1 /*
2 * Copyright (c) 2015, 2024, 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 "asm/macroAssembler.hpp"
26 #include "classfile/javaClasses.hpp"
27 #include "gc/z/c2/zBarrierSetC2.hpp"
28 #include "gc/z/zBarrierSet.hpp"
29 #include "gc/z/zBarrierSetAssembler.hpp"
30 #include "gc/z/zBarrierSetRuntime.hpp"
31 #include "opto/arraycopynode.hpp"
32 #include "opto/addnode.hpp"
33 #include "opto/block.hpp"
34 #include "opto/compile.hpp"
35 #include "opto/graphKit.hpp"
36 #include "opto/machnode.hpp"
37 #include "opto/macro.hpp"
38 #include "opto/memnode.hpp"
39 #include "opto/node.hpp"
40 #include "opto/output.hpp"
41 #include "opto/regalloc.hpp"
42 #include "opto/rootnode.hpp"
43 #include "opto/runtime.hpp"
44 #include "opto/type.hpp"
45 #include "utilities/debug.hpp"
46 #include "utilities/growableArray.hpp"
47 #include "utilities/macros.hpp"
48
49 template<typename K, typename V, size_t _table_size>
50 class ZArenaHashtable : public ResourceObj {
51 class ZArenaHashtableEntry : public ResourceObj {
52 public:
53 ZArenaHashtableEntry* _next;
54 K _key;
55 V _value;
56 };
57
58 static const size_t _table_mask = _table_size - 1;
59
60 Arena* _arena;
61 ZArenaHashtableEntry* _table[_table_size];
62
63 public:
64 class Iterator {
65 ZArenaHashtable* _table;
66 ZArenaHashtableEntry* _current_entry;
67 size_t _current_index;
68
69 public:
70 Iterator(ZArenaHashtable* table)
71 : _table(table),
72 _current_entry(table->_table[0]),
73 _current_index(0) {
74 if (_current_entry == nullptr) {
75 next();
76 }
77 }
78
79 bool has_next() { return _current_entry != nullptr; }
80 K key() { return _current_entry->_key; }
81 V value() { return _current_entry->_value; }
82
83 void next() {
84 if (_current_entry != nullptr) {
85 _current_entry = _current_entry->_next;
86 }
87 while (_current_entry == nullptr && ++_current_index < _table_size) {
88 _current_entry = _table->_table[_current_index];
89 }
90 }
91 };
92
93 ZArenaHashtable(Arena* arena)
94 : _arena(arena),
95 _table() {
96 Copy::zero_to_bytes(&_table, sizeof(_table));
97 }
98
99 void add(K key, V value) {
100 ZArenaHashtableEntry* entry = new (_arena) ZArenaHashtableEntry();
101 entry->_key = key;
102 entry->_value = value;
103 entry->_next = _table[key & _table_mask];
104 _table[key & _table_mask] = entry;
105 }
106
107 V* get(K key) const {
108 for (ZArenaHashtableEntry* e = _table[key & _table_mask]; e != nullptr; e = e->_next) {
109 if (e->_key == key) {
110 return &(e->_value);
111 }
112 }
113 return nullptr;
114 }
115
116 Iterator iterator() {
117 return Iterator(this);
118 }
119 };
120
121 typedef ZArenaHashtable<intptr_t, bool, 4> ZOffsetTable;
122
123 class ZBarrierSetC2State : public ArenaObj {
124 private:
125 GrowableArray<ZBarrierStubC2*>* _stubs;
126 Node_Array _live;
127 int _trampoline_stubs_count;
128 int _stubs_start_offset;
129
130 public:
131 ZBarrierSetC2State(Arena* arena)
132 : _stubs(new (arena) GrowableArray<ZBarrierStubC2*>(arena, 8, 0, nullptr)),
133 _live(arena),
134 _trampoline_stubs_count(0),
135 _stubs_start_offset(0) {}
136
137 GrowableArray<ZBarrierStubC2*>* stubs() {
138 return _stubs;
139 }
140
141 RegMask* live(const Node* node) {
142 if (!node->is_Mach()) {
143 // Don't need liveness for non-MachNodes
144 return nullptr;
145 }
146
147 const MachNode* const mach = node->as_Mach();
148 if (mach->barrier_data() == ZBarrierElided) {
149 // Don't need liveness data for nodes without barriers
150 return nullptr;
151 }
152
153 RegMask* live = (RegMask*)_live[node->_idx];
154 if (live == nullptr) {
155 live = new (Compile::current()->comp_arena()->AmallocWords(sizeof(RegMask))) RegMask();
156 _live.map(node->_idx, (Node*)live);
157 }
158
159 return live;
160 }
161
162 void inc_trampoline_stubs_count() {
163 assert(_trampoline_stubs_count != INT_MAX, "Overflow");
164 ++_trampoline_stubs_count;
165 }
166
