if (deepb_dom_depth < (int) inb->_dom_depth) {
441 // The new inb must be dominated by the previous deepb.
442 // The various inputs must be linearly ordered in the dom
443 // tree, or else there will not be a unique deepest block.
444 DEBUG_ONLY(assert_dom(deepb, inb, load, cfg));
445 deepb = inb; // Save deepest block
446 deepb_dom_depth = deepb->_dom_depth;
447 }
448 }
449 early = deepb;
450 }
451
452 return early;
453 }
454
455 //--------------------------insert_anti_dependences---------------------------
456 // A load may need to witness memory that nearby stores can overwrite.
457 // For each nearby store, either insert an "anti-dependence" edge
458 // from the load to the store, or else move LCA upward to force the
459 // load to (eventually) be scheduled in a block above the store.
460 //
461 // Do not add edges to stores on distinct control-flow paths;
462 // only add edges to stores which might interfere.
463 //
464 // Return the (updated) LCA. There will not be any possibly interfering
465 // store between the load's "early block" and the updated LCA.
466 // Any stores in the updated LCA will have new precedence edges
467 // back to the load. The caller is expected to schedule the load
468 // in the LCA, in which case the precedence edges will make LCM
469 // preserve anti-dependences. The caller may also hoist the load
470 // above the LCA, if it is not the early block.
471 Block* PhaseCFG::insert_anti_dependences(Block* LCA, Node* load, bool verify) {
472 assert(load->needs_anti_dependence_check(), "must be a load of some sort");
473 assert(LCA != NULL, "");
474 DEBUG_ONLY(Block* LCA_orig = LCA);
475
476 // Compute the alias index. Loads and stores with different alias indices
477 // do not need anti-dependence edges.
478 uint load_alias_idx = C->get_alias_index(load->adr_type());
479 #ifdef ASSERT
480 if (load_alias_idx == Compile::AliasIdxBot && C->AliasLevel() > 0 &&
481 (PrintOpto || VerifyAliases ||
482 PrintMiscellaneous && (WizardMode || Verbose))) {
483 // Load nodes should not consume all of memory.
484 // Reporting a bottom type indicates a bug in adlc.
485 // If some particular type of node validly consumes all of memory,
486 // sharpen the preceding "if" to exclude it, so we can catch bugs here.
487 tty->print_cr("*** Possible Anti-Dependence Bug: Load consumes all of memory.");
488 load->dump(2);
489 if (VerifyAliases) assert(load_alias_idx != Compile::AliasIdxBot, "");
490 }
491 #endif
492 assert(load_alias_idx || (load->is_Mach() && load->as_Mach()->ideal_Opcode() == Op_StrComp),
493 "String compare is only known 'load' that does not conflict with any stores");
494 assert(load_alias_idx || (load->is_Mach() && load->as_Mach()->ideal_Opcode() == Op_StrEquals),
495 "String equals is a 'load' that does not conflict with any stores");
496 assert(load_alias_idx || (load->is_Mach() && load->as_Mach()->ideal_Opcode() == Op_StrIndexOf),
497 "String indexOf is a 'load' that does not conflict with any stores");
498 assert(load_alias_idx || (load->is_Mach() && load->as_Mach()->ideal_Opcode() == Op_AryEq),
499 "Arrays equals is a 'load' that do not conflict with any stores");
500
501 if (!C->alias_type(load_alias_idx)->is_rewritable()) {
502 // It is impossible to spoil this load by putting stores before it,
503 // because we know that the stores will never update the value
504 // which 'load' must witness.
505 return LCA;
506 }
507
508 node_idx_t load_index = load->_idx;
509
510 // Note the earliest legal placement of 'load', as determined by
511 // by the unique point in the dom tree where all memory effects
512 // and other inputs are first available. (Computed by schedule_early.)
513 // For normal loads, 'early' is the shallowest place (dom graph wise)
514 // to look for anti-deps between this load and any store.
515 Block* early = get_block_for_node(load);
516
517 // If we are subsuming loads, compute an "early" block that only considers
518 // memory or address inputs. This block may be different than the
519 // schedule_early block in that it could be at an even shallower depth in the
520 // dominator tree, and allow for a broader discovery of anti-dependences.
521 if (C->subsume_loads()) {
522 early = memory_early_block(load, early, this);
523 }
524
525 ResourceArea *area = Thread::current()->resource_area();
526 Node_List worklist_mem(area); // prior memory state to store
527 Node_List worklist_store(area); // possible-def to explore
528 Node_List worklist_visited(area); // visited mergemem nodes
529 Node_List non_early_stores(area); // all relevant stores outside of early
530 bool must_raise_LCA = false;
531
532 #ifdef TRACK_PHI_INPUTS
533 // %%% This extra checking fails because MergeMem nodes are not GVNed.
534 // Provide "phi_inputs" to check if every input to a PhiNode is from the
535 // original memory state. This indicates a PhiNode for which should not
536 // prevent the load from sinking. For such a block, set_raise_LCA_mark
537 // may be overly conservative.
538 // Mechanism: count inputs seen for each Phi encountered in worklist_store.
539 DEBUG_ONLY(GrowableArray<uint> phi_inputs(area, C->unique(),0,0));
540 #endif
541
542 // 'load' uses some memory state; look for users of the same state.
543 // Recurse through MergeMem nodes to the stores that use them.
544
545 // Each of these stores is a possible definition of memory
546 // that 'load' needs to use. We need to force 'load'
547 // to occur before each such store. When the store is in
548 // the same block as 'load', we insert an anti-dependence
549 // edge load->store.
550
551 // The relevant stores "nearby" the load consist of a tree rooted
552 // at initial_mem, with internal nodes of type MergeMem.
553 // Therefore, the branches visited by the worklist are of this form:
554 // initial_mem -> (MergeMem ->)* store
555 // The anti-dependence constraints apply only to the fringe of this tree.
556
557 Node* initial_mem = load->in(MemNode::Memory);
558 worklist_store.push(initial_mem);
559 worklist_visited.push(initial_mem);
560 worklist_mem.push(NULL);
561 while (worklist_store.size() > 0) {
562 // Examine a nearby store to see if it might interfere with our load.
563 Node* mem = worklist_mem.pop();
564 Node* store = worklist_store.pop();
565 uint op = store->Opcode();
566
567 // MergeMems do not directly have anti-deps.
568 // Treat them as internal nodes in a forward tree of memory states,
569 // the leaves of which are each a 'possible-def'.
570 if (store == initial_mem // root (exclusive) of tree we are searching
571 || op == Op_MergeMem // internal node of tree we are searching
572 ) {
573 mem = store; // It's not a possibly interfering store.
574 if (store == initial_mem)
575 initial_mem = NULL; // only process initial memory once
576
577 for (DUIterator_Fast imax, i = mem->fast_outs(imax); i < imax; i++) {
578 store = mem->fast_out(i);
579 if (store->is_MergeMem()) {
580 // Be sure we don't get into combinatorial problems.
581 // (Allow phis to be repeated; they can merge two relevant states.)
582 uint j = worklist_visited.size();
583 for (; j > 0; j--) {
584 if (worklist_visited.at(j-1) == store) break;
585 }
586 if (j > 0) continue; // already on work list; do not repeat
587 worklist_visited.push(store);
588 }
589 worklist_mem.push(mem);
590 worklist_store.push(store);
591 }
592 continue;
593 }
594
595 if (op == Op_MachProj || op == Op_Catch) continue;
596 if (store->needs_anti_dependence_check()) continue; // not really a store
597
598 // Compute the alias index. Loads and stores with different alias
599 // indices do not need anti-dependence edges. Wide MemBar's are
600 // anti-dependent on everything (except immutable memories).
601 const TypePtr* adr_type = store->adr_type();
602 if (!C->can_alias(adr_type, load_alias_idx)) continue;
603
604 // Most slow-path runtime calls do NOT modify Java memory, but
605 // they can block and so write Raw memory.
