1 /* 2 * Copyright (c) 1998, 2011, Oracle and/or its affiliates. All rights reserved. 3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. 4 * 5 * This code is free software; you can redistribute it and/or modify it 6 * under the terms of the GNU General Public License version 2 only, as 7 * published by the Free Software Foundation. 8 * 9 * This code is distributed in the hope that it will be useful, but WITHOUT 10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 12 * version 2 for more details (a copy is included in the LICENSE file that 13 * accompanied this code). 14 * 15 * You should have received a copy of the GNU General Public License version 16 * 2 along with this work; if not, write to the Free Software Foundation, 17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. 18 * 19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA 20 * or visit www.oracle.com if you need additional information or have any 21 * questions. 22 * 23 */ 24 25 #include "precompiled.hpp" 26 #include "asm/assembler.inline.hpp" 27 #include "code/debugInfo.hpp" 28 #include "code/debugInfoRec.hpp" 29 #include "compiler/compileBroker.hpp" 30 #include "compiler/oopMap.hpp" 31 #include "memory/allocation.inline.hpp" 32 #include "opto/callnode.hpp" 33 #include "opto/cfgnode.hpp" 34 #include "opto/locknode.hpp" 35 #include "opto/machnode.hpp" 36 #include "opto/output.hpp" 37 #include "opto/regalloc.hpp" 38 #include "opto/runtime.hpp" 39 #include "opto/subnode.hpp" 40 #include "opto/type.hpp" 41 #include "runtime/handles.inline.hpp" 42 #include "utilities/xmlstream.hpp" 43 44 extern uint size_java_to_interp(); 45 extern uint reloc_java_to_interp(); 46 extern uint size_exception_handler(); 47 extern uint size_deopt_handler(); 48 49 #ifndef PRODUCT 50 #define DEBUG_ARG(x) , x 51 #else 52 #define DEBUG_ARG(x) 53 #endif 54 55 extern int emit_exception_handler(CodeBuffer &cbuf); 56 extern int emit_deopt_handler(CodeBuffer &cbuf); 57 58 //------------------------------Output----------------------------------------- 59 // Convert Nodes to instruction bits and pass off to the VM 60 void Compile::Output() { 61 // RootNode goes 62 assert( _cfg->_broot->_nodes.size() == 0, "" ); 63 64 // The number of new nodes (mostly MachNop) is proportional to 65 // the number of java calls and inner loops which are aligned. 66 if ( C->check_node_count((NodeLimitFudgeFactor + C->java_calls()*3 + 67 C->inner_loops()*(OptoLoopAlignment-1)), 68 "out of nodes before code generation" ) ) { 69 return; 70 } 71 // Make sure I can find the Start Node 72 Block_Array& bbs = _cfg->_bbs; 73 Block *entry = _cfg->_blocks[1]; 74 Block *broot = _cfg->_broot; 75 76 const StartNode *start = entry->_nodes[0]->as_Start(); 77 78 // Replace StartNode with prolog 79 MachPrologNode *prolog = new (this) MachPrologNode(); 80 entry->_nodes.map( 0, prolog ); 81 bbs.map( prolog->_idx, entry ); 82 bbs.map( start->_idx, NULL ); // start is no longer in any block 83 84 // Virtual methods need an unverified entry point 85 86 if( is_osr_compilation() ) { 87 if( PoisonOSREntry ) { 88 // TODO: Should use a ShouldNotReachHereNode... 89 _cfg->insert( broot, 0, new (this) MachBreakpointNode() ); 90 } 91 } else { 92 if( _method && !_method->flags().is_static() ) { 93 // Insert unvalidated entry point 94 _cfg->insert( broot, 0, new (this) MachUEPNode() ); 95 } 96 97 } 98 99 100 // Break before main entry point 101 if( (_method && _method->break_at_execute()) 102 #ifndef PRODUCT 103 ||(OptoBreakpoint && is_method_compilation()) 104 ||(OptoBreakpointOSR && is_osr_compilation()) 105 ||(OptoBreakpointC2R && !_method) 106 #endif 107 ) { 108 // checking for _method means that OptoBreakpoint does not apply to 109 // runtime stubs or frame converters 110 _cfg->insert( entry, 1, new (this) MachBreakpointNode() ); 111 } 112 113 // Insert epilogs before every return 114 for( uint i=0; i<_cfg->_num_blocks; i++ ) { 115 Block *b = _cfg->_blocks[i]; 116 if( !b->is_connector() && b->non_connector_successor(0) == _cfg->_broot ) { // Found a program exit point? 117 Node *m = b->end(); 118 if( m->is_Mach() && m->as_Mach()->ideal_Opcode() != Op_Halt ) { 119 MachEpilogNode *epilog = new (this) MachEpilogNode(m->as_Mach()->ideal_Opcode() == Op_Return); 120 b->add_inst( epilog ); 121 bbs.map(epilog->_idx, b); 122 //_regalloc->set_bad(epilog->_idx); // Already initialized this way. 123 } 124 } 125 } 126 127 # ifdef ENABLE_ZAP_DEAD_LOCALS 128 if ( ZapDeadCompiledLocals ) Insert_zap_nodes(); 129 # endif 130 131 uint* blk_starts = NEW_RESOURCE_ARRAY(uint,_cfg->_num_blocks+1); 132 blk_starts[0] = 0; 133 134 // Initialize code buffer and process short branches. 135 CodeBuffer* cb = init_buffer(blk_starts); 136 137 if (cb == NULL || failing()) return; 138 139 ScheduleAndBundle(); 140 141 #ifndef PRODUCT 142 if (trace_opto_output()) { 143 tty->print("\n---- After ScheduleAndBundle ----\n"); 144 for (uint i = 0; i < _cfg->_num_blocks; i++) { 145 tty->print("\nBB#%03d:\n", i); 146 Block *bb = _cfg->_blocks[i]; 147 for (uint j = 0; j < bb->_nodes.size(); j++) { 148 Node *n = bb->_nodes[j]; 149 OptoReg::Name reg = _regalloc->get_reg_first(n); 150 tty->print(" %-6s ", reg >= 0 && reg < REG_COUNT ? Matcher::regName[reg] : ""); 151 n->dump(); 152 } 153 } 154 } 155 #endif 156 157 if (failing()) return; 158 159 BuildOopMaps(); 160 161 if (failing()) return; 162 163 fill_buffer(cb, blk_starts); 164 } 165 166 bool Compile::need_stack_bang(int frame_size_in_bytes) const { 167 // Determine if we need to generate a stack overflow check. 168 // Do it if the method is not a stub function and 169 // has java calls or has frame size > vm_page_size/8. 170 return (UseStackBanging && stub_function() == NULL && 171 (has_java_calls() || frame_size_in_bytes > os::vm_page_size()>>3)); 172 } 173 174 bool Compile::need_register_stack_bang() const { 175 // Determine if we need to generate a register stack overflow check. 176 // This is only used on architectures which have split register 177 // and memory stacks (ie. IA64). 178 // Bang if the method is not a stub function and has java calls 179 return (stub_function() == NULL && has_java_calls()); 180 } 181 182 # ifdef ENABLE_ZAP_DEAD_LOCALS 183 184 185 // In order to catch compiler oop-map bugs, we have implemented 186 // a debugging mode called ZapDeadCompilerLocals. 187 // This mode causes the compiler to insert a call to a runtime routine, 188 // "zap_dead_locals", right before each place in compiled code 189 // that could potentially be a gc-point (i.e., a safepoint or oop map point). 190 // The runtime routine checks that locations mapped as oops are really 191 // oops, that locations mapped as values do not look like oops, 192 // and that locations mapped as dead are not used later 193 // (by zapping them to an invalid address). 194 195 int Compile::_CompiledZap_count = 0; 196 197 void Compile::Insert_zap_nodes() { 198 bool skip = false; 199 200 201 // Dink with static counts because code code without the extra 202 // runtime calls is MUCH faster for debugging purposes 203 204 if ( CompileZapFirst == 0 ) ; // nothing special 205 else if ( CompileZapFirst > CompiledZap_count() ) skip = true; 206 else if ( CompileZapFirst == CompiledZap_count() ) 207 warning("starting zap compilation after skipping"); 208 209 if ( CompileZapLast == -1 ) ; // nothing special 210 else if ( CompileZapLast < CompiledZap_count() ) skip = true; 211 else if ( CompileZapLast == CompiledZap_count() ) 212 warning("about to compile last zap"); 213 214 ++_CompiledZap_count; // counts skipped zaps, too 215 216 if ( skip ) return; 217 218 219 if ( _method == NULL ) 220 return; // no safepoints/oopmaps emitted for calls in stubs,so we don't care 221 222 // Insert call to zap runtime stub before every node with an oop map 223 for( uint i=0; i<_cfg->_num_blocks; i++ ) { 224 Block *b = _cfg->_blocks[i]; 225 for ( uint j = 0; j < b->_nodes.size(); ++j ) { 226 Node *n = b->_nodes[j]; 227 228 // Determining if we should insert a zap-a-lot node in output. 229 // We do that for all nodes that has oopmap info, except for calls 230 // to allocation. Calls to allocation passes in the old top-of-eden pointer 231 // and expect the C code to reset it. Hence, there can be no safepoints between 232 // the inlined-allocation and the call to new_Java, etc. 233 // We also cannot zap monitor calls, as they must hold the microlock 234 // during the call to Zap, which also wants to grab the microlock. 235 bool insert = n->is_MachSafePoint() && (n->as_MachSafePoint()->oop_map() != NULL); 236 if ( insert ) { // it is MachSafePoint 237 if ( !n->is_MachCall() ) { 238 insert = false; 239 } else if ( n->is_MachCall() ) { 240 MachCallNode* call = n->as_MachCall(); 241 if (call->entry_point() == OptoRuntime::new_instance_Java() || 242 call->entry_point() == OptoRuntime::new_array_Java() || 243 call->entry_point() == OptoRuntime::multianewarray2_Java() || 244 call->entry_point() == OptoRuntime::multianewarray3_Java() || 245 call->entry_point() == OptoRuntime::multianewarray4_Java() || 246 call->entry_point() == OptoRuntime::multianewarray5_Java() || 247 call->entry_point() == OptoRuntime::slow_arraycopy_Java() || 248 call->entry_point() == OptoRuntime::complete_monitor_locking_Java() 249 ) { 250 insert = false; 251 } 252 } 253 if (insert) { 254 Node *zap = call_zap_node(n->as_MachSafePoint(), i); 255 b->_nodes.insert( j, zap ); 256 _cfg->_bbs.map( zap->_idx, b ); 257 ++j; 258 } 259 } 260 } 261 } 262 } 263 264 265 Node* Compile::call_zap_node(MachSafePointNode* node_to_check, int block_no) { 266 const TypeFunc *tf = OptoRuntime::zap_dead_locals_Type(); 267 CallStaticJavaNode* ideal_node = 268 new (this) CallStaticJavaNode( tf, 269 OptoRuntime::zap_dead_locals_stub(_method->flags().is_native()), 270 "call zap dead locals stub", 0, TypePtr::BOTTOM); 271 // We need to copy the OopMap from the site we're zapping at. 272 // We have to make a copy, because the zap site might not be 273 // a call site, and zap_dead is a call site. 274 OopMap* clone = node_to_check->oop_map()->deep_copy(); 275 276 // Add the cloned OopMap to the zap node 277 ideal_node->set_oop_map(clone); 278 return _matcher->match_sfpt(ideal_node); 279 } 280 281 //------------------------------is_node_getting_a_safepoint-------------------- 282 bool Compile::is_node_getting_a_safepoint( Node* n) { 283 // This code duplicates the logic prior to the call of add_safepoint 284 // below in this file. 285 if( n->is_MachSafePoint() ) return true; 286 return false; 287 } 288 289 # endif // ENABLE_ZAP_DEAD_LOCALS 290 291 //------------------------------compute_loop_first_inst_sizes------------------ 292 // Compute the size of first NumberOfLoopInstrToAlign instructions at the top 293 // of a loop. When aligning a loop we need to provide enough instructions 294 // in cpu's fetch buffer to feed decoders. The loop alignment could be 295 // avoided if we have enough instructions in fetch buffer at the head of a loop. 296 // By default, the size is set to 999999 by Block's constructor so that 297 // a loop will be aligned if the size is not reset here. 298 // 299 // Note: Mach instructions could contain several HW instructions 300 // so the size is estimated only. 301 // 302 void Compile::compute_loop_first_inst_sizes() { 303 // The next condition is used to gate the loop alignment optimization. 304 // Don't aligned a loop if there are enough instructions at the head of a loop 305 // or alignment padding is larger then MaxLoopPad. By default, MaxLoopPad 306 // is equal to OptoLoopAlignment-1 except on new Intel cpus, where it is 307 // equal to 11 bytes which is the largest address NOP instruction. 308 if( MaxLoopPad < OptoLoopAlignment-1 ) { 309 uint last_block = _cfg->_num_blocks-1; 310 for( uint i=1; i <= last_block; i++ ) { 311 Block *b = _cfg->_blocks[i]; 312 // Check the first loop's block which requires an alignment. 313 if( b->loop_alignment() > (uint)relocInfo::addr_unit() ) { 314 uint sum_size = 0; 315 uint inst_cnt = NumberOfLoopInstrToAlign; 316 inst_cnt = b->compute_first_inst_size(sum_size, inst_cnt, _regalloc); 317 318 // Check subsequent fallthrough blocks if the loop's first 319 // block(s) does not have enough instructions. 320 Block *nb = b; 321 while( inst_cnt > 0 && 322 i < last_block && 323 !_cfg->_blocks[i+1]->has_loop_alignment() && 324 !nb->has_successor(b) ) { 325 i++; 326 nb = _cfg->_blocks[i]; 327 inst_cnt = nb->compute_first_inst_size(sum_size, inst_cnt, _regalloc); 328 } // while( inst_cnt > 0 && i < last_block ) 329 330 b->set_first_inst_size(sum_size); 331 } // f( b->head()->is_Loop() ) 332 } // for( i <= last_block ) 333 } // if( MaxLoopPad < OptoLoopAlignment-1 ) 334 } 335 336 //----------------------shorten_branches--------------------------------------- 337 // The architecture description provides short branch variants for some long 338 // branch instructions. Replace eligible long branches with short branches. 339 void Compile::shorten_branches(uint* blk_starts, int& code_size, int& reloc_size, int& stub_size) { 340 341 // ------------------ 342 // Compute size of each block, method size, and relocation information size 343 uint nblocks = _cfg->_num_blocks; 344 345 uint* jmp_offset = NEW_RESOURCE_ARRAY(uint,nblocks); 346 uint* jmp_size = NEW_RESOURCE_ARRAY(uint,nblocks); 347 int* jmp_nidx = NEW_RESOURCE_ARRAY(int ,nblocks); 348 DEBUG_ONLY( uint *jmp_target = NEW_RESOURCE_ARRAY(uint,nblocks); ) 349 DEBUG_ONLY( uint *jmp_rule = NEW_RESOURCE_ARRAY(uint,nblocks); ) 350 351 bool has_short_branch_candidate = false; 352 353 // Initialize the sizes to 0 354 code_size = 0; // Size in bytes of generated code 355 stub_size = 0; // Size in bytes of all stub entries 356 // Size in bytes of all relocation entries, including those in local stubs. 357 // Start with 2-bytes of reloc info for the unvalidated entry point 358 reloc_size = 1; // Number of relocation entries 359 360 // Make three passes. The first computes pessimistic blk_starts, 361 // relative jmp_offset and reloc_size information. The second performs 362 // short branch substitution using the pessimistic sizing. The 363 // third inserts nops where needed. 