167 int trampoline_stubs_count() {
168 return _trampoline_stubs_count;
169 }
170
171 void set_stubs_start_offset(int offset) {
172 _stubs_start_offset = offset;
173 }
174
175 int stubs_start_offset() {
176 return _stubs_start_offset;
177 }
178 };
179
180 static ZBarrierSetC2State* barrier_set_state() {
181 return reinterpret_cast<ZBarrierSetC2State*>(Compile::current()->barrier_set_state());
182 }
183
184 void ZBarrierStubC2::register_stub(ZBarrierStubC2* stub) {
185 if (!Compile::current()->output()->in_scratch_emit_size()) {
186 barrier_set_state()->stubs()->append(stub);
187 }
188 }
189
190 void ZBarrierStubC2::inc_trampoline_stubs_count() {
191 if (!Compile::current()->output()->in_scratch_emit_size()) {
192 barrier_set_state()->inc_trampoline_stubs_count();
193 }
194 }
195
196 int ZBarrierStubC2::trampoline_stubs_count() {
197 return barrier_set_state()->trampoline_stubs_count();
198 }
199
200 int ZBarrierStubC2::stubs_start_offset() {
201 return barrier_set_state()->stubs_start_offset();
202 }
203
204 ZBarrierStubC2::ZBarrierStubC2(const MachNode* node)
205 : _node(node),
206 _entry(),
207 _continuation() {}
208
209 Register ZBarrierStubC2::result() const {
210 return noreg;
211 }
212
213 RegMask& ZBarrierStubC2::live() const {
214 return *barrier_set_state()->live(_node);
215 }
216
217 Label* ZBarrierStubC2::entry() {
218 // The _entry will never be bound when in_scratch_emit_size() is true.
219 // However, we still need to return a label that is not bound now, but
220 // will eventually be bound. Any eventually bound label will do, as it
221 // will only act as a placeholder, so we return the _continuation label.
222 return Compile::current()->output()->in_scratch_emit_size() ? &_continuation : &_entry;
223 }
224
225 Label* ZBarrierStubC2::continuation() {
226 return &_continuation;
227 }
228
229 ZLoadBarrierStubC2* ZLoadBarrierStubC2::create(const MachNode* node, Address ref_addr, Register ref) {
230 AARCH64_ONLY(fatal("Should use ZLoadBarrierStubC2Aarch64::create"));
231 ZLoadBarrierStubC2* const stub = new (Compile::current()->comp_arena()) ZLoadBarrierStubC2(node, ref_addr, ref);
232 register_stub(stub);
233
234 return stub;
235 }
236
237 ZLoadBarrierStubC2::ZLoadBarrierStubC2(const MachNode* node, Address ref_addr, Register ref)
238 : ZBarrierStubC2(node),
239 _ref_addr(ref_addr),
240 _ref(ref) {
241 assert_different_registers(ref, ref_addr.base());
242 assert_different_registers(ref, ref_addr.index());
243 }
244
245 Address ZLoadBarrierStubC2::ref_addr() const {
246 return _ref_addr;
247 }
248
249 Register ZLoadBarrierStubC2::ref() const {
250 return _ref;
251 }
252
253 Register ZLoadBarrierStubC2::result() const {
254 return ref();
255 }
256
257 address ZLoadBarrierStubC2::slow_path() const {
258 const uint8_t barrier_data = _node->barrier_data();
259 DecoratorSet decorators = DECORATORS_NONE;
260 if (barrier_data & ZBarrierStrong) {
261 decorators |= ON_STRONG_OOP_REF;
262 }
263 if (barrier_data & ZBarrierWeak) {
264 decorators |= ON_WEAK_OOP_REF;
265 }
266 if (barrier_data & ZBarrierPhantom) {
267 decorators |= ON_PHANTOM_OOP_REF;
268 }
269 if (barrier_data & ZBarrierNoKeepalive) {
270 decorators |= AS_NO_KEEPALIVE;
271 }
272 return ZBarrierSetRuntime::load_barrier_on_oop_field_preloaded_addr(decorators);
273 }
274
275 void ZLoadBarrierStubC2::emit_code(MacroAssembler& masm) {
276 ZBarrierSet::assembler()->generate_c2_load_barrier_stub(&masm, static_cast<ZLoadBarrierStubC2*>(this));
277 }
278
279 ZStoreBarrierStubC2* ZStoreBarrierStubC2::create(const MachNode* node, Address ref_addr, Register new_zaddress, Register new_zpointer, bool is_native, bool is_atomic) {
280 AARCH64_ONLY(fatal("Should use ZStoreBarrierStubC2Aarch64::create"));
281 ZStoreBarrierStubC2* const stub = new (Compile::current()->comp_arena()) ZStoreBarrierStubC2(node, ref_addr, new_zaddress, new_zpointer, is_native, is_atomic);
