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 "classfile/javaClasses.hpp"
  26 #include "gc/z/c2/zBarrierSetC2.hpp"
  27 #include "gc/z/zBarrierSet.hpp"
  28 #include "gc/z/zBarrierSetAssembler.hpp"
  29 #include "gc/z/zBarrierSetRuntime.hpp"
  30 #include "opto/block.hpp"
  31 #include "opto/compile.hpp"
  32 #include "opto/graphKit.hpp"
  33 #include "opto/machnode.hpp"
  34 #include "opto/memnode.hpp"
  35 #include "opto/node.hpp"
  36 #include "opto/regalloc.hpp"
  37 #include "opto/rootnode.hpp"
  38 #include "utilities/growableArray.hpp"
  39 #include "utilities/macros.hpp"
  40 
  41 class ZBarrierSetC2State : public ResourceObj {
  42 private:
  43   GrowableArray<ZLoadBarrierStubC2*>* _stubs;
  44   Node_Array                          _live;
  45 
  46 public:
  47   ZBarrierSetC2State(Arena* arena) :
  48     _stubs(new (arena) GrowableArray<ZLoadBarrierStubC2*>(arena, 8,  0, NULL)),
  49     _live(arena) {}
  50 
  51   GrowableArray<ZLoadBarrierStubC2*>* stubs() {
  52     return _stubs;
  53   }
  54 
  55   RegMask* live(const Node* node) {
  56     if (!node->is_Mach()) {
  57       // Don't need liveness for non-MachNodes
  58       return NULL;
  59     }
  60 
  61     const MachNode* const mach = node->as_Mach();
  62     if (mach->barrier_data() != ZLoadBarrierStrong &&
  63         mach->barrier_data() != ZLoadBarrierWeak) {
  64       // Don't need liveness data for nodes without barriers
  65       return NULL;
  66     }
  67 
  68     RegMask* live = (RegMask*)_live[node->_idx];
  69     if (live == NULL) {
  70       live = new (Compile::current()->comp_arena()->Amalloc_D(sizeof(RegMask))) RegMask();
  71       _live.map(node->_idx, (Node*)live);
  72     }
  73 
  74     return live;
  75   }
  76 };
  77 
  78 static ZBarrierSetC2State* barrier_set_state() {
  79   return reinterpret_cast<ZBarrierSetC2State*>(Compile::current()->barrier_set_state());
  80 }
  81 
  82 ZLoadBarrierStubC2* ZLoadBarrierStubC2::create(const MachNode* node, Address ref_addr, Register ref, Register tmp, bool weak) {
  83   ZLoadBarrierStubC2* const stub = new (Compile::current()->comp_arena()) ZLoadBarrierStubC2(node, ref_addr, ref, tmp, weak);
  84   if (!Compile::current()->in_scratch_emit_size()) {
  85     barrier_set_state()->stubs()->append(stub);
  86   }
  87 
  88   return stub;
  89 }
  90 
  91 ZLoadBarrierStubC2::ZLoadBarrierStubC2(const MachNode* node, Address ref_addr, Register ref, Register tmp, bool weak) :
  92     _node(node),
  93     _ref_addr(ref_addr),
  94     _ref(ref),
  95     _tmp(tmp),
  96     _weak(weak),
  97     _entry(),
  98     _continuation() {
  99   assert_different_registers(ref, ref_addr.base());
 100   assert_different_registers(ref, ref_addr.index());
 101 }
 102 
 103 Address ZLoadBarrierStubC2::ref_addr() const {
 104   return _ref_addr;
 105 }
 106 
 107 Register ZLoadBarrierStubC2::ref() const {
 108   return _ref;
 109 }
 110 
 111 Register ZLoadBarrierStubC2::tmp() const {
 112   return _tmp;
 113 }
 114 
 115 address ZLoadBarrierStubC2::slow_path() const {
 116   const DecoratorSet decorators = _weak ? ON_WEAK_OOP_REF : ON_STRONG_OOP_REF;
 117   return ZBarrierSetRuntime::load_barrier_on_oop_field_preloaded_addr(decorators);
 118 }
 119 
 120 RegMask& ZLoadBarrierStubC2::live() const {
 121   return *barrier_set_state()->live(_node);
 122 }
 123 
 124 Label* ZLoadBarrierStubC2::entry() {
 125   // The _entry will never be bound when in_scratch_emit_size() is true.