364 365 // Step one, perform a pessimistic sizing pass. 366 uint last_call_adr = max_uint; 367 uint last_avoid_back_to_back_adr = max_uint; 368 uint nop_size = (new (this) MachNopNode())->size(_regalloc); 369 for (uint i = 0; i < nblocks; i++) { // For all blocks 370 Block *b = _cfg->_blocks[i]; 371 372 // During short branch replacement, we store the relative (to blk_starts) 373 // offset of jump in jmp_offset, rather than the absolute offset of jump. 374 // This is so that we do not need to recompute sizes of all nodes when 375 // we compute correct blk_starts in our next sizing pass. 376 jmp_offset[i] = 0; 377 jmp_size[i] = 0; 378 jmp_nidx[i] = -1; 379 DEBUG_ONLY( jmp_target[i] = 0; ) 380 DEBUG_ONLY( jmp_rule[i] = 0; ) 381 382 // Sum all instruction sizes to compute block size 383 uint last_inst = b->_nodes.size(); 384 uint blk_size = 0; 385 for (uint j = 0; j < last_inst; j++) { 386 Node* nj = b->_nodes[j]; 387 // Handle machine instruction nodes 388 if (nj->is_Mach()) { 389 MachNode *mach = nj->as_Mach(); 390 blk_size += (mach->alignment_required() - 1) * relocInfo::addr_unit(); // assume worst case padding 391 reloc_size += mach->reloc(); 392 if( mach->is_MachCall() ) { 393 MachCallNode *mcall = mach->as_MachCall(); 394 // This destination address is NOT PC-relative 395 396 mcall->method_set((intptr_t)mcall->entry_point()); 397 398 if( mcall->is_MachCallJava() && mcall->as_MachCallJava()->_method ) { 399 stub_size += size_java_to_interp(); 400 reloc_size += reloc_java_to_interp(); 401 } 402 } else if (mach->is_MachSafePoint()) { 403 // If call/safepoint are adjacent, account for possible 404 // nop to disambiguate the two safepoints. 405 // ScheduleAndBundle() can rearrange nodes in a block, 406 // check for all offsets inside this block. 407 if (last_call_adr >= blk_starts[i]) { 408 blk_size += nop_size; 409 } 410 } 411 if (mach->avoid_back_to_back()) { 412 // Nop is inserted between "avoid back to back" instructions. 413 // ScheduleAndBundle() can rearrange nodes in a block, 414 // check for all offsets inside this block. 415 if (last_avoid_back_to_back_adr >= blk_starts[i]) { 416 blk_size += nop_size; 417 } 418 } 419 if (mach->may_be_short_branch()) { 420 if (!nj->is_MachBranch()) { 421 #ifndef PRODUCT 422 nj->dump(3); 423 #endif 424 Unimplemented(); 425 } 426 assert(jmp_nidx[i] == -1, "block should have only one branch"); 427 jmp_offset[i] = blk_size; 428 jmp_size[i] = nj->size(_regalloc); 429 jmp_nidx[i] = j; 430 has_short_branch_candidate = true; 431 } 432 } 433 blk_size += nj->size(_regalloc); 434 // Remember end of call offset 435 if (nj->is_MachCall() && !nj->is_MachCallLeaf()) { 436 last_call_adr = blk_starts[i]+blk_size; 437 } 438 // Remember end of avoid_back_to_back offset 439 if (nj->is_Mach() && nj->as_Mach()->avoid_back_to_back()) { 440 last_avoid_back_to_back_adr = blk_starts[i]+blk_size; 441 } 442 } 443 444 // When the next block starts a loop, we may insert pad NOP 445 // instructions. Since we cannot know our future alignment, 446 // assume the worst. 447 if (i< nblocks-1) { 448 Block *nb = _cfg->_blocks[i+1]; 449 int max_loop_pad = nb->code_alignment()-relocInfo::addr_unit(); 450 if (max_loop_pad > 0) { 451 assert(is_power_of_2(max_loop_pad+relocInfo::addr_unit()), ""); 452 // Adjust last_call_adr and/or last_avoid_back_to_back_adr. 453 // If either is the last instruction in this block, bump by 454 // max_loop_pad in lock-step with blk_size, so sizing 455 // calculations in subsequent blocks still can conservatively 456 // detect that it may the last instruction in this block. 457 if (last_call_adr == blk_starts[i]+blk_size) { 458 last_call_adr += max_loop_pad; 459 } 460 if (last_avoid_back_to_back_adr == blk_starts[i]+blk_size) { 461 last_avoid_back_to_back_adr += max_loop_pad; 462 } 463 blk_size += max_loop_pad; 464 } 465 } 466 467 // Save block size; update total method size 468 blk_starts[i+1] = blk_starts[i]+blk_size; 469 } 470 471 // Step two, replace eligible long jumps. 472 bool progress = true; 473 uint last_may_be_short_branch_adr = max_uint; 474 while (has_short_branch_candidate && progress) { 475 progress = false; 476 has_short_branch_candidate = false; 477 int adjust_block_start = 0; 478 for (uint i = 0; i < nblocks; i++) { 479 Block *b = _cfg->_blocks[i]; 480 int idx = jmp_nidx[i]; 481 MachNode* mach = (idx == -1) ? NULL: b->_nodes[idx]->as_Mach(); 482 if (mach != NULL && mach->may_be_short_branch()) { 483 #ifdef ASSERT 484 assert(jmp_size[i] > 0 && mach->is_MachBranch(), "sanity"); 485 int j; 486 // Find the branch; ignore trailing NOPs. 487 for (j = b->_nodes.size()-1; j>=0; j--) { 488 Node* n = b->_nodes[j]; 489 if (!n->is_Mach() || n->as_Mach()->ideal_Opcode() != Op_Con) 490 break; 491 } 492 assert(j >= 0 && j == idx && b->_nodes[j] == (Node*)mach, "sanity"); 493 #endif 494 int br_size = jmp_size[i]; 495 int br_offs = blk_starts[i] + jmp_offset[i]; 496 497 // This requires the TRUE branch target be in succs[0] 498 uint bnum = b->non_connector_successor(0)->_pre_order; 499 int offset = blk_starts[bnum] - br_offs; 500 if (bnum > i) { // adjust following block's offset 501 offset -= adjust_block_start; 502 } 503 // In the following code a nop could be inserted before 504 // the branch which will increase the backward distance. 505 bool needs_padding = ((uint)br_offs == last_may_be_short_branch_adr); 506 if (needs_padding && offset <= 0) 507 offset -= nop_size; 508 509 if (_matcher->is_short_branch_offset(mach->rule(), br_size, offset)) { 510 // We've got a winner. Replace this branch. 511 MachNode* replacement = mach->as_MachBranch()->short_branch_version(this); 512 513 // Update the jmp_size. 514 int new_size = replacement->size(_regalloc); 515 int diff = br_size - new_size; 516 assert(diff >= (int)nop_size, "short_branch size should be smaller"); 517 // Conservatively take into accound padding between 518 // avoid_back_to_back branches. Previous branch could be 519 // converted into avoid_back_to_back branch during next 520 // rounds. 521 if (needs_padding && replacement->avoid_back_to_back()) { 522 jmp_offset[i] += nop_size; 523 diff -= nop_size; 524 } 525 adjust_block_start += diff; 526 b->_nodes.map(idx, replacement); 527 mach->subsume_by(replacement, C); 528 mach = replacement; 529 progress = true; 530 531 jmp_size[i] = new_size; 532 DEBUG_ONLY( jmp_target[i] = bnum; ); 533 DEBUG_ONLY( jmp_rule[i] = mach->rule(); ); 534 } else { 535 // The jump distance is not short, try again during next iteration. 536 has_short_branch_candidate = true; 537 } 538 } // (mach->may_be_short_branch()) 539 if (mach != NULL && (mach->may_be_short_branch() || 540 mach->avoid_back_to_back())) { 541 last_may_be_short_branch_adr = blk_starts[i] + jmp_offset[i] + jmp_size[i]; 542 } 543 blk_starts[i+1] -= adjust_block_start; 544 } 545 } 546 547 #ifdef ASSERT 548 for (uint i = 0; i < nblocks; i++) { // For all blocks 549 if (jmp_target[i] != 0) { 550 int br_size = jmp_size[i]; 551 int offset = blk_starts[jmp_target[i]]-(blk_starts[i] + jmp_offset[i]); 552 if (!_matcher->is_short_branch_offset(jmp_rule[i], br_size, offset)) { 553 tty->print_cr("target (%d) - jmp_offset(%d) = offset (%d), jump_size(%d), jmp_block B%d, target_block B%d", blk_starts[jmp_target[i]], blk_starts[i] + jmp_offset[i], offset, br_size, i, jmp_target[i]); 554 } 555 assert(_matcher->is_short_branch_offset(jmp_rule[i], br_size, offset), "Displacement too large for short jmp"); 556 } 557 } 558 #endif 559 560 // Step 3, compute the offsets of all blocks, will be done in fill_buffer() 561 // after ScheduleAndBundle(). 562 563 // ------------------ 564 // Compute size for code buffer 565 code_size = blk_starts[nblocks]; 566 567 // Relocation records 568 reloc_size += 1; // Relo entry for exception handler 569 570 // Adjust reloc_size to number of record of relocation info 571 // Min is 2 bytes, max is probably 6 or 8, with a tax up to 25% for 572 // a relocation index. 573 // The CodeBuffer will expand the locs array if this estimate is too low. 574 reloc_size *= 10 / sizeof(relocInfo); 575 } 576 577 //------------------------------FillLocArray----------------------------------- 578 // Create a bit of debug info and append it to the array. The mapping is from 579 // Java local or expression stack to constant, register or stack-slot. For 580 // doubles, insert 2 mappings and return 1 (to tell the caller that the next 581 // entry has been taken care of and caller should skip it). 582 static LocationValue *new_loc_value( PhaseRegAlloc *ra, OptoReg::Name regnum, Location::Type l_type ) { 583 // This should never have accepted Bad before 584 assert(OptoReg::is_valid(regnum), "location must be valid"); 585 return (OptoReg::is_reg(regnum)) 586 ? new LocationValue(Location::new_reg_loc(l_type, OptoReg::as_VMReg(regnum)) ) 587 : new LocationValue(Location::new_stk_loc(l_type, ra->reg2offset(regnum))); 588 } 589 590 591 ObjectValue* 592 Compile::sv_for_node_id(GrowableArray<ScopeValue*> *objs, int id) { 593 for (int i = 0; i < objs->length(); i++) { 594 assert(objs->at(i)->is_object(), "corrupt object cache"); 595 ObjectValue* sv = (ObjectValue*) objs->at(i); 596 if (sv->id() == id) { 597 return sv; 598 } 599 } 600 // Otherwise.. 601 return NULL; 602 } 603 604 void Compile::set_sv_for_object_node(GrowableArray<ScopeValue*> *objs, 605 ObjectValue* sv ) { 606 assert(sv_for_node_id(objs, sv->id()) == NULL, "Precondition"); 607 objs->append(sv); 608 } 609 610 611 void Compile::FillLocArray( int idx, MachSafePointNode* sfpt, Node *local, 612 GrowableArray<ScopeValue*> *array, 613 GrowableArray<ScopeValue*> *objs ) { 614 assert( local, "use _top instead of null" ); 615 if (array->length() != idx) { 616 assert(array->length() == idx + 1, "Unexpected array count"); 617 // Old functionality: 618 // return 619 // New functionality: 620 // Assert if the local is not top. In product mode let the new node 621 // override the old entry. 622 assert(local == top(), "LocArray collision"); 623 if (local == top()) { 624 return; 625 } 626 array->pop(); 627 } 628 const Type *t = local->bottom_type(); 629 630 // Is it a safepoint scalar object node? 631 if (local->is_SafePointScalarObject()) { 632 SafePointScalarObjectNode* spobj = local->as_SafePointScalarObject(); 633 634 ObjectValue* sv = Compile::sv_for_node_id(objs, spobj->_idx); 635 if (sv == NULL) { 636 ciKlass* cik = t->is_oopptr()->klass(); 637 assert(cik->is_instance_klass() || 638 cik->is_array_klass(), "Not supported allocation."); 639 sv = new ObjectValue(spobj->_idx, 640 new ConstantOopWriteValue(cik->constant_encoding())); 641 Compile::set_sv_for_object_node(objs, sv); 642 643 uint first_ind = spobj->first_index(); 644 for (uint i = 0; i < spobj->n_fields(); i++) { 645 Node* fld_node = sfpt->in(first_ind+i); 646 (void)FillLocArray(sv->field_values()->length(), sfpt, fld_node, sv->field_values(), objs); 647 } 648 } 649 array->append(sv); 650 return; 651 } 652 653 // Grab the register number for the local 654 OptoReg::Name regnum = _regalloc->get_reg_first(local); 655 if( OptoReg::is_valid(regnum) ) {// Got a register/stack? 656 // Record the double as two float registers. 657 // The register mask for such a value always specifies two adjacent 658 // float registers, with the lower register number even. 659 // Normally, the allocation of high and low words to these registers 660 // is irrelevant, because nearly all operations on register pairs 661 // (e.g., StoreD) treat them as a single unit. 662 // Here, we assume in addition that the words in these two registers 663 // stored "naturally" (by operations like StoreD and double stores 664 // within the interpreter) such that the lower-numbered register 665 // is written to the lower memory address. This may seem like 666 // a machine dependency, but it is not--it is a requirement on 667 // the author of the <arch>.ad file to ensure that, for every 668 // even/odd double-register pair to which a double may be allocated, 669 // the word in the even single-register is stored to the first 670 // memory word. (Note that register numbers are completely 671 // arbitrary, and are not tied to any machine-level encodings.) 672 #ifdef _LP64 673 if( t->base() == Type::DoubleBot || t->base() == Type::DoubleCon ) { 674 array->append(new ConstantIntValue(0)); 675 array->append(new_loc_value( _regalloc, regnum, Location::dbl )); 676 } else if ( t->base() == Type::Long ) { 677 array->append(new ConstantIntValue(0)); 678 array->append(new_loc_value( _regalloc, regnum, Location::lng )); 679 } else if ( t->base() == Type::RawPtr ) { 680 // jsr/ret return address which must be restored into a the full 681 // width 64-bit stack slot. 682 array->append(new_loc_value( _regalloc, regnum, Location::lng )); 683 } 684 #else //_LP64 685 #ifdef SPARC 686 if (t->base() == Type::Long && OptoReg::is_reg(regnum)) { 687 // For SPARC we have to swap high and low words for 688 // long values stored in a single-register (g0-g7). 689 array->append(new_loc_value( _regalloc, regnum , Location::normal )); 690 array->append(new_loc_value( _regalloc, OptoReg::add(regnum,1), Location::normal )); 691 } else 692 #endif //SPARC 693 if( t->base() == Type::DoubleBot || t->base() == Type::DoubleCon || t->base() == Type::Long ) { 694 // Repack the double/long as two jints. 695 // The convention the interpreter uses is that the second local 696 // holds the first raw word of the native double representation. 697 // This is actually reasonable, since locals and stack arrays 698 // grow downwards in all implementations. 699 // (If, on some machine, the interpreter's Java locals or stack 700 // were to grow upwards, the embedded doubles would be word-swapped.) 