282 register_stub(stub);
283
284 return stub;
285 }
286
287 ZStoreBarrierStubC2::ZStoreBarrierStubC2(const MachNode* node, Address ref_addr, Register new_zaddress, Register new_zpointer, bool is_native, bool is_atomic)
288 : ZBarrierStubC2(node),
289 _ref_addr(ref_addr),
290 _new_zaddress(new_zaddress),
291 _new_zpointer(new_zpointer),
292 _is_native(is_native),
293 _is_atomic(is_atomic) {}
294
295 Address ZStoreBarrierStubC2::ref_addr() const {
296 return _ref_addr;
297 }
298
299 Register ZStoreBarrierStubC2::new_zaddress() const {
300 return _new_zaddress;
301 }
302
303 Register ZStoreBarrierStubC2::new_zpointer() const {
304 return _new_zpointer;
305 }
306
307 bool ZStoreBarrierStubC2::is_native() const {
308 return _is_native;
309 }
310
311 bool ZStoreBarrierStubC2::is_atomic() const {
312 return _is_atomic;
313 }
314
315 Register ZStoreBarrierStubC2::result() const {
316 return noreg;
317 }
318
319 void ZStoreBarrierStubC2::emit_code(MacroAssembler& masm) {
320 ZBarrierSet::assembler()->generate_c2_store_barrier_stub(&masm, static_cast<ZStoreBarrierStubC2*>(this));
321 }
322
323 uint ZBarrierSetC2::estimated_barrier_size(const Node* node) const {
324 uint8_t barrier_data = MemNode::barrier_data(node);
325 assert(barrier_data != 0, "should be a barrier node");
326 uint uncolor_or_color_size = node->is_Load() ? 1 : 2;
327 if ((barrier_data & ZBarrierElided) != 0) {
328 return uncolor_or_color_size;
329 }
330 // A compare and branch corresponds to approximately four fast-path Ideal
331 // nodes (Cmp, Bool, If, If projection). The slow path (If projection and
332 // runtime call) is excluded since the corresponding code is laid out
333 // separately and does not directly affect performance.
334 return uncolor_or_color_size + 4;
335 }
336
337 void* ZBarrierSetC2::create_barrier_state(Arena* comp_arena) const {
338 return new (comp_arena) ZBarrierSetC2State(comp_arena);
339 }
340
341 void ZBarrierSetC2::late_barrier_analysis() const {
342 compute_liveness_at_stubs();
343 analyze_dominating_barriers();
344 }
345
346 void ZBarrierSetC2::emit_stubs(CodeBuffer& cb) const {
347 MacroAssembler masm(&cb);
348 GrowableArray<ZBarrierStubC2*>* const stubs = barrier_set_state()->stubs();
349 barrier_set_state()->set_stubs_start_offset(masm.offset());
350
351 for (int i = 0; i < stubs->length(); i++) {
352 // Make sure there is enough space in the code buffer
353 if (cb.insts()->maybe_expand_to_ensure_remaining(PhaseOutput::MAX_inst_size) && cb.blob() == nullptr) {
354 ciEnv::current()->record_failure("CodeCache is full");
355 return;
356 }
357
358 stubs->at(i)->emit_code(masm);
359 }
360
361 masm.flush();
362 }
363
364 int ZBarrierSetC2::estimate_stub_size() const {
365 Compile* const C = Compile::current();
366 BufferBlob* const blob = C->output()->scratch_buffer_blob();
367 GrowableArray<ZBarrierStubC2*>* const stubs = barrier_set_state()->stubs();
368 int size = 0;
369
370 for (int i = 0; i < stubs->length(); i++) {
371 CodeBuffer cb(blob->content_begin(), (address)C->output()->scratch_locs_memory() - blob->content_begin());
372 MacroAssembler masm(&cb);
373 stubs->at(i)->emit_code(masm);
374 size += cb.insts_size();
375 }
376
377 return size;
378 }
379
380 static void set_barrier_data(C2Access& access) {
381 if (!ZBarrierSet::barrier_needed(access.decorators(), access.type())) {
382 return;
383 }
384
385 if (access.decorators() & C2_TIGHTLY_COUPLED_ALLOC) {
386 access.set_barrier_data(ZBarrierElided);
387 return;
388 }
389
390 uint8_t barrier_data = 0;
391
392 if (access.decorators() & ON_PHANTOM_OOP_REF) {
393 barrier_data |= ZBarrierPhantom;
394 } else if (access.decorators() & ON_WEAK_OOP_REF) {
395 barrier_data |= ZBarrierWeak;
396 } else {
397 barrier_data |= ZBarrierStrong;
398 }
399
400 if (access.decorators() & IN_NATIVE) {
401 barrier_data |= ZBarrierNative;
402 }
403