 126   // However, we still need to return a label that is not bound now, but
 127   // will eventually be bound. Any lable will do, as it will only act as
 128   // a placeholder, so we return the _continuation label.
 129   return Compile::current()->in_scratch_emit_size() ? &_continuation : &_entry;
 130 }
 131 
 132 Label* ZLoadBarrierStubC2::continuation() {
 133   return &_continuation;
 134 }
 135 
 136 void* ZBarrierSetC2::create_barrier_state(Arena* comp_arena) const {
 137   return new (comp_arena) ZBarrierSetC2State(comp_arena);
 138 }
 139 
 140 void ZBarrierSetC2::late_barrier_analysis() const {
 141   analyze_dominating_barriers();
 142   compute_liveness_at_stubs();
 143 }
 144 
 145 void ZBarrierSetC2::emit_stubs(CodeBuffer& cb) const {
 146   MacroAssembler masm(&cb);
 147   GrowableArray<ZLoadBarrierStubC2*>* const stubs = barrier_set_state()->stubs();
 148 
 149   for (int i = 0; i < stubs->length(); i++) {
 150     // Make sure there is enough space in the code buffer
 151     if (cb.insts()->maybe_expand_to_ensure_remaining(Compile::MAX_inst_size) && cb.blob() == NULL) {
 152       ciEnv::current()->record_failure("CodeCache is full");
 153       return;
 154     }
 155 
 156     ZBarrierSet::assembler()->generate_c2_load_barrier_stub(&masm, stubs->at(i));
 157   }
 158 
 159   masm.flush();
 160 }
 161 
 162 size_t ZBarrierSetC2::estimate_stub_size() const {
 163   Compile* const C = Compile::current();
 164   BufferBlob* const blob = C->scratch_buffer_blob();
 165   GrowableArray<ZLoadBarrierStubC2*>* const stubs = barrier_set_state()->stubs();
 166   size_t size = 0;
 167 
 168   for (int i = 0; i < stubs->length(); i++) {
 169     CodeBuffer cb(blob->content_begin(), (address)C->scratch_locs_memory() - blob->content_begin());
 170     MacroAssembler masm(&cb);
 171     ZBarrierSet::assembler()->generate_c2_load_barrier_stub(&masm, stubs->at(i));
 172     size += cb.insts_size();
 173   }
 174 
 175   return size;
 176 }
 177 
 178 static bool barrier_needed(C2Access& access) {
 179   return ZBarrierSet::barrier_needed(access.decorators(), access.type());
 180 }
 181 
 182 Node* ZBarrierSetC2::load_at_resolved(C2Access& access, const Type* val_type) const {
 183   Node* result = BarrierSetC2::load_at_resolved(access, val_type);
 184   if (barrier_needed(access) && access.raw_access()->is_Mem()) {
 185     if ((access.decorators() & ON_WEAK_OOP_REF) != 0) {
 186       access.raw_access()->as_Load()->set_barrier_data(ZLoadBarrierWeak);
 187     } else {
 188       access.raw_access()->as_Load()->set_barrier_data(ZLoadBarrierStrong);
 189     }
 190   }
 191 
 192   return result;
 193 }
 194 
 195 Node* ZBarrierSetC2::atomic_cmpxchg_val_at_resolved(C2AtomicParseAccess& access, Node* expected_val,
 196                                                     Node* new_val, const Type* val_type) const {
 197   Node* result = BarrierSetC2::atomic_cmpxchg_val_at_resolved(access, expected_val, new_val, val_type);
 198   if (barrier_needed(access)) {
 199     access.raw_access()->as_LoadStore()->set_barrier_data(ZLoadBarrierStrong);
 200   }
 201   return result;
 202 }
 203 
 204 Node* ZBarrierSetC2::atomic_cmpxchg_bool_at_resolved(C2AtomicParseAccess& access, Node* expected_val,
 205                                                      Node* new_val, const Type* value_type) const {
 206   Node* result = BarrierSetC2::atomic_cmpxchg_bool_at_resolved(access, expected_val, new_val, value_type);
 207   if (barrier_needed(access)) {
 208     access.raw_access()->as_LoadStore()->set_barrier_data(ZLoadBarrierStrong);
 209   }
 210   return result;
 211 }
 212 
 213 Node* ZBarrierSetC2::atomic_xchg_at_resolved(C2AtomicParseAccess& access, Node* new_val, const Type* val_type) const {
 214   Node* result = BarrierSetC2::atomic_xchg_at_resolved(access, new_val, val_type);
 215   if (barrier_needed(access)) {
 216     access.raw_access()->as_LoadStore()->set_barrier_data(ZLoadBarrierStrong);