701 array->append(new_loc_value( _regalloc, OptoReg::add(regnum,1), Location::normal )); 702 array->append(new_loc_value( _regalloc, regnum , Location::normal )); 703 } 704 #endif //_LP64 705 else if( (t->base() == Type::FloatBot || t->base() == Type::FloatCon) && 706 OptoReg::is_reg(regnum) ) { 707 array->append(new_loc_value( _regalloc, regnum, Matcher::float_in_double() 708 ? Location::float_in_dbl : Location::normal )); 709 } else if( t->base() == Type::Int && OptoReg::is_reg(regnum) ) { 710 array->append(new_loc_value( _regalloc, regnum, Matcher::int_in_long 711 ? Location::int_in_long : Location::normal )); 712 } else if( t->base() == Type::NarrowOop ) { 713 array->append(new_loc_value( _regalloc, regnum, Location::narrowoop )); 714 } else { 715 array->append(new_loc_value( _regalloc, regnum, _regalloc->is_oop(local) ? Location::oop : Location::normal )); 716 } 717 return; 718 } 719 720 // No register. It must be constant data. 721 switch (t->base()) { 722 case Type::Half: // Second half of a double 723 ShouldNotReachHere(); // Caller should skip 2nd halves 724 break; 725 case Type::AnyPtr: 726 array->append(new ConstantOopWriteValue(NULL)); 727 break; 728 case Type::AryPtr: 729 case Type::InstPtr: 730 case Type::KlassPtr: // fall through 731 array->append(new ConstantOopWriteValue(t->isa_oopptr()->const_oop()->constant_encoding())); 732 break; 733 case Type::NarrowOop: 734 if (t == TypeNarrowOop::NULL_PTR) { 735 array->append(new ConstantOopWriteValue(NULL)); 736 } else { 737 array->append(new ConstantOopWriteValue(t->make_ptr()->isa_oopptr()->const_oop()->constant_encoding())); 738 } 739 break; 740 case Type::Int: 741 array->append(new ConstantIntValue(t->is_int()->get_con())); 742 break; 743 case Type::RawPtr: 744 // A return address (T_ADDRESS). 745 assert((intptr_t)t->is_ptr()->get_con() < (intptr_t)0x10000, "must be a valid BCI"); 746 #ifdef _LP64 747 // Must be restored to the full-width 64-bit stack slot. 748 array->append(new ConstantLongValue(t->is_ptr()->get_con())); 749 #else 750 array->append(new ConstantIntValue(t->is_ptr()->get_con())); 751 #endif 752 break; 753 case Type::FloatCon: { 754 float f = t->is_float_constant()->getf(); 755 array->append(new ConstantIntValue(jint_cast(f))); 756 break; 757 } 758 case Type::DoubleCon: { 759 jdouble d = t->is_double_constant()->getd(); 760 #ifdef _LP64 761 array->append(new ConstantIntValue(0)); 762 array->append(new ConstantDoubleValue(d)); 763 #else 764 // Repack the double as two jints. 765 // The convention the interpreter uses is that the second local 766 // holds the first raw word of the native double representation. 767 // This is actually reasonable, since locals and stack arrays 768 // grow downwards in all implementations. 769 // (If, on some machine, the interpreter's Java locals or stack 770 // were to grow upwards, the embedded doubles would be word-swapped.) 771 jint *dp = (jint*)&d; 772 array->append(new ConstantIntValue(dp[1])); 773 array->append(new ConstantIntValue(dp[0])); 774 #endif 775 break; 776 } 777 case Type::Long: { 778 jlong d = t->is_long()->get_con(); 779 #ifdef _LP64 780 array->append(new ConstantIntValue(0)); 781 array->append(new ConstantLongValue(d)); 782 #else 783 // Repack the long as two jints. 784 // The convention the interpreter uses is that the second local 785 // holds the first raw word of the native double representation. 786 // This is actually reasonable, since locals and stack arrays 787 // grow downwards in all implementations. 788 // (If, on some machine, the interpreter's Java locals or stack 789 // were to grow upwards, the embedded doubles would be word-swapped.) 790 jint *dp = (jint*)&d; 791 array->append(new ConstantIntValue(dp[1])); 792 array->append(new ConstantIntValue(dp[0])); 793 #endif 794 break; 795 } 796 case Type::Top: // Add an illegal value here 797 array->append(new LocationValue(Location())); 798 break; 799 default: 800 ShouldNotReachHere(); 801 break; 802 } 803 } 804 805 // Determine if this node starts a bundle 806 bool Compile::starts_bundle(const Node *n) const { 807 return (_node_bundling_limit > n->_idx && 808 _node_bundling_base[n->_idx].starts_bundle()); 809 } 810 811 //--------------------------Process_OopMap_Node-------------------------------- 812 void Compile::Process_OopMap_Node(MachNode *mach, int current_offset) { 813 814 // Handle special safepoint nodes for synchronization 815 MachSafePointNode *sfn = mach->as_MachSafePoint(); 816 MachCallNode *mcall; 817 818 #ifdef ENABLE_ZAP_DEAD_LOCALS 819 assert( is_node_getting_a_safepoint(mach), "logic does not match; false negative"); 820 #endif 821 822 int safepoint_pc_offset = current_offset; 823 bool is_method_handle_invoke = false; 824 bool return_oop = false; 825 826 // Add the safepoint in the DebugInfoRecorder 827 if( !mach->is_MachCall() ) { 828 mcall = NULL; 829 debug_info()->add_safepoint(safepoint_pc_offset, sfn->_oop_map); 830 } else { 831 mcall = mach->as_MachCall(); 832 833 // Is the call a MethodHandle call? 834 if (mcall->is_MachCallJava()) { 835 if (mcall->as_MachCallJava()->_method_handle_invoke) { 836 assert(has_method_handle_invokes(), "must have been set during call generation"); 837 is_method_handle_invoke = true; 838 } 839 } 840 841 // Check if a call returns an object. 842 if (mcall->return_value_is_used() && 843 mcall->tf()->range()->field_at(TypeFunc::Parms)->isa_ptr()) { 844 return_oop = true; 845 } 846 safepoint_pc_offset += mcall->ret_addr_offset(); 847 debug_info()->add_safepoint(safepoint_pc_offset, mcall->_oop_map); 848 } 849 850 // Loop over the JVMState list to add scope information 851 // Do not skip safepoints with a NULL method, they need monitor info 852 JVMState* youngest_jvms = sfn->jvms(); 853 int max_depth = youngest_jvms->depth(); 854 855 // Allocate the object pool for scalar-replaced objects -- the map from 856 // small-integer keys (which can be recorded in the local and ostack 857 // arrays) to descriptions of the object state. 858 GrowableArray<ScopeValue*> *objs = new GrowableArray<ScopeValue*>(); 859 860 // Visit scopes from oldest to youngest. 861 for (int depth = 1; depth <= max_depth; depth++) { 862 JVMState* jvms = youngest_jvms->of_depth(depth); 863 int idx; 864 ciMethod* method = jvms->has_method() ? jvms->method() : NULL; 865 // Safepoints that do not have method() set only provide oop-map and monitor info 866 // to support GC; these do not support deoptimization. 867 int num_locs = (method == NULL) ? 0 : jvms->loc_size(); 868 int num_exps = (method == NULL) ? 0 : jvms->stk_size(); 869 int num_mon = jvms->nof_monitors(); 870 assert(method == NULL || jvms->bci() < 0 || num_locs == method->max_locals(), 871 "JVMS local count must match that of the method"); 872 873 // Add Local and Expression Stack Information 874 875 // Insert locals into the locarray 876 GrowableArray<ScopeValue*> *locarray = new GrowableArray<ScopeValue*>(num_locs); 877 for( idx = 0; idx < num_locs; idx++ ) { 878 FillLocArray( idx, sfn, sfn->local(jvms, idx), locarray, objs ); 879 } 880 881 // Insert expression stack entries into the exparray 882 GrowableArray<ScopeValue*> *exparray = new GrowableArray<ScopeValue*>(num_exps); 883 for( idx = 0; idx < num_exps; idx++ ) { 884 FillLocArray( idx, sfn, sfn->stack(jvms, idx), exparray, objs ); 885 } 886 887 // Add in mappings of the monitors 888 assert( !method || 889 !method->is_synchronized() || 890 method->is_native() || 891 num_mon > 0 || 892 !GenerateSynchronizationCode, 893 "monitors must always exist for synchronized methods"); 894 895 // Build the growable array of ScopeValues for exp stack 896 GrowableArray<MonitorValue*> *monarray = new GrowableArray<MonitorValue*>(num_mon); 897 898 // Loop over monitors and insert into array 899 for(idx = 0; idx < num_mon; idx++) { 900 // Grab the node that defines this monitor 901 Node* box_node = sfn->monitor_box(jvms, idx); 902 Node* obj_node = sfn->monitor_obj(jvms, idx); 903 904 // Create ScopeValue for object 905 ScopeValue *scval = NULL; 906 907 if( obj_node->is_SafePointScalarObject() ) { 908 SafePointScalarObjectNode* spobj = obj_node->as_SafePointScalarObject(); 909 scval = Compile::sv_for_node_id(objs, spobj->_idx); 910 if (scval == NULL) { 911 const Type *t = obj_node->bottom_type(); 912 ciKlass* cik = t->is_oopptr()->klass(); 913 assert(cik->is_instance_klass() || 914 cik->is_array_klass(), "Not supported allocation."); 915 ObjectValue* sv = new ObjectValue(spobj->_idx, 916 new ConstantOopWriteValue(cik->constant_encoding())); 917 Compile::set_sv_for_object_node(objs, sv); 918 919 uint first_ind = spobj->first_index(); 920 for (uint i = 0; i < spobj->n_fields(); i++) { 921 Node* fld_node = sfn->in(first_ind+i); 922 (void)FillLocArray(sv->field_values()->length(), sfn, fld_node, sv->field_values(), objs); 923 } 924 scval = sv; 925 } 926 } else if( !obj_node->is_Con() ) { 927 OptoReg::Name obj_reg = _regalloc->get_reg_first(obj_node); 928 if( obj_node->bottom_type()->base() == Type::NarrowOop ) { 929 scval = new_loc_value( _regalloc, obj_reg, Location::narrowoop ); 930 } else { 931 scval = new_loc_value( _regalloc, obj_reg, Location::oop ); 932 } 933 } else { 934 const TypePtr *tp = obj_node->get_ptr_type(); 935 scval = new ConstantOopWriteValue(tp->is_oopptr()->const_oop()->constant_encoding()); 936 } 937 938 OptoReg::Name box_reg = BoxLockNode::reg(box_node); 939 Location basic_lock = Location::new_stk_loc(Location::normal,_regalloc->reg2offset(box_reg)); 940 bool eliminated = (box_node->is_BoxLock() && box_node->as_BoxLock()->is_eliminated()); 941 monarray->append(new MonitorValue(scval, basic_lock, eliminated)); 942 } 943 944 // We dump the object pool first, since deoptimization reads it in first. 945 debug_info()->dump_object_pool(objs); 946 947 // Build first class objects to pass to scope 948 DebugToken *locvals = debug_info()->create_scope_values(locarray); 949 DebugToken *expvals = debug_info()->create_scope_values(exparray); 950 DebugToken *monvals = debug_info()->create_monitor_values(monarray); 951 952 // Make method available for all Safepoints 953 ciMethod* scope_method = method ? method : _method; 954 // Describe the scope here 955 assert(jvms->bci() >= InvocationEntryBci && jvms->bci() <= 0x10000, "must be a valid or entry BCI"); 956 assert(!jvms->should_reexecute() || depth == max_depth, "reexecute allowed only for the youngest"); 957 // Now we can describe the scope. 958 debug_info()->describe_scope(safepoint_pc_offset, scope_method, jvms->bci(), jvms->should_reexecute(), is_method_handle_invoke, return_oop, locvals, expvals, monvals); 959 } // End jvms loop 960 961 // Mark the end of the scope set. 962 debug_info()->end_safepoint(safepoint_pc_offset); 963 } 964 965 966 967 // A simplified version of Process_OopMap_Node, to handle non-safepoints. 968 class NonSafepointEmitter { 969 Compile* C; 970 JVMState* _pending_jvms; 971 int _pending_offset; 972 973 void emit_non_safepoint(); 974 975 public: 976 NonSafepointEmitter(Compile* compile) { 977 this->C = compile; 978 _pending_jvms = NULL; 979 _pending_offset = 0; 980 } 981 982 void observe_instruction(Node* n, int pc_offset) { 983 if (!C->debug_info()->recording_non_safepoints()) return; 984 985 Node_Notes* nn = C->node_notes_at(n->_idx); 986 if (nn == NULL || nn->jvms() == NULL) return; 987 if (_pending_jvms != NULL && 988 _pending_jvms->same_calls_as(nn->jvms())) { 989 // Repeated JVMS? Stretch it up here. 990 _pending_offset = pc_offset; 991 } else { 992 if (_pending_jvms != NULL && 993 _pending_offset < pc_offset) { 994 emit_non_safepoint(); 995 } 996 _pending_jvms = NULL; 997 if (pc_offset > C->debug_info()->last_pc_offset()) { 998 // This is the only way _pending_jvms can become non-NULL: 999 _pending_jvms = nn->jvms(); 1000 _pending_offset = pc_offset; 1001 } 1002 } 1003 } 1004 1005 // Stay out of the way of real safepoints: 1006 void observe_safepoint(JVMState* jvms, int pc_offset) { 1007 if (_pending_jvms != NULL && 1008 !_pending_jvms->same_calls_as(jvms) && 1009 _pending_offset < pc_offset) { 1010 emit_non_safepoint(); 1011 } 1012 _pending_jvms = NULL; 1013 } 1014 1015 void flush_at_end() { 1016 if (_pending_jvms != NULL) { 1017 emit_non_safepoint(); 1018 } 1019 _pending_jvms = NULL; 1020 } 1021 }; 1022 1023 void NonSafepointEmitter::emit_non_safepoint() { 1024 JVMState* youngest_jvms = _pending_jvms; 1025 int pc_offset = _pending_offset; 1026 1027 // Clear it now: 1028 _pending_jvms = NULL; 1029 1030 DebugInformationRecorder* debug_info = C->debug_info(); 1031 assert(debug_info->recording_non_safepoints(), "sanity"); 1032 1033 debug_info->add_non_safepoint(pc_offset); 1034 int max_depth = youngest_jvms->depth(); 1035 1036 // Visit scopes from oldest to youngest. 1037 for (int depth = 1; depth <= max_depth; depth++) { 1038 JVMState* jvms = youngest_jvms->of_depth(depth); 1039 ciMethod* method = jvms->has_method() ? jvms->method() : NULL; 1040 assert(!jvms->should_reexecute() || depth==max_depth, "reexecute allowed only for the youngest"); 1041 debug_info->describe_scope(pc_offset, method, jvms->bci(), jvms->should_reexecute()); 1042 } 1043 1044 // Mark the end of the scope set. 1045 debug_info->end_non_safepoint(pc_offset); 1046 } 1047 1048 1049 1050 // helper for fill_buffer bailout logic 1051 static void turn_off_compiler(Compile* C) { 1052 if (CodeCache::largest_free_block() >= CodeCacheMinimumFreeSpace*10) { 1053 // Do not turn off compilation if a single giant method has 1054 // blown the code cache size. 1055 C->record_failure("excessive request to CodeCache"); 1056 } else { 1057 // Let CompilerBroker disable further compilations. 1058 C->record_failure("CodeCache is full"); 1059 } 1060 } 1061 1062 1063 //------------------------------init_buffer------------------------------------ 1064 CodeBuffer* Compile::init_buffer(uint* blk_starts) { 1065 1066 // Set the initially allocated size 1067 int code_req = initial_code_capacity; 1068 int locs_req = initial_locs_capacity; 1069 int stub_req = TraceJumps ? initial_stub_capacity * 10 : initial_stub_capacity; 1070 int const_req = initial_const_capacity; 1071 1072 int pad_req = NativeCall::instruction_size; 1073 // The extra spacing after the code is necessary on some platforms. 