404 if (access.decorators() & AS_NO_KEEPALIVE) {
405 barrier_data |= ZBarrierNoKeepalive;
406 }
407
408 access.set_barrier_data(barrier_data);
409 }
410
411 Node* ZBarrierSetC2::store_at_resolved(C2Access& access, C2AccessValue& val) const {
412 set_barrier_data(access);
413 return BarrierSetC2::store_at_resolved(access, val);
414 }
415
416 Node* ZBarrierSetC2::load_at_resolved(C2Access& access, const Type* val_type) const {
417 set_barrier_data(access);
418 return BarrierSetC2::load_at_resolved(access, val_type);
419 }
420
421 Node* ZBarrierSetC2::atomic_cmpxchg_val_at_resolved(C2AtomicParseAccess& access, Node* expected_val,
422 Node* new_val, const Type* val_type) const {
423 set_barrier_data(access);
424 return BarrierSetC2::atomic_cmpxchg_val_at_resolved(access, expected_val, new_val, val_type);
425 }
426
427 Node* ZBarrierSetC2::atomic_cmpxchg_bool_at_resolved(C2AtomicParseAccess& access, Node* expected_val,
428 Node* new_val, const Type* value_type) const {
429 set_barrier_data(access);
430 return BarrierSetC2::atomic_cmpxchg_bool_at_resolved(access, expected_val, new_val, value_type);
431 }
432
433 Node* ZBarrierSetC2::atomic_xchg_at_resolved(C2AtomicParseAccess& access, Node* new_val, const Type* val_type) const {
434 set_barrier_data(access);
435 return BarrierSetC2::atomic_xchg_at_resolved(access, new_val, val_type);
436 }
437
438 bool ZBarrierSetC2::array_copy_requires_gc_barriers(bool tightly_coupled_alloc, BasicType type,
439 bool is_clone, bool is_clone_instance,
440 ArrayCopyPhase phase) const {
441 if (phase == ArrayCopyPhase::Parsing) {
442 return false;
443 }
444 if (phase == ArrayCopyPhase::Optimization) {
445 return is_clone_instance;
446 }
447 // else ArrayCopyPhase::Expansion
448 return type == T_OBJECT || type == T_ARRAY;
449 }
450
451 // This TypeFunc assumes a 64bit system
452 static const TypeFunc* clone_type() {
453 // Create input type (domain)
454 const Type** const domain_fields = TypeTuple::fields(4);
455 domain_fields[TypeFunc::Parms + 0] = TypeInstPtr::NOTNULL; // src
456 domain_fields[TypeFunc::Parms + 1] = TypeInstPtr::NOTNULL; // dst
457 domain_fields[TypeFunc::Parms + 2] = TypeLong::LONG; // size lower
458 domain_fields[TypeFunc::Parms + 3] = Type::HALF; // size upper
459 const TypeTuple* const domain = TypeTuple::make(TypeFunc::Parms + 4, domain_fields);
460
461 // Create result type (range)
462 const Type** const range_fields = TypeTuple::fields(0);
463 const TypeTuple* const range = TypeTuple::make(TypeFunc::Parms + 0, range_fields);
464
465 return TypeFunc::make(domain, range);
466 }
467
468 #define XTOP LP64_ONLY(COMMA phase->top())
469
470 void ZBarrierSetC2::clone_at_expansion(PhaseMacroExpand* phase, ArrayCopyNode* ac) const {
471 Node* const src = ac->in(ArrayCopyNode::Src);
472 const TypeAryPtr* const ary_ptr = src->get_ptr_type()->isa_aryptr();
473
474 if (ac->is_clone_array() && ary_ptr != nullptr) {
475 BasicType bt = ary_ptr->elem()->array_element_basic_type();
476 if (is_reference_type(bt)) {
477 // Clone object array
478 bt = T_OBJECT;
479 } else {
480 // Clone primitive array
481 bt = T_LONG;
482 }
483
484 Node* const ctrl = ac->in(TypeFunc::Control);
485 Node* const mem = ac->in(TypeFunc::Memory);
486 Node* const src = ac->in(ArrayCopyNode::Src);
487 Node* src_offset = ac->in(ArrayCopyNode::SrcPos);
488 Node* const dest = ac->in(ArrayCopyNode::Dest);
489 Node* dest_offset = ac->in(ArrayCopyNode::DestPos);
490 Node* length = ac->in(ArrayCopyNode::Length);
491
492 if (bt == T_OBJECT) {
493 // BarrierSetC2::clone sets the offsets via BarrierSetC2::arraycopy_payload_base_offset
494 // which 8-byte aligns them to allow for word size copies. Make sure the offsets point
495 // to the first element in the array when cloning object arrays. Otherwise, load
496 // barriers are applied to parts of the header. Also adjust the length accordingly.
