1 /*
2 * Copyright (c) 2015, 2019, 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 #include "precompiled.hpp"
25 #include "opto/castnode.hpp"
26 #include "opto/compile.hpp"
27 #include "opto/escape.hpp"
28 #include "opto/graphKit.hpp"
29 #include "opto/loopnode.hpp"
30 #include "opto/machnode.hpp"
31 #include "opto/macro.hpp"
32 #include "opto/memnode.hpp"
33 #include "opto/movenode.hpp"
34 #include "opto/node.hpp"
35 #include "opto/phase.hpp"
36 #include "opto/phaseX.hpp"
37 #include "opto/rootnode.hpp"
38 #include "opto/type.hpp"
39 #include "utilities/copy.hpp"
40 #include "utilities/growableArray.hpp"
41 #include "utilities/macros.hpp"
42 #include "gc/z/zBarrierSet.hpp"
43 #include "gc/z/c2/zBarrierSetC2.hpp"
44 #include "gc/z/zThreadLocalData.hpp"
45 #include "gc/z/zBarrierSetRuntime.hpp"
46
47 ZBarrierSetC2State::ZBarrierSetC2State(Arena* comp_arena) :
48 _load_barrier_nodes(new (comp_arena) GrowableArray<LoadBarrierNode*>(comp_arena, 8, 0, NULL)) {}
49
50 int ZBarrierSetC2State::load_barrier_count() const {
51 return _load_barrier_nodes->length();
52 }
53
54 void ZBarrierSetC2State::add_load_barrier_node(LoadBarrierNode * n) {
55 assert(!_load_barrier_nodes->contains(n), " duplicate entry in expand list");
56 _load_barrier_nodes->append(n);
57 }
58
59 void ZBarrierSetC2State::remove_load_barrier_node(LoadBarrierNode * n) {
60 // this function may be called twice for a node so check
61 // that the node is in the array before attempting to remove it
62 if (_load_barrier_nodes->contains(n)) {
63 _load_barrier_nodes->remove(n);
64 }
65 }
66
67 LoadBarrierNode* ZBarrierSetC2State::load_barrier_node(int idx) const {
68 return _load_barrier_nodes->at(idx);
69 }
70
71 void* ZBarrierSetC2::create_barrier_state(Arena* comp_arena) const {
72 return new(comp_arena) ZBarrierSetC2State(comp_arena);
73 }
74
75 ZBarrierSetC2State* ZBarrierSetC2::state() const {
76 return reinterpret_cast<ZBarrierSetC2State*>(Compile::current()->barrier_set_state());
77 }
78
79 bool ZBarrierSetC2::is_gc_barrier_node(Node* node) const {
80 // 1. This step follows potential oop projections of a load barrier before expansion
81 if (node->is_Proj()) {
82 node = node->in(0);
83 }
84
85 // 2. This step checks for unexpanded load barriers
86 if (node->is_LoadBarrier()) {
87 return true;
88 }
89
90 // 3. This step checks for the phi corresponding to an optimized load barrier expansion
91 if (node->is_Phi()) {
92 PhiNode* phi = node->as_Phi();
93 Node* n = phi->in(1);
94 if (n != NULL && n->is_LoadBarrierSlowReg()) {
95 return true;
96 }
97 }
98
99 return false;
100 }
101
102 void ZBarrierSetC2::register_potential_barrier_node(Node* node) const {
103 if (node->is_LoadBarrier()) {
104 state()->add_load_barrier_node(node->as_LoadBarrier());
105 }
106 }
107
108 void ZBarrierSetC2::unregister_potential_barrier_node(Node* node) const {
109 if (node->is_LoadBarrier()) {
110 state()->remove_load_barrier_node(node->as_LoadBarrier());
111 }
112 }
113
114 void ZBarrierSetC2::eliminate_useless_gc_barriers(Unique_Node_List &useful, Compile* C) const {
115 // Remove useless LoadBarrier nodes
116 ZBarrierSetC2State* s = state();
117 for (int i = s->load_barrier_count()-1; i >= 0; i--) {
118 LoadBarrierNode* n = s->load_barrier_node(i);
119 if (!useful.member(n)) {
120 unregister_potential_barrier_node(n);
121 }
122 }
123 }
124
125 void ZBarrierSetC2::enqueue_useful_gc_barrier(PhaseIterGVN* igvn, Node* node) const {
126 if (node->is_LoadBarrier() && !node->as_LoadBarrier()->has_true_uses()) {
127 igvn->_worklist.push(node);
128 }
129 }
130
131 const uint NoBarrier = 0;
132 const uint RequireBarrier = 1;
133 const uint WeakBarrier = 2;
134 const uint ExpandedBarrier = 4;
135
136 static bool load_require_barrier(LoadNode* load) { return (load->barrier_data() & RequireBarrier) == RequireBarrier; }
137 static bool load_has_weak_barrier(LoadNode* load) { return (load->barrier_data() & WeakBarrier) == WeakBarrier; }
138 static bool load_has_expanded_barrier(LoadNode* load) { return (load->barrier_data() & ExpandedBarrier) == ExpandedBarrier; }
139 static void load_set_expanded_barrier(LoadNode* load) { return load->set_barrier_data(ExpandedBarrier); }
140
141 static void load_set_barrier(LoadNode* load, bool weak) {
142 if (weak) {
143 load->set_barrier_data(RequireBarrier | WeakBarrier);
144 } else {
145 load->set_barrier_data(RequireBarrier);
146 }
147 }
148
149 // == LoadBarrierNode ==
150
151 LoadBarrierNode::LoadBarrierNode(Compile* C,
152 Node* c,
153 Node* mem,
154 Node* val,
155 Node* adr,
156 bool weak) :
157 MultiNode(Number_of_Inputs),
158 _weak(weak) {
159 init_req(Control, c);
160 init_req(Memory, mem);
161 init_req(Oop, val);
162 init_req(Address, adr);
163 init_req(Similar, C->top());
164
165 init_class_id(Class_LoadBarrier);
166 BarrierSetC2* bs = BarrierSet::barrier_set()->barrier_set_c2();
167 bs->register_potential_barrier_node(this);
168 }
169
170 uint LoadBarrierNode::size_of() const {
171 return sizeof(*this);
172 }
173
174 bool LoadBarrierNode::cmp(const Node& n) const {
175 ShouldNotReachHere();
176 return false;
177 }
178
179 const Type *LoadBarrierNode::bottom_type() const {
180 const Type** floadbarrier = (const Type **)(Compile::current()->type_arena()->Amalloc_4((Number_of_Outputs)*sizeof(Type*)));
181 Node* in_oop = in(Oop);
182 floadbarrier[Control] = Type::CONTROL;
183 floadbarrier[Memory] = Type::MEMORY;
184 floadbarrier[Oop] = in_oop == NULL ? Type::TOP : in_oop->bottom_type();
185 return TypeTuple::make(Number_of_Outputs, floadbarrier);
186 }
187
188 const TypePtr* LoadBarrierNode::adr_type() const {
189 ShouldNotReachHere();
190 return NULL;
191 }
192
193 const Type *LoadBarrierNode::Value(PhaseGVN *phase) const {
194 const Type** floadbarrier = (const Type **)(phase->C->type_arena()->Amalloc_4((Number_of_Outputs)*sizeof(Type*)));
195 const Type* val_t = phase->type(in(Oop));
196 floadbarrier[Control] = Type::CONTROL;
197 floadbarrier[Memory] = Type::MEMORY;
198 floadbarrier[Oop] = val_t;
199 return TypeTuple::make(Number_of_Outputs, floadbarrier);
200 }
201
202 bool LoadBarrierNode::is_dominator(PhaseIdealLoop* phase, bool linear_only, Node *d, Node *n) {
203 if (phase != NULL) {
204 return phase->is_dominator(d, n);
205 }
206
207 for (int i = 0; i < 10 && n != NULL; i++) {
208 n = IfNode::up_one_dom(n, linear_only);
209 if (n == d) {
210 return true;
211 }
212 }
213
214 return false;
215 }
216
217 LoadBarrierNode* LoadBarrierNode::has_dominating_barrier(PhaseIdealLoop* phase, bool linear_only, bool look_for_similar) {
218 if (is_weak()) {
219 // Weak barriers can't be eliminated
220 return NULL;
221 }
222
223 Node* val = in(LoadBarrierNode::Oop);
224 if (in(Similar)->is_Proj() && in(Similar)->in(0)->is_LoadBarrier()) {
225 LoadBarrierNode* lb = in(Similar)->in(0)->as_LoadBarrier();
226 assert(lb->in(Address) == in(Address), "");
227 // Load barrier on Similar edge dominates so if it now has the Oop field it can replace this barrier.