606 if (store->is_Mach()) {
607 MachNode* mstore = store->as_Mach();
608 if (load_alias_idx != Compile::AliasIdxRaw) {
609 // Check for call into the runtime using the Java calling
610 // convention (and from there into a wrapper); it has no
611 // _method. Can't do this optimization for Native calls because
612 // they CAN write to Java memory.
613 if (mstore->ideal_Opcode() == Op_CallStaticJava) {
614 assert(mstore->is_MachSafePoint(), "");
615 MachSafePointNode* ms = (MachSafePointNode*) mstore;
616 assert(ms->is_MachCallJava(), "");
617 MachCallJavaNode* mcj = (MachCallJavaNode*) ms;
618 if (mcj->_method == NULL) {
619 // These runtime calls do not write to Java visible memory
620 // (other than Raw) and so do not require anti-dependence edges.
621 continue;
622 }
623 }
624 // Same for SafePoints: they read/write Raw but only read otherwise.
625 // This is basically a workaround for SafePoints only defining control
626 // instead of control + memory.
627 if (mstore->ideal_Opcode() == Op_SafePoint)
628 continue;
629 } else {
630 // Some raw memory, such as the load of "top" at an allocation,
631 // can be control dependent on the previous safepoint. See
632 // comments in GraphKit::allocate_heap() about control input.
633 // Inserting an anti-dep between such a safepoint and a use
634 // creates a cycle, and will cause a subsequent failure in
635 // local scheduling. (BugId 4919904)
636 // (%%% How can a control input be a safepoint and not a projection??)
637 if (mstore->ideal_Opcode() == Op_SafePoint && load->in(0) == mstore)
638 continue;
639 }
640 }
641
642 // Identify a block that the current load must be above,
643 // or else observe that 'store' is all the way up in the
644 // earliest legal block for 'load'. In the latter case,
645 // immediately insert an anti-dependence edge.
646 Block* store_block = get_block_for_node(store);
647 assert(store_block != NULL, "unused killing projections skipped above");
648
649 if (store->is_Phi()) {
650 // 'load' uses memory which is one (or more) of the Phi's inputs.
651 // It must be scheduled not before the Phi, but rather before
652 // each of the relevant Phi inputs.
653 //
654 // Instead of finding the LCA of all inputs to a Phi that match 'mem',
655 // we mark each corresponding predecessor block and do a combined
656 // hoisting operation later (raise_LCA_above_marks).
657 //
658 // Do not assert(store_block != early, "Phi merging memory after access")
659 // PhiNode may be at start of block 'early' with backedge to 'early'
660 DEBUG_ONLY(bool found_match = false);
661 for (uint j = PhiNode::Input, jmax = store->req(); j < jmax; j++) {
662 if (store->in(j) == mem) { // Found matching input?
663 DEBUG_ONLY(found_match = true);
664 Block* pred_block = get_block_for_node(store_block->pred(j));
665 if (pred_block != early) {
666 // If any predecessor of the Phi matches the load's "early block",
667 // we do not need a precedence edge between the Phi and 'load'
668 // since the load will be forced into a block preceding the Phi.
669 pred_block->set_raise_LCA_mark(load_index);
670 assert(!LCA_orig->dominates(pred_block) ||
671 early->dominates(pred_block), "early is high enough");
672 must_raise_LCA = true;
673 } else {
674 // anti-dependent upon PHI pinned below 'early', no edge needed
675 LCA = early; // but can not schedule below 'early'
676 }
677 }
678 }
679 assert(found_match, "no worklist bug");
680 #ifdef TRACK_PHI_INPUTS
681 #ifdef ASSERT
682 // This assert asks about correct handling of PhiNodes, which may not
683 // have all input edges directly from 'mem'. See BugId 4621264
684 int num_mem_inputs = phi_inputs.at_grow(store->_idx,0) + 1;
685 // Increment by exactly one even if there are multiple copies of 'mem'
686 // coming into the phi, because we will run this block several times
687 // if there are several copies of 'mem'. (That's how DU iterators work.)
688 phi_inputs.at_put(store->_idx, num_mem_inputs);
689 assert(PhiNode::Input + num_mem_inputs < store->req(),
690 "Expect at least one phi input will not be from original memory state");
691 #endif //ASSERT
692 #endif //TRACK_PHI_INPUTS
693 } else if (store_block != early) {
694 // 'store' is between the current LCA and earliest possible block.
695 // Label its block, and decide later on how to raise the LCA
696 // to include the effect on LCA of this store.
697 // If this store's block gets chosen as the raised LCA, we
698 // will find him on the non_early_stores list and stick him
699 // with a precedence edge.
700 // (But, don't bother if LCA is already raised all the way.)
701 if (LCA != early) {
702 store_block->set_raise_LCA_mark(load_index);
703 must_raise_LCA = true;
704 non_early_stores.push(store);
705 }
706 } else {
707 // Found a possibly-interfering store in the load's 'early' block.
708 // This means 'load' cannot sink at all in the dominator tree.
709 // Add an anti-dep edge, and squeeze 'load' into the highest block.
710 assert(store != load->in(0), "dependence cycle found");
711 if (verify) {
712 assert(store->find_edge(load) != -1, "missing precedence edge");
713 } else {
714 store->add_prec(load);
715 }
716 LCA = early;
717 // This turns off the process of gathering non_early_stores.
718 }
719 }
720 // (Worklist is now empty; all nearby stores have been visited.)
721
722 // Finished if 'load' must be scheduled in its 'early' block.
723 // If we found any stores there, they have already been given
724 // precedence edges.
725 if (LCA == early) return LCA;
726
727 // We get here only if there are no possibly-interfering stores
728 // in the load's 'early' block. Move LCA up above all predecessors
729 // which contain stores we have noted.
730 //
731 // The raised LCA block can be a home to such interfering stores,
732 // but its predecessors must not contain any such stores.
733 //
734 // The raised LCA will be a lower bound for placing the load,
735 // preventing the load from sinking past any block containing
736 // a store that may invalidate the memory state required by 'load'.
737 if (must_raise_LCA)
738 LCA = raise_LCA_above_marks(LCA, load->_idx, early, this);
739 if (LCA == early) return LCA;
740
741 // Insert anti-dependence edges from 'load' to each store
742 // in the non-early LCA block.
743 // Mine the non_early_stores list for such stores.
744 if (LCA->raise_LCA_mark() == load_index) {
745 while (non_early_stores.size() > 0) {
746 Node* store = non_early_stores.pop();
747 Block* store_block = get_block_for_node(store);
748 if (store_block == LCA) {
749 // add anti_dependence from store to load in its own block
750 assert(store != load->in(0), "dependence cycle found");
751 if (verify) {
752 assert(store->find_edge(load) != -1, "missing precedence edge");
753 } else {
754 store->add_prec(load);
755 }
756 } else {
757 assert(store_block->raise_LCA_mark() == load_index, "block was marked");
758 // Any other stores we found must be either inside the new LCA
759 // or else outside the original LCA. In the latter case, they
760 // did not interfere with any use of 'load'.
761 assert(LCA->dominates(store_block)
762 || !LCA_orig->dominates(store_block), "no stray stores");
763 }
764 }
765 }
766
767 // Return the highest block containing stores; any stores
768 // within that block have been given anti-dependence edges.
769 return LCA;
770 }
771
772 // This class is used to iterate backwards over the nodes in the graph.
773
774 class Node_Backward_Iterator {
775
776 private:
777 Node_Backward_Iterator();
778
779 public:
780 // Constructor for the iterator
781 Node_Backward_Iterator(Node *root, VectorSet &visited, Node_List &stack, PhaseCFG &cfg);
782
783 // Postincrement operator to iterate over the nodes
784 Node *next();
785
786 private:
787 VectorSet &_visited;
788 Node_List &_stack;
789 PhaseCFG &_cfg;
790 };
791
792 // Constructor for the Node_Backward_Iterator
793 Node_Backward_Iterator::Node_Backward_Iterator( Node *root, VectorSet &visited, Node_List &stack, PhaseCFG &cfg)
794 : _visited(visited), _stack(stack), _cfg(cfg) {
795 // The stack should contain exactly the root
796 stack.clear();
797 stack.push(root);
798
799 // Clear the visited bits
800 visited.Clear();
801 }
802
803 // Iterator for the Node_Backward_Iterator
804 Node *Node_Backward_Iterator::next() {
805
806 // If the _stack is empty, then just return NULL: finished.