497 assert(src_offset == dest_offset, "should be equal");
498 const jlong offset = src_offset->get_long();
499 if (offset != arrayOopDesc::base_offset_in_bytes(T_OBJECT)) {
500 assert(!UseCompressedClassPointers || UseCompactObjectHeaders, "should only happen without compressed class pointers");
501 assert((arrayOopDesc::base_offset_in_bytes(T_OBJECT) - offset) == BytesPerLong, "unexpected offset");
502 length = phase->transform_later(new SubLNode(length, phase->longcon(1))); // Size is in longs
503 src_offset = phase->longcon(arrayOopDesc::base_offset_in_bytes(T_OBJECT));
504 dest_offset = src_offset;
505 }
506 }
507 Node* const payload_src = phase->basic_plus_adr(src, src_offset);
508 Node* const payload_dst = phase->basic_plus_adr(dest, dest_offset);
509
510 const char* copyfunc_name = "arraycopy";
511 const address copyfunc_addr = phase->basictype2arraycopy(bt, nullptr, nullptr, true, copyfunc_name, true);
512
513 const TypePtr* const raw_adr_type = TypeRawPtr::BOTTOM;
514 const TypeFunc* const call_type = OptoRuntime::fast_arraycopy_Type();
515
516 Node* const call = phase->make_leaf_call(ctrl, mem, call_type, copyfunc_addr, copyfunc_name, raw_adr_type, payload_src, payload_dst, length XTOP);
517 phase->transform_later(call);
518
519 phase->igvn().replace_node(ac, call);
520 return;
521 }
522
523 // Clone instance
524 Node* const ctrl = ac->in(TypeFunc::Control);
525 Node* const mem = ac->in(TypeFunc::Memory);
526 Node* const dst = ac->in(ArrayCopyNode::Dest);
527 Node* const size = ac->in(ArrayCopyNode::Length);
528
529 assert(size->bottom_type()->is_long(), "Should be long");
530
531 // The native clone we are calling here expects the instance size in words
532 // Add header/offset size to payload size to get instance size.
533 Node* const base_offset = phase->longcon(arraycopy_payload_base_offset(ac->is_clone_array()) >> LogBytesPerLong);
534 Node* const full_size = phase->transform_later(new AddLNode(size, base_offset));
535
536 Node* const call = phase->make_leaf_call(ctrl,
537 mem,
538 clone_type(),
539 ZBarrierSetRuntime::clone_addr(),
540 "ZBarrierSetRuntime::clone",
541 TypeRawPtr::BOTTOM,
542 src,
543 dst,
544 full_size,
545 phase->top());
546 phase->transform_later(call);
547 phase->igvn().replace_node(ac, call);
548 }
549
550 #undef XTOP
551
552 // == Dominating barrier elision ==
553
554 static bool block_has_safepoint(const Block* block, uint from, uint to) {
555 for (uint i = from; i < to; i++) {
556 if (block->get_node(i)->is_MachSafePoint()) {
557 // Safepoint found
558 return true;
559 }
560 }
561
562 // Safepoint not found
563 return false;
564 }
565
566 static bool block_has_safepoint(const Block* block) {
567 return block_has_safepoint(block, 0, block->number_of_nodes());
568 }
569
570 static uint block_index(const Block* block, const Node* node) {
571 for (uint j = 0; j < block->number_of_nodes(); ++j) {
572 if (block->get_node(j) == node) {
573 return j;
574 }
575 }
576 ShouldNotReachHere();
577 return 0;
578 }
579
580 // Look through various node aliases
581 static const Node* look_through_node(const Node* node) {
582 while (node != nullptr) {
583 const Node* new_node = node;
584 if (node->is_Mach()) {
585 const MachNode* const node_mach = node->as_Mach();
586 if (node_mach->ideal_Opcode() == Op_CheckCastPP) {
587 new_node = node->in(1);
588 }
589 if (node_mach->is_SpillCopy()) {
590 new_node = node->in(1);
591 }
592 }
593 if (new_node == node || new_node == nullptr) {
594 break;
595 } else {
596 node = new_node;
597 }
598 }
599
600 return node;
601 }
602
603 // Whether the given offset is undefined.
604 static bool is_undefined(intptr_t offset) {
605 return offset == Type::OffsetTop;
606 }
607
608 // Whether the given offset is unknown.
609 static bool is_unknown(intptr_t offset) {
610 return offset == Type::OffsetBot;
611 }
612
613 // Whether the given offset is concrete (defined and compile-time known).
614 static bool is_concrete(intptr_t offset) {
615 return !is_undefined(offset) && !is_unknown(offset);
616 }
617
618 // Compute base + offset components of the memory address accessed by mach.
619 // Return a node representing the base address, or null if the base cannot be
620 // found or the offset is undefined or a concrete negative value. If a non-null
621 // base is returned, the offset is a concrete, nonnegative value or unknown.