228 if (lb->in(Oop) == in(Oop)) {
229 return lb;
230 }
231 // Follow chain of load barrier through Similar edges
232 while (!lb->in(Similar)->is_top()) {
233 lb = lb->in(Similar)->in(0)->as_LoadBarrier();
234 assert(lb->in(Address) == in(Address), "");
235 }
236 if (lb != in(Similar)->in(0)) {
237 return lb;
238 }
239 }
240 for (DUIterator_Fast imax, i = val->fast_outs(imax); i < imax; i++) {
241 Node* u = val->fast_out(i);
242 if (u != this && u->is_LoadBarrier() && u->in(Oop) == val && u->as_LoadBarrier()->has_true_uses()) {
243 Node* this_ctrl = in(LoadBarrierNode::Control);
244 Node* other_ctrl = u->in(LoadBarrierNode::Control);
245 if (is_dominator(phase, linear_only, other_ctrl, this_ctrl)) {
246 return u->as_LoadBarrier();
247 }
248 }
249 }
250
251 if (can_be_eliminated()) {
252 return NULL;
253 }
254
255 if (!look_for_similar) {
256 return NULL;
257 }
258
259 Node* addr = in(LoadBarrierNode::Address);
260 for (DUIterator_Fast imax, i = addr->fast_outs(imax); i < imax; i++) {
261 Node* u = addr->fast_out(i);
262 if (u != this && u->is_LoadBarrier() && u->as_LoadBarrier()->has_true_uses()) {
263 Node* this_ctrl = in(LoadBarrierNode::Control);
264 Node* other_ctrl = u->in(LoadBarrierNode::Control);
265 if (is_dominator(phase, linear_only, other_ctrl, this_ctrl)) {
266 ResourceMark rm;
267 Unique_Node_List wq;
268 wq.push(in(LoadBarrierNode::Control));
269 bool ok = true;
270 bool dom_found = false;
271 for (uint next = 0; next < wq.size(); ++next) {
272 Node *n = wq.at(next);
273 if (n->is_top()) {
274 return NULL;
275 }
276 assert(n->is_CFG(), "");
277 if (n->is_SafePoint()) {
278 ok = false;
279 break;
280 }
281 if (n == u) {
282 dom_found = true;
283 continue;
284 }
285 if (n->is_Region()) {
286 for (uint i = 1; i < n->req(); i++) {
287 Node* m = n->in(i);
288 if (m != NULL) {
289 wq.push(m);
290 }
291 }
292 } else {
293 Node* m = n->in(0);
294 if (m != NULL) {
295 wq.push(m);
296 }
297 }
298 }
299 if (ok) {
300 assert(dom_found, "");
301 return u->as_LoadBarrier();
302 }
303 break;
304 }
305 }
306 }
307
308 return NULL;
309 }
310
311 void LoadBarrierNode::push_dominated_barriers(PhaseIterGVN* igvn) const {
312 // Change to that barrier may affect a dominated barrier so re-push those
313 assert(!is_weak(), "sanity");
314 Node* val = in(LoadBarrierNode::Oop);
315
316 for (DUIterator_Fast imax, i = val->fast_outs(imax); i < imax; i++) {
317 Node* u = val->fast_out(i);
318 if (u != this && u->is_LoadBarrier() && u->in(Oop) == val) {
319 Node* this_ctrl = in(Control);
320 Node* other_ctrl = u->in(Control);
321 if (is_dominator(NULL, false, this_ctrl, other_ctrl)) {
322 igvn->_worklist.push(u);
323 }
324 }
325
326 Node* addr = in(LoadBarrierNode::Address);
327 for (DUIterator_Fast imax, i = addr->fast_outs(imax); i < imax; i++) {
328 Node* u = addr->fast_out(i);
329 if (u != this && u->is_LoadBarrier() && u->in(Similar)->is_top()) {
330 Node* this_ctrl = in(Control);
331 Node* other_ctrl = u->in(Control);
332 if (is_dominator(NULL, false, this_ctrl, other_ctrl)) {
333 igvn->_worklist.push(u);
334 }
335 }
336 }
337 }
338 }
339
340 Node *LoadBarrierNode::Identity(PhaseGVN *phase) {
341 LoadBarrierNode* dominating_barrier = has_dominating_barrier(NULL, true, false);
342 if (dominating_barrier != NULL) {
343 assert(!is_weak(), "Weak barriers cant be eliminated");
344 assert(dominating_barrier->in(Oop) == in(Oop), "");
345 return dominating_barrier;
346 }
347
348 return this;
349 }
350
351 Node *LoadBarrierNode::Ideal(PhaseGVN *phase, bool can_reshape) {
352 if (remove_dead_region(phase, can_reshape)) {
353 return this;
354 }
355
356 Node *val = in(Oop);
357 Node *mem = in(Memory);
358 Node *ctrl = in(Control);
359
360 assert(val->Opcode() != Op_LoadN, "");
361 assert(val->Opcode() != Op_DecodeN, "");
362
363 if (mem->is_MergeMem()) {
364 Node *new_mem = mem->as_MergeMem()->memory_at(Compile::AliasIdxRaw);
365 set_req(Memory, new_mem);
366 if (mem->outcnt() == 0 && can_reshape) {
367 phase->is_IterGVN()->_worklist.push(mem);
368 }
369 return this;
370 }
371
372 LoadBarrierNode *dominating_barrier = NULL;
373 if (!is_weak()) {
374 dominating_barrier = has_dominating_barrier(NULL, !can_reshape, !phase->C->major_progress());
375 if (dominating_barrier != NULL && dominating_barrier->in(Oop) != in(Oop)) {
376 assert(in(Address) == dominating_barrier->in(Address), "");
377 set_req(Similar, dominating_barrier->proj_out(Oop));
378 return this;
379 }
380 }
381
382 bool eliminate = can_reshape && (dominating_barrier != NULL || !has_true_uses());
383 if (eliminate) {
384 if (can_reshape) {
385 PhaseIterGVN* igvn = phase->is_IterGVN();
386 Node* out_ctrl = proj_out_or_null(Control);
387 Node* out_res = proj_out_or_null(Oop);
388
389 if (out_ctrl != NULL) {
390 igvn->replace_node(out_ctrl, ctrl);
391 }
392
393 // That transformation may cause the Similar edge on the load barrier to be invalid
394 fix_similar_in_uses(igvn);
395 if (out_res != NULL) {
396 if (dominating_barrier != NULL) {
397 assert(!is_weak(), "Sanity");
398 igvn->replace_node(out_res, dominating_barrier->proj_out(Oop));
399 } else {
400 igvn->replace_node(out_res, val);
401 }
402 }
403 }
404 return new ConINode(TypeInt::ZERO);
405 }
406
407 // If the Similar edge is no longer a load barrier, clear it
408 Node* similar = in(Similar);
409 if (!similar->is_top() && !(similar->is_Proj() && similar->in(0)->is_LoadBarrier())) {
410 set_req(Similar, phase->C->top());
411 return this;
412 }
413
414 if (can_reshape && !is_weak()) {
415 // If this barrier is linked through the Similar edge by a
416 // dominated barrier and both barriers have the same Oop field,
417 // the dominated barrier can go away, so push it for reprocessing.
418 // We also want to avoid a barrier to depend on another dominating
419 // barrier through its Similar edge that itself depend on another
420 // barrier through its Similar edge and rather have the first
421 // depend on the third.