807 if ( !_stack.size() )
808 return NULL;
809
810 // '_stack' is emulating a real _stack. The 'visit-all-users' loop has been
811 // made stateless, so I do not need to record the index 'i' on my _stack.
812 // Instead I visit all users each time, scanning for unvisited users.
813 // I visit unvisited not-anti-dependence users first, then anti-dependent
814 // children next.
815 Node *self = _stack.pop();
816
817 // I cycle here when I am entering a deeper level of recursion.
818 // The key variable 'self' was set prior to jumping here.
819 while( 1 ) {
820
821 _visited.set(self->_idx);
822
823 // Now schedule all uses as late as possible.
824 const Node* src = self->is_Proj() ? self->in(0) : self;
825 uint src_rpo = _cfg.get_block_for_node(src)->_rpo;
826
827 // Schedule all nodes in a post-order visit
828 Node *unvisited = NULL; // Unvisited anti-dependent Node, if any
829
830 // Scan for unvisited nodes
831 for (DUIterator_Fast imax, i = self->fast_outs(imax); i < imax; i++) {
832 // For all uses, schedule late
833 Node* n = self->fast_out(i); // Use
834
835 // Skip already visited children
836 if ( _visited.test(n->_idx) )
837 continue;
838
839 // do not traverse backward control edges
840 Node *use = n->is_Proj() ? n->in(0) : n;
841 uint use_rpo = _cfg.get_block_for_node(use)->_rpo;
842
843 if ( use_rpo < src_rpo )
844 continue;
845
846 // Phi nodes always precede uses in a basic block
847 if ( use_rpo == src_rpo && use->is_Phi() )
848 continue;
849
850 unvisited = n; // Found unvisited
851
852 // Check for possible-anti-dependent
853 if( !n->needs_anti_dependence_check() )
854 break; // Not visited, not anti-dep; schedule it NOW
855 }
856
857 // Did I find an unvisited not-anti-dependent Node?
858 if ( !unvisited )
859 break; // All done with children; post-visit 'self'
860
861 // Visit the unvisited Node. Contains the obvious push to
862 // indicate I'm entering a deeper level of recursion. I push the
863 // old state onto the _stack and set a new state and loop (recurse).
864 _stack.push(self);
865 self = unvisited;
866 } // End recursion loop
867
868 return self;
869 }
870
871 //------------------------------ComputeLatenciesBackwards----------------------
872 // Compute the latency of all the instructions.
873 void PhaseCFG::compute_latencies_backwards(VectorSet &visited, Node_List &stack) {
874 #ifndef PRODUCT
875 if (trace_opto_pipelining())
876 tty->print("\n#---- ComputeLatenciesBackwards ----\n");
877 #endif
878
879 Node_Backward_Iterator iter((Node *)_root, visited, stack, *this);
880 Node *n;
881
882 // Walk over all the nodes from last to first
883 while (n = iter.next()) {
884 // Set the latency for the definitions of this instruction
885 partial_latency_of_defs(n);
886 }
887 } // end ComputeLatenciesBackwards
888
889 //------------------------------partial_latency_of_defs------------------------
890 // Compute the latency impact of this node on all defs. This computes
891 // a number that increases as we approach the beginning of the routine.
892 void PhaseCFG::partial_latency_of_defs(Node *n) {
893 // Set the latency for this instruction
894 #ifndef PRODUCT
895 if (trace_opto_pipelining()) {
896 tty->print("# latency_to_inputs: node_latency[%d] = %d for node", n->_idx, get_latency_for_node(n));
897 dump();
898 }
899 #endif
900
901 if (n->is_Proj()) {
902 n = n->in(0);
903 }
904
905 if (n->is_Root()) {
906 return;
907 }
908
909 uint nlen = n->len();
910 uint use_latency = get_latency_for_node(n);
911 uint use_pre_order = get_block_for_node(n)->_pre_order;
912
913 for (uint j = 0; j < nlen; j++) {
914 Node *def = n->in(j);
915
916 if (!def || def == n) {
917 continue;
918 }
919
920 // Walk backwards thru projections
921 if (def->is_Proj()) {
922 def = def->in(0);
923 }
924
925 #ifndef PRODUCT
926 if (trace_opto_pipelining()) {
927 tty->print("# in(%2d): ", j);
928 def->dump();
929 }
930 #endif
931
932 // If the defining block is not known, assume it is ok
933 Block *def_block = get_block_for_node(def);
934 uint def_pre_order = def_block ? def_block->_pre_order : 0;
935
936 if ((use_pre_order < def_pre_order) || (use_pre_order == def_pre_order && n->is_Phi())) {
937 continue;
938 }
939
940 uint delta_latency = n->latency(j);
941 uint current_latency = delta_latency + use_latency;
942
943 if (get_latency_for_node(def) < current_latency) {
944 set_latency_for_node(def, current_latency);
945 }
946
947 #ifndef PRODUCT
948 if (trace_opto_pipelining()) {
949 tty->print_cr("# %d + edge_latency(%d) == %d -> %d, node_latency[%d] = %d", use_latency, j, delta_latency, current_latency, def->_idx, get_latency_for_node(def));
950 }
951 #endif
952 }
953 }
954
955 //------------------------------latency_from_use-------------------------------
956 // Compute the latency of a specific use
957 int PhaseCFG::latency_from_use(Node *n, const Node *def, Node *use) {
958 // If self-reference, return no latency
959 if (use == n || use->is_Root()) {
960 return 0;
961 }
962
963 uint def_pre_order = get_block_for_node(def)->_pre_order;
964 uint latency = 0;
965
966 // If the use is not a projection, then it is simple...
967 if (!use->is_Proj()) {
968 #ifndef PRODUCT
969 if (trace_opto_pipelining()) {
970 tty->print("# out(): ");
971 use->dump();
972 }
973 #endif
974
975 uint use_pre_order = get_block_for_node(use)->_pre_order;
976
977 if (use_pre_order < def_pre_order)
978 return 0;
979
980 if (use_pre_order == def_pre_order && use->is_Phi())
981 return 0;
982
983 uint nlen = use->len();
984 uint nl = get_latency_for_node(use);
985
986 for ( uint j=0; j<nlen; j++ ) {
987 if (use->in(j) == n) {
988 // Change this if we want local latencies
989 uint ul = use->latency(j);
990 uint l = ul + nl;
991 if (latency < l) latency = l;
992 #ifndef PRODUCT
993 if (trace_opto_pipelining()) {
994 tty->print_cr("# %d + edge_latency(%d) == %d -> %d, latency = %d",
995 nl, j, ul, l, latency);
996 }
997 #endif
998 }
999 }
1000 } else {
1001 // This is a projection, just grab the latency of the use(s)
1002 for (DUIterator_Fast jmax, j = use->fast_outs(jmax); j < jmax; j++) {
1003 uint l = latency_from_use(use, def, use->fast_out(j));
1004 if (latency < l) latency = l;
1005 }
1006 }
1007
1008 return latency;
1009 }
1010
1011 //------------------------------latency_from_uses------------------------------
1012 // Compute the latency of this instruction relative to all of it's uses.
1013 // This computes a number that increases as we approach the beginning of the
1014 // routine.
1015 void PhaseCFG::latency_from_uses(Node *n) {
1016 // Set the latency for this instruction
1017 #ifndef PRODUCT
1018 if (trace_opto_pipelining()) {
1019 tty->print("# latency_from_outputs: node_latency[%d] = %d for node", n->_idx, get_latency_for_node(n));
1020 dump();
1021 }
1022 #endif
1023 uint latency=0;
1024 const Node *def = n->is_Proj() ? n->in(0): n;
1025
1026 for (DUIterator_Fast imax, i = n->fast_outs(imax); i < imax; i++) {
1027 uint l = latency_from_use(n, def, n->fast_out(i));
1028
1029 if (latency < l) latency = l;
1030 }
1031
1032 set_latency_for_node(n, latency);
1033 }
1034
1035 //------------------------------hoist_to_cheaper_block-------------------------
1036 // Pick a block for node self, between early and LCA, that is a cheaper
1037 // alternative to LCA.