 217   }
 218   return result;
 219 }
 220 
 221 bool ZBarrierSetC2::array_copy_requires_gc_barriers(bool tightly_coupled_alloc, BasicType type,
 222                                                     bool is_clone, ArrayCopyPhase phase) const {
 223   return type == T_OBJECT || type == T_ARRAY;
 224 }
 225 
 226 // == Dominating barrier elision ==
 227 
 228 static bool block_has_safepoint(const Block* block, uint from, uint to) {
 229   for (uint i = from; i < to; i++) {
 230     if (block->get_node(i)->is_MachSafePoint()) {
 231       // Safepoint found
 232       return true;
 233     }
 234   }
 235 
 236   // Safepoint not found
 237   return false;
 238 }
 239 
 240 static bool block_has_safepoint(const Block* block) {
 241   return block_has_safepoint(block, 0, block->number_of_nodes());
 242 }
 243 
 244 static uint block_index(const Block* block, const Node* node) {
 245   for (uint j = 0; j < block->number_of_nodes(); ++j) {
 246     if (block->get_node(j) == node) {
 247       return j;
 248     }
 249   }
 250   ShouldNotReachHere();
 251   return 0;
 252 }
 253 
 254 void ZBarrierSetC2::analyze_dominating_barriers() const {
 255   ResourceMark rm;
 256   Compile* const C = Compile::current();
 257   PhaseCFG* const cfg = C->cfg();
 258   Block_List worklist;
 259   Node_List mem_ops;
 260   Node_List barrier_loads;
 261 
 262   // Step 1 - Find accesses, and track them in lists
 263   for (uint i = 0; i < cfg->number_of_blocks(); ++i) {
 264     const Block* const block = cfg->get_block(i);
 265     for (uint j = 0; j < block->number_of_nodes(); ++j) {
 266       const Node* const node = block->get_node(j);
 267       if (!node->is_Mach()) {
 268         continue;
 269       }
 270 
 271       MachNode* const mach = node->as_Mach();
 272       switch (mach->ideal_Opcode()) {
 273       case Op_LoadP:
 274       case Op_CompareAndExchangeP:
 275       case Op_CompareAndSwapP:
 276       case Op_GetAndSetP:
 277         if (mach->barrier_data() == ZLoadBarrierStrong) {
 278           barrier_loads.push(mach);
 279         }
 280       case Op_StoreP:
 281         mem_ops.push(mach);
 282         break;
 283 
 284       default:
 285         break;
 286       }
 287     }
 288   }
 289 
 290   // Step 2 - Find dominating accesses for each load
 291   for (uint i = 0; i < barrier_loads.size(); i++) {
 292     MachNode* const load = barrier_loads.at(i)->as_Mach();
 293     const TypePtr* load_adr_type = NULL;
 294     intptr_t load_offset = 0;
 295     const Node* const load_obj = load->get_base_and_disp(load_offset, load_adr_type);
 296     Block* const load_block = cfg->get_block_for_node(load);
 297     const uint load_index = block_index(load_block, load);
 298 
 299     for (uint j = 0; j < mem_ops.size(); j++) {
 300       MachNode* mem = mem_ops.at(j)->as_Mach();
 301       const TypePtr* mem_adr_type = NULL;
 302       intptr_t mem_offset = 0;
 303       const Node* mem_obj = mem_obj = mem->get_base_and_disp(mem_offset, mem_adr_type);
 304       Block* mem_block = cfg->get_block_for_node(mem);
 305       uint mem_index = block_index(mem_block, mem);
 306 
 307       if (load_obj == NodeSentinel || mem_obj == NodeSentinel ||
 308           load_obj == NULL || mem_obj == NULL ||
 309           load_offset < 0 || mem_offset < 0) {
 310         continue;
 311       }
 312 
 313       if (mem_obj != load_obj || mem_offset != load_offset) {
 314         // Not the same addresses, not a candidate
 315         continue;
 316       }
 317 
 318       if (load_block == mem_block) {
 319         // Earlier accesses in the same block
 320         if (mem_index < load_index && !block_has_safepoint(mem_block, mem_index + 1, load_index)) {
 321           load->set_barrier_data(ZLoadBarrierElided);
 322         }
 323       } else if (mem_block->dominates(load_block)) {
 324         // Dominating block? Look around for safepoints