422 PhaseIterGVN* igvn = phase->is_IterGVN();
423 Node* out_res = proj_out(Oop);
424 for (DUIterator_Fast imax, i = out_res->fast_outs(imax); i < imax; i++) {
425 Node* u = out_res->fast_out(i);
426 if (u->is_LoadBarrier() && u->in(Similar) == out_res &&
427 (u->in(Oop) == val || !u->in(Similar)->is_top())) {
428 assert(!u->as_LoadBarrier()->is_weak(), "Sanity");
429 igvn->_worklist.push(u);
430 }
431 }
432 push_dominated_barriers(igvn);
433 }
434
435 return NULL;
436 }
437
438 uint LoadBarrierNode::match_edge(uint idx) const {
439 ShouldNotReachHere();
440 return 0;
441 }
442
443 void LoadBarrierNode::fix_similar_in_uses(PhaseIterGVN* igvn) {
444 Node* out_res = proj_out_or_null(Oop);
445 if (out_res == NULL) {
446 return;
447 }
448
449 for (DUIterator_Fast imax, i = out_res->fast_outs(imax); i < imax; i++) {
450 Node* u = out_res->fast_out(i);
451 if (u->is_LoadBarrier() && u->in(Similar) == out_res) {
452 igvn->replace_input_of(u, Similar, igvn->C->top());
453 --i;
454 --imax;
455 }
456 }
457 }
458
459 bool LoadBarrierNode::has_true_uses() const {
460 Node* out_res = proj_out_or_null(Oop);
461 if (out_res != NULL) {
462 for (DUIterator_Fast imax, i = out_res->fast_outs(imax); i < imax; i++) {
463 Node *u = out_res->fast_out(i);
464 if (!u->is_LoadBarrier() || u->in(Similar) != out_res) {
465 return true;
466 }
467 }
468 }
469 return false;
470 }
471
472 static bool barrier_needed(C2Access& access) {
473 return ZBarrierSet::barrier_needed(access.decorators(), access.type());
474 }
475
476 Node* ZBarrierSetC2::load_at_resolved(C2Access& access, const Type* val_type) const {
477 Node* p = BarrierSetC2::load_at_resolved(access, val_type);
478 if (!barrier_needed(access)) {
479 return p;
480 }
481
482 bool weak = (access.decorators() & ON_WEAK_OOP_REF) != 0;
483 if (p->isa_Load()) {
484 load_set_barrier(p->as_Load(), weak);
485 }
486 return p;
487 }
488
489 Node* ZBarrierSetC2::atomic_cmpxchg_val_at_resolved(C2AtomicParseAccess& access, Node* expected_val,
490 Node* new_val, const Type* val_type) const {
491 Node* result = BarrierSetC2::atomic_cmpxchg_val_at_resolved(access, expected_val, new_val, val_type);
492 LoadStoreNode* lsn = result->as_LoadStore();
493 if (barrier_needed(access)) {
494 lsn->set_has_barrier();
495 }
496 return lsn;
497 }
498
499 Node* ZBarrierSetC2::atomic_cmpxchg_bool_at_resolved(C2AtomicParseAccess& access, Node* expected_val,
500 Node* new_val, const Type* value_type) const {
501 Node* result = BarrierSetC2::atomic_cmpxchg_bool_at_resolved(access, expected_val, new_val, value_type);
502 LoadStoreNode* lsn = result->as_LoadStore();
503 if (barrier_needed(access)) {
504 lsn->set_has_barrier();
505 }
506 return lsn;
507 }
508
509 Node* ZBarrierSetC2::atomic_xchg_at_resolved(C2AtomicParseAccess& access, Node* new_val, const Type* val_type) const {
510 Node* result = BarrierSetC2::atomic_xchg_at_resolved(access, new_val, val_type);
511 LoadStoreNode* lsn = result->as_LoadStore();
512 if (barrier_needed(access)) {
513 lsn->set_has_barrier();
514 }
515 return lsn;
516 }
517
518 // == Macro Expansion ==
519
520 // Optimized, low spill, loadbarrier variant using stub specialized on register used
521 void ZBarrierSetC2::expand_loadbarrier_node(PhaseMacroExpand* phase, LoadBarrierNode* barrier) const {
522 PhaseIterGVN &igvn = phase->igvn();
523 float unlikely = PROB_UNLIKELY(0.999);
524
525 Node* in_ctrl = barrier->in(LoadBarrierNode::Control);
526 Node* in_mem = barrier->in(LoadBarrierNode::Memory);
527 Node* in_val = barrier->in(LoadBarrierNode::Oop);
528 Node* in_adr = barrier->in(LoadBarrierNode::Address);
529
530 Node* out_ctrl = barrier->proj_out(LoadBarrierNode::Control);
531 Node* out_res = barrier->proj_out(LoadBarrierNode::Oop);
532
533 assert(barrier->in(LoadBarrierNode::Oop) != NULL, "oop to loadbarrier node cannot be null");
534
535 Node* jthread = igvn.transform(new ThreadLocalNode());
536 Node* adr = phase->basic_plus_adr(jthread, in_bytes(ZThreadLocalData::address_bad_mask_offset()));
537 Node* bad_mask = igvn.transform(LoadNode::make(igvn, in_ctrl, in_mem, adr,
538 TypeRawPtr::BOTTOM, TypeX_X, TypeX_X->basic_type(),
539 MemNode::unordered));
540 Node* cast = igvn.transform(new CastP2XNode(in_ctrl, in_val));
541 Node* obj_masked = igvn.transform(new AndXNode(cast, bad_mask));
542 Node* cmp = igvn.transform(new CmpXNode(obj_masked, igvn.zerocon(TypeX_X->basic_type())));
543 Node *bol = igvn.transform(new BoolNode(cmp, BoolTest::ne))->as_Bool();
544 IfNode* iff = igvn.transform(new IfNode(in_ctrl, bol, unlikely, COUNT_UNKNOWN))->as_If();
545 Node* then = igvn.transform(new IfTrueNode(iff));
546 Node* elsen = igvn.transform(new IfFalseNode(iff));
547
548 Node* new_loadp = igvn.transform(new LoadBarrierSlowRegNode(then, in_adr, in_val,
549 (const TypePtr*) in_val->bottom_type(), barrier->is_weak()));
550
551 // Create the final region/phi pair to converge cntl/data paths to downstream code
552 Node* result_region = igvn.transform(new RegionNode(3));
553 result_region->set_req(1, then);
554 result_region->set_req(2, elsen);
555
556 Node* result_phi = igvn.transform(new PhiNode(result_region, TypeInstPtr::BOTTOM));
557 result_phi->set_req(1, new_loadp);
558 result_phi->set_req(2, barrier->in(LoadBarrierNode::Oop));
559
560 igvn.replace_node(out_ctrl, result_region);
561 igvn.replace_node(out_res, result_phi);
562
563 assert(barrier->outcnt() == 0,"LoadBarrier macro node has non-null outputs after expansion!");
564
565 igvn.remove_dead_node(barrier);
566 igvn.remove_dead_node(out_ctrl);
567 igvn.remove_dead_node(out_res);
568
569 assert(is_gc_barrier_node(result_phi), "sanity");
570 assert(step_over_gc_barrier(result_phi) == in_val, "sanity");
571
572 phase->C->print_method(PHASE_BARRIER_EXPANSION, 4, barrier->_idx);
573 }
574
575 bool ZBarrierSetC2::expand_barriers(Compile* C, PhaseIterGVN& igvn) const {
576 ZBarrierSetC2State* s = state();
577 if (s->load_barrier_count() > 0) {
578 PhaseMacroExpand macro(igvn);
579
580 int skipped = 0;
581 while (s->load_barrier_count() > skipped) {
582 int load_barrier_count = s->load_barrier_count();
583 LoadBarrierNode * n = s->load_barrier_node(load_barrier_count-1-skipped);
584 if (igvn.type(n) == Type::TOP || (n->in(0) != NULL && n->in(0)->is_top())) {
585 // Node is unreachable, so don't try to expand it
586 s->remove_load_barrier_node(n);
587 continue;
588 }
589 if (!n->can_be_eliminated()) {
590 skipped++;
591 continue;
592 }
593 expand_loadbarrier_node(¯o, n);
594 assert(s->load_barrier_count() < load_barrier_count, "must have deleted a node from load barrier list");
595 if (C->failing()) {
596 return true;
597 }
598 }
599 while (s->load_barrier_count() > 0) {
600 int load_barrier_count = s->load_barrier_count();
601 LoadBarrierNode* n = s->load_barrier_node(load_barrier_count - 1);
602 assert(!(igvn.type(n) == Type::TOP || (n->in(0) != NULL && n->in(0)->is_top())), "should have been processed already");
603 assert(!n->can_be_eliminated(), "should have been processed already");
604 expand_loadbarrier_node(¯o, n);
605 assert(s->load_barrier_count() < load_barrier_count, "must have deleted a node from load barrier list");
606 if (C->failing()) {
607 return true;
608 }
609 }
610 igvn.set_delay_transform(false);
611 igvn.optimize();
612 if (C->failing()) {
613 return true;
614 }
615 }
616
617 return false;
618 }
619
620 Node* ZBarrierSetC2::step_over_gc_barrier(Node* c) const {
621 Node* node = c;
622
623 // 1. This step follows potential oop projections of a load barrier before expansion
624 if (node->is_Proj()) {
625 node = node->in(0);
626 }
627
628 // 2. This step checks for unexpanded load barriers
629 if (node->is_LoadBarrier()) {
630 return node->in(LoadBarrierNode::Oop);
631 }
632
633 // 3. This step checks for the phi corresponding to an optimized load barrier expansion
634 if (node->is_Phi()) {
635 PhiNode* phi = node->as_Phi();
636 Node* n = phi->in(1);
637 if (n != NULL && n->is_LoadBarrierSlowReg()) {
638 assert(c == node, "projections from step 1 should only be seen before macro expansion");
639 return phi->in(2);
640 }
641 }
642
643 return c;
644 }
645
646 Node* ZBarrierSetC2::step_over_gc_barrier_ctrl(Node* c) const {
647 Node* node = c;
648
649 // 1. This step follows potential ctrl projections of a load barrier before expansion
650 if (node->is_Proj()) {
651 node = node->in(0);
652 }
653
654 // 2. This step checks for unexpanded load barriers
655 if (node->is_LoadBarrier()) {
656 return node->in(LoadBarrierNode::Control);
657 }
658
659 return c;
660 }
661
662 bool ZBarrierSetC2::array_copy_requires_gc_barriers(bool tightly_coupled_alloc, BasicType type, bool is_clone, ArrayCopyPhase phase) const {
663 return type == T_OBJECT || type == T_ARRAY;
664 }
665
666 bool ZBarrierSetC2::final_graph_reshaping(Compile* compile, Node* n, uint opcode) const {
667 switch (opcode) {
668 case Op_LoadBarrier:
669 assert(0, "There should be no load barriers left");
670 case Op_ZGetAndSetP:
671 case Op_ZCompareAndExchangeP:
672 case Op_ZCompareAndSwapP:
673 case Op_ZWeakCompareAndSwapP:
674 #ifdef ASSERT
675 if (VerifyOptoOopOffsets) {
676 MemNode *mem = n->as_Mem();
677 // Check to see if address types have grounded out somehow.