622 static const Node* get_base_and_offset(const MachNode* mach, intptr_t& offset) {
623 const TypePtr* adr_type = nullptr;
624 offset = 0;
625 const Node* base = mach->get_base_and_disp(offset, adr_type);
626
627 if (base == nullptr || base == NodeSentinel) {
628 return nullptr;
629 }
630
631 if (offset == 0 && base->is_Mach() && base->as_Mach()->ideal_Opcode() == Op_AddP) {
632 // The memory address is computed by 'base' and fed to 'mach' via an
633 // indirect memory operand (indicated by offset == 0). The ultimate base and
634 // offset can be fetched directly from the inputs and Ideal type of 'base'.
635 offset = base->bottom_type()->isa_oopptr()->offset();
636 // Even if 'base' is not an Ideal AddP node anymore, Matcher::ReduceInst()
637 // guarantees that the base address is still available at the same slot.
638 base = base->in(AddPNode::Base);
639 assert(base != nullptr, "");
640 }
641
642 if (is_undefined(offset) || (is_concrete(offset) && offset < 0)) {
643 return nullptr;
644 }
645
646 return look_through_node(base);
647 }
648
649 // Whether a phi node corresponds to an array allocation.
650 // This test is incomplete: in some edge cases, it might return false even
651 // though the node does correspond to an array allocation.
652 static bool is_array_allocation(const Node* phi) {
653 precond(phi->is_Phi());
654 // Check whether phi has a successor cast (CheckCastPP) to Java array pointer,
655 // possibly below spill copies and other cast nodes. Limit the exploration to
656 // a single path from the phi node consisting of these node types.
657 const Node* current = phi;
658 while (true) {
659 const Node* next = nullptr;
660 for (DUIterator_Fast imax, i = current->fast_outs(imax); i < imax; i++) {
661 if (!current->fast_out(i)->isa_Mach()) {
662 continue;
663 }
664 const MachNode* succ = current->fast_out(i)->as_Mach();
665 if (succ->ideal_Opcode() == Op_CheckCastPP) {
666 if (succ->get_ptr_type()->isa_aryptr()) {
667 // Cast to Java array pointer: phi corresponds to an array allocation.
668 return true;
669 }
670 // Other cast: record as candidate for further exploration.
671 next = succ;
672 } else if (succ->is_SpillCopy() && next == nullptr) {
673 // Spill copy, and no better candidate found: record as candidate.
674 next = succ;
675 }
676 }
677 if (next == nullptr) {
678 // No evidence found that phi corresponds to an array allocation, and no
679 // candidates available to continue exploring.
680 return false;
681 }
682 // Continue exploring from the best candidate found.
683 current = next;
684 }
685 ShouldNotReachHere();
686 }
687
688 // Match the phi node that connects a TLAB allocation fast path with its slowpath
689 static bool is_allocation(const Node* node) {
690 if (node->req() != 3) {
691 return false;
692 }
693 const Node* const fast_node = node->in(2);
694 if (!fast_node->is_Mach()) {
695 return false;
696 }
697 const MachNode* const fast_mach = fast_node->as_Mach();
698 if (fast_mach->ideal_Opcode() != Op_LoadP) {
699 return false;
700 }
701 const TypePtr* const adr_type = nullptr;
702 intptr_t offset;
703 const Node* const base = get_base_and_offset(fast_mach, offset);
704 if (base == nullptr || !base->is_Mach() || !is_concrete(offset)) {
705 return false;
706 }
707 const MachNode* const base_mach = base->as_Mach();
708 if (base_mach->ideal_Opcode() != Op_ThreadLocal) {
709 return false;
710 }
711 return offset == in_bytes(Thread::tlab_top_offset());
712 }
713
714 static void elide_mach_barrier(MachNode* mach) {
715 mach->set_barrier_data(ZBarrierElided);
716 }
717
718 void ZBarrierSetC2::analyze_dominating_barriers_impl(Node_List& accesses, Node_List& access_dominators) const {
719 Compile* const C = Compile::current();
720 PhaseCFG* const cfg = C->cfg();
721
722 for (uint i = 0; i < accesses.size(); i++) {
723 MachNode* const access = accesses.at(i)->as_Mach();
724 intptr_t access_offset;
725 const Node* const access_obj = get_base_and_offset(access, access_offset);
726 Block* const access_block = cfg->get_block_for_node(access);
727 const uint access_index = block_index(access_block, access);
728
729 if (access_obj == nullptr) {
730 // No information available
731 continue;
732 }
733
734 for (uint j = 0; j < access_dominators.size(); j++) {
735 const Node* const mem = access_dominators.at(j);
736 if (mem->is_Phi()) {
737 // Allocation node
738 if (mem != access_obj) {
739 continue;
740 }
741 if (is_unknown(access_offset) && !is_array_allocation(mem)) {
742 // The accessed address has an unknown offset, but the allocated
743 // object cannot be determined to be an array. Avoid eliding in this
744 // case, to be on the safe side.