1038 Block* PhaseCFG::hoist_to_cheaper_block(Block* LCA, Block* early, Node* self) {
1039 const double delta = 1+PROB_UNLIKELY_MAG(4);
1040 Block* least = LCA;
1041 double least_freq = least->_freq;
1042 uint target = get_latency_for_node(self);
1043 uint start_latency = get_latency_for_node(LCA->head());
1044 uint end_latency = get_latency_for_node(LCA->get_node(LCA->end_idx()));
1045 bool in_latency = (target <= start_latency);
1046 const Block* root_block = get_block_for_node(_root);
1047
1048 // Turn off latency scheduling if scheduling is just plain off
1049 if (!C->do_scheduling())
1050 in_latency = true;
1051
1052 // Do not hoist (to cover latency) instructions which target a
1053 // single register. Hoisting stretches the live range of the
1054 // single register and may force spilling.
1055 MachNode* mach = self->is_Mach() ? self->as_Mach() : NULL;
1056 if (mach && mach->out_RegMask().is_bound1() && mach->out_RegMask().is_NotEmpty())
1057 in_latency = true;
1058
1059 #ifndef PRODUCT
1060 if (trace_opto_pipelining()) {
1061 tty->print("# Find cheaper block for latency %d: ", get_latency_for_node(self));
1062 self->dump();
1063 tty->print_cr("# B%d: start latency for [%4d]=%d, end latency for [%4d]=%d, freq=%g",
1064 LCA->_pre_order,
1065 LCA->head()->_idx,
1066 start_latency,
1067 LCA->get_node(LCA->end_idx())->_idx,
1068 end_latency,
1069 least_freq);
1070 }
1071 #endif
1072
1073 int cand_cnt = 0; // number of candidates tried
1074
1075 // Walk up the dominator tree from LCA (Lowest common ancestor) to
1076 // the earliest legal location. Capture the least execution frequency.
1077 while (LCA != early) {
1078 LCA = LCA->_idom; // Follow up the dominator tree
1079
1080 if (LCA == NULL) {
1081 // Bailout without retry
1082 C->record_method_not_compilable("late schedule failed: LCA == NULL");
1083 return least;
1084 }
1085
1086 // Don't hoist machine instructions to the root basic block
1087 if (mach && LCA == root_block)
1088 break;
1089
1090 uint start_lat = get_latency_for_node(LCA->head());
1091 uint end_idx = LCA->end_idx();
1092 uint end_lat = get_latency_for_node(LCA->get_node(end_idx));
1093 double LCA_freq = LCA->_freq;
1094 #ifndef PRODUCT
1095 if (trace_opto_pipelining()) {
1096 tty->print_cr("# B%d: start latency for [%4d]=%d, end latency for [%4d]=%d, freq=%g",
1097 LCA->_pre_order, LCA->head()->_idx, start_lat, end_idx, end_lat, LCA_freq);
1098 }
1099 #endif
1100 cand_cnt++;
1101 if (LCA_freq < least_freq || // Better Frequency
1102 (StressGCM && Compile::randomized_select(cand_cnt)) || // Should be randomly accepted in stress mode
1103 (!StressGCM && // Otherwise, choose with latency
1104 !in_latency && // No block containing latency
1105 LCA_freq < least_freq * delta && // No worse frequency
1106 target >= end_lat && // within latency range
1107 !self->is_iteratively_computed() ) // But don't hoist IV increments
1108 // because they may end up above other uses of their phi forcing
1109 // their result register to be different from their input.
1110 ) {
1111 least = LCA; // Found cheaper block
1112 least_freq = LCA_freq;
1113 start_latency = start_lat;
1114 end_latency = end_lat;
1115 if (target <= start_lat)
1116 in_latency = true;
1117 }
1118 }
1119
1120 #ifndef PRODUCT
1121 if (trace_opto_pipelining()) {
1122 tty->print_cr("# Choose block B%d with start latency=%d and freq=%g",
1123 least->_pre_order, start_latency, least_freq);
1124 }
1125 #endif
1126
1127 // See if the latency needs to be updated
1128 if (target < end_latency) {
1129 #ifndef PRODUCT
1130 if (trace_opto_pipelining()) {
1131 tty->print_cr("# Change latency for [%4d] from %d to %d", self->_idx, target, end_latency);
1132 }
1133 #endif
1134 set_latency_for_node(self, end_latency);
1135 partial_latency_of_defs(self);
1136 }
1137
1138 return least;
1139 }
1140
1141
1142 //------------------------------schedule_late-----------------------------------
1143 // Now schedule all codes as LATE as possible. This is the LCA in the
1144 // dominator tree of all USES of a value. Pick the block with the least
1145 // loop nesting depth that is lowest in the dominator tree.
1146 extern const char must_clone[];
1147 void PhaseCFG::schedule_late(VectorSet &visited, Node_List &stack) {
1148 #ifndef PRODUCT
1149 if (trace_opto_pipelining())
1150 tty->print("\n#---- schedule_late ----\n");
1151 #endif
1152
1153 Node_Backward_Iterator iter((Node *)_root, visited, stack, *this);
1154 Node *self;
1155
1156 // Walk over all the nodes from last to first
1157 while (self = iter.next()) {
1158 Block* early = get_block_for_node(self); // Earliest legal placement
1159
1160 if (self->is_top()) {
1161 // Top node goes in bb #2 with other constants.
1162 // It must be special-cased, because it has no out edges.
1163 early->add_inst(self);
1164 continue;
1165 }
1166
1167 // No uses, just terminate
1168 if (self->outcnt() == 0) {
1169 assert(self->is_MachProj(), "sanity");
1170 continue; // Must be a dead machine projection
1171 }
1172
1173 // If node is pinned in the block, then no scheduling can be done.
1174 if( self->pinned() ) // Pinned in block?
1175 continue;
1176
1177 MachNode* mach = self->is_Mach() ? self->as_Mach() : NULL;
1178 if (mach) {
1179 switch (mach->ideal_Opcode()) {
1180 case Op_CreateEx:
1181 // Don't move exception creation
1182 early->add_inst(self);
1183 continue;
1184 break;
1185 case Op_CheckCastPP:
1186 // Don't move CheckCastPP nodes away from their input, if the input
1187 // is a rawptr (5071820).
1188 Node *def = self->in(1);
1189 if (def != NULL && def->bottom_type()->base() == Type::RawPtr) {
1190 early->add_inst(self);
1191 #ifdef ASSERT
1192 _raw_oops.push(def);
1193 #endif
1194 continue;
1195 }
1196 break;
1197 }
1198 }
1199
1200 // Gather LCA of all uses
1201 Block *LCA = NULL;
1202 {
1203 for (DUIterator_Fast imax, i = self->fast_outs(imax); i < imax; i++) {
1204 // For all uses, find LCA
1205 Node* use = self->fast_out(i);
1206 LCA = raise_LCA_above_use(LCA, use, self, this);
1207 }
1208 } // (Hide defs of imax, i from rest of block.)
1209
1210 // Place temps in the block of their use. This isn't a
1211 // requirement for correctness but it reduces useless
1212 // interference between temps and other nodes.
1213 if (mach != NULL && mach->is_MachTemp()) {
1214 map_node_to_block(self, LCA);
1215 LCA->add_inst(self);
1216 continue;
1217 }
1218
1219 // Check if 'self' could be anti-dependent on memory
1220 if (self->needs_anti_dependence_check()) {
1221 // Hoist LCA above possible-defs and insert anti-dependences to
1222 // defs in new LCA block.
1223 LCA = insert_anti_dependences(LCA, self);
1224 }
1225
1226 if (early->_dom_depth > LCA->_dom_depth) {
1227 // Somehow the LCA has moved above the earliest legal point.
1228 // (One way this can happen is via memory_early_block.)