678 const TypeInstPtr *tp = mem->in(MemNode::Address)->bottom_type()->isa_instptr();
679 ciInstanceKlass *k = tp->klass()->as_instance_klass();
680 bool oop_offset_is_sane = k->contains_field_offset(tp->offset());
681 assert(!tp || oop_offset_is_sane, "");
682 }
683 #endif
684 return true;
685 default:
686 return false;
687 }
688 }
689
690 bool ZBarrierSetC2::matcher_find_shared_visit(Matcher* matcher, Matcher::MStack& mstack, Node* n, uint opcode, bool& mem_op, int& mem_addr_idx) const {
691 switch(opcode) {
692 case Op_CallLeaf:
693 if (n->as_Call()->entry_point() == ZBarrierSetRuntime::load_barrier_on_oop_field_preloaded_addr() ||
694 n->as_Call()->entry_point() == ZBarrierSetRuntime::load_barrier_on_weak_oop_field_preloaded_addr()) {
695 mem_op = true;
696 mem_addr_idx = TypeFunc::Parms + 1;
697 return true;
698 }
699 return false;
700 default:
701 return false;
702 }
703 }
704
705 bool ZBarrierSetC2::matcher_find_shared_post_visit(Matcher* matcher, Node* n, uint opcode) const {
706 switch(opcode) {
707 case Op_ZCompareAndExchangeP:
708 case Op_ZCompareAndSwapP:
709 case Op_ZWeakCompareAndSwapP: {
710 Node *mem = n->in(MemNode::Address);
711 Node *keepalive = n->in(5);
712 Node *pair1 = new BinaryNode(mem, keepalive);
713
714 Node *newval = n->in(MemNode::ValueIn);
715 Node *oldval = n->in(LoadStoreConditionalNode::ExpectedIn);
716 Node *pair2 = new BinaryNode(oldval, newval);
717
718 n->set_req(MemNode::Address, pair1);
719 n->set_req(MemNode::ValueIn, pair2);
720 n->del_req(5);
721 n->del_req(LoadStoreConditionalNode::ExpectedIn);
722 return true;
723 }
724 case Op_ZGetAndSetP: {
725 Node *keepalive = n->in(4);
726 Node *newval = n->in(MemNode::ValueIn);
727 Node *pair = new BinaryNode(newval, keepalive);
728 n->set_req(MemNode::ValueIn, pair);
729 n->del_req(4);
730 return true;
731 }
732
733 default:
734 return false;
735 }
736 }
737
738 // == Verification ==
739
740 #ifdef ASSERT
741
742 static void verify_slippery_safepoints_internal(Node* ctrl) {
743 // Given a CFG node, make sure it does not contain both safepoints and loads
744 // that have expanded barriers.
745 bool found_safepoint = false;
746 bool found_load = false;
747
748 for (DUIterator_Fast imax, i = ctrl->fast_outs(imax); i < imax; i++) {
749 Node* node = ctrl->fast_out(i);
750 if (node->in(0) != ctrl) {
751 // Skip outgoing precedence edges from ctrl.
752 continue;
753 }
754 if (node->is_SafePoint()) {
755 found_safepoint = true;
756 }
757 if (node->is_Load() && load_require_barrier(node->as_Load()) &&
758 load_has_expanded_barrier(node->as_Load())) {
759 found_load = true;
760 }
761 }
762 assert(!found_safepoint || !found_load, "found load and safepoint in same block");
763 }
764
765 static void verify_slippery_safepoints(Compile* C) {
766 ResourceArea *area = Thread::current()->resource_area();
767 Unique_Node_List visited(area);
768 Unique_Node_List checked(area);
769
770 // Recursively walk the graph.
771 visited.push(C->root());
772 while (visited.size() > 0) {
773 Node* node = visited.pop();
774
775 Node* ctrl = node;
776 if (!node->is_CFG()) {
777 ctrl = node->in(0);
778 }
779
780 if (ctrl != NULL && !checked.member(ctrl)) {
781 // For each block found in the graph, verify that it does not
782 // contain both a safepoint and a load requiring barriers.
783 verify_slippery_safepoints_internal(ctrl);
784
785 checked.push(ctrl);
786 }
787
788 checked.push(node);
789
790 for (DUIterator_Fast imax, i = node->fast_outs(imax); i < imax; i++) {
791 Node* use = node->fast_out(i);
792 if (checked.member(use)) continue;
793 if (visited.member(use)) continue;
794 visited.push(use);
795 }
796 }
797 }
798
799 void ZBarrierSetC2::verify_gc_barriers(Compile* compile, CompilePhase phase) const {
800 switch(phase) {
801 case BarrierSetC2::BeforeOptimize:
802 case BarrierSetC2::BeforeLateInsertion:
803 assert(state()->load_barrier_count() == 0, "No barriers inserted yet");
804 break;
805 case BarrierSetC2::BeforeMacroExpand:
806 // Barrier placement should be set by now.
807 verify_gc_barriers(false /*post_parse*/);
808 break;
809 case BarrierSetC2::BeforeCodeGen:
810 // Barriers has been fully expanded.
811 assert(state()->load_barrier_count() == 0, "No more macro barriers");
812 verify_slippery_safepoints(compile);
813 break;
814 default:
815 assert(0, "Phase without verification");
816 }
817 }
818
819 // post_parse implies that there might be load barriers without uses after parsing
820 // That only applies when adding barriers at parse time.
821 void ZBarrierSetC2::verify_gc_barriers(bool post_parse) const {
822 ZBarrierSetC2State* s = state();
823 Compile* C = Compile::current();
824 ResourceMark rm;
825 VectorSet visited(Thread::current()->resource_area());
826
827 for (int i = 0; i < s->load_barrier_count(); i++) {
828 LoadBarrierNode* n = s->load_barrier_node(i);
829
830 // The dominating barrier on the same address if it exists and
831 // this barrier must not be applied on the value from the same
832 // load otherwise the value is not reloaded before it's used the
833 // second time.
834 assert(n->in(LoadBarrierNode::Similar)->is_top() ||
835 (n->in(LoadBarrierNode::Similar)->in(0)->is_LoadBarrier() &&
836 n->in(LoadBarrierNode::Similar)->in(0)->in(LoadBarrierNode::Address) == n->in(LoadBarrierNode::Address) &&
837 n->in(LoadBarrierNode::Similar)->in(0)->in(LoadBarrierNode::Oop) != n->in(LoadBarrierNode::Oop)),
838 "broken similar edge");
839
840 assert(n->as_LoadBarrier()->has_true_uses(),
841 "found unneeded load barrier");
842
843 // Several load barrier nodes chained through their Similar edge
844 // break the code that remove the barriers in final graph reshape.
845 assert(n->in(LoadBarrierNode::Similar)->is_top() ||
846 (n->in(LoadBarrierNode::Similar)->in(0)->is_LoadBarrier() &&
847 n->in(LoadBarrierNode::Similar)->in(0)->in(LoadBarrierNode::Similar)->is_top()),
848 "chain of Similar load barriers");
849
850 if (!n->in(LoadBarrierNode::Similar)->is_top()) {
851 ResourceMark rm;
852 Unique_Node_List wq;
853 Node* other = n->in(LoadBarrierNode::Similar)->in(0);
854 wq.push(n);
855 for (uint next = 0; next < wq.size(); ++next) {
856 Node *nn = wq.at(next);
857 assert(nn->is_CFG(), "");
858 assert(!nn->is_SafePoint(), "");
859
860 if (nn == other) {
861 continue;
862 }
863
864 if (nn->is_Region()) {
865 for (uint i = 1; i < nn->req(); i++) {
866 Node* m = nn->in(i);
867 if (m != NULL) {
868 wq.push(m);
869 }
870 }
871 } else {
872 Node* m = nn->in(0);
873 if (m != NULL) {
874 wq.push(m);
875 }
876 }
877 }
878 }
879 }
880 }
881
882 #endif // end verification code
883
884 static void call_catch_cleanup_one(PhaseIdealLoop* phase, LoadNode* load, Node* ctrl);
885
886 // This code is cloning all uses of a load that is between a call and the catch blocks,
887 // to each use.
888
889 static bool fixup_uses_in_catch(PhaseIdealLoop *phase, Node *start_ctrl, Node *node) {
890
891 if (!phase->has_ctrl(node)) {
892 // This node is floating - doesn't need to be cloned.
893 assert(node != start_ctrl, "check");
894 return false;
895 }
896
897 Node* ctrl = phase->get_ctrl(node);
898 if (ctrl != start_ctrl) {
899 // We are in a successor block - the node is ok.