745 continue;
746 }
747 assert((is_concrete(access_offset) && access_offset >= 0) || (is_unknown(access_offset) && is_array_allocation(mem)),
748 "candidate allocation-dominated access offsets must be either concrete and nonnegative, or unknown (for array allocations only)");
749 } else {
750 // Access node
751 const MachNode* const mem_mach = mem->as_Mach();
752 intptr_t mem_offset;
753 const Node* const mem_obj = get_base_and_offset(mem_mach, mem_offset);
754
755 if (mem_obj == nullptr ||
756 !is_concrete(access_offset) ||
757 !is_concrete(mem_offset)) {
758 // No information available
759 continue;
760 }
761
762 if (mem_obj != access_obj || mem_offset != access_offset) {
763 // Not the same addresses, not a candidate
764 continue;
765 }
766 assert(is_concrete(access_offset) && access_offset >= 0,
767 "candidate non-allocation-dominated access offsets must be concrete and nonnegative");
768 }
769
770 Block* mem_block = cfg->get_block_for_node(mem);
771 const uint mem_index = block_index(mem_block, mem);
772
773 if (access_block == mem_block) {
774 // Earlier accesses in the same block
775 if (mem_index < access_index && !block_has_safepoint(mem_block, mem_index + 1, access_index)) {
776 elide_mach_barrier(access);
777 }
778 } else if (mem_block->dominates(access_block)) {
779 // Dominating block? Look around for safepoints
780 ResourceMark rm;
781 Block_List stack;
782 VectorSet visited;
783 stack.push(access_block);
784 bool safepoint_found = block_has_safepoint(access_block);
785 while (!safepoint_found && stack.size() > 0) {
786 const Block* const block = stack.pop();
787 if (visited.test_set(block->_pre_order)) {
788 continue;
789 }
790 if (block_has_safepoint(block)) {
791 safepoint_found = true;
792 break;
793 }
794 if (block == mem_block) {
795 continue;
796 }
797
798 // Push predecessor blocks
799 for (uint p = 1; p < block->num_preds(); ++p) {
800 Block* const pred = cfg->get_block_for_node(block->pred(p));
801 stack.push(pred);
802 }
803 }
804
805 if (!safepoint_found) {
806 elide_mach_barrier(access);
807 }
808 }
809 }
810 }
811 }
812
813 void ZBarrierSetC2::analyze_dominating_barriers() const {
814 ResourceMark rm;
815 Compile* const C = Compile::current();
816 PhaseCFG* const cfg = C->cfg();
817
818 Node_List loads;
819 Node_List load_dominators;
820
821 Node_List stores;
822 Node_List store_dominators;
823
824 Node_List atomics;
825 Node_List atomic_dominators;
826
827 // Step 1 - Find accesses and allocations, and track them in lists
828 for (uint i = 0; i < cfg->number_of_blocks(); ++i) {
829 const Block* const block = cfg->get_block(i);
830 for (uint j = 0; j < block->number_of_nodes(); ++j) {
831 Node* const node = block->get_node(j);
832 if (node->is_Phi()) {
833 if (is_allocation(node)) {
834 load_dominators.push(node);
835 store_dominators.push(node);
836 // An allocation can't be considered to "dominate" an atomic operation.
837 // For example a CAS requires the memory location to be store-good.
838 // When you have a dominating store or atomic instruction, that is
839 // indeed ensured to be the case. However, as for allocations, the
840 // initialized memory location could be raw null, which isn't store-good.