1229 if (C->subsume_loads() == true && !C->failing()) {
1230 // Retry with subsume_loads == false
1231 // If this is the first failure, the sentinel string will "stick"
1232 // to the Compile object, and the C2Compiler will see it and retry.
1233 C->record_failure(C2Compiler::retry_no_subsuming_loads());
1234 } else {
1235 // Bailout without retry when (early->_dom_depth > LCA->_dom_depth)
1236 C->record_method_not_compilable("late schedule failed: incorrect graph");
1237 }
1238 return;
1239 }
1240
1241 // If there is no opportunity to hoist, then we're done.
1242 // In stress mode, try to hoist even the single operations.
1243 bool try_to_hoist = StressGCM || (LCA != early);
1244
1245 // Must clone guys stay next to use; no hoisting allowed.
1246 // Also cannot hoist guys that alter memory or are otherwise not
1247 // allocatable (hoisting can make a value live longer, leading to
1248 // anti and output dependency problems which are normally resolved
1249 // by the register allocator giving everyone a different register).
1250 if (mach != NULL && must_clone[mach->ideal_Opcode()])
1251 try_to_hoist = false;
1252
1253 Block* late = NULL;
1254 if (try_to_hoist) {
1255 // Now find the block with the least execution frequency.
1256 // Start at the latest schedule and work up to the earliest schedule
1257 // in the dominator tree. Thus the Node will dominate all its uses.
1258 late = hoist_to_cheaper_block(LCA, early, self);
1259 } else {
1260 // Just use the LCA of the uses.
1261 late = LCA;
1262 }
1263
1264 // Put the node into target block
1265 schedule_node_into_block(self, late);
1266
1267 #ifdef ASSERT
1268 if (self->needs_anti_dependence_check()) {
1269 // since precedence edges are only inserted when we're sure they
1270 // are needed make sure that after placement in a block we don't
1271 // need any new precedence edges.
1272 verify_anti_dependences(late, self);
1273 }
1274 #endif
1275 } // Loop until all nodes have been visited
1276
1277 } // end ScheduleLate
1278
1279 //------------------------------GlobalCodeMotion-------------------------------
1280 void PhaseCFG::global_code_motion() {
1281 ResourceMark rm;
1282
1283 #ifndef PRODUCT
1284 if (trace_opto_pipelining()) {
1285 tty->print("\n---- Start GlobalCodeMotion ----\n");
1286 }
1287 #endif
1288
1289 // Initialize the node to block mapping for things on the proj_list
1290 for (uint i = 0; i < _matcher.number_of_projections(); i++) {
1291 unmap_node_from_block(_matcher.get_projection(i));
1292 }
1293
1294 // Set the basic block for Nodes pinned into blocks
1295 Arena* arena = Thread::current()->resource_area();
1296 VectorSet visited(arena);
1297 schedule_pinned_nodes(visited);
1298
1299 // Find the earliest Block any instruction can be placed in. Some
1300 // instructions are pinned into Blocks. Unpinned instructions can
1301 // appear in last block in which all their inputs occur.
1302 visited.Clear();
1303 Node_List stack(arena);
1304 // Pre-grow the list
1305 stack.map((C->unique() >> 1) + 16, NULL);
1306 if (!schedule_early(visited, stack)) {
1307 // Bailout without retry
1308 C->record_method_not_compilable("early schedule failed");
1309 return;
1310 }
1311
1312 // Build Def-Use edges.
1313 // Compute the latency information (via backwards walk) for all the
1314 // instructions in the graph
1315 _node_latency = new GrowableArray<uint>(); // resource_area allocation
1316
1317 if (C->do_scheduling()) {
1318 compute_latencies_backwards(visited, stack);
1319 }
1320
1321 // Now schedule all codes as LATE as possible. This is the LCA in the
1322 // dominator tree of all USES of a value. Pick the block with the least
1323 // loop nesting depth that is lowest in the dominator tree.
1324 // ( visited.Clear() called in schedule_late()->Node_Backward_Iterator() )
1325 schedule_late(visited, stack);
1326 if (C->failing()) {
1327 // schedule_late fails only when graph is incorrect.
1328 assert(!VerifyGraphEdges, "verification should have failed");
1329 return;
1330 }
1331
1332 #ifndef PRODUCT
1333 if (trace_opto_pipelining()) {
1334 tty->print("\n---- Detect implicit null checks ----\n");
1335 }
1336 #endif
1337
1338 // Detect implicit-null-check opportunities. Basically, find NULL checks
1339 // with suitable memory ops nearby. Use the memory op to do the NULL check.
1340 // I can generate a memory op if there is not one nearby.
1341 if (C->is_method_compilation()) {
1342 // By reversing the loop direction we get a very minor gain on mpegaudio.
1343 // Feel free to revert to a forward loop for clarity.
1344 // for( int i=0; i < (int)matcher._null_check_tests.size(); i+=2 ) {
1345 for (int i = _matcher._null_check_tests.size() - 2; i >= 0; i -= 2) {
1346 Node* proj = _matcher._null_check_tests[i];
1347 Node* val = _matcher._null_check_tests[i + 1];
1348 Block* block = get_block_for_node(proj);
1349 implicit_null_check(block, proj, val, C->allowed_deopt_reasons());
1350 // The implicit_null_check will only perform the transformation
1351 // if the null branch is truly uncommon, *and* it leads to an
1352 // uncommon trap. Combined with the too_many_traps guards
1353 // above, this prevents SEGV storms reported in 6366351,
1354 // by recompiling offending methods without this optimization.
1355 }
1356 }
1357
1358 #ifndef PRODUCT
1359 if (trace_opto_pipelining()) {
1360 tty->print("\n---- Start Local Scheduling ----\n");
1361 }
1362 #endif
1363
1364 // Schedule locally. Right now a simple topological sort.
1365 // Later, do a real latency aware scheduler.
1366 GrowableArray<int> ready_cnt(C->unique(), C->unique(), -1);
1367 visited.Clear();
1368 for (uint i = 0; i < number_of_blocks(); i++) {
1369 Block* block = get_block(i);
1370 if (!schedule_local(block, ready_cnt, visited)) {
1371 if (!C->failure_reason_is(C2Compiler::retry_no_subsuming_loads())) {
1372 C->record_method_not_compilable("local schedule failed");
1373 }
1374 return;
1375 }
1376 }
1377
1378 // If we inserted any instructions between a Call and his CatchNode,
1379 // clone the instructions on all paths below the Catch.
1380 for (uint i = 0; i < number_of_blocks(); i++) {
1381 Block* block = get_block(i);
1382 call_catch_cleanup(block);
1383 }
1384
1385 #ifndef PRODUCT
1386 if (trace_opto_pipelining()) {
1387 tty->print("\n---- After GlobalCodeMotion ----\n");
1388 for (uint i = 0; i < number_of_blocks(); i++) {
1389 Block* block = get_block(i);
1390 block->dump();
1391 }
1392 }
1393 #endif
1394 // Dead.
1395 _node_latency = (GrowableArray<uint> *)0xdeadbeef;
1396 }
1397
1398 bool PhaseCFG::do_global_code_motion() {
1399
1400 build_dominator_tree();
1401 if (C->failing()) {
1402 return false;
1403 }
1404
1405 NOT_PRODUCT( C->verify_graph_edges(); )
1406
1407 estimate_block_frequency();
1408
1409 global_code_motion();
1410
1411 if (C->failing()) {
1412 return false;
1413 }
1414
1415 return true;
1416 }
1417
1418 //------------------------------Estimate_Block_Frequency-----------------------
1419 // Estimate block frequencies based on IfNode probabilities.
1420 void PhaseCFG::estimate_block_frequency() {
1421
1422 // Force conditional branches leading to uncommon traps to be unlikely,
1423 // not because we get to the uncommon_trap with less relative frequency,
1424 // but because an uncommon_trap typically causes a deopt, so we only get
1425 // there once.