1074 // Sometimes we need to patch in a jump after the last instruction, 1075 // if the nmethod has been deoptimized. (See 4932387, 4894843.) 1076 1077 // Compute the byte offset where we can store the deopt pc. 1078 if (fixed_slots() != 0) { 1079 _orig_pc_slot_offset_in_bytes = _regalloc->reg2offset(OptoReg::stack2reg(_orig_pc_slot)); 1080 } 1081 1082 // Compute prolog code size 1083 _method_size = 0; 1084 _frame_slots = OptoReg::reg2stack(_matcher->_old_SP)+_regalloc->_framesize; 1085 #ifdef IA64 1086 if (save_argument_registers()) { 1087 // 4815101: this is a stub with implicit and unknown precision fp args. 1088 // The usual spill mechanism can only generate stfd's in this case, which 1089 // doesn't work if the fp reg to spill contains a single-precision denorm. 1090 // Instead, we hack around the normal spill mechanism using stfspill's and 1091 // ldffill's in the MachProlog and MachEpilog emit methods. We allocate 1092 // space here for the fp arg regs (f8-f15) we're going to thusly spill. 1093 // 1094 // If we ever implement 16-byte 'registers' == stack slots, we can 1095 // get rid of this hack and have SpillCopy generate stfspill/ldffill 1096 // instead of stfd/stfs/ldfd/ldfs. 1097 _frame_slots += 8*(16/BytesPerInt); 1098 } 1099 #endif 1100 assert(_frame_slots >= 0 && _frame_slots < 1000000, "sanity check"); 1101 1102 if (has_mach_constant_base_node()) { 1103 // Fill the constant table. 1104 // Note: This must happen before shorten_branches. 1105 for (uint i = 0; i < _cfg->_num_blocks; i++) { 1106 Block* b = _cfg->_blocks[i]; 1107 1108 for (uint j = 0; j < b->_nodes.size(); j++) { 1109 Node* n = b->_nodes[j]; 1110 1111 // If the node is a MachConstantNode evaluate the constant 1112 // value section. 1113 if (n->is_MachConstant()) { 1114 MachConstantNode* machcon = n->as_MachConstant(); 1115 machcon->eval_constant(C); 1116 } 1117 } 1118 } 1119 1120 // Calculate the offsets of the constants and the size of the 1121 // constant table (including the padding to the next section). 1122 constant_table().calculate_offsets_and_size(); 1123 const_req = constant_table().size(); 1124 } 1125 1126 // Initialize the space for the BufferBlob used to find and verify 1127 // instruction size in MachNode::emit_size() 1128 init_scratch_buffer_blob(const_req); 1129 if (failing()) return NULL; // Out of memory 1130 1131 // Pre-compute the length of blocks and replace 1132 // long branches with short if machine supports it. 1133 shorten_branches(blk_starts, code_req, locs_req, stub_req); 1134 1135 // nmethod and CodeBuffer count stubs & constants as part of method's code. 1136 int exception_handler_req = size_exception_handler(); 1137 int deopt_handler_req = size_deopt_handler(); 1138 exception_handler_req += MAX_stubs_size; // add marginal slop for handler 1139 deopt_handler_req += MAX_stubs_size; // add marginal slop for handler 1140 stub_req += MAX_stubs_size; // ensure per-stub margin 1141 code_req += MAX_inst_size; // ensure per-instruction margin 1142 1143 if (StressCodeBuffers) 1144 code_req = const_req = stub_req = exception_handler_req = deopt_handler_req = 0x10; // force expansion 1145 1146 int total_req = 1147 const_req + 1148 code_req + 1149 pad_req + 1150 stub_req + 1151 exception_handler_req + 1152 deopt_handler_req; // deopt handler 1153 1154 if (has_method_handle_invokes()) 1155 total_req += deopt_handler_req; // deopt MH handler 1156 1157 CodeBuffer* cb = code_buffer(); 1158 cb->initialize(total_req, locs_req); 1159 1160 // Have we run out of code space? 1161 if ((cb->blob() == NULL) || (!CompileBroker::should_compile_new_jobs())) { 1162 turn_off_compiler(this); 1163 return NULL; 1164 } 1165 // Configure the code buffer. 1166 cb->initialize_consts_size(const_req); 1167 cb->initialize_stubs_size(stub_req); 1168 cb->initialize_oop_recorder(env()->oop_recorder()); 1169 1170 // fill in the nop array for bundling computations 1171 MachNode *_nop_list[Bundle::_nop_count]; 1172 Bundle::initialize_nops(_nop_list, this); 1173 1174 return cb; 1175 } 1176 1177 //------------------------------fill_buffer------------------------------------ 1178 void Compile::fill_buffer(CodeBuffer* cb, uint* blk_starts) { 1179 // blk_starts[] contains offsets calculated during short branches processing, 1180 // offsets should not be increased during following steps. 1181 1182 // Compute the size of first NumberOfLoopInstrToAlign instructions at head 1183 // of a loop. It is used to determine the padding for loop alignment. 1184 compute_loop_first_inst_sizes(); 1185 1186 // Create oopmap set. 1187 _oop_map_set = new OopMapSet(); 1188 1189 // !!!!! This preserves old handling of oopmaps for now 1190 debug_info()->set_oopmaps(_oop_map_set); 1191 1192 uint nblocks = _cfg->_num_blocks; 1193 // Count and start of implicit null check instructions 1194 uint inct_cnt = 0; 1195 uint *inct_starts = NEW_RESOURCE_ARRAY(uint, nblocks+1); 1196 1197 // Count and start of calls 1198 uint *call_returns = NEW_RESOURCE_ARRAY(uint, nblocks+1); 1199 1200 uint return_offset = 0; 1201 int nop_size = (new (this) MachNopNode())->size(_regalloc); 1202 1203 int previous_offset = 0; 1204 int current_offset = 0; 1205 int last_call_offset = -1; 1206 int last_avoid_back_to_back_offset = -1; 1207 #ifdef ASSERT 1208 uint* jmp_target = NEW_RESOURCE_ARRAY(uint,nblocks); 1209 uint* jmp_offset = NEW_RESOURCE_ARRAY(uint,nblocks); 1210 uint* jmp_size = NEW_RESOURCE_ARRAY(uint,nblocks); 1211 uint* jmp_rule = NEW_RESOURCE_ARRAY(uint,nblocks); 1212 #endif 1213 1214 // Create an array of unused labels, one for each basic block, if printing is enabled 1215 #ifndef PRODUCT 1216 int *node_offsets = NULL; 1217 uint node_offset_limit = unique(); 1218 1219 if (print_assembly()) 1220 node_offsets = NEW_RESOURCE_ARRAY(int, node_offset_limit); 1221 #endif 1222 1223 NonSafepointEmitter non_safepoints(this); // emit non-safepoints lazily 1224 1225 // Emit the constant table. 1226 if (has_mach_constant_base_node()) { 1227 constant_table().emit(*cb); 1228 } 1229 1230 // Create an array of labels, one for each basic block 1231 Label *blk_labels = NEW_RESOURCE_ARRAY(Label, nblocks+1); 1232 for (uint i=0; i <= nblocks; i++) { 1233 blk_labels[i].init(); 1234 } 1235 1236 // ------------------ 1237 // Now fill in the code buffer 1238 Node *delay_slot = NULL; 1239 1240 for (uint i=0; i < nblocks; i++) { 1241 Block *b = _cfg->_blocks[i]; 1242 1243 Node *head = b->head(); 1244 1245 // If this block needs to start aligned (i.e, can be reached other 1246 // than by falling-thru from the previous block), then force the 1247 // start of a new bundle. 1248 if (Pipeline::requires_bundling() && starts_bundle(head)) 1249 cb->flush_bundle(true); 1250 1251 #ifdef ASSERT 1252 if (!b->is_connector()) { 1253 stringStream st; 1254 b->dump_head(&_cfg->_bbs, &st); 1255 MacroAssembler(cb).block_comment(st.as_string()); 1256 } 1257 jmp_target[i] = 0; 1258 jmp_offset[i] = 0; 1259 jmp_size[i] = 0; 1260 jmp_rule[i] = 0; 1261 #endif 1262 int blk_offset = current_offset; 1263 1264 // Define the label at the beginning of the basic block 1265 MacroAssembler(cb).bind(blk_labels[b->_pre_order]); 1266 1267 uint last_inst = b->_nodes.size(); 1268 1269 // Emit block normally, except for last instruction. 1270 // Emit means "dump code bits into code buffer". 1271 for (uint j = 0; j<last_inst; j++) { 1272 1273 // Get the node 1274 Node* n = b->_nodes[j]; 1275 1276 // See if delay slots are supported 1277 if (valid_bundle_info(n) && 1278 node_bundling(n)->used_in_unconditional_delay()) { 1279 assert(delay_slot == NULL, "no use of delay slot node"); 1280 assert(n->size(_regalloc) == Pipeline::instr_unit_size(), "delay slot instruction wrong size"); 1281 1282 delay_slot = n; 1283 continue; 1284 } 1285 1286 // If this starts a new instruction group, then flush the current one 1287 // (but allow split bundles) 1288 if (Pipeline::requires_bundling() && starts_bundle(n)) 1289 cb->flush_bundle(false); 1290 1291 // The following logic is duplicated in the code ifdeffed for 1292 // ENABLE_ZAP_DEAD_LOCALS which appears above in this file. It 1293 // should be factored out. Or maybe dispersed to the nodes? 1294 1295 // Special handling for SafePoint/Call Nodes 1296 bool is_mcall = false; 1297 if (n->is_Mach()) { 1298 MachNode *mach = n->as_Mach(); 1299 is_mcall = n->is_MachCall(); 1300 bool is_sfn = n->is_MachSafePoint(); 1301 1302 // If this requires all previous instructions be flushed, then do so 1303 if (is_sfn || is_mcall || mach->alignment_required() != 1) { 1304 cb->flush_bundle(true); 1305 current_offset = cb->insts_size(); 1306 } 1307 1308 // A padding may be needed again since a previous instruction 1309 // could be moved to delay slot. 1310 1311 // align the instruction if necessary 1312 int padding = mach->compute_padding(current_offset); 1313 // Make sure safepoint node for polling is distinct from a call's 1314 // return by adding a nop if needed. 1315 if (is_sfn && !is_mcall && padding == 0 && current_offset == last_call_offset) { 1316 padding = nop_size; 1317 } 1318 if (padding == 0 && mach->avoid_back_to_back() && 1319 current_offset == last_avoid_back_to_back_offset) { 1320 // Avoid back to back some instructions. 1321 padding = nop_size; 1322 } 1323 1324 if(padding > 0) { 1325 assert((padding % nop_size) == 0, "padding is not a multiple of NOP size"); 1326 int nops_cnt = padding / nop_size; 1327 MachNode *nop = new (this) MachNopNode(nops_cnt); 1328 b->_nodes.insert(j++, nop); 1329 last_inst++; 1330 _cfg->_bbs.map( nop->_idx, b ); 1331 nop->emit(*cb, _regalloc); 1332 cb->flush_bundle(true); 1333 current_offset = cb->insts_size(); 1334 } 1335 1336 // Remember the start of the last call in a basic block 1337 if (is_mcall) { 1338 MachCallNode *mcall = mach->as_MachCall(); 1339 1340 // This destination address is NOT PC-relative 1341 mcall->method_set((intptr_t)mcall->entry_point()); 1342 1343 // Save the return address 1344 call_returns[b->_pre_order] = current_offset + mcall->ret_addr_offset(); 1345 1346 if (mcall->is_MachCallLeaf()) { 1347 is_mcall = false; 1348 is_sfn = false; 1349 } 1350 } 1351 1352 // sfn will be valid whenever mcall is valid now because of inheritance 1353 if (is_sfn || is_mcall) { 1354 1355 // Handle special safepoint nodes for synchronization 1356 if (!is_mcall) { 1357 MachSafePointNode *sfn = mach->as_MachSafePoint(); 1358 // !!!!! Stubs only need an oopmap right now, so bail out 1359 if (sfn->jvms()->method() == NULL) { 1360 // Write the oopmap directly to the code blob??!! 1361 # ifdef ENABLE_ZAP_DEAD_LOCALS 1362 assert( !is_node_getting_a_safepoint(sfn), "logic does not match; false positive"); 1363 # endif 1364 continue; 1365 } 1366 } // End synchronization 1367 1368 non_safepoints.observe_safepoint(mach->as_MachSafePoint()->jvms(), 1369 current_offset); 1370 Process_OopMap_Node(mach, current_offset); 1371 } // End if safepoint 1372 1373 // If this is a null check, then add the start of the previous instruction to the list 1374 else if( mach->is_MachNullCheck() ) { 1375 inct_starts[inct_cnt++] = previous_offset; 1376 } 1377 1378 // If this is a branch, then fill in the label with the target BB's label 1379 else if (mach->is_MachBranch()) { 1380 // This requires the TRUE branch target be in succs[0] 1381 uint block_num = b->non_connector_successor(0)->_pre_order; 1382 1383 // Try to replace long branch if delay slot is not used, 1384 // it is mostly for back branches since forward branch's 1385 // distance is not updated yet. 1386 bool delay_slot_is_used = valid_bundle_info(n) && 1387 node_bundling(n)->use_unconditional_delay(); 1388 if (!delay_slot_is_used && mach->may_be_short_branch()) { 1389 assert(delay_slot == NULL, "not expecting delay slot node"); 1390 int br_size = n->size(_regalloc); 1391 int offset = blk_starts[block_num] - current_offset; 1392 if (block_num >= i) { 1393 // Current and following block's offset are not 1394 // finilized yet, adjust distance by the difference 1395 // between calculated and final offsets of current block. 1396 offset -= (blk_starts[i] - blk_offset); 1397 } 1398 // In the following code a nop could be inserted before 1399 // the branch which will increase the backward distance. 1400 bool needs_padding = (current_offset == last_avoid_back_to_back_offset); 1401 if (needs_padding && offset <= 0) 1402 offset -= nop_size; 1403 1404 if (_matcher->is_short_branch_offset(mach->rule(), br_size, offset)) { 1405 // We've got a winner. Replace this branch. 1406 MachNode* replacement = mach->as_MachBranch()->short_branch_version(this); 1407 1408 // Update the jmp_size. 1409 int new_size = replacement->size(_regalloc); 1410 assert((br_size - new_size) >= (int)nop_size, "short_branch size should be smaller"); 1411 // Insert padding between avoid_back_to_back branches. 1412 if (needs_padding && replacement->avoid_back_to_back()) { 1413 MachNode *nop = new (this) MachNopNode(); 1414 b->_nodes.insert(j++, nop); 1415 _cfg->_bbs.map(nop->_idx, b); 1416 last_inst++; 1417 nop->emit(*cb, _regalloc); 1418 cb->flush_bundle(true); 1419 current_offset = cb->insts_size(); 1420 } 1421 #ifdef ASSERT 1422 jmp_target[i] = block_num; 1423 jmp_offset[i] = current_offset - blk_offset; 1424 jmp_size[i] = new_size; 1425 jmp_rule[i] = mach->rule(); 1426 #endif 1427 b->_nodes.map(j, replacement); 1428 mach->subsume_by(replacement, C); 1429 n = replacement; 1430 mach = replacement; 1431 } 1432 } 1433 mach->as_MachBranch()->label_set( &blk_labels[block_num], block_num ); 1434 } else if (mach->ideal_Opcode() == Op_Jump) { 1435 for (uint h = 0; h < b->_num_succs; h++) { 1436 Block* succs_block = b->_succs[h]; 1437 for (uint j = 1; j < succs_block->num_preds(); j++) { 1438 Node* jpn = succs_block->pred(j); 1439 if (jpn->is_JumpProj() && jpn->in(0) == mach) { 1440 uint block_num = succs_block->non_connector()->_pre_order; 1441 Label *blkLabel = &blk_labels[block_num]; 1442 mach->add_case_label(jpn->as_JumpProj()->proj_no(), blkLabel); 1443 } 1444 } 1445 } 1446 } 1447 1448 #ifdef ASSERT 1449 // Check that oop-store precedes the card-mark 1450 else if (mach->ideal_Opcode() == Op_StoreCM) { 1451 uint storeCM_idx = j; 1452 int count = 0; 1453 for (uint prec = mach->req(); prec < mach->len(); prec++) { 1454 Node *oop_store = mach->in(prec); // Precedence edge 1455 if (oop_store == NULL) continue; 1456 count++; 1457 uint i4; 1458 for( i4 = 0; i4 < last_inst; ++i4 ) { 1459 if( b->_nodes[i4] == oop_store ) break; 1460 } 1461 // Note: This test can provide a false failure if other precedence 1462 // edges have been added to the storeCMNode. 