900 return false; // Unwind
901 }
902
903 // Process successor nodes
904 int outcnt = node->outcnt();
905 for (int i = 0; i < outcnt; i++) {
906 Node* n = node->raw_out(0);
907 assert(!n->is_LoadBarrier(), "Sanity");
908 // Calling recursively, visiting leafs first
909 fixup_uses_in_catch(phase, start_ctrl, n);
910 }
911
912 // Now all successors are outside
913 // - Clone this node to both successors
914 int no_succs = node->outcnt();
915 assert(!node->is_Store(), "Stores not expected here");
916
917 // In some very rare cases a load that doesn't need a barrier will end up here
918 // Treat it as a LoadP and the insertion of phis will be done correctly.
919 if (node->is_Load()) {
920 call_catch_cleanup_one(phase, node->as_Load(), phase->get_ctrl(node));
921 } else {
922 for (DUIterator_Fast jmax, i = node->fast_outs(jmax); i < jmax; i++) {
923 Node* use = node->fast_out(i);
924 Node* clone = node->clone();
925 assert(clone->outcnt() == 0, "");
926
927 assert(use->find_edge(node) != -1, "check");
928 phase->igvn().rehash_node_delayed(use);
929 use->replace_edge(node, clone);
930
931 Node* new_ctrl;
932 if (use->is_block_start()) {
933 new_ctrl = use;
934 } else if (use->is_CFG()) {
935 new_ctrl = use->in(0);
936 assert (new_ctrl != NULL, "");
937 } else {
938 new_ctrl = phase->get_ctrl(use);
939 }
940
941 phase->set_ctrl(clone, new_ctrl);
942
943 if (phase->C->directive()->ZTraceLoadBarriersOption) tty->print_cr(" Clone op %i as %i to control %i", node->_idx, clone->_idx, new_ctrl->_idx);
944 phase->igvn().register_new_node_with_optimizer(clone);
945 --i, --jmax;
946 }
947 assert(node->outcnt() == 0, "must be empty now");
948
949 // Node node is dead.
950 phase->igvn().remove_dead_node(node);
951 }
952 return true; // unwind - return if a use was processed
953 }
954
955 // Clone a load to a specific catch_proj
956 static Node* clone_load_to_catchproj(PhaseIdealLoop* phase, Node* load, Node* catch_proj) {
957 Node* cloned_load = load->clone();
958 cloned_load->set_req(0, catch_proj); // set explicit control
959 phase->set_ctrl(cloned_load, catch_proj); // update
960 if (phase->C->directive()->ZTraceLoadBarriersOption) tty->print_cr(" Clone LOAD %i as %i to control %i", load->_idx, cloned_load->_idx, catch_proj->_idx);
961 phase->igvn().register_new_node_with_optimizer(cloned_load);
962 return cloned_load;
963 }
964
965 static Node* get_dominating_region(PhaseIdealLoop* phase, Node* node, Node* stop) {
966 Node* region = node;
967 while (!region->isa_Region()) {
968 Node *up = phase->idom(region);
969 assert(up != region, "Must not loop");
970 assert(up != stop, "Must not find original control");
971 region = up;
972 }
973 return region;
974 }
975
976 // Clone this load to each catch block
977 static void call_catch_cleanup_one(PhaseIdealLoop* phase, LoadNode* load, Node* ctrl) {
978 bool trace = phase->C->directive()->ZTraceLoadBarriersOption;
979 phase->igvn().set_delay_transform(true);
980
981 // Verify pre conditions
982 assert(ctrl->isa_Proj() && ctrl->in(0)->isa_Call(), "Must be a call proj");
983 assert(ctrl->raw_out(0)->isa_Catch(), "Must be a catch");
984
985 if (ctrl->raw_out(0)->isa_Catch()->outcnt() == 1) {
986 if (trace) tty->print_cr("Cleaning up catch: Skipping load %i, call with single catch", load->_idx);
987 return;
988 }
989
990 // Process the loads successor nodes - if any is between
991 // the call and the catch blocks, they need to be cloned to.
992 // This is done recursively
993 for (uint i = 0; i < load->outcnt();) {
994 Node *n = load->raw_out(i);
995 assert(!n->is_LoadBarrier(), "Sanity");
996 if (!fixup_uses_in_catch(phase, ctrl, n)) {
997 // if no successor was cloned, progress to next out.
998 i++;
999 }
1000 }
1001
1002 // Now all the loads uses has been cloned down
1003 // Only thing left is to clone the loads, but they must end up
1004 // first in the catch blocks.
1005
1006 // We clone the loads oo the catch blocks only when needed.
1007 // An array is used to map the catch blocks to each lazily cloned load.
1008 // In that way no extra unnecessary loads are cloned.
1009
1010 // Any use dominated by original block must have an phi and a region added
1011
1012 Node* catch_node = ctrl->raw_out(0);
1013 int number_of_catch_projs = catch_node->outcnt();
1014 Node** proj_to_load_mapping = NEW_RESOURCE_ARRAY(Node*, number_of_catch_projs);
1015 Copy::zero_to_bytes(proj_to_load_mapping, sizeof(Node*) * number_of_catch_projs);
1016
1017 // The phi_map is used to keep track of where phis have already been inserted
1018 int phi_map_len = phase->C->unique();
1019 Node** phi_map = NEW_RESOURCE_ARRAY(Node*, phi_map_len);
1020 Copy::zero_to_bytes(phi_map, sizeof(Node*) * phi_map_len);
1021
1022 for (unsigned int i = 0; i < load->outcnt(); i++) {
1023 Node* load_use_control = NULL;
1024 Node* load_use = load->raw_out(i);
1025
1026 if (phase->has_ctrl(load_use)) {
1027 load_use_control = phase->get_ctrl(load_use);
1028 } else {
1029 load_use_control = load_use->in(0);
1030 }
1031 assert(load_use_control != NULL, "sanity");
1032 if (trace) tty->print_cr(" Handling use: %i, with control: %i", load_use->_idx, load_use_control->_idx);
1033
1034 // Some times the loads use is a phi. For them we need to determine from which catch block
1035 // the use is defined.
1036 bool load_use_is_phi = false;
1037 unsigned int load_use_phi_index = 0;
1038 Node* phi_ctrl = NULL;
1039 if (load_use->is_Phi()) {
1040 // Find phi input that matches load
1041 for (unsigned int u = 1; u < load_use->req(); u++) {
1042 if (load_use->in(u) == load) {
1043 load_use_is_phi = true;
1044 load_use_phi_index = u;
1045 assert(load_use->in(0)->is_Region(), "Region or broken");
1046 phi_ctrl = load_use->in(0)->in(u);
1047 assert(phi_ctrl->is_CFG(), "check");
1048 assert(phi_ctrl != load, "check");
1049 break;
1050 }
1051 }
1052 assert(load_use_is_phi, "must find");
1053 assert(load_use_phi_index > 0, "sanity");
1054 }
1055
1056 // For each load use, see which catch projs dominates, create load clone lazily and reconnect
1057 bool found_dominating_catchproj = false;
1058 for (int c = 0; c < number_of_catch_projs; c++) {
1059 Node* catchproj = catch_node->raw_out(c);
1060 assert(catchproj != NULL && catchproj->isa_CatchProj(), "Sanity");
1061
1062 if (!phase->is_dominator(catchproj, load_use_control)) {
1063 if (load_use_is_phi && phase->is_dominator(catchproj, phi_ctrl)) {
1064 // The loads use is local to the catchproj.
1065 // fall out and replace load with catch-local load clone.
1066 } else {
1067 continue;
1068 }
1069 }
1070 assert(!found_dominating_catchproj, "Max one should match");
1071
1072 // Clone loads to catch projs
1073 Node* load_clone = proj_to_load_mapping[c];
1074 if (load_clone == NULL) {
1075 load_clone = clone_load_to_catchproj(phase, load, catchproj);
1076 proj_to_load_mapping[c] = load_clone;
1077 }
1078 phase->igvn().rehash_node_delayed(load_use);
1079
1080 if (load_use_is_phi) {
1081 // phis are special - the load is defined from a specific control flow
1082 load_use->set_req(load_use_phi_index, load_clone);
1083 } else {
1084 // Multipe edges can be replaced at once - on calls for example
1085 load_use->replace_edge(load, load_clone);
1086 }
1087 --i; // more than one edge can have been removed, but the next is in later iterations
1088
1089 // We could break the for-loop after finding a dominating match.
1090 // But keep iterating to catch any bad idom early.