1426 if (C->do_freq_based_layout()) {
1427 Block_List worklist;
1428 Block* root_blk = get_block(0);
1429 for (uint i = 1; i < root_blk->num_preds(); i++) {
1430 Block *pb = get_block_for_node(root_blk->pred(i));
1431 if (pb->has_uncommon_code()) {
1432 worklist.push(pb);
1433 }
1434 }
1435 while (worklist.size() > 0) {
1436 Block* uct = worklist.pop();
1437 if (uct == get_root_block()) {
1438 continue;
1439 }
1440 for (uint i = 1; i < uct->num_preds(); i++) {
1441 Block *pb = get_block_for_node(uct->pred(i));
1442 if (pb->_num_succs == 1) {
1443 worklist.push(pb);
1444 } else if (pb->num_fall_throughs() == 2) {
1445 pb->update_uncommon_branch(uct);
1446 }
1447 }
1448 }
1449 }
1450
1451 // Create the loop tree and calculate loop depth.
1452 _root_loop = create_loop_tree();
1453 _root_loop->compute_loop_depth(0);
1454
1455 // Compute block frequency of each block, relative to a single loop entry.
1456 _root_loop->compute_freq();
1457
1458 // Adjust all frequencies to be relative to a single method entry
1459 _root_loop->_freq = 1.0;
1460 _root_loop->scale_freq();
1461
1462 // Save outmost loop frequency for LRG frequency threshold
1463 _outer_loop_frequency = _root_loop->outer_loop_freq();
1464
1465 // force paths ending at uncommon traps to be infrequent
1466 if (!C->do_freq_based_layout()) {
1467 Block_List worklist;
1468 Block* root_blk = get_block(0);
1469 for (uint i = 1; i < root_blk->num_preds(); i++) {
1470 Block *pb = get_block_for_node(root_blk->pred(i));
1471 if (pb->has_uncommon_code()) {
1472 worklist.push(pb);
1473 }
1474 }
1475 while (worklist.size() > 0) {
1476 Block* uct = worklist.pop();
1477 uct->_freq = PROB_MIN;
1478 for (uint i = 1; i < uct->num_preds(); i++) {
1479 Block *pb = get_block_for_node(uct->pred(i));
1480 if (pb->_num_succs == 1 && pb->_freq > PROB_MIN) {
1481 worklist.push(pb);
1482 }
1483 }
1484 }
1485 }
1486
1487 #ifdef ASSERT
1488 for (uint i = 0; i < number_of_blocks(); i++) {
1489 Block* b = get_block(i);
1490 assert(b->_freq >= MIN_BLOCK_FREQUENCY, "Register Allocator requires meaningful block frequency");
1491 }
1492 #endif
1493
1494 #ifndef PRODUCT
1495 if (PrintCFGBlockFreq) {
1496 tty->print_cr("CFG Block Frequencies");
1497 _root_loop->dump_tree();
1498 if (Verbose) {
1499 tty->print_cr("PhaseCFG dump");
1500 dump();
1501 tty->print_cr("Node dump");
1502 _root->dump(99999);
1503 }
1504 }
1505 #endif
1506 }
1507
1508 //----------------------------create_loop_tree--------------------------------
1509 // Create a loop tree from the CFG
1510 CFGLoop* PhaseCFG::create_loop_tree() {
1511
1512 #ifdef ASSERT
1513 assert(get_block(0) == get_root_block(), "first block should be root block");
1514 for (uint i = 0; i < number_of_blocks(); i++) {
1515 Block* block = get_block(i);
1516 // Check that _loop field are clear...we could clear them if not.
1517 assert(block->_loop == NULL, "clear _loop expected");
1518 // Sanity check that the RPO numbering is reflected in the _blocks array.
1519 // It doesn't have to be for the loop tree to be built, but if it is not,
1520 // then the blocks have been reordered since dom graph building...which
1521 // may question the RPO numbering
1522 assert(block->_rpo == i, "unexpected reverse post order number");
1523 }
1524 #endif
1525
1526 int idct = 0;
1527 CFGLoop* root_loop = new CFGLoop(idct++);
1528
1529 Block_List worklist;
1530
1531 // Assign blocks to loops
1532 for(uint i = number_of_blocks() - 1; i > 0; i-- ) { // skip Root block
1533 Block* block = get_block(i);
1534
1535 if (block->head()->is_Loop()) {
1536 Block* loop_head = block;
1537 assert(loop_head->num_preds() - 1 == 2, "loop must have 2 predecessors");
1538 Node* tail_n = loop_head->pred(LoopNode::LoopBackControl);
1539 Block* tail = get_block_for_node(tail_n);
1540
1541 // Defensively filter out Loop nodes for non-single-entry loops.
1542 // For all reasonable loops, the head occurs before the tail in RPO.
1543 if (i <= tail->_rpo) {
1544
1545 // The tail and (recursive) predecessors of the tail
1546 // are made members of a new loop.
1547
1548 assert(worklist.size() == 0, "nonempty worklist");
1549 CFGLoop* nloop = new CFGLoop(idct++);
1550 assert(loop_head->_loop == NULL, "just checking");
1551 loop_head->_loop = nloop;
1552 // Add to nloop so push_pred() will skip over inner loops
1553 nloop->add_member(loop_head);
1554 nloop->push_pred(loop_head, LoopNode::LoopBackControl, worklist, this);
1555
1556 while (worklist.size() > 0) {
1557 Block* member = worklist.pop();
1558 if (member != loop_head) {
1559 for (uint j = 1; j < member->num_preds(); j++) {
1560 nloop->push_pred(member, j, worklist, this);
1561 }
1562 }
1563 }
1564 }
1565 }
1566 }
1567
1568 // Create a member list for each loop consisting
1569 // of both blocks and (immediate child) loops.
1570 for (uint i = 0; i < number_of_blocks(); i++) {
1571 Block* block = get_block(i);
1572 CFGLoop* lp = block->_loop;
1573 if (lp == NULL) {
1574 // Not assigned to a loop. Add it to the method's pseudo loop.
1575 block->_loop = root_loop;
1576 lp = root_loop;
1577 }
1578 if (lp == root_loop || block != lp->head()) { // loop heads are already members
1579 lp->add_member(block);
1580 }
1581 if (lp != root_loop) {
1582 if (lp->parent() == NULL) {
1583 // Not a nested loop. Make it a child of the method's pseudo loop.
1584 root_loop->add_nested_loop(lp);
1585 }
1586 if (block == lp->head()) {
1587 // Add nested loop to member list of parent loop.
1588 lp->parent()->add_member(lp);
1589 }
1590 }
1591 }
1592
1593 return root_loop;
1594 }
1595
1596 //------------------------------push_pred--------------------------------------
1597 void CFGLoop::push_pred(Block* blk, int i, Block_List& worklist, PhaseCFG* cfg) {
1598 Node* pred_n = blk->pred(i);
1599 Block* pred = cfg->get_block_for_node(pred_n);
1600 CFGLoop *pred_loop = pred->_loop;
1601 if (pred_loop == NULL) {
1602 // Filter out blocks for non-single-entry loops.
1603 // For all reasonable loops, the head occurs before the tail in RPO.
1604 if (pred->_rpo > head()->_rpo) {
1605 pred->_loop = this;
1606 worklist.push(pred);
1607 }
1608 } else if (pred_loop != this) {
1609 // Nested loop.
1610 while (pred_loop->_parent != NULL && pred_loop->_parent != this) {
1611 pred_loop = pred_loop->_parent;
1612 }
1613 // Make pred's loop be a child
1614 if (pred_loop->_parent == NULL) {
1615 add_nested_loop(pred_loop);
1616 // Continue with loop entry predecessor.
1617 Block* pred_head = pred_loop->head();
1618 assert(pred_head->num_preds() - 1 == 2, "loop must have 2 predecessors");
1619 assert(pred_head != head(), "loop head in only one loop");
1620 push_pred(pred_head, LoopNode::EntryControl, worklist, cfg);
1621 } else {
1622 assert(pred_loop->_parent == this && _parent == NULL, "just checking");
1623 }
1624 }
1625 }
1626
1627 //------------------------------add_nested_loop--------------------------------
1628 // Make cl a child of the current loop in the loop tree.