1463 assert( i4 == last_inst || i4 < storeCM_idx, "CM card-mark executes before oop-store"); 1464 } 1465 assert(count > 0, "storeCM expects at least one precedence edge"); 1466 } 1467 #endif 1468 1469 else if (!n->is_Proj()) { 1470 // Remember the beginning of the previous instruction, in case 1471 // it's followed by a flag-kill and a null-check. Happens on 1472 // Intel all the time, with add-to-memory kind of opcodes. 1473 previous_offset = current_offset; 1474 } 1475 } 1476 1477 // Verify that there is sufficient space remaining 1478 cb->insts()->maybe_expand_to_ensure_remaining(MAX_inst_size); 1479 if ((cb->blob() == NULL) || (!CompileBroker::should_compile_new_jobs())) { 1480 turn_off_compiler(this); 1481 return; 1482 } 1483 1484 // Save the offset for the listing 1485 #ifndef PRODUCT 1486 if (node_offsets && n->_idx < node_offset_limit) 1487 node_offsets[n->_idx] = cb->insts_size(); 1488 #endif 1489 1490 // "Normal" instruction case 1491 DEBUG_ONLY( uint instr_offset = cb->insts_size(); ) 1492 n->emit(*cb, _regalloc); 1493 current_offset = cb->insts_size(); 1494 1495 #ifdef ASSERT 1496 if (n->size(_regalloc) < (current_offset-instr_offset)) { 1497 n->dump(); 1498 assert(false, "wrong size of mach node"); 1499 } 1500 #endif 1501 non_safepoints.observe_instruction(n, current_offset); 1502 1503 // mcall is last "call" that can be a safepoint 1504 // record it so we can see if a poll will directly follow it 1505 // in which case we'll need a pad to make the PcDesc sites unique 1506 // see 5010568. This can be slightly inaccurate but conservative 1507 // in the case that return address is not actually at current_offset. 1508 // This is a small price to pay. 1509 1510 if (is_mcall) { 1511 last_call_offset = current_offset; 1512 } 1513 1514 if (n->is_Mach() && n->as_Mach()->avoid_back_to_back()) { 1515 // Avoid back to back some instructions. 1516 last_avoid_back_to_back_offset = current_offset; 1517 } 1518 1519 // See if this instruction has a delay slot 1520 if (valid_bundle_info(n) && node_bundling(n)->use_unconditional_delay()) { 1521 assert(delay_slot != NULL, "expecting delay slot node"); 1522 1523 // Back up 1 instruction 1524 cb->set_insts_end(cb->insts_end() - Pipeline::instr_unit_size()); 1525 1526 // Save the offset for the listing 1527 #ifndef PRODUCT 1528 if (node_offsets && delay_slot->_idx < node_offset_limit) 1529 node_offsets[delay_slot->_idx] = cb->insts_size(); 1530 #endif 1531 1532 // Support a SafePoint in the delay slot 1533 if (delay_slot->is_MachSafePoint()) { 1534 MachNode *mach = delay_slot->as_Mach(); 1535 // !!!!! Stubs only need an oopmap right now, so bail out 1536 if (!mach->is_MachCall() && mach->as_MachSafePoint()->jvms()->method() == NULL) { 1537 // Write the oopmap directly to the code blob??!! 1538 # ifdef ENABLE_ZAP_DEAD_LOCALS 1539 assert( !is_node_getting_a_safepoint(mach), "logic does not match; false positive"); 1540 # endif 1541 delay_slot = NULL; 1542 continue; 1543 } 1544 1545 int adjusted_offset = current_offset - Pipeline::instr_unit_size(); 1546 non_safepoints.observe_safepoint(mach->as_MachSafePoint()->jvms(), 1547 adjusted_offset); 1548 // Generate an OopMap entry 1549 Process_OopMap_Node(mach, adjusted_offset); 1550 } 1551 1552 // Insert the delay slot instruction 1553 delay_slot->emit(*cb, _regalloc); 1554 1555 // Don't reuse it 1556 delay_slot = NULL; 1557 } 1558 1559 } // End for all instructions in block 1560 1561 // If the next block is the top of a loop, pad this block out to align 1562 // the loop top a little. Helps prevent pipe stalls at loop back branches. 1563 if (i < nblocks-1) { 1564 Block *nb = _cfg->_blocks[i+1]; 1565 int padding = nb->alignment_padding(current_offset); 1566 if( padding > 0 ) { 1567 MachNode *nop = new (this) MachNopNode(padding / nop_size); 1568 b->_nodes.insert( b->_nodes.size(), nop ); 1569 _cfg->_bbs.map( nop->_idx, b ); 1570 nop->emit(*cb, _regalloc); 1571 current_offset = cb->insts_size(); 1572 } 1573 } 1574 // Verify that the distance for generated before forward 1575 // short branches is still valid. 1576 guarantee((int)(blk_starts[i+1] - blk_starts[i]) >= (current_offset - blk_offset), "shouldn't increase block size"); 1577 1578 // Save new block start offset 1579 blk_starts[i] = blk_offset; 1580 } // End of for all blocks 1581 blk_starts[nblocks] = current_offset; 1582 1583 non_safepoints.flush_at_end(); 1584 1585 // Offset too large? 1586 if (failing()) return; 1587 1588 // Define a pseudo-label at the end of the code 1589 MacroAssembler(cb).bind( blk_labels[nblocks] ); 1590 1591 // Compute the size of the first block 1592 _first_block_size = blk_labels[1].loc_pos() - blk_labels[0].loc_pos(); 1593 1594 assert(cb->insts_size() < 500000, "method is unreasonably large"); 1595 1596 #ifdef ASSERT 1597 for (uint i = 0; i < nblocks; i++) { // For all blocks 1598 if (jmp_target[i] != 0) { 1599 int br_size = jmp_size[i]; 1600 int offset = blk_starts[jmp_target[i]]-(blk_starts[i] + jmp_offset[i]); 1601 if (!_matcher->is_short_branch_offset(jmp_rule[i], br_size, offset)) { 1602 tty->print_cr("target (%d) - jmp_offset(%d) = offset (%d), jump_size(%d), jmp_block B%d, target_block B%d", blk_starts[jmp_target[i]], blk_starts[i] + jmp_offset[i], offset, br_size, i, jmp_target[i]); 1603 assert(false, "Displacement too large for short jmp"); 1604 } 1605 } 1606 } 1607 #endif 1608 1609 // ------------------ 1610 1611 #ifndef PRODUCT 1612 // Information on the size of the method, without the extraneous code 1613 Scheduling::increment_method_size(cb->insts_size()); 1614 #endif 1615 1616 // ------------------ 1617 // Fill in exception table entries. 1618 FillExceptionTables(inct_cnt, call_returns, inct_starts, blk_labels); 1619 1620 // Only java methods have exception handlers and deopt handlers 1621 if (_method) { 1622 // Emit the exception handler code. 1623 _code_offsets.set_value(CodeOffsets::Exceptions, emit_exception_handler(*cb)); 1624 // Emit the deopt handler code. 1625 _code_offsets.set_value(CodeOffsets::Deopt, emit_deopt_handler(*cb)); 1626 1627 // Emit the MethodHandle deopt handler code (if required). 1628 if (has_method_handle_invokes()) { 1629 // We can use the same code as for the normal deopt handler, we 1630 // just need a different entry point address. 1631 _code_offsets.set_value(CodeOffsets::DeoptMH, emit_deopt_handler(*cb)); 1632 } 1633 } 1634 1635 // One last check for failed CodeBuffer::expand: 1636 if ((cb->blob() == NULL) || (!CompileBroker::should_compile_new_jobs())) { 1637 turn_off_compiler(this); 1638 return; 1639 } 1640 1641 #ifndef PRODUCT 1642 // Dump the assembly code, including basic-block numbers 1643 if (print_assembly()) { 1644 ttyLocker ttyl; // keep the following output all in one block 1645 if (!VMThread::should_terminate()) { // test this under the tty lock 1646 // This output goes directly to the tty, not the compiler log. 1647 // To enable tools to match it up with the compilation activity, 1648 // be sure to tag this tty output with the compile ID. 1649 if (xtty != NULL) { 1650 xtty->head("opto_assembly compile_id='%d'%s", compile_id(), 1651 is_osr_compilation() ? " compile_kind='osr'" : 1652 ""); 1653 } 1654 if (method() != NULL) { 1655 method()->print_oop(); 1656 print_codes(); 1657 } 1658 dump_asm(node_offsets, node_offset_limit); 1659 if (xtty != NULL) { 1660 xtty->tail("opto_assembly"); 1661 } 1662 } 1663 } 1664 #endif 1665 1666 } 1667 1668 void Compile::FillExceptionTables(uint cnt, uint *call_returns, uint *inct_starts, Label *blk_labels) { 1669 _inc_table.set_size(cnt); 1670 1671 uint inct_cnt = 0; 1672 for( uint i=0; i<_cfg->_num_blocks; i++ ) { 1673 Block *b = _cfg->_blocks[i]; 1674 Node *n = NULL; 1675 int j; 1676 1677 // Find the branch; ignore trailing NOPs. 1678 for( j = b->_nodes.size()-1; j>=0; j-- ) { 1679 n = b->_nodes[j]; 1680 if( !n->is_Mach() || n->as_Mach()->ideal_Opcode() != Op_Con ) 1681 break; 1682 } 1683 1684 // If we didn't find anything, continue 1685 if( j < 0 ) continue; 1686 1687 // Compute ExceptionHandlerTable subtable entry and add it 1688 // (skip empty blocks) 1689 if( n->is_Catch() ) { 1690 1691 // Get the offset of the return from the call 1692 uint call_return = call_returns[b->_pre_order]; 1693 #ifdef ASSERT 1694 assert( call_return > 0, "no call seen for this basic block" ); 1695 while( b->_nodes[--j]->is_MachProj() ) ; 1696 assert( b->_nodes[j]->is_MachCall(), "CatchProj must follow call" ); 1697 #endif 1698 // last instruction is a CatchNode, find it's CatchProjNodes 1699 int nof_succs = b->_num_succs; 1700 // allocate space 1701 GrowableArray<intptr_t> handler_bcis(nof_succs); 1702 GrowableArray<intptr_t> handler_pcos(nof_succs); 1703 // iterate through all successors 1704 for (int j = 0; j < nof_succs; j++) { 1705 Block* s = b->_succs[j]; 1706 bool found_p = false; 1707 for( uint k = 1; k < s->num_preds(); k++ ) { 1708 Node *pk = s->pred(k); 1709 if( pk->is_CatchProj() && pk->in(0) == n ) { 1710 const CatchProjNode* p = pk->as_CatchProj(); 1711 found_p = true; 1712 // add the corresponding handler bci & pco information 1713 if( p->_con != CatchProjNode::fall_through_index ) { 1714 // p leads to an exception handler (and is not fall through) 1715 assert(s == _cfg->_blocks[s->_pre_order],"bad numbering"); 1716 // no duplicates, please 1717 if( !handler_bcis.contains(p->handler_bci()) ) { 1718 uint block_num = s->non_connector()->_pre_order; 1719 handler_bcis.append(p->handler_bci()); 1720 handler_pcos.append(blk_labels[block_num].loc_pos()); 1721 } 1722 } 1723 } 1724 } 1725 assert(found_p, "no matching predecessor found"); 1726 // Note: Due to empty block removal, one block may have 1727 // several CatchProj inputs, from the same Catch. 1728 } 1729 1730 // Set the offset of the return from the call 1731 _handler_table.add_subtable(call_return, &handler_bcis, NULL, &handler_pcos); 1732 continue; 1733 } 1734 1735 // Handle implicit null exception table updates 1736 if( n->is_MachNullCheck() ) { 1737 uint block_num = b->non_connector_successor(0)->_pre_order; 1738 _inc_table.append( inct_starts[inct_cnt++], blk_labels[block_num].loc_pos() ); 1739 continue; 1740 } 1741 } // End of for all blocks fill in exception table entries 1742 } 1743 1744 // Static Variables 1745 #ifndef PRODUCT 1746 uint Scheduling::_total_nop_size = 0; 1747 uint Scheduling::_total_method_size = 0; 1748 uint Scheduling::_total_branches = 0; 1749 uint Scheduling::_total_unconditional_delays = 0; 1750 uint Scheduling::_total_instructions_per_bundle[Pipeline::_max_instrs_per_cycle+1]; 1751 #endif 1752 1753 // Initializer for class Scheduling 1754 1755 Scheduling::Scheduling(Arena *arena, Compile &compile) 1756 : _arena(arena), 1757 _cfg(compile.cfg()), 1758 _bbs(compile.cfg()->_bbs), 1759 _regalloc(compile.regalloc()), 1760 _reg_node(arena), 1761 _bundle_instr_count(0), 1762 _bundle_cycle_number(0), 1763 _scheduled(arena), 1764 _available(arena), 1765 _next_node(NULL), 1766 _bundle_use(0, 0, resource_count, &_bundle_use_elements[0]), 1767 _pinch_free_list(arena) 1768 #ifndef PRODUCT 1769 , _branches(0) 1770 , _unconditional_delays(0) 1771 #endif 1772 { 1773 // Create a MachNopNode 1774 _nop = new (&compile) MachNopNode(); 1775 1776 // Now that the nops are in the array, save the count 1777 // (but allow entries for the nops) 1778 _node_bundling_limit = compile.unique(); 1779 uint node_max = _regalloc->node_regs_max_index(); 1780 1781 compile.set_node_bundling_limit(_node_bundling_limit); 1782 1783 // This one is persistent within the Compile class 1784 _node_bundling_base = NEW_ARENA_ARRAY(compile.comp_arena(), Bundle, node_max); 1785 1786 // Allocate space for fixed-size arrays 1787 _node_latency = NEW_ARENA_ARRAY(arena, unsigned short, node_max); 1788 _uses = NEW_ARENA_ARRAY(arena, short, node_max); 1789 _current_latency = NEW_ARENA_ARRAY(arena, unsigned short, node_max); 1790 1791 // Clear the arrays 1792 memset(_node_bundling_base, 0, node_max * sizeof(Bundle)); 1793 memset(_node_latency, 0, node_max * sizeof(unsigned short)); 1794 memset(_uses, 0, node_max * sizeof(short)); 1795 memset(_current_latency, 0, node_max * sizeof(unsigned short)); 1796 1797 // Clear the bundling information 1798 memcpy(_bundle_use_elements, 1799 Pipeline_Use::elaborated_elements, 1800 sizeof(Pipeline_Use::elaborated_elements)); 1801 1802 // Get the last node 1803 Block *bb = _cfg->_blocks[_cfg->_blocks.size()-1]; 1804 1805 _next_node = bb->_nodes[bb->_nodes.size()-1]; 1806 } 1807 1808 #ifndef PRODUCT 1809 // Scheduling destructor 1810 Scheduling::~Scheduling() { 1811 _total_branches += _branches; 1812 _total_unconditional_delays += _unconditional_delays; 1813 } 1814 #endif 1815 1816 // Step ahead "i" cycles 1817 void Scheduling::step(uint i) { 1818 1819 Bundle *bundle = node_bundling(_next_node); 1820 bundle->set_starts_bundle(); 1821 1822 // Update the bundle record, but leave the flags information alone 1823 if (_bundle_instr_count > 0) { 1824 bundle->set_instr_count(_bundle_instr_count); 1825 bundle->set_resources_used(_bundle_use.resourcesUsed()); 1826 } 1827 1828 // Update the state information 1829 _bundle_instr_count = 0; 1830 _bundle_cycle_number += i; 1831 _bundle_use.step(i); 1832 } 1833 1834 void Scheduling::step_and_clear() { 1835 Bundle *bundle = node_bundling(_next_node); 1836 bundle->set_starts_bundle(); 1837 1838 // Update the bundle record 1839 if (_bundle_instr_count > 0) { 1840 bundle->set_instr_count(_bundle_instr_count); 1841 bundle->set_resources_used(_bundle_use.resourcesUsed()); 1842 1843 _bundle_cycle_number += 1; 1844 } 1845 1846 // Clear the bundling information 1847 _bundle_instr_count = 0; 1848 _bundle_use.reset(); 1849 1850 memcpy(_bundle_use_elements, 1851 Pipeline_Use::elaborated_elements, 1852 sizeof(Pipeline_Use::elaborated_elements)); 1853 } 1854 1855 //------------------------------ScheduleAndBundle------------------------------ 1856 // Perform instruction scheduling and bundling over the sequence of 1857 // instructions in backwards order. 