1091 found_dominating_catchproj = true;
1092 }
1093
1094 // We found no single catchproj that dominated the use - The use is at a point after
1095 // where control flow from multiple catch projs have merged. We will have to create
1096 // phi nodes before the use and tie the output from the cloned loads together. It
1097 // can be a single phi or a number of chained phis, depending on control flow
1098 if (!found_dominating_catchproj) {
1099
1100 // Use phi-control if use is a phi
1101 if (load_use_is_phi) {
1102 load_use_control = phi_ctrl;
1103 }
1104 assert(phase->is_dominator(ctrl, load_use_control), "Common use but no dominator");
1105
1106 // Clone a load on all paths
1107 for (int c = 0; c < number_of_catch_projs; c++) {
1108 Node* catchproj = catch_node->raw_out(c);
1109 Node* load_clone = proj_to_load_mapping[c];
1110 if (load_clone == NULL) {
1111 load_clone = clone_load_to_catchproj(phase, load, catchproj);
1112 proj_to_load_mapping[c] = load_clone;
1113 }
1114 }
1115
1116 // Move up dominator tree from use until dom front is reached
1117 Node* next_region = get_dominating_region(phase, load_use_control, ctrl);
1118 while (phase->idom(next_region) != catch_node) {
1119 next_region = phase->idom(next_region);
1120 if (trace) tty->print_cr("Moving up idom to region ctrl %i", next_region->_idx);
1121 }
1122 assert(phase->is_dominator(catch_node, next_region), "Sanity");
1123
1124 // Create or reuse phi node that collect all cloned loads and feed it to the use.
1125 Node* test_phi = phi_map[next_region->_idx];
1126 if ((test_phi != NULL) && test_phi->is_Phi()) {
1127 // Reuse an already created phi
1128 if (trace) tty->print_cr(" Using cached Phi %i on load_use %i", test_phi->_idx, load_use->_idx);
1129 phase->igvn().rehash_node_delayed(load_use);
1130 load_use->replace_edge(load, test_phi);
1131 // Now this use is done
1132 } else {
1133 // Otherwise we need to create one or more phis
1134 PhiNode* next_phi = new PhiNode(next_region, load->type());
1135 phi_map[next_region->_idx] = next_phi; // cache new phi
1136 phase->igvn().rehash_node_delayed(load_use);
1137 load_use->replace_edge(load, next_phi);
1138
1139 int dominators_of_region = 0;
1140 do {
1141 // New phi, connect to region and add all loads as in.
1142 Node* region = next_region;
1143 assert(region->isa_Region() && region->req() > 2, "Catch dead region nodes");
1144 PhiNode* new_phi = next_phi;
1145
1146 if (trace) tty->print_cr("Created Phi %i on load %i with control %i", new_phi->_idx, load->_idx, region->_idx);
1147
1148 // Need to add all cloned loads to the phi, taking care that the right path is matched
1149 dominators_of_region = 0; // reset for new region
1150 for (unsigned int reg_i = 1; reg_i < region->req(); reg_i++) {
1151 Node* region_pred = region->in(reg_i);
1152 assert(region_pred->is_CFG(), "check");
1153 bool pred_has_dominator = false;
1154 for (int c = 0; c < number_of_catch_projs; c++) {
1155 Node* catchproj = catch_node->raw_out(c);
1156 if (phase->is_dominator(catchproj, region_pred)) {
1157 new_phi->set_req(reg_i, proj_to_load_mapping[c]);
1158 if (trace) tty->print_cr(" - Phi in(%i) set to load %i", reg_i, proj_to_load_mapping[c]->_idx);
1159 pred_has_dominator = true;
1160 dominators_of_region++;
1161 break;
1162 }
1163 }
1164
1165 // Sometimes we need to chain several phis.
1166 if (!pred_has_dominator) {
1167 assert(dominators_of_region <= 1, "More than one region can't require extra phi");
1168 if (trace) tty->print_cr(" - Region %i pred %i not dominated by catch proj", region->_idx, region_pred->_idx);
1169 // Continue search on on this region_pred
1170 // - walk up to next region
1171 // - create a new phi and connect to first new_phi
1172 next_region = get_dominating_region(phase, region_pred, ctrl);
1173
1174 // Lookup if there already is a phi, create a new otherwise
1175 Node* test_phi = phi_map[next_region->_idx];
1176 if ((test_phi != NULL) && test_phi->is_Phi()) {
1177 next_phi = test_phi->isa_Phi();
1178 dominators_of_region++; // record that a match was found and that we are done
1179 if (trace) tty->print_cr(" Using cached phi Phi %i on control %i", next_phi->_idx, next_region->_idx);
1180 } else {
1181 next_phi = new PhiNode(next_region, load->type());
1182 phi_map[next_region->_idx] = next_phi;
1183 }
1184 new_phi->set_req(reg_i, next_phi);
1185 }
1186 }
1187
1188 new_phi->set_req(0, region);
1189 phase->igvn().register_new_node_with_optimizer(new_phi);
1190 phase->set_ctrl(new_phi, region);
1191
1192 assert(dominators_of_region != 0, "Must have found one this iteration");
1193 } while (dominators_of_region == 1);
1194 }
1195 --i;
1196 }
1197 } // end of loop over uses
1198
1199 assert(load->outcnt() == 0, "All uses should be handled");
1200 phase->igvn().remove_dead_node(load);
1201 phase->C->print_method(PHASE_CALL_CATCH_CLEANUP, 4, load->_idx);
1202
1203 // Now we should be home
1204 phase->igvn().set_delay_transform(false);
1205 }
1206
1207 // Sort out the loads that are between a call ant its catch blocks
1208 static void process_catch_cleanup_candidate(PhaseIdealLoop* phase, LoadNode* load) {
1209 bool trace = phase->C->directive()->ZTraceLoadBarriersOption;
1210
1211 Node* ctrl = phase->get_ctrl(load);
1212 if (!ctrl->is_Proj() || (ctrl->in(0) == NULL) || !ctrl->in(0)->isa_Call()) {
1213 return;
1214 }
1215
1216 Node* catch_node = ctrl->isa_Proj()->raw_out(0);
1217 if (catch_node->is_Catch()) {
1218 if (catch_node->outcnt() > 1) {
1219 call_catch_cleanup_one(phase, load, ctrl);
1220 } else {
1221 if (trace) tty->print_cr("Call catch cleanup with only one catch: load %i ", load->_idx);
1222 }
1223 }
1224 }
1225
1226 void ZBarrierSetC2::barrier_insertion_phase(Compile* C, PhaseIterGVN& igvn) const {
1227 PhaseIdealLoop::optimize(igvn, LoopOptsZBarrierInsertion);
1228 if (C->failing()) return;
1229 }
1230
1231 bool ZBarrierSetC2::optimize_loops(PhaseIdealLoop* phase, LoopOptsMode mode, VectorSet& visited, Node_Stack& nstack, Node_List& worklist) const {
1232
1233 if (mode == LoopOptsZBarrierInsertion) {
1234 // First make sure all loads between call and catch are moved to the catch block
1235 clean_catch_blocks(phase);
1236
1237 // Then expand barriers on all loads
1238 insert_load_barriers(phase);
1239
1240 // Handle all Unsafe that need barriers.
1241 insert_barriers_on_unsafe(phase);
1242
1243 phase->C->clear_major_progress();
1244 return true;
1245 } else {
1246 return false;
1247 }
1248 }
1249
1250 static bool can_simplify_cas(LoadStoreNode* node) {
1251 if (node->isa_LoadStoreConditional()) {
1252 Node *expected_in = node->as_LoadStoreConditional()->in(LoadStoreConditionalNode::ExpectedIn);
1253 return (expected_in->get_ptr_type() == TypePtr::NULL_PTR);
1254 } else {
1255 return false;
1256 }
1257 }
1258
1259 static void insert_barrier_before_unsafe(PhaseIdealLoop* phase, LoadStoreNode* old_node) {
1260
1261 Compile *C = phase->C;
1262 PhaseIterGVN &igvn = phase->igvn();
1263 LoadStoreNode* zclone = NULL;
1264
1265 Node *in_ctrl = old_node->in(MemNode::Control);
1266 Node *in_mem = old_node->in(MemNode::Memory);
1267 Node *in_adr = old_node->in(MemNode::Address);
1268 Node *in_val = old_node->in(MemNode::ValueIn);
1269 const TypePtr *adr_type = old_node->adr_type();
1270 const TypePtr* load_type = TypeOopPtr::BOTTOM; // The type for the load we are adding
1271
1272 switch (old_node->Opcode()) {
1273 case Op_CompareAndExchangeP: {
1274 zclone = new ZCompareAndExchangePNode(in_ctrl, in_mem, in_adr, in_val, old_node->in(LoadStoreConditionalNode::ExpectedIn),
1275 adr_type, old_node->get_ptr_type(), ((CompareAndExchangeNode*)old_node)->order());
1276 load_type = old_node->bottom_type()->is_ptr();
1277 break;
1278 }
1279 case Op_WeakCompareAndSwapP: {
1280 if (can_simplify_cas(old_node)) {
1281 break;
1282 }
1283 zclone = new ZWeakCompareAndSwapPNode(in_ctrl, in_mem, in_adr, in_val, old_node->in(LoadStoreConditionalNode::ExpectedIn),
1284 ((CompareAndSwapNode*)old_node)->order());
1285 adr_type = TypePtr::BOTTOM;
1286 break;
1287 }
1288 case Op_CompareAndSwapP: {
1289 if (can_simplify_cas(old_node)) {
1290 break;
1291 }
1292 zclone = new ZCompareAndSwapPNode(in_ctrl, in_mem, in_adr, in_val, old_node->in(LoadStoreConditionalNode::ExpectedIn),
1293 ((CompareAndSwapNode*)old_node)->order());
1294 adr_type = TypePtr::BOTTOM;
1295 break;
1296 }
1297 case Op_GetAndSetP: {
1298 zclone = new ZGetAndSetPNode(in_ctrl, in_mem, in_adr, in_val, old_node->adr_type(), old_node->get_ptr_type());
1299 load_type = old_node->bottom_type()->is_ptr();
1300 break;
1301 }
1302 }
1303 if (zclone != NULL) {
1304 igvn.register_new_node_with_optimizer(zclone, old_node);
1305
1306 // Make load
1307 LoadPNode *load = new LoadPNode(NULL, in_mem, in_adr, adr_type, load_type, MemNode::unordered,
1308 LoadNode::DependsOnlyOnTest);
1309 load_set_expanded_barrier(load);
1310 igvn.register_new_node_with_optimizer(load);
1311 igvn.replace_node(old_node, zclone);
1312
1313 Node *barrier = new LoadBarrierNode(C, NULL, in_mem, load, in_adr, false /* weak */);
1314 Node *barrier_val = new ProjNode(barrier, LoadBarrierNode::Oop);
1315 Node *barrier_ctrl = new ProjNode(barrier, LoadBarrierNode::Control);
1316
1317 igvn.register_new_node_with_optimizer(barrier);
1318 igvn.register_new_node_with_optimizer(barrier_val);
1319 igvn.register_new_node_with_optimizer(barrier_ctrl);
1320
1321 // loop over all of in_ctrl usages and move to barrier_ctrl
1322 for (DUIterator_Last imin, i = in_ctrl->last_outs(imin); i >= imin; --i) {
1323 Node *use = in_ctrl->last_out(i);
1324 uint l;
1325 for (l = 0; use->in(l) != in_ctrl; l++) {}
1326 igvn.replace_input_of(use, l, barrier_ctrl);
1327 }
1328
1329 load->set_req(MemNode::Control, in_ctrl);
1330 barrier->set_req(LoadBarrierNode::Control, in_ctrl);
1331 zclone->add_req(barrier_val); // add req as keep alive.