1629 void CFGLoop::add_nested_loop(CFGLoop* cl) {
1630 assert(_parent == NULL, "no parent yet");
1631 assert(cl != this, "not my own parent");
1632 cl->_parent = this;
1633 CFGLoop* ch = _child;
1634 if (ch == NULL) {
1635 _child = cl;
1636 } else {
1637 while (ch->_sibling != NULL) { ch = ch->_sibling; }
1638 ch->_sibling = cl;
1639 }
1640 }
1641
1642 //------------------------------compute_loop_depth-----------------------------
1643 // Store the loop depth in each CFGLoop object.
1644 // Recursively walk the children to do the same for them.
1645 void CFGLoop::compute_loop_depth(int depth) {
1646 _depth = depth;
1647 CFGLoop* ch = _child;
1648 while (ch != NULL) {
1649 ch->compute_loop_depth(depth + 1);
1650 ch = ch->_sibling;
1651 }
1652 }
1653
1654 //------------------------------compute_freq-----------------------------------
1655 // Compute the frequency of each block and loop, relative to a single entry
1656 // into the dominating loop head.
1657 void CFGLoop::compute_freq() {
1658 // Bottom up traversal of loop tree (visit inner loops first.)
1659 // Set loop head frequency to 1.0, then transitively
1660 // compute frequency for all successors in the loop,
1661 // as well as for each exit edge. Inner loops are
1662 // treated as single blocks with loop exit targets
1663 // as the successor blocks.
1664
1665 // Nested loops first
1666 CFGLoop* ch = _child;
1667 while (ch != NULL) {
1668 ch->compute_freq();
1669 ch = ch->_sibling;
1670 }
1671 assert (_members.length() > 0, "no empty loops");
1672 Block* hd = head();
1673 hd->_freq = 1.0;
1674 for (int i = 0; i < _members.length(); i++) {
1675 CFGElement* s = _members.at(i);
1676 double freq = s->_freq;
1677 if (s->is_block()) {
1678 Block* b = s->as_Block();
1679 for (uint j = 0; j < b->_num_succs; j++) {
1680 Block* sb = b->_succs[j];
1681 update_succ_freq(sb, freq * b->succ_prob(j));
1682 }
1683 } else {
1684 CFGLoop* lp = s->as_CFGLoop();
1685 assert(lp->_parent == this, "immediate child");
1686 for (int k = 0; k < lp->_exits.length(); k++) {
1687 Block* eb = lp->_exits.at(k).get_target();
1688 double prob = lp->_exits.at(k).get_prob();
1689 update_succ_freq(eb, freq * prob);
1690 }
1691 }
1692 }
1693
1694 // For all loops other than the outer, "method" loop,
1695 // sum and normalize the exit probability. The "method" loop
1696 // should keep the initial exit probability of 1, so that
1697 // inner blocks do not get erroneously scaled.
1698 if (_depth != 0) {
1699 // Total the exit probabilities for this loop.
1700 double exits_sum = 0.0f;
1701 for (int i = 0; i < _exits.length(); i++) {
1702 exits_sum += _exits.at(i).get_prob();
1703 }
1704
1705 // Normalize the exit probabilities. Until now, the
1706 // probabilities estimate the possibility of exit per
1707 // a single loop iteration; afterward, they estimate
1708 // the probability of exit per loop entry.
1709 for (int i = 0; i < _exits.length(); i++) {
1710 Block* et = _exits.at(i).get_target();
1711 float new_prob = 0.0f;
1712 if (_exits.at(i).get_prob() > 0.0f) {
1713 new_prob = _exits.at(i).get_prob() / exits_sum;
1714 }
1715 BlockProbPair bpp(et, new_prob);
1716 _exits.at_put(i, bpp);
1717 }
1718
1719 // Save the total, but guard against unreasonable probability,
1720 // as the value is used to estimate the loop trip count.
1721 // An infinite trip count would blur relative block
1722 // frequencies.
1723 if (exits_sum > 1.0f) exits_sum = 1.0;
1724 if (exits_sum < PROB_MIN) exits_sum = PROB_MIN;
1725 _exit_prob = exits_sum;
1726 }
1727 }
1728
1729 //------------------------------succ_prob-------------------------------------
1730 // Determine the probability of reaching successor 'i' from the receiver block.
1731 float Block::succ_prob(uint i) {
1732 int eidx = end_idx();
1733 Node *n = get_node(eidx); // Get ending Node
1734
1735 int op = n->Opcode();
1736 if (n->is_Mach()) {
1737 if (n->is_MachNullCheck()) {
1738 // Can only reach here if called after lcm. The original Op_If is gone,
1739 // so we attempt to infer the probability from one or both of the
1740 // successor blocks.
1741 assert(_num_succs == 2, "expecting 2 successors of a null check");
1742 // If either successor has only one predecessor, then the
1743 // probability estimate can be derived using the
1744 // relative frequency of the successor and this block.
1745 if (_succs[i]->num_preds() == 2) {
1746 return _succs[i]->_freq / _freq;
1747 } else if (_succs[1-i]->num_preds() == 2) {
1748 return 1 - (_succs[1-i]->_freq / _freq);
1749 } else {
1750 // Estimate using both successor frequencies
1751 float freq = _succs[i]->_freq;
1752 return freq / (freq + _succs[1-i]->_freq);
1753 }
1754 }
1755 op = n->as_Mach()->ideal_Opcode();
1756 }
1757
1758
1759 // Switch on branch type
1760 switch( op ) {
1761 case Op_CountedLoopEnd:
1762 case Op_If: {
1763 assert (i < 2, "just checking");
1764 // Conditionals pass on only part of their frequency
1765 float prob = n->as_MachIf()->_prob;
1766 assert(prob >= 0.0 && prob <= 1.0, "out of range probability");
1767 // If succ[i] is the FALSE branch, invert path info
1768 if( get_node(i + eidx + 1)->Opcode() == Op_IfFalse ) {
1769 return 1.0f - prob; // not taken
1770 } else {
1771 return prob; // taken
1772 }
1773 }
1774
1775 case Op_Jump:
1776 // Divide the frequency between all successors evenly
1777 return 1.0f/_num_succs;
1778
1779 case Op_Catch: {
1780 const CatchProjNode *ci = get_node(i + eidx + 1)->as_CatchProj();
1781 if (ci->_con == CatchProjNode::fall_through_index) {
1782 // Fall-thru path gets the lion's share.
1783 return 1.0f - PROB_UNLIKELY_MAG(5)*_num_succs;
1784 } else {
1785 // Presume exceptional paths are equally unlikely
1786 return PROB_UNLIKELY_MAG(5);
1787 }
1788 }
1789
1790 case Op_Root:
1791 case Op_Goto:
1792 // Pass frequency straight thru to target
1793 return 1.0f;
1794
1795 case Op_NeverBranch:
1796 return 0.0f;
1797
1798 case Op_TailCall:
1799 case Op_TailJump:
1800 case Op_Return:
1801 case Op_Halt:
1802 case Op_Rethrow:
1803 // Do not push out freq to root block
1804 return 0.0f;
1805
1806 default:
1807 ShouldNotReachHere();
1808 }
1809
1810 return 0.0f;
1811 }
1812
1813 //------------------------------num_fall_throughs-----------------------------
1814 // Return the number of fall-through candidates for a block
1815 int Block::num_fall_throughs() {
1816 int eidx = end_idx();
1817 Node *n = get_node(eidx); // Get ending Node
1818
1819 int op = n->Opcode();
1820 if (n->is_Mach()) {
1821 if (n->is_MachNullCheck()) {
1822 // In theory, either side can fall-thru, for simplicity sake,
1823 // let's say only the false branch can now.