1858 void Compile::ScheduleAndBundle() { 1859 1860 // Don't optimize this if it isn't a method 1861 if (!_method) 1862 return; 1863 1864 // Don't optimize this if scheduling is disabled 1865 if (!do_scheduling()) 1866 return; 1867 1868 // Scheduling code works only with pairs (8 bytes) maximum. 1869 if (max_vector_size() > 8) 1870 return; 1871 1872 NOT_PRODUCT( TracePhase t2("isched", &_t_instrSched, TimeCompiler); ) 1873 1874 // Create a data structure for all the scheduling information 1875 Scheduling scheduling(Thread::current()->resource_area(), *this); 1876 1877 // Walk backwards over each basic block, computing the needed alignment 1878 // Walk over all the basic blocks 1879 scheduling.DoScheduling(); 1880 } 1881 1882 //------------------------------ComputeLocalLatenciesForward------------------- 1883 // Compute the latency of all the instructions. This is fairly simple, 1884 // because we already have a legal ordering. Walk over the instructions 1885 // from first to last, and compute the latency of the instruction based 1886 // on the latency of the preceding instruction(s). 1887 void Scheduling::ComputeLocalLatenciesForward(const Block *bb) { 1888 #ifndef PRODUCT 1889 if (_cfg->C->trace_opto_output()) 1890 tty->print("# -> ComputeLocalLatenciesForward\n"); 1891 #endif 1892 1893 // Walk over all the schedulable instructions 1894 for( uint j=_bb_start; j < _bb_end; j++ ) { 1895 1896 // This is a kludge, forcing all latency calculations to start at 1. 1897 // Used to allow latency 0 to force an instruction to the beginning 1898 // of the bb 1899 uint latency = 1; 1900 Node *use = bb->_nodes[j]; 1901 uint nlen = use->len(); 1902 1903 // Walk over all the inputs 1904 for ( uint k=0; k < nlen; k++ ) { 1905 Node *def = use->in(k); 1906 if (!def) 1907 continue; 1908 1909 uint l = _node_latency[def->_idx] + use->latency(k); 1910 if (latency < l) 1911 latency = l; 1912 } 1913 1914 _node_latency[use->_idx] = latency; 1915 1916 #ifndef PRODUCT 1917 if (_cfg->C->trace_opto_output()) { 1918 tty->print("# latency %4d: ", latency); 1919 use->dump(); 1920 } 1921 #endif 1922 } 1923 1924 #ifndef PRODUCT 1925 if (_cfg->C->trace_opto_output()) 1926 tty->print("# <- ComputeLocalLatenciesForward\n"); 1927 #endif 1928 1929 } // end ComputeLocalLatenciesForward 1930 1931 // See if this node fits into the present instruction bundle 1932 bool Scheduling::NodeFitsInBundle(Node *n) { 1933 uint n_idx = n->_idx; 1934 1935 // If this is the unconditional delay instruction, then it fits 1936 if (n == _unconditional_delay_slot) { 1937 #ifndef PRODUCT 1938 if (_cfg->C->trace_opto_output()) 1939 tty->print("# NodeFitsInBundle [%4d]: TRUE; is in unconditional delay slot\n", n->_idx); 1940 #endif 1941 return (true); 1942 } 1943 1944 // If the node cannot be scheduled this cycle, skip it 1945 if (_current_latency[n_idx] > _bundle_cycle_number) { 1946 #ifndef PRODUCT 1947 if (_cfg->C->trace_opto_output()) 1948 tty->print("# NodeFitsInBundle [%4d]: FALSE; latency %4d > %d\n", 1949 n->_idx, _current_latency[n_idx], _bundle_cycle_number); 1950 #endif 1951 return (false); 1952 } 1953 1954 const Pipeline *node_pipeline = n->pipeline(); 1955 1956 uint instruction_count = node_pipeline->instructionCount(); 1957 if (node_pipeline->mayHaveNoCode() && n->size(_regalloc) == 0) 1958 instruction_count = 0; 1959 else if (node_pipeline->hasBranchDelay() && !_unconditional_delay_slot) 1960 instruction_count++; 1961 1962 if (_bundle_instr_count + instruction_count > Pipeline::_max_instrs_per_cycle) { 1963 #ifndef PRODUCT 1964 if (_cfg->C->trace_opto_output()) 1965 tty->print("# NodeFitsInBundle [%4d]: FALSE; too many instructions: %d > %d\n", 1966 n->_idx, _bundle_instr_count + instruction_count, Pipeline::_max_instrs_per_cycle); 1967 #endif 1968 return (false); 1969 } 1970 1971 // Don't allow non-machine nodes to be handled this way 1972 if (!n->is_Mach() && instruction_count == 0) 1973 return (false); 1974 1975 // See if there is any overlap 1976 uint delay = _bundle_use.full_latency(0, node_pipeline->resourceUse()); 1977 1978 if (delay > 0) { 1979 #ifndef PRODUCT 1980 if (_cfg->C->trace_opto_output()) 1981 tty->print("# NodeFitsInBundle [%4d]: FALSE; functional units overlap\n", n_idx); 1982 #endif 1983 return false; 1984 } 1985 1986 #ifndef PRODUCT 1987 if (_cfg->C->trace_opto_output()) 1988 tty->print("# NodeFitsInBundle [%4d]: TRUE\n", n_idx); 1989 #endif 1990 1991 return true; 1992 } 1993 1994 Node * Scheduling::ChooseNodeToBundle() { 1995 uint siz = _available.size(); 1996 1997 if (siz == 0) { 1998 1999 #ifndef PRODUCT 2000 if (_cfg->C->trace_opto_output()) 2001 tty->print("# ChooseNodeToBundle: NULL\n"); 2002 #endif 2003 return (NULL); 2004 } 2005 2006 // Fast path, if only 1 instruction in the bundle 2007 if (siz == 1) { 2008 #ifndef PRODUCT 2009 if (_cfg->C->trace_opto_output()) { 2010 tty->print("# ChooseNodeToBundle (only 1): "); 2011 _available[0]->dump(); 2012 } 2013 #endif 2014 return (_available[0]); 2015 } 2016 2017 // Don't bother, if the bundle is already full 2018 if (_bundle_instr_count < Pipeline::_max_instrs_per_cycle) { 2019 for ( uint i = 0; i < siz; i++ ) { 2020 Node *n = _available[i]; 2021 2022 // Skip projections, we'll handle them another way 2023 if (n->is_Proj()) 2024 continue; 2025 2026 // This presupposed that instructions are inserted into the 2027 // available list in a legality order; i.e. instructions that 2028 // must be inserted first are at the head of the list 2029 if (NodeFitsInBundle(n)) { 2030 #ifndef PRODUCT 2031 if (_cfg->C->trace_opto_output()) { 2032 tty->print("# ChooseNodeToBundle: "); 2033 n->dump(); 2034 } 2035 #endif 2036 return (n); 2037 } 2038 } 2039 } 2040 2041 // Nothing fits in this bundle, choose the highest priority 2042 #ifndef PRODUCT 2043 if (_cfg->C->trace_opto_output()) { 2044 tty->print("# ChooseNodeToBundle: "); 2045 _available[0]->dump(); 2046 } 2047 #endif 2048 2049 return _available[0]; 2050 } 2051 2052 //------------------------------AddNodeToAvailableList------------------------- 2053 void Scheduling::AddNodeToAvailableList(Node *n) { 2054 assert( !n->is_Proj(), "projections never directly made available" ); 2055 #ifndef PRODUCT 2056 if (_cfg->C->trace_opto_output()) { 2057 tty->print("# AddNodeToAvailableList: "); 2058 n->dump(); 2059 } 2060 #endif 2061 2062 int latency = _current_latency[n->_idx]; 2063 2064 // Insert in latency order (insertion sort) 2065 uint i; 2066 for ( i=0; i < _available.size(); i++ ) 2067 if (_current_latency[_available[i]->_idx] > latency) 2068 break; 2069 2070 // Special Check for compares following branches 2071 if( n->is_Mach() && _scheduled.size() > 0 ) { 2072 int op = n->as_Mach()->ideal_Opcode(); 2073 Node *last = _scheduled[0]; 2074 if( last->is_MachIf() && last->in(1) == n && 2075 ( op == Op_CmpI || 2076 op == Op_CmpU || 2077 op == Op_CmpP || 2078 op == Op_CmpF || 2079 op == Op_CmpD || 2080 op == Op_CmpL ) ) { 2081 2082 // Recalculate position, moving to front of same latency 2083 for ( i=0 ; i < _available.size(); i++ ) 2084 if (_current_latency[_available[i]->_idx] >= latency) 2085 break; 2086 } 2087 } 2088 2089 // Insert the node in the available list 2090 _available.insert(i, n); 2091 2092 #ifndef PRODUCT 2093 if (_cfg->C->trace_opto_output()) 2094 dump_available(); 2095 #endif 2096 } 2097 2098 //------------------------------DecrementUseCounts----------------------------- 2099 void Scheduling::DecrementUseCounts(Node *n, const Block *bb) { 2100 for ( uint i=0; i < n->len(); i++ ) { 2101 Node *def = n->in(i); 2102 if (!def) continue; 2103 if( def->is_Proj() ) // If this is a machine projection, then 2104 def = def->in(0); // propagate usage thru to the base instruction 2105 2106 if( _bbs[def->_idx] != bb ) // Ignore if not block-local 2107 continue; 2108 2109 // Compute the latency 2110 uint l = _bundle_cycle_number + n->latency(i); 2111 if (_current_latency[def->_idx] < l) 2112 _current_latency[def->_idx] = l; 2113 2114 // If this does not have uses then schedule it 2115 if ((--_uses[def->_idx]) == 0) 2116 AddNodeToAvailableList(def); 2117 } 2118 } 2119 2120 //------------------------------AddNodeToBundle-------------------------------- 2121 void Scheduling::AddNodeToBundle(Node *n, const Block *bb) { 2122 #ifndef PRODUCT 2123 if (_cfg->C->trace_opto_output()) { 2124 tty->print("# AddNodeToBundle: "); 2125 n->dump(); 2126 } 2127 #endif 2128 2129 // Remove this from the available list 2130 uint i; 2131 for (i = 0; i < _available.size(); i++) 2132 if (_available[i] == n) 2133 break; 2134 assert(i < _available.size(), "entry in _available list not found"); 2135 _available.remove(i); 2136 2137 // See if this fits in the current bundle 2138 const Pipeline *node_pipeline = n->pipeline(); 2139 const Pipeline_Use& node_usage = node_pipeline->resourceUse(); 2140 2141 // Check for instructions to be placed in the delay slot. We 2142 // do this before we actually schedule the current instruction, 2143 // because the delay slot follows the current instruction. 2144 if (Pipeline::_branch_has_delay_slot && 2145 node_pipeline->hasBranchDelay() && 2146 !_unconditional_delay_slot) { 2147 2148 uint siz = _available.size(); 2149 2150 // Conditional branches can support an instruction that 2151 // is unconditionally executed and not dependent by the 2152 // branch, OR a conditionally executed instruction if 2153 // the branch is taken. In practice, this means that 2154 // the first instruction at the branch target is 2155 // copied to the delay slot, and the branch goes to 2156 // the instruction after that at the branch target 2157 if ( n->is_MachBranch() ) { 2158 2159 assert( !n->is_MachNullCheck(), "should not look for delay slot for Null Check" ); 2160 assert( !n->is_Catch(), "should not look for delay slot for Catch" ); 2161 2162 #ifndef PRODUCT 2163 _branches++; 2164 #endif 2165 2166 // At least 1 instruction is on the available list 2167 // that is not dependent on the branch 2168 for (uint i = 0; i < siz; i++) { 2169 Node *d = _available[i]; 2170 const Pipeline *avail_pipeline = d->pipeline(); 2171 2172 // Don't allow safepoints in the branch shadow, that will 2173 // cause a number of difficulties 2174 if ( avail_pipeline->instructionCount() == 1 && 2175 !avail_pipeline->hasMultipleBundles() && 2176 !avail_pipeline->hasBranchDelay() && 2177 Pipeline::instr_has_unit_size() && 2178 d->size(_regalloc) == Pipeline::instr_unit_size() && 2179 NodeFitsInBundle(d) && 2180 !node_bundling(d)->used_in_delay()) { 2181 2182 if (d->is_Mach() && !d->is_MachSafePoint()) { 2183 // A node that fits in the delay slot was found, so we need to 2184 // set the appropriate bits in the bundle pipeline information so 2185 // that it correctly indicates resource usage. Later, when we 2186 // attempt to add this instruction to the bundle, we will skip 2187 // setting the resource usage. 2188 _unconditional_delay_slot = d; 2189 node_bundling(n)->set_use_unconditional_delay(); 2190 node_bundling(d)->set_used_in_unconditional_delay(); 2191 _bundle_use.add_usage(avail_pipeline->resourceUse()); 2192 _current_latency[d->_idx] = _bundle_cycle_number; 2193 _next_node = d; 2194 ++_bundle_instr_count; 2195 #ifndef PRODUCT 2196 _unconditional_delays++; 2197 #endif 2198 break; 2199 } 2200 } 2201 } 2202 } 2203 2204 // No delay slot, add a nop to the usage 2205 if (!_unconditional_delay_slot) { 2206 // See if adding an instruction in the delay slot will overflow 2207 // the bundle. 2208 if (!NodeFitsInBundle(_nop)) { 2209 #ifndef PRODUCT 2210 if (_cfg->C->trace_opto_output()) 2211 tty->print("# *** STEP(1 instruction for delay slot) ***\n"); 2212 #endif 2213 step(1); 2214 } 2215 2216 _bundle_use.add_usage(_nop->pipeline()->resourceUse()); 2217 _next_node = _nop; 2218 ++_bundle_instr_count; 2219 } 2220 2221 // See if the instruction in the delay slot requires a 2222 // step of the bundles 2223 if (!NodeFitsInBundle(n)) { 2224 #ifndef PRODUCT 2225 if (_cfg->C->trace_opto_output()) 2226 tty->print("# *** STEP(branch won't fit) ***\n"); 2227 #endif 2228 // Update the state information 2229 _bundle_instr_count = 0; 2230 _bundle_cycle_number += 1; 2231 _bundle_use.step(1); 2232 } 2233 } 2234 2235 // Get the number of instructions 2236 uint instruction_count = node_pipeline->instructionCount(); 2237 if (node_pipeline->mayHaveNoCode() && n->size(_regalloc) == 0) 2238 instruction_count = 0; 2239 2240 // Compute the latency information 2241 uint delay = 0; 2242 2243 if (instruction_count > 0 || !node_pipeline->mayHaveNoCode()) { 2244 int relative_latency = _current_latency[n->_idx] - _bundle_cycle_number; 2245 if (relative_latency < 0) 2246 relative_latency = 0; 2247 2248 delay = _bundle_use.full_latency(relative_latency, node_usage); 2249 2250 // Does not fit in this bundle, start a new one 2251 if (delay > 0) { 2252 step(delay); 2253 2254 #ifndef PRODUCT 2255 if (_cfg->C->trace_opto_output()) 2256 tty->print("# *** STEP(%d) ***\n", delay); 2257 #endif 2258 } 2259 } 2260 2261 // If this was placed in the delay slot, ignore it 2262 if (n != _unconditional_delay_slot) { 2263 2264 if (delay == 0) { 2265 if (node_pipeline->hasMultipleBundles()) { 2266 #ifndef PRODUCT 2267 if (_cfg->C->trace_opto_output()) 2268 tty->print("# *** STEP(multiple instructions) ***\n"); 2269 #endif 2270 step(1); 2271 } 2272 2273 else if (instruction_count + _bundle_instr_count > Pipeline::_max_instrs_per_cycle) { 2274 #ifndef PRODUCT 2275 if (_cfg->C->trace_opto_output()) 2276 tty->print("# *** STEP(%d >= %d instructions) ***\n", 2277 instruction_count + _bundle_instr_count, 2278 Pipeline::_max_instrs_per_cycle); 2279 #endif 2280 step(1); 2281 } 2282 } 2283 2284 if (node_pipeline->hasBranchDelay() && !