 325         ResourceMark rm;
 326         Block_List stack;
 327         VectorSet visited(Thread::current()->resource_area());
 328         stack.push(load_block);
 329         bool safepoint_found = block_has_safepoint(load_block);
 330         while (!safepoint_found && stack.size() > 0) {
 331           Block* block = stack.pop();
 332           if (visited.test_set(block->_pre_order)) {
 333             continue;
 334           }
 335           if (block_has_safepoint(block)) {
 336             safepoint_found = true;
 337             break;
 338           }
 339           if (block == mem_block) {
 340             continue;
 341           }
 342 
 343           // Push predecessor blocks
 344           for (uint p = 1; p < block->num_preds(); ++p) {
 345             Block* pred = cfg->get_block_for_node(block->pred(p));
 346             stack.push(pred);
 347           }
 348         }
 349 
 350         if (!safepoint_found) {
 351           load->set_barrier_data(ZLoadBarrierElided);
 352         }
 353       }
 354     }
 355   }
 356 }
 357 
 358 // == Reduced spilling optimization ==
 359 
 360 void ZBarrierSetC2::compute_liveness_at_stubs() const {
 361   ResourceMark rm;
 362   Compile* const C = Compile::current();
 363   Arena* const A = Thread::current()->resource_area();
 364   PhaseCFG* const cfg = C->cfg();
 365   PhaseRegAlloc* const regalloc = C->regalloc();
 366   RegMask* const live = NEW_ARENA_ARRAY(A, RegMask, cfg->number_of_blocks() * sizeof(RegMask));
 367   ZBarrierSetAssembler* const bs = ZBarrierSet::assembler();
 368   Block_List worklist;
 369 
 370   for (uint i = 0; i < cfg->number_of_blocks(); ++i) {
 371     new ((void*)(live + i)) RegMask();
 372     worklist.push(cfg->get_block(i));
 373   }
 374 
 375   while (worklist.size() > 0) {
 376     const Block* const block = worklist.pop();
 377     RegMask& old_live = live[block->_pre_order];
 378     RegMask new_live;
 379 
 380     // Initialize to union of successors
 381     for (uint i = 0; i < block->_num_succs; i++) {
 382       const uint succ_id = block->_succs[i]->_pre_order;
 383       new_live.OR(live[succ_id]);
 384     }
 385 
 386     // Walk block backwards, computing liveness
 387     for (int i = block->number_of_nodes() - 1; i >= 0; --i) {
 388       const Node* const node = block->get_node(i);
 389 
 390       // Remove def bits
 391       const OptoReg::Name first = bs->refine_register(node, regalloc->get_reg_first(node));
 392       const OptoReg::Name second = bs->refine_register(node, regalloc->get_reg_second(node));
 393       if (first != OptoReg::Bad) {
 394         new_live.Remove(first);
 395       }
 396       if (second != OptoReg::Bad) {
 397         new_live.Remove(second);
 398       }
 399 
 400       // Add use bits
 401       for (uint j = 1; j < node->req(); ++j) {
 402         const Node* const use = node->in(j);
 403         const OptoReg::Name first = bs->refine_register(use, regalloc->get_reg_first(use));
 404         const OptoReg::Name second = bs->refine_register(use, regalloc->get_reg_second(use));
 405         if (first != OptoReg::Bad) {
 406           new_live.Insert(first);
 407         }
 408         if (second != OptoReg::Bad) {
 409           new_live.Insert(second);
 410         }
 411       }
 412 
 413       // If this node tracks liveness, update it
 414       RegMask* const regs = barrier_set_state()->live(node);
 415       if (regs != NULL) {
 416         regs->OR(new_live);
 417       }
 418     }
 419 
 420     // Now at block top, see if we have any changes
 421     new_live.SUBTRACT(old_live);
 422     if (new_live.is_NotEmpty()) {
 423       // Liveness has refined, update and propagate to prior blocks
 424       old_live.OR(new_live);
 425       for (uint i = 1; i < block->num_preds(); ++i) {
 426         Block* const pred = cfg->get_block_for_node(block->pred(i));
 427         worklist.push(pred);
 428       }
 429     }
 430   }
 431 }