1332
1333 C->print_method(PHASE_ADD_UNSAFE_BARRIER, 4, zclone->_idx);
1334 }
1335 }
1336
1337 void ZBarrierSetC2::insert_barriers_on_unsafe(PhaseIdealLoop* phase) const {
1338 Compile *C = phase->C;
1339 PhaseIterGVN &igvn = phase->igvn();
1340 uint new_ids = C->unique();
1341 VectorSet visited(Thread::current()->resource_area());
1342 GrowableArray<Node *> nodeStack(Thread::current()->resource_area(), 0, 0, NULL);
1343 nodeStack.push(C->root());
1344 visited.test_set(C->root()->_idx);
1345
1346 // Traverse all nodes, visit all unsafe ops that require a barrier
1347 while (nodeStack.length() > 0) {
1348 Node *n = nodeStack.pop();
1349
1350 bool is_old_node = (n->_idx < new_ids); // don't process nodes that were created during cleanup
1351 if (is_old_node) {
1352 if (n->is_LoadStore()) {
1353 LoadStoreNode* lsn = n->as_LoadStore();
1354 if (lsn->has_barrier()) {
1355 BasicType bt = lsn->in(MemNode::Address)->bottom_type()->basic_type();
1356 assert ((bt == T_OBJECT || bt == T_ARRAY), "Sanity test");
1357 insert_barrier_before_unsafe(phase, lsn);
1358 }
1359 }
1360 }
1361 for (uint i = 0; i < n->len(); i++) {
1362 if (n->in(i)) {
1363 if (!visited.test_set(n->in(i)->_idx)) {
1364 nodeStack.push(n->in(i));
1365 }
1366 }
1367 }
1368 }
1369
1370 igvn.optimize();
1371 C->print_method(PHASE_ADD_UNSAFE_BARRIER, 2);
1372 }
1373
1374 // The purpose of ZBarrierSetC2::clean_catch_blocks is to prepare the IR for
1375 // splicing in load barrier nodes.
1376 //
1377 // The problem is that we might have instructions between a call and its catch nodes.
1378 // (This is usually handled in PhaseCFG:call_catch_cleanup, which clones mach nodes in
1379 // already scheduled blocks.) We can't have loads that require barriers there,
1380 // because we need to splice in new control flow, and that would violate the IR.
1381 //
1382 // clean_catch_blocks find all Loads that require a barrier and clone them and any
1383 // dependent instructions to each use. The loads must be in the beginning of the catch block
1384 // before any store.
1385 //
1386 // Sometimes the loads use will be at a place dominated by all catch blocks, then we need
1387 // a load in each catch block, and a Phi at the dominated use.
1388
1389 void ZBarrierSetC2::clean_catch_blocks(PhaseIdealLoop* phase) const {
1390
1391 Compile *C = phase->C;
1392 uint new_ids = C->unique();
1393 PhaseIterGVN &igvn = phase->igvn();
1394 VectorSet visited(Thread::current()->resource_area());
1395 GrowableArray<Node *> nodeStack(Thread::current()->resource_area(), 0, 0, NULL);
1396 nodeStack.push(C->root());
1397 visited.test_set(C->root()->_idx);
1398
1399 // Traverse all nodes, visit all loads that require a barrier
1400 while(nodeStack.length() > 0) {
1401 Node *n = nodeStack.pop();
1402
1403 for (uint i = 0; i < n->len(); i++) {
1404 if (n->in(i)) {
1405 if (!visited.test_set(n->in(i)->_idx)) {
1406 nodeStack.push(n->in(i));
1407 }
1408 }
1409 }
1410
1411 bool is_old_node = (n->_idx < new_ids); // don't process nodes that were created during cleanup
1412 if (n->is_Load() && is_old_node) {
1413 LoadNode* load = n->isa_Load();
1414 // only care about loads that will have a barrier
1415 if (load_require_barrier(load)) {
1416 process_catch_cleanup_candidate(phase, load);
1417 }
1418 }
1419 }
1420
1421 C->print_method(PHASE_CALL_CATCH_CLEANUP, 2);
1422 }
1423
1424 class DomDepthCompareClosure : public CompareClosure<LoadNode*> {
1425 PhaseIdealLoop* _phase;
1426
1427 public:
1428 DomDepthCompareClosure(PhaseIdealLoop* phase) : _phase(phase) { }
1429
1430 int do_compare(LoadNode* const &n1, LoadNode* const &n2) {
1431 int d1 = _phase->dom_depth(_phase->get_ctrl(n1));
1432 int d2 = _phase->dom_depth(_phase->get_ctrl(n2));
1433 if (d1 == d2) {
1434 // Compare index if the depth is the same, ensures all entries are unique.