_unconditional_delay_slot) 2285 _bundle_instr_count++; 2286 2287 // Set the node's latency 2288 _current_latency[n->_idx] = _bundle_cycle_number; 2289 2290 // Now merge the functional unit information 2291 if (instruction_count > 0 || !node_pipeline->mayHaveNoCode()) 2292 _bundle_use.add_usage(node_usage); 2293 2294 // Increment the number of instructions in this bundle 2295 _bundle_instr_count += instruction_count; 2296 2297 // Remember this node for later 2298 if (n->is_Mach()) 2299 _next_node = n; 2300 } 2301 2302 // It's possible to have a BoxLock in the graph and in the _bbs mapping but 2303 // not in the bb->_nodes array. This happens for debug-info-only BoxLocks. 2304 // 'Schedule' them (basically ignore in the schedule) but do not insert them 2305 // into the block. All other scheduled nodes get put in the schedule here. 2306 int op = n->Opcode(); 2307 if( (op == Op_Node && n->req() == 0) || // anti-dependence node OR 2308 (op != Op_Node && // Not an unused antidepedence node and 2309 // not an unallocated boxlock 2310 (OptoReg::is_valid(_regalloc->get_reg_first(n)) || op != Op_BoxLock)) ) { 2311 2312 // Push any trailing projections 2313 if( bb->_nodes[bb->_nodes.size()-1] != n ) { 2314 for (DUIterator_Fast imax, i = n->fast_outs(imax); i < imax; i++) { 2315 Node *foi = n->fast_out(i); 2316 if( foi->is_Proj() ) 2317 _scheduled.push(foi); 2318 } 2319 } 2320 2321 // Put the instruction in the schedule list 2322 _scheduled.push(n); 2323 } 2324 2325 #ifndef PRODUCT 2326 if (_cfg->C->trace_opto_output()) 2327 dump_available(); 2328 #endif 2329 2330 // Walk all the definitions, decrementing use counts, and 2331 // if a definition has a 0 use count, place it in the available list. 2332 DecrementUseCounts(n,bb); 2333 } 2334 2335 //------------------------------ComputeUseCount-------------------------------- 2336 // This method sets the use count within a basic block. We will ignore all 2337 // uses outside the current basic block. As we are doing a backwards walk, 2338 // any node we reach that has a use count of 0 may be scheduled. This also 2339 // avoids the problem of cyclic references from phi nodes, as long as phi 2340 // nodes are at the front of the basic block. This method also initializes 2341 // the available list to the set of instructions that have no uses within this 2342 // basic block. 2343 void Scheduling::ComputeUseCount(const Block *bb) { 2344 #ifndef PRODUCT 2345 if (_cfg->C->trace_opto_output()) 2346 tty->print("# -> ComputeUseCount\n"); 2347 #endif 2348 2349 // Clear the list of available and scheduled instructions, just in case 2350 _available.clear(); 2351 _scheduled.clear(); 2352 2353 // No delay slot specified 2354 _unconditional_delay_slot = NULL; 2355 2356 #ifdef ASSERT 2357 for( uint i=0; i < bb->_nodes.size(); i++ ) 2358 assert( _uses[bb->_nodes[i]->_idx] == 0, "_use array not clean" ); 2359 #endif 2360 2361 // Force the _uses count to never go to zero for unscheduable pieces 2362 // of the block 2363 for( uint k = 0; k < _bb_start; k++ ) 2364 _uses[bb->_nodes[k]->_idx] = 1; 2365 for( uint l = _bb_end; l < bb->_nodes.size(); l++ ) 2366 _uses[bb->_nodes[l]->_idx] = 1; 2367 2368 // Iterate backwards over the instructions in the block. Don't count the 2369 // branch projections at end or the block header instructions. 2370 for( uint j = _bb_end-1; j >= _bb_start; j-- ) { 2371 Node *n = bb->_nodes[j]; 2372 if( n->is_Proj() ) continue; // Projections handled another way 2373 2374 // Account for all uses 2375 for ( uint k = 0; k < n->len(); k++ ) { 2376 Node *inp = n->in(k); 2377 if (!inp) continue; 2378 assert(inp != n, "no cycles allowed" ); 2379 if( _bbs[inp->_idx] == bb ) { // Block-local use? 2380 if( inp->is_Proj() ) // Skip through Proj's 2381 inp = inp->in(0); 2382 ++_uses[inp->_idx]; // Count 1 block-local use 2383 } 2384 } 2385 2386 // If this instruction has a 0 use count, then it is available 2387 if (!_uses[n->_idx]) { 2388 _current_latency[n->_idx] = _bundle_cycle_number; 2389 AddNodeToAvailableList(n); 2390 } 2391 2392 #ifndef PRODUCT 2393 if (_cfg->C->trace_opto_output()) { 2394 tty->print("# uses: %3d: ", _uses[n->_idx]); 2395 n->dump(); 2396 } 2397 #endif 2398 } 2399 2400 #ifndef PRODUCT 2401 if (_cfg->C->trace_opto_output()) 2402 tty->print("# <- ComputeUseCount\n"); 2403 #endif 2404 } 2405 2406 // This routine performs scheduling on each basic block in reverse order, 2407 // using instruction latencies and taking into account function unit 2408 // availability. 2409 void Scheduling::DoScheduling() { 2410 #ifndef PRODUCT 2411 if (_cfg->C->trace_opto_output()) 2412 tty->print("# -> DoScheduling\n"); 2413 #endif 2414 2415 Block *succ_bb = NULL; 2416 Block *bb; 2417 2418 // Walk over all the basic blocks in reverse order 2419 for( int i=_cfg->_num_blocks-1; i >= 0; succ_bb = bb, i-- ) { 2420 bb = _cfg->_blocks[i]; 2421 2422 #ifndef PRODUCT 2423 if (_cfg->C->trace_opto_output()) { 2424 tty->print("# Schedule BB#%03d (initial)\n", i); 2425 for (uint j = 0; j < bb->_nodes.size(); j++) 2426 bb->_nodes[j]->dump(); 2427 } 2428 #endif 2429 2430 // On the head node, skip processing 2431 if( bb == _cfg->_broot ) 2432 continue; 2433 2434 // Skip empty, connector blocks 2435 if (bb->is_connector()) 2436 continue; 2437 2438 // If the following block is not the sole successor of 2439 // this one, then reset the pipeline information 2440 if (bb->_num_succs != 1 || bb->non_connector_successor(0) != succ_bb) { 2441 #ifndef PRODUCT 2442 if (_cfg->C->trace_opto_output()) { 2443 tty->print("*** bundle start of next BB, node %d, for %d instructions\n", 2444 _next_node->_idx, _bundle_instr_count); 2445 } 2446 #endif 2447 step_and_clear(); 2448 } 2449 2450 // Leave untouched the starting instruction, any Phis, a CreateEx node 2451 // or Top. bb->_nodes[_bb_start] is the first schedulable instruction. 2452 _bb_end = bb->_nodes.size()-1; 2453 for( _bb_start=1; _bb_start <= _bb_end; _bb_start++ ) { 2454 Node *n = bb->_nodes[_bb_start]; 2455 // Things not matched, like Phinodes and ProjNodes don't get scheduled. 2456 // Also, MachIdealNodes do not get scheduled 2457 if( !n->is_Mach() ) continue; // Skip non-machine nodes 2458 MachNode *mach = n->as_Mach(); 2459 int iop = mach->ideal_Opcode(); 2460 if( iop == Op_CreateEx ) continue; // CreateEx is pinned 2461 if( iop == Op_Con ) continue; // Do not schedule Top 2462 if( iop == Op_Node && // Do not schedule PhiNodes, ProjNodes 2463 mach->pipeline() == MachNode::pipeline_class() && 2464 !n->is_SpillCopy() ) // Breakpoints, Prolog, etc 2465 continue; 2466 break; // Funny loop structure to be sure... 2467 } 2468 // Compute last "interesting" instruction in block - last instruction we 2469 // might schedule. _bb_end points just after last schedulable inst. We 2470 // normally schedule conditional branches (despite them being forced last 2471 // in the block), because they have delay slots we can fill. Calls all 2472 // have their delay slots filled in the template expansions, so we don't 2473 // bother scheduling them. 2474 Node *last = bb->_nodes[_bb_end]; 2475 // Ignore trailing NOPs. 2476 while (_bb_end > 0 && last->is_Mach() && 2477 last->as_Mach()->ideal_Opcode() == Op_Con) { 2478 last = bb->_nodes[--_bb_end]; 2479 } 2480 assert(!last->is_Mach() || last->as_Mach()->ideal_Opcode() != Op_Con, ""); 2481 if( last->is_Catch() || 2482 // Exclude unreachable path case when Halt node is in a separate block. 2483 (_bb_end > 1 && last->is_Mach() && last->as_Mach()->ideal_Opcode() == Op_Halt) ) { 2484 // There must be a prior call. Skip it. 2485 while( !bb->_nodes[--_bb_end]->is_MachCall() ) { 2486 assert( bb->_nodes[_bb_end]->is_MachProj(), "skipping projections after expected call" ); 2487 } 2488 } else if( last->is_MachNullCheck() ) { 2489 // Backup so the last null-checked memory instruction is 2490 // outside the schedulable range. Skip over the nullcheck, 2491 // projection, and the memory nodes. 2492 Node *mem = last->in(1); 2493 do { 2494 _bb_end--; 2495 } while (mem != bb->_nodes[_bb_end]); 2496 } else { 2497 // Set _bb_end to point after last schedulable inst. 2498 _bb_end++; 2499 } 2500 2501 assert( _bb_start <= _bb_end, "inverted block ends" ); 2502 2503 // Compute the register antidependencies for the basic block 2504 ComputeRegisterAntidependencies(bb); 2505 if (_cfg->C->failing()) return; // too many D-U pinch points 2506 2507 // Compute intra-bb latencies for the nodes 2508 ComputeLocalLatenciesForward(bb); 2509 2510 // Compute the usage within the block, and set the list of all nodes 2511 // in the block that have no uses within the block. 2512 ComputeUseCount(bb); 2513 2514 // Schedule the remaining instructions in the block 2515 while ( _available.size() > 0 ) { 2516 Node *n = ChooseNodeToBundle(); 2517 AddNodeToBundle(n,bb); 2518 } 2519 2520 assert( _scheduled.size() == _bb_end - _bb_start, "wrong number of instructions" ); 2521 #ifdef ASSERT 2522 for( uint l = _bb_start; l < _bb_end; l++ ) { 2523 Node *n = bb->_nodes[l]; 2524 uint m; 2525 for( m = 0; m < _bb_end-_bb_start; m++ ) 2526 if( _scheduled[m] == n ) 2527 break; 2528 assert( m < _bb_end-_bb_start, "instruction missing in schedule" ); 2529 } 2530 #endif 2531 2532 // Now copy the instructions (in reverse order) back to the block 2533 for ( uint k = _bb_start; k < _bb_end; k++ ) 2534 bb->_nodes.map(k, _scheduled[_bb_end-k-1]); 2535 2536 #ifndef PRODUCT 2537 if (_cfg->C->trace_opto_output()) { 2538 tty->print("# Schedule BB#%03d (final)\n", i); 2539 uint current = 0; 2540 for (uint j = 0; j < bb->_nodes.size(); j++) { 2541 Node *n = bb->_nodes[j]; 2542 if( valid_bundle_info(n) ) { 2543 Bundle *bundle = node_bundling(n); 2544 if (bundle->instr_count() > 0 || bundle->flags() > 0) { 2545 tty->print("*** Bundle: "); 2546 bundle->dump(); 2547 } 2548 n->dump(); 2549 } 2550 } 2551 } 2552 #endif 2553 #ifdef ASSERT 2554 verify_good_schedule(bb,"after block local scheduling"); 2555 #endif 2556 } 2557 2558 #ifndef PRODUCT 2559 if (_cfg->C->trace_opto_output()) 2560 tty->print("# <- DoScheduling\n"); 2561 #endif 2562 2563 // Record final node-bundling array location 2564 _regalloc->C->set_node_bundling_base(_node_bundling_base); 2565 2566 } // end DoScheduling 2567 2568 //------------------------------verify_good_schedule--------------------------- 2569 // Verify that no live-range used in the block is killed in the block by a 2570 // wrong DEF. This doesn't verify live-ranges that span blocks. 2571 2572 // Check for edge existence. Used to avoid adding redundant precedence edges. 2573 static bool edge_from_to( Node *from, Node *to ) { 2574 for( uint i=0; i<from->len(); i++ ) 2575 if( from->in(i) == to ) 2576 return true; 2577 return false; 2578 } 2579 2580 #ifdef ASSERT 2581 //------------------------------verify_do_def---------------------------------- 2582 void Scheduling::verify_do_def( Node *n, OptoReg::Name def, const char *msg ) { 2583 // Check for bad kills 2584 if( OptoReg::is_valid(def) ) { // Ignore stores & control flow 2585 Node *prior_use = _reg_node[def]; 2586 if( prior_use && !edge_from_to(prior_use,n) ) { 2587 tty->print("%s = ",OptoReg::as_VMReg(def)->name()); 2588 n->dump(); 2589 tty->print_cr("..."); 2590 prior_use->dump(); 2591 assert(edge_from_to(prior_use,n),msg); 2592 } 2593 _reg_node.map(def,NULL); // Kill live USEs 2594 } 2595 } 2596 2597 //------------------------------verify_good_schedule--------------------------- 2598 void Scheduling::verify_good_schedule( Block *b, const char *msg ) { 2599 2600 // Zap to something reasonable for the verify code 2601 _reg_node.clear(); 2602 2603 // Walk over the block backwards. Check to make sure each DEF doesn't 2604 // kill a live value (other than the one it's supposed to). Add each 2605 // USE to the live set. 2606 for( uint i = b->_nodes.size()-1; i >= _bb_start; i-- ) { 2607 Node *n = b->_nodes[i]; 2608 int n_op = n->Opcode(); 2609 if( n_op == Op_MachProj && n->ideal_reg() == MachProjNode::fat_proj ) { 2610 // Fat-proj kills a slew of registers 2611 RegMask rm = n->out_RegMask();// Make local copy 2612 while( rm.is_NotEmpty() ) { 2613 OptoReg::Name kill = rm.find_first_elem(); 2614 rm.Remove(kill); 2615 verify_do_def( n, kill, msg ); 2616 } 2617 } else if( n_op != Op_Node ) { // Avoid brand new antidependence nodes 2618 // Get DEF'd registers the normal way 2619 verify_do_def( n, _regalloc->get_reg_first(n), msg ); 2620 verify_do_def( n, _regalloc->get_reg_second(n), msg ); 2621 } 2622 2623 // Now make all USEs live 2624 for( uint i=1; i<n->req(); i++ ) { 2625 Node *def = n->in(i); 2626 assert(def != 0, "input edge required"); 2627 OptoReg::Name reg_lo = _regalloc->get_reg_first(def); 2628 OptoReg::Name reg_hi = _regalloc->get_reg_second(def); 2629 if( OptoReg::is_valid(reg_lo) ) { 2630 assert(!_reg_node[reg_lo] || edge_from_to(_reg_node[reg_lo],def), msg); 2631 _reg_node.map(reg_lo,n); 2632 } 2633 if( OptoReg::is_valid(reg_hi) ) { 2634 assert(!_reg_node[reg_hi] || edge_from_to(_reg_node[reg_hi],def), msg); 2635 _reg_node.map(reg_hi,n); 2636 } 2637 } 2638 2639 } 2640 2641 // Zap to something reasonable for the Antidependence code 2642 _reg_node.clear(); 2643 } 2644 #endif 2645 2646 // Conditionally add precedence edges. Avoid putting edges on Projs. 