1824 return 1;
1825 }
1826 op = n->as_Mach()->ideal_Opcode();
1827 }
1828
1829 // Switch on branch type
1830 switch( op ) {
1831 case Op_CountedLoopEnd:
1832 case Op_If:
1833 return 2;
1834
1835 case Op_Root:
1836 case Op_Goto:
1837 return 1;
1838
1839 case Op_Catch: {
1840 for (uint i = 0; i < _num_succs; i++) {
1841 const CatchProjNode *ci = get_node(i + eidx + 1)->as_CatchProj();
1842 if (ci->_con == CatchProjNode::fall_through_index) {
1843 return 1;
1844 }
1845 }
1846 return 0;
1847 }
1848
1849 case Op_Jump:
1850 case Op_NeverBranch:
1851 case Op_TailCall:
1852 case Op_TailJump:
1853 case Op_Return:
1854 case Op_Halt:
1855 case Op_Rethrow:
1856 return 0;
1857
1858 default:
1859 ShouldNotReachHere();
1860 }
1861
1862 return 0;
1863 }
1864
1865 //------------------------------succ_fall_through-----------------------------
1866 // Return true if a specific successor could be fall-through target.
1867 bool Block::succ_fall_through(uint i) {
1868 int eidx = end_idx();
1869 Node *n = get_node(eidx); // Get ending Node
1870
1871 int op = n->Opcode();
1872 if (n->is_Mach()) {
1873 if (n->is_MachNullCheck()) {
1874 // In theory, either side can fall-thru, for simplicity sake,
1875 // let's say only the false branch can now.
1876 return get_node(i + eidx + 1)->Opcode() == Op_IfFalse;
1877 }
1878 op = n->as_Mach()->ideal_Opcode();
1879 }
1880
1881 // Switch on branch type
1882 switch( op ) {
1883 case Op_CountedLoopEnd:
1884 case Op_If:
1885 case Op_Root:
1886 case Op_Goto:
1887 return true;
1888
1889 case Op_Catch: {
1890 const CatchProjNode *ci = get_node(i + eidx + 1)->as_CatchProj();
1891 return ci->_con == CatchProjNode::fall_through_index;
1892 }
1893
1894 case Op_Jump:
1895 case Op_NeverBranch:
1896 case Op_TailCall:
1897 case Op_TailJump:
1898 case Op_Return:
1899 case Op_Halt:
1900 case Op_Rethrow:
1901 return false;
1902
1903 default:
1904 ShouldNotReachHere();
1905 }
1906
1907 return false;
1908 }
1909
1910 //------------------------------update_uncommon_branch------------------------
1911 // Update the probability of a two-branch to be uncommon
1912 void Block::update_uncommon_branch(Block* ub) {
1913 int eidx = end_idx();
1914 Node *n = get_node(eidx); // Get ending Node
1915
1916 int op = n->as_Mach()->ideal_Opcode();
1917
1918 assert(op == Op_CountedLoopEnd || op == Op_If, "must be a If");
1919 assert(num_fall_throughs() == 2, "must be a two way branch block");
1920
1921 // Which successor is ub?
1922 uint s;
1923 for (s = 0; s <_num_succs; s++) {
1924 if (_succs[s] == ub) break;
1925 }
1926 assert(s < 2, "uncommon successor must be found");
1927
1928 // If ub is the true path, make the proability small, else
1929 // ub is the false path, and make the probability large
1930 bool invert = (get_node(s + eidx + 1)->Opcode() == Op_IfFalse);
1931
1932 // Get existing probability
1933 float p = n->as_MachIf()->_prob;
1934
1935 if (invert) p = 1.0 - p;
1936 if (p > PROB_MIN) {
1937 p = PROB_MIN;
1938 }
1939 if (invert) p = 1.0 - p;
1940
1941 n->as_MachIf()->_prob = p;
1942 }
1943
1944 //------------------------------update_succ_freq-------------------------------
1945 // Update the appropriate frequency associated with block 'b', a successor of
1946 // a block in this loop.
1947 void CFGLoop::update_succ_freq(Block* b, double freq) {
1948 if (b->_loop == this) {
1949 if (b == head()) {
1950 // back branch within the loop
1951 // Do nothing now, the loop carried frequency will be
1952 // adjust later in scale_freq().
1953 } else {
1954 // simple branch within the loop
1955 b->_freq += freq;
1956 }
1957 } else if (!in_loop_nest(b)) {
1958 // branch is exit from this loop
1959 BlockProbPair bpp(b, freq);
1960 _exits.append(bpp);
1961 } else {
1962 // branch into nested loop
1963 CFGLoop* ch = b->_loop;
1964 ch->_freq += freq;
1965 }
1966 }
1967
1968 //------------------------------in_loop_nest-----------------------------------
1969 // Determine if block b is in the receiver's loop nest.
1970 bool CFGLoop::in_loop_nest(Block* b) {
1971 int depth = _depth;
1972 CFGLoop* b_loop = b->_loop;
1973 int b_depth = b_loop->_depth;
1974 if (depth == b_depth) {
1975 return true;
1976 }
1977 while (b_depth > depth) {
1978 b_loop = b_loop->_parent;
1979 b_depth = b_loop->_depth;
1980 }
1981 return b_loop == this;
1982 }
1983
1984 //------------------------------scale_freq-------------------------------------
1985 // Scale frequency of loops and blocks by trip counts from outer loops
1986 // Do a top down traversal of loop tree (visit outer loops first.)
1987 void CFGLoop::scale_freq() {
1988 double loop_freq = _freq * trip_count();
1989 _freq = loop_freq;
1990 for (int i = 0; i < _members.length(); i++) {
1991 CFGElement* s = _members.at(i);
1992 double block_freq = s->_freq * loop_freq;
1993 if (g_isnan(block_freq) || block_freq < MIN_BLOCK_FREQUENCY)
1994 block_freq = MIN_BLOCK_FREQUENCY;
1995 s->_freq = block_freq;
1996 }
1997 CFGLoop* ch = _child;
1998 while (ch != NULL) {
1999 ch->scale_freq();
2000 ch = ch->_sibling;
2001 }
2002 }
2003
2004 // Frequency of outer loop
2005 double CFGLoop::outer_loop_freq() const {
2006 if (_child != NULL) {
2007 return _child->_freq;
2008 }
2009 return _freq;
2010 }
2011
2012 #ifndef PRODUCT
2013 //------------------------------dump_tree--------------------------------------
2014 void CFGLoop::dump_tree() const {
2015 dump();
2016 if (_child != NULL) _child->dump_tree();
2017 if (_sibling != NULL) _sibling->dump_tree();
2018 }
2019
2020 //------------------------------dump-------------------------------------------
2021 void CFGLoop::dump() const {
2022 for (int i = 0; i < _depth; i++) tty->print(" ");
2023 tty->print("%s: %d trip_count: %6.0f freq: %6.0f\n",
2024 _depth == 0 ? "Method" : "Loop", _id, trip_count(), _freq);
2025 for (int i = 0; i < _depth; i++) tty->print(" ");
2026 tty->print(" members:");
2027 int k = 0;
2028 for (int i = 0; i < _members.length(); i++) {
2029 if (k++ >= 6) {
2030 tty->print("\n ");
2031 for (int j = 0; j < _depth+1; j++) tty->print(" ");
2032 k = 0;
2033 }
2034 CFGElement *s = _members.at(i);
2035 if (s->is_block()) {
2036 Block *b = s->as_Block();
2037 tty->print(" B%d(%6.3f)", b->_pre_order, b->_freq);
2038 } else {
2039 CFGLoop* lp = s->as_CFGLoop();
2040 tty->print(" L%d(%6.3f)", lp->_id, lp->_freq);
2041 }
2042 }
2043 tty->print("\n");
2044 for (int i = 0; i < _depth; i++) tty->print(" ");
2045 tty->print(" exits: ");
2046 k = 0;
2047 for (int i = 0; i < _exits.length(); i++) {
2048 if (k++ >= 7) {
2049 tty->print("\n ");
2050 for (int j = 0; j < _depth+1; j++) tty->print(" ");
2051 k = 0;
2052 }
2053 Block *blk = _exits.at(i).get_target();
2054 double prob = _exits.at(i).get_prob();
2055 tty->print(" ->%d@%d%%", blk->_pre_order, (int)(prob*100));
2056 }
2057 tty->print("\n");
2058 }
2059 #endif