1435 return n1->_idx - n2->_idx;
1436 } else {
1437 return d2 - d1;
1438 }
1439 }
1440 };
1441
1442 // Traverse graph and add all loadPs to list, sorted by dom depth
1443 void gather_loadnodes_sorted(PhaseIdealLoop* phase, GrowableArray<LoadNode*>* loadList) {
1444
1445 VectorSet visited(Thread::current()->resource_area());
1446 GrowableArray<Node *> nodeStack(Thread::current()->resource_area(), 0, 0, NULL);
1447 DomDepthCompareClosure ddcc(phase);
1448
1449 nodeStack.push(phase->C->root());
1450 while(nodeStack.length() > 0) {
1451 Node *n = nodeStack.pop();
1452 if (visited.test(n->_idx)) {
1453 continue;
1454 }
1455
1456 if (n->isa_Load()) {
1457 LoadNode *load = n->as_Load();
1458 if (load_require_barrier(load)) {
1459 assert(phase->get_ctrl(load) != NULL, "sanity");
1460 assert(phase->dom_depth(phase->get_ctrl(load)) != 0, "sanity");
1461 loadList->insert_sorted(&ddcc, load);
1462 }
1463 }
1464
1465 visited.set(n->_idx);
1466 for (uint i = 0; i < n->req(); i++) {
1467 if (n->in(i)) {
1468 if (!visited.test(n->in(i)->_idx)) {
1469 nodeStack.push(n->in(i));
1470 }
1471 }
1472 }
1473 }
1474 }
1475
1476 // Add LoadBarriers to all LoadPs
1477 void ZBarrierSetC2::insert_load_barriers(PhaseIdealLoop* phase) const {
1478
1479 bool trace = phase->C->directive()->ZTraceLoadBarriersOption;
1480 GrowableArray<LoadNode *> loadList(Thread::current()->resource_area(), 0, 0, NULL);
1481 gather_loadnodes_sorted(phase, &loadList);
1482
1483 PhaseIterGVN &igvn = phase->igvn();
1484 int count = 0;
1485
1486 for (GrowableArrayIterator<LoadNode *> loadIter = loadList.begin(); loadIter != loadList.end(); ++loadIter) {
1487 LoadNode *load = *loadIter;
1488
1489 if (load_has_expanded_barrier(load)) {
1490 continue;
1491 }
1492
1493 do {
1494 // Insert a barrier on a loadP
1495 // if another load is found that needs to be expanded first, retry on that one
1496 LoadNode* result = insert_one_loadbarrier(phase, load, phase->get_ctrl(load));
1497 while (result != NULL) {
1498 result = insert_one_loadbarrier(phase, result, phase->get_ctrl(result));
1499 }
1500 } while (!load_has_expanded_barrier(load));
1501 }
1502
1503 phase->C->print_method(PHASE_INSERT_BARRIER, 2);
1504 }
1505
1506 void push_antidependent_stores(PhaseIdealLoop* phase, Node_Stack& nodestack, LoadNode* start_load) {
1507 // push all stores on the same mem, that can_alias
1508 // Any load found must be handled first
1509 PhaseIterGVN &igvn = phase->igvn();
1510 int load_alias_idx = igvn.C->get_alias_index(start_load->adr_type());
1511
1512 Node *mem = start_load->in(1);
1513 for (DUIterator_Fast imax, u = mem->fast_outs(imax); u < imax; u++) {
1514 Node *mem_use = mem->fast_out(u);
1515
1516 if (mem_use == start_load) continue;
1517 if (!mem_use->is_Store()) continue;
1518 if (!phase->has_ctrl(mem_use)) continue;
1519 if (phase->get_ctrl(mem_use) != phase->get_ctrl(start_load)) continue;
1520
1521 // add any aliasing store in this block
1522 StoreNode *store = mem_use->isa_Store();
1523 const TypePtr *adr_type = store->adr_type();
1524 if (igvn.C->can_alias(adr_type, load_alias_idx)) {
1525 nodestack.push(store, 0);
1526 }
1527 }
1528 }
1529
1530 LoadNode* ZBarrierSetC2::insert_one_loadbarrier(PhaseIdealLoop* phase, LoadNode* start_load, Node* ctrl) const {
1531 bool trace = phase->C->directive()->ZTraceLoadBarriersOption;
1532 PhaseIterGVN &igvn = phase->igvn();
1533
1534 // Check for other loadPs at the same loop depth that is reachable by a DFS
1535 // - if found - return it. It needs to be inserted first
1536 // - otherwise proceed and insert barrier
1537
1538 VectorSet visited(Thread::current()->resource_area());
1539 Node_Stack nodestack(100);
1540
1541 nodestack.push(start_load, 0);
1542 push_antidependent_stores(phase, nodestack, start_load);
1543
1544 while(!nodestack.is_empty()) {
1545 Node* n = nodestack.node(); // peek
1546 nodestack.pop();
1547 if (visited.test(n->_idx)) {
1548 continue;
1549 }
1550
1551 if (n->is_Load() && n != start_load && load_require_barrier(n->as_Load()) && !load_has_expanded_barrier(n->as_Load())) {
1552 // Found another load that needs a barrier in the same block. Must expand later loads first.
1553 if (trace) tty->print_cr(" * Found LoadP %i on DFS", n->_idx);
1554 return n->as_Load(); // return node that should be expanded first
1555 }
1556
1557 if (!phase->has_ctrl(n)) continue;
1558 if (phase->get_ctrl(n) != phase->get_ctrl(start_load)) continue;
1559 if (n->is_Phi()) continue;
1560
1561 visited.set(n->_idx);
1562 // push all children
1563 for (DUIterator_Fast imax, ii = n->fast_outs(imax); ii < imax; ii++) {
1564 Node* c = n->fast_out(ii);
1565 if (c != NULL) {
1566 nodestack.push(c, 0);
1567 }
1568 }
1569 }
1570
1571 insert_one_loadbarrier_inner(phase, start_load, ctrl, visited);
1572 return NULL;
1573 }
1574
1575 void ZBarrierSetC2::insert_one_loadbarrier_inner(PhaseIdealLoop* phase, LoadNode* load, Node* ctrl, VectorSet visited2) const {
1576 PhaseIterGVN &igvn = phase->igvn();
1577 Compile* C = igvn.C;
1578 bool trace = C->directive()->ZTraceLoadBarriersOption;
1579
1580 // create barrier
1581 Node* barrier = new LoadBarrierNode(C, NULL, load->in(LoadNode::Memory), NULL, load->in(LoadNode::Address), load_has_weak_barrier(load));
1582 Node* barrier_val = new ProjNode(barrier, LoadBarrierNode::Oop);
1583 Node* barrier_ctrl = new ProjNode(barrier, LoadBarrierNode::Control);
1584
1585 if (trace) tty->print_cr("Insert load %i with barrier: %i and ctrl : %i", load->_idx, barrier->_idx, ctrl->_idx);
1586
1587 // Splice control
1588 // - insert barrier control diamond between loads ctrl and ctrl successor on path to block end.
1589 // - If control successor is a catch, step over to next.
1590 Node* ctrl_succ = NULL;
1591 for (DUIterator_Fast imax, j = ctrl->fast_outs(imax); j < imax; j++) {
1592 Node* tmp = ctrl->fast_out(j);
1593
1594 // - CFG nodes is the ones we are going to splice (1 only!)
1595 // - Phi nodes will continue to hang from the region node!
1596 // - self loops should be skipped
1597 if (tmp->is_Phi() || tmp == ctrl) {
1598 continue;
1599 }
1600
1601 if (tmp->is_CFG()) {
1602 assert(ctrl_succ == NULL, "There can be only one");
1603 ctrl_succ = tmp;
1604 continue;
1605 }
1606 }
1607
1608 // Now splice control
1609 assert(ctrl_succ != load, "sanity");
1610 assert(ctrl_succ != NULL, "Broken IR");
1611 bool found = false;
1612 for(uint k = 0; k < ctrl_succ->req(); k++) {
1613 if (ctrl_succ->in(k) == ctrl) {
1614 assert(!found, "sanity");
1615 if (trace) tty->print_cr(" Move CFG ctrl_succ %i to barrier_ctrl", ctrl_succ->_idx);
1616 igvn.replace_input_of(ctrl_succ, k, barrier_ctrl);
1617 found = true;
1618 k--;
1619 }
1620 }
1621
1622 // For all successors of ctrl - move all visited to become successors of barrier_ctrl instead
1623 for (DUIterator_Fast imax, r = ctrl->fast_outs(imax); r < imax; r++) {
1624 Node* tmp = ctrl->fast_out(r);
1625 if (tmp->is_SafePoint() || (visited2.test(tmp->_idx) && (tmp != load))) {
1626 if (trace) tty->print_cr(" Move ctrl_succ %i to barrier_ctrl", tmp->_idx);
1627 igvn.replace_input_of(tmp, 0, barrier_ctrl);
1628 --r; --imax;
1629 }
1630 }
1631
1632 // Move the loads user to the barrier
1633 for (DUIterator_Fast imax, i = load->fast_outs(imax); i < imax; i++) {
1634 Node* u = load->fast_out(i);
1635 if (u->isa_LoadBarrier()) {
1636 continue;
1637 }
1638
1639 // find correct input - replace with iterator?
1640 for(uint j = 0; j < u->req(); j++) {
1641 if (u->in(j) == load) {
1642 igvn.replace_input_of(u, j, barrier_val);
1643 --i; --imax; // Adjust the iterator of the *outer* loop
1644 break; // some nodes (calls) might have several uses from the same node
1645 }
1646 }
1647 }
1648
1649 // Connect barrier to load and control
1650 barrier->set_req(LoadBarrierNode::Oop, load);
1651 barrier->set_req(LoadBarrierNode::Control, ctrl);
1652
1653 igvn.replace_input_of(load, MemNode::Control, ctrl);
1654 load->pin();
1655
1656 igvn.rehash_node_delayed(load);
1657 igvn.register_new_node_with_optimizer(barrier);
1658 igvn.register_new_node_with_optimizer(barrier_val);
1659 igvn.register_new_node_with_optimizer(barrier_ctrl);
1660 load_set_expanded_barrier(load);
1661
1662 C->print_method(PHASE_INSERT_BARRIER, 3, load->_idx);
1663 }
1664
1665 // The bad_mask in the ThreadLocalData shouldn't have an anti-dep-check.
1666 // The bad_mask address if of type TypeRawPtr, but that will alias
1667 // InitializeNodes until the type system is expanded.
1668 bool ZBarrierSetC2::needs_anti_dependence_check(const Node* node) const {
1669 MachNode* mnode = node->as_Mach();
1670 if (mnode != NULL) {
1671 intptr_t offset = 0;
1672 const TypePtr *adr_type2 = NULL;
1673 const Node* base = mnode->get_base_and_disp(offset, adr_type2);
1674 if ((base != NULL) &&
1675 (base->is_Mach() && base->as_Mach()->ideal_Opcode() == Op_ThreadLocal) &&
1676 (offset == in_bytes(ZThreadLocalData::address_bad_mask_offset()))) {
1677 return false;
1678 }
1679 }
1680 return true;
1681 }