2647 static void add_prec_edge_from_to( Node *from, Node *to ) { 2648 if( from->is_Proj() ) { // Put precedence edge on Proj's input 2649 assert( from->req() == 1 && (from->len() == 1 || from->in(1)==0), "no precedence edges on projections" ); 2650 from = from->in(0); 2651 } 2652 if( from != to && // No cycles (for things like LD L0,[L0+4] ) 2653 !edge_from_to( from, to ) ) // Avoid duplicate edge 2654 from->add_prec(to); 2655 } 2656 2657 //------------------------------anti_do_def------------------------------------ 2658 void Scheduling::anti_do_def( Block *b, Node *def, OptoReg::Name def_reg, int is_def ) { 2659 if( !OptoReg::is_valid(def_reg) ) // Ignore stores & control flow 2660 return; 2661 2662 Node *pinch = _reg_node[def_reg]; // Get pinch point 2663 if( !pinch || _bbs[pinch->_idx] != b || // No pinch-point yet? 2664 is_def ) { // Check for a true def (not a kill) 2665 _reg_node.map(def_reg,def); // Record def/kill as the optimistic pinch-point 2666 return; 2667 } 2668 2669 Node *kill = def; // Rename 'def' to more descriptive 'kill' 2670 debug_only( def = (Node*)0xdeadbeef; ) 2671 2672 // After some number of kills there _may_ be a later def 2673 Node *later_def = NULL; 2674 2675 // Finding a kill requires a real pinch-point. 2676 // Check for not already having a pinch-point. 2677 // Pinch points are Op_Node's. 2678 if( pinch->Opcode() != Op_Node ) { // Or later-def/kill as pinch-point? 2679 later_def = pinch; // Must be def/kill as optimistic pinch-point 2680 if ( _pinch_free_list.size() > 0) { 2681 pinch = _pinch_free_list.pop(); 2682 } else { 2683 pinch = new (_cfg->C) Node(1); // Pinch point to-be 2684 } 2685 if (pinch->_idx >= _regalloc->node_regs_max_index()) { 2686 _cfg->C->record_method_not_compilable("too many D-U pinch points"); 2687 return; 2688 } 2689 _bbs.map(pinch->_idx,b); // Pretend it's valid in this block (lazy init) 2690 _reg_node.map(def_reg,pinch); // Record pinch-point 2691 //_regalloc->set_bad(pinch->_idx); // Already initialized this way. 2692 if( later_def->outcnt() == 0 || later_def->ideal_reg() == MachProjNode::fat_proj ) { // Distinguish def from kill 2693 pinch->init_req(0, _cfg->C->top()); // set not NULL for the next call 2694 add_prec_edge_from_to(later_def,pinch); // Add edge from kill to pinch 2695 later_def = NULL; // and no later def 2696 } 2697 pinch->set_req(0,later_def); // Hook later def so we can find it 2698 } else { // Else have valid pinch point 2699 if( pinch->in(0) ) // If there is a later-def 2700 later_def = pinch->in(0); // Get it 2701 } 2702 2703 // Add output-dependence edge from later def to kill 2704 if( later_def ) // If there is some original def 2705 add_prec_edge_from_to(later_def,kill); // Add edge from def to kill 2706 2707 // See if current kill is also a use, and so is forced to be the pinch-point. 2708 if( pinch->Opcode() == Op_Node ) { 2709 Node *uses = kill->is_Proj() ? kill->in(0) : kill; 2710 for( uint i=1; i<uses->req(); i++ ) { 2711 if( _regalloc->get_reg_first(uses->in(i)) == def_reg || 2712 _regalloc->get_reg_second(uses->in(i)) == def_reg ) { 2713 // Yes, found a use/kill pinch-point 2714 pinch->set_req(0,NULL); // 2715 pinch->replace_by(kill); // Move anti-dep edges up 2716 pinch = kill; 2717 _reg_node.map(def_reg,pinch); 2718 return; 2719 } 2720 } 2721 } 2722 2723 // Add edge from kill to pinch-point 2724 add_prec_edge_from_to(kill,pinch); 2725 } 2726 2727 //------------------------------anti_do_use------------------------------------ 2728 void Scheduling::anti_do_use( Block *b, Node *use, OptoReg::Name use_reg ) { 2729 if( !OptoReg::is_valid(use_reg) ) // Ignore stores & control flow 2730 return; 2731 Node *pinch = _reg_node[use_reg]; // Get pinch point 2732 // Check for no later def_reg/kill in block 2733 if( pinch && _bbs[pinch->_idx] == b && 2734 // Use has to be block-local as well 2735 _bbs[use->_idx] == b ) { 2736 if( pinch->Opcode() == Op_Node && // Real pinch-point (not optimistic?) 2737 pinch->req() == 1 ) { // pinch not yet in block? 2738 pinch->del_req(0); // yank pointer to later-def, also set flag 2739 // Insert the pinch-point in the block just after the last use 2740 b->_nodes.insert(b->find_node(use)+1,pinch); 2741 _bb_end++; // Increase size scheduled region in block 2742 } 2743 2744 add_prec_edge_from_to(pinch,use); 2745 } 2746 } 2747 2748 //------------------------------ComputeRegisterAntidependences----------------- 2749 // We insert antidependences between the reads and following write of 2750 // allocated registers to prevent illegal code motion. Hopefully, the 2751 // number of added references should be fairly small, especially as we 2752 // are only adding references within the current basic block. 2753 void Scheduling::ComputeRegisterAntidependencies(Block *b) { 2754 2755 #ifdef ASSERT 2756 verify_good_schedule(b,"before block local scheduling"); 2757 #endif 2758 2759 // A valid schedule, for each register independently, is an endless cycle 2760 // of: a def, then some uses (connected to the def by true dependencies), 2761 // then some kills (defs with no uses), finally the cycle repeats with a new 2762 // def. The uses are allowed to float relative to each other, as are the 2763 // kills. No use is allowed to slide past a kill (or def). This requires 2764 // antidependencies between all uses of a single def and all kills that 2765 // follow, up to the next def. More edges are redundant, because later defs 2766 // & kills are already serialized with true or antidependencies. To keep 2767 // the edge count down, we add a 'pinch point' node if there's more than 2768 // one use or more than one kill/def. 2769 2770 // We add dependencies in one bottom-up pass. 2771 2772 // For each instruction we handle it's DEFs/KILLs, then it's USEs. 2773 2774 // For each DEF/KILL, we check to see if there's a prior DEF/KILL for this 2775 // register. If not, we record the DEF/KILL in _reg_node, the 2776 // register-to-def mapping. If there is a prior DEF/KILL, we insert a 2777 // "pinch point", a new Node that's in the graph but not in the block. 2778 // We put edges from the prior and current DEF/KILLs to the pinch point. 2779 // We put the pinch point in _reg_node. If there's already a pinch point 2780 // we merely add an edge from the current DEF/KILL to the pinch point. 2781 2782 // After doing the DEF/KILLs, we handle USEs. For each used register, we 2783 // put an edge from the pinch point to the USE. 2784 2785 // To be expedient, the _reg_node array is pre-allocated for the whole 2786 // compilation. _reg_node is lazily initialized; it either contains a NULL, 2787 // or a valid def/kill/pinch-point, or a leftover node from some prior 2788 // block. Leftover node from some prior block is treated like a NULL (no 2789 // prior def, so no anti-dependence needed). Valid def is distinguished by 2790 // it being in the current block. 2791 bool fat_proj_seen = false; 2792 uint last_safept = _bb_end-1; 2793 Node* end_node = (_bb_end-1 >= _bb_start) ? b->_nodes[last_safept] : NULL; 2794 Node* last_safept_node = end_node; 2795 for( uint i = _bb_end-1; i >= _bb_start; i-- ) { 2796 Node *n = b->_nodes[i]; 2797 int is_def = n->outcnt(); // def if some uses prior to adding precedence edges 2798 if( n->is_MachProj() && n->ideal_reg() == MachProjNode::fat_proj ) { 2799 // Fat-proj kills a slew of registers 2800 // This can add edges to 'n' and obscure whether or not it was a def, 2801 // hence the is_def flag. 2802 fat_proj_seen = true; 2803 RegMask rm = n->out_RegMask();// Make local copy 2804 while( rm.is_NotEmpty() ) { 2805 OptoReg::Name kill = rm.find_first_elem(); 2806 rm.Remove(kill); 2807 anti_do_def( b, n, kill, is_def ); 2808 } 2809 } else { 2810 // Get DEF'd registers the normal way 2811 anti_do_def( b, n, _regalloc->get_reg_first(n), is_def ); 2812 anti_do_def( b, n, _regalloc->get_reg_second(n), is_def ); 2813 } 2814 2815 // Kill projections on a branch should appear to occur on the 2816 // branch, not afterwards, so grab the masks from the projections 2817 // and process them. 2818 if (n->is_MachBranch() || n->is_Mach() && n->as_Mach()->ideal_Opcode() == Op_Jump) { 2819 for (DUIterator_Fast imax, i = n->fast_outs(imax); i < imax; i++) { 2820 Node* use = n->fast_out(i); 2821 if (use->is_Proj()) { 2822 RegMask rm = use->out_RegMask();// Make local copy 2823 while( rm.is_NotEmpty() ) { 2824 OptoReg::Name kill = rm.find_first_elem(); 2825 rm.Remove(kill); 2826 anti_do_def( b, n, kill, false ); 2827 } 2828 } 2829 } 2830 } 2831 2832 // Check each register used by this instruction for a following DEF/KILL 2833 // that must occur afterward and requires an anti-dependence edge. 2834 for( uint j=0; j<n->req(); j++ ) { 2835 Node *def = n->in(j); 2836 if( def ) { 2837 assert( !def->is_MachProj() || def->ideal_reg() != MachProjNode::fat_proj, "" ); 2838 anti_do_use( b, n, _regalloc->get_reg_first(def) ); 2839 anti_do_use( b, n, _regalloc->get_reg_second(def) ); 2840 } 2841 } 2842 // Do not allow defs of new derived values to float above GC 2843 // points unless the base is definitely available at the GC point. 2844 2845 Node *m = b->_nodes[i]; 2846 2847 // Add precedence edge from following safepoint to use of derived pointer 2848 if( last_safept_node != end_node && 2849 m != last_safept_node) { 2850 for (uint k = 1; k < m->req(); k++) { 2851 const Type *t = m->in(k)->bottom_type(); 2852 if( t->isa_oop_ptr() && 2853 t->is_ptr()->offset() != 0 ) { 2854 last_safept_node->add_prec( m ); 2855 break; 2856 } 2857 } 2858 } 2859 2860 if( n->jvms() ) { // Precedence edge from derived to safept 2861 // Check if last_safept_node was moved by pinch-point insertion in anti_do_use() 2862 if( b->_nodes[last_safept] != last_safept_node ) { 2863 last_safept = b->find_node(last_safept_node); 2864 } 2865 for( uint j=last_safept; j > i; j-- ) { 2866 Node *mach = b->_nodes[j]; 2867 if( mach->is_Mach() && mach->as_Mach()->ideal_Opcode() == Op_AddP ) 2868 mach->add_prec( n ); 2869 } 2870 last_safept = i; 2871 last_safept_node = m; 2872 } 2873 } 2874 2875 if (fat_proj_seen) { 2876 // Garbage collect pinch nodes that were not consumed. 2877 // They are usually created by a fat kill MachProj for a call. 2878 garbage_collect_pinch_nodes(); 2879 } 2880 } 2881 2882 //------------------------------garbage_collect_pinch_nodes------------------------------- 2883 2884 // Garbage collect pinch nodes for reuse by other blocks. 2885 // 2886 // The block scheduler's insertion of anti-dependence 2887 // edges creates many pinch nodes when the block contains 2888 // 2 or more Calls. A pinch node is used to prevent a 2889 // combinatorial explosion of edges. If a set of kills for a 2890 // register is anti-dependent on a set of uses (or defs), rather 2891 // than adding an edge in the graph between each pair of kill 2892 // and use (or def), a pinch is inserted between them: 2893 // 2894 // use1 use2 use3 2895 // \ | / 2896 // \ | / 2897 // pinch 2898 // / | \ 2899 // / | \ 2900 // kill1 kill2 kill3 2901 // 2902 // One pinch node is created per register killed when 2903 // the second call is encountered during a backwards pass 2904 // over the block. Most of these pinch nodes are never 2905 // wired into the graph because the register is never 2906 // used or def'ed in the block. 2907 // 2908 void Scheduling::garbage_collect_pinch_nodes() { 2909 #ifndef PRODUCT 2910 if (_cfg->C->trace_opto_output()) tty->print("Reclaimed pinch nodes:"); 2911 #endif 2912 int trace_cnt = 0; 2913 for (uint k = 0; k < _reg_node.Size(); k++) { 2914 Node* pinch = _reg_node[k]; 2915 if (pinch != NULL && pinch->Opcode() == Op_Node && 2916 // no predecence input edges 2917 (pinch->req() == pinch->len() || pinch->in(pinch->req()) == NULL) ) { 2918 cleanup_pinch(pinch); 2919 _pinch_free_list.push(pinch); 2920 _reg_node.map(k, NULL); 2921 #ifndef PRODUCT 2922 if (_cfg->C->trace_opto_output()) { 2923 trace_cnt++; 2924 if (trace_cnt > 40) { 2925 tty->print("\n"); 2926 trace_cnt = 0; 2927 } 2928 tty->print(" %d", pinch->_idx); 2929 } 2930 #endif 2931 } 2932 } 2933 #ifndef PRODUCT 2934 if (_cfg->C->trace_opto_output()) tty->print("\n"); 2935 #endif 2936 } 2937 2938 // Clean up a pinch node for reuse. 2939 void Scheduling::cleanup_pinch( Node *pinch ) { 2940 assert (pinch && pinch->Opcode() == Op_Node && pinch->req() == 1, "just checking"); 2941 2942 for (DUIterator_Last imin, i = pinch->last_outs(imin); i >= imin; ) { 2943 Node* use = pinch->last_out(i); 2944 uint uses_found = 0; 2945 for (uint j = use->req(); j < use->len(); j++) { 2946 if (use->in(j) == pinch) { 2947 use->rm_prec(j); 2948 uses_found++; 2949 } 2950 } 2951 assert(uses_found > 0, "must be a precedence edge"); 2952 i -= uses_found; // we deleted 1 or more copies of this edge 2953 } 2954 // May have a later_def entry 2955 pinch->set_req(0, NULL); 2956 } 2957 2958 //------------------------------print_statistics------------------------------- 2959 #ifndef PRODUCT 2960 2961 void Scheduling::dump_available() const { 2962 tty->print("#Availist "); 2963 for (uint i = 0; i < _available.size(); i++) 2964 tty->print(" N%d/l%d", _available[i]->_idx,_current_latency[_available[i]->_idx]); 2965 tty->cr(); 2966 } 2967 2968 // Print Scheduling Statistics 2969 void Scheduling::print_statistics() { 2970 // Print the size added by nops for bundling 2971 tty->print("Nops added %d bytes to total of %d bytes", 2972 _total_nop_size, _total_method_size); 2973 if (_total_method_size > 0) 2974 tty->print(", for %.2f%%", 2975 ((double)_total_nop_size) / ((double) _total_method_size) * 100.0); 2976 tty->print("\n"); 2977 2978 // Print the number of branch shadows filled 2979 if (Pipeline::_branch_has_delay_slot) { 2980 tty->print("Of %d branches, %d had unconditional delay slots filled", 2981 _total_branches, _total_unconditional_delays); 2982 if (_total_branches > 0) 2983 tty->print(", for %.2f%%", 2984 ((double)_total_unconditional_delays) / ((double)_total_branches) * 100.0); 2985 tty->print("\n"); 2986 } 2987 2988 uint total_instructions = 0, total_bundles = 0; 2989 2990 for (uint i = 1; i <= Pipeline::_max_instrs_per_cycle; i++) { 2991 uint bundle_count = _total_instructions_per_bundle[i]; 2992 total_instructions += bundle_count * i; 2993 total_bundles += bundle_count; 2994 } 2995 2996 if (total_bundles > 0) 2997 tty->print("Average ILP (excluding nops) is %.2f\n", 2998 ((double)total_instructions) / ((double)total_bundles)); 2999 } 3000 #endif