841 }
842 continue;
843 } else if (!node->is_Mach()) {
844 continue;
845 }
846
847 MachNode* const mach = node->as_Mach();
848 switch (mach->ideal_Opcode()) {
849 case Op_LoadP:
850 if ((mach->barrier_data() & ZBarrierStrong) != 0 &&
851 (mach->barrier_data() & ZBarrierNoKeepalive) == 0) {
852 loads.push(mach);
853 load_dominators.push(mach);
854 }
855 break;
856 case Op_StoreP:
857 if (mach->barrier_data() != 0) {
858 stores.push(mach);
859 load_dominators.push(mach);
860 store_dominators.push(mach);
861 atomic_dominators.push(mach);
862 }
863 break;
864 case Op_CompareAndExchangeP:
865 case Op_CompareAndSwapP:
866 case Op_GetAndSetP:
867 if (mach->barrier_data() != 0) {
868 atomics.push(mach);
869 load_dominators.push(mach);
870 store_dominators.push(mach);
871 atomic_dominators.push(mach);
872 }
873 break;
874
875 default:
876 break;
877 }
878 }
879 }
880
881 // Step 2 - Find dominating accesses or allocations for each access
882 analyze_dominating_barriers_impl(loads, load_dominators);
883 analyze_dominating_barriers_impl(stores, store_dominators);
884 analyze_dominating_barriers_impl(atomics, atomic_dominators);
885 }
886
887 // == Reduced spilling optimization ==
888
889 void ZBarrierSetC2::compute_liveness_at_stubs() const {
890 ResourceMark rm;
891 Compile* const C = Compile::current();
892 Arena* const A = Thread::current()->resource_area();
893 PhaseCFG* const cfg = C->cfg();
894 PhaseRegAlloc* const regalloc = C->regalloc();
895 RegMask* const live = NEW_ARENA_ARRAY(A, RegMask, cfg->number_of_blocks() * sizeof(RegMask));
896 ZBarrierSetAssembler* const bs = ZBarrierSet::assembler();
897 Block_List worklist;
898
899 for (uint i = 0; i < cfg->number_of_blocks(); ++i) {
900 new ((void*)(live + i)) RegMask();
901 worklist.push(cfg->get_block(i));
902 }
903
904 while (worklist.size() > 0) {
905 const Block* const block = worklist.pop();
906 RegMask& old_live = live[block->_pre_order];
907 RegMask new_live;
908
909 // Initialize to union of successors
910 for (uint i = 0; i < block->_num_succs; i++) {
911 const uint succ_id = block->_succs[i]->_pre_order;
912 new_live.OR(live[succ_id]);
913 }
914
915 // Walk block backwards, computing liveness
916 for (int i = block->number_of_nodes() - 1; i >= 0; --i) {
917 const Node* const node = block->get_node(i);
918
919 // Remove def bits
920 const OptoReg::Name first = bs->refine_register(node, regalloc->get_reg_first(node));
921 const OptoReg::Name second = bs->refine_register(node, regalloc->get_reg_second(node));
922 if (first != OptoReg::Bad) {
923 new_live.Remove(first);
924 }
925 if (second != OptoReg::Bad) {
926 new_live.Remove(second);
927 }
928
929 // Add use bits
930 for (uint j = 1; j < node->req(); ++j) {
931 const Node* const use = node->in(j);
932 const OptoReg::Name first = bs->refine_register(use, regalloc->get_reg_first(use));
933 const OptoReg::Name second = bs->refine_register(use, regalloc->get_reg_second(use));
934 if (first != OptoReg::Bad) {
935 new_live.Insert(first);
936 }
937 if (second != OptoReg::Bad) {
938 new_live.Insert(second);
939 }
940 }
941
942 // If this node tracks liveness, update it
943 RegMask* const regs = barrier_set_state()->live(node);
944 if (regs != nullptr) {
945 regs->OR(new_live);
946 }
947 }
948
949 // Now at block top, see if we have any changes
950 new_live.SUBTRACT(old_live);
951 if (new_live.is_NotEmpty()) {
952 // Liveness has refined, update and propagate to prior blocks
953 old_live.OR(new_live);
954 for (uint i = 1; i < block->num_preds(); ++i) {
955 Block* const pred = cfg->get_block_for_node(block->pred(i));
956 worklist.push(pred);
957 }
958 }
959 }
960 }
961
962 void ZBarrierSetC2::eliminate_gc_barrier(PhaseMacroExpand* macro, Node* node) const {
963 eliminate_gc_barrier_data(node);
964 }
965
966 void ZBarrierSetC2::eliminate_gc_barrier_data(Node* node) const {
967 if (node->is_LoadStore()) {
968 LoadStoreNode* loadstore = node->as_LoadStore();
969 loadstore->set_barrier_data(ZBarrierElided);
970 } else if (node->is_Mem()) {
971 MemNode* mem = node->as_Mem();
972 mem->set_barrier_data(ZBarrierElided);
973 }
974 }
975
976 #ifndef PRODUCT
977 void ZBarrierSetC2::dump_barrier_data(const MachNode* mach, outputStream* st) const {
978 if ((mach->barrier_data() & ZBarrierStrong) != 0) {
979 st->print("strong ");
980 }
981 if ((mach->barrier_data() & ZBarrierWeak) != 0) {
982 st->print("weak ");
983 }
984 if ((mach->barrier_data() & ZBarrierPhantom) != 0) {
985 st->print("phantom ");
986 }
987 if ((mach->barrier_data() & ZBarrierNoKeepalive) != 0) {
988 st->print("nokeepalive ");
989 }
990 if ((mach->barrier_data() & ZBarrierNative) != 0) {
991 st->print("native ");
992 }
993 if ((mach->barrier_data() & ZBarrierElided) != 0) {
994 st->print("elided ");
995 }
996 }
997 #endif // !PRODUCT