| 1 | /* |
| 2 | * Copyright (c) 1997, 2017, 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 "memory/allocation.inline.hpp" |
| 27 | #include "memory/resourceArea.hpp" |
| 28 | #include "opto/callnode.hpp" |
| 29 | #include "opto/chaitin.hpp" |
| 30 | #include "opto/live.hpp" |
| 31 | #include "opto/machnode.hpp" |
| 32 | |
| 33 | |
| 34 | // Compute live-in/live-out. We use a totally incremental algorithm. The LIVE |
| 35 | // problem is monotonic. The steady-state solution looks like this: pull a |
| 36 | // block from the worklist. It has a set of delta's - values which are newly |
| 37 | // live-in from the block. Push these to the live-out sets of all predecessor |
| 38 | // blocks. At each predecessor, the new live-out values are ANDed with what is |
| 39 | // already live-out (extra stuff is added to the live-out sets). Then the |
| 40 | // remaining new live-out values are ANDed with what is locally defined. |
| 41 | // Leftover bits become the new live-in for the predecessor block, and the pred |
| 42 | // block is put on the worklist. |
| 43 | // The locally live-in stuff is computed once and added to predecessor |
| 44 | // live-out sets. This separate compilation is done in the outer loop below. |
| 45 | PhaseLive::PhaseLive(const PhaseCFG &cfg, const LRG_List &names, Arena *arena, bool keep_deltas) |
| 46 | : Phase(LIVE), |
| 47 | _live(0), |
| 48 | _livein(0), |
| 49 | _cfg(cfg), |
| 50 | _names(names), |
| 51 | _arena(arena), |
| 52 | _keep_deltas(keep_deltas) { |
| 53 | } |
| 54 | |
| 55 | void PhaseLive::compute(uint maxlrg) { |
| 56 | _maxlrg = maxlrg; |
| 57 | _worklist = new (_arena) Block_List(); |
| 58 | |
| 59 | // Init the sparse live arrays. This data is live on exit from here! |
| 60 | // The _live info is the live-out info. |
| 61 | _live = (IndexSet*)_arena->Amalloc(sizeof(IndexSet) * _cfg.number_of_blocks()); |
| 62 | uint i; |
| 63 | for (i = 0; i < _cfg.number_of_blocks(); i++) { |
| 64 | _live[i].initialize(_maxlrg); |
| 65 | } |
| 66 | |
| 67 | if (_keep_deltas) { |
| 68 | _livein = (IndexSet*)_arena->Amalloc(sizeof(IndexSet) * _cfg.number_of_blocks()); |
| 69 | for (i = 0; i < _cfg.number_of_blocks(); i++) { |
| 70 | _livein[i].initialize(_maxlrg); |
| 71 | } |
| 72 | } |
| 73 | |
| 74 | // Init the sparse arrays for delta-sets. |
| 75 | ResourceMark rm; // Nuke temp storage on exit |
| 76 | |
| 77 | // Does the memory used by _defs and _deltas get reclaimed? Does it matter? TT |
| 78 | |
| 79 | // Array of values defined locally in blocks |
| 80 | _defs = NEW_RESOURCE_ARRAY(IndexSet,_cfg.number_of_blocks()); |
| 81 | for (i = 0; i < _cfg.number_of_blocks(); i++) { |
| 82 | _defs[i].initialize(_maxlrg); |
| 83 | } |
| 84 | |
| 85 | // Array of delta-set pointers, indexed by block pre_order-1. |
| 86 | _deltas = NEW_RESOURCE_ARRAY(IndexSet*,_cfg.number_of_blocks()); |
| 87 | memset( _deltas, 0, sizeof(IndexSet*)* _cfg.number_of_blocks()); |
| 88 | |
| 89 | _free_IndexSet = NULL; |
| 90 | |
| 91 | // Blocks having done pass-1 |
| 92 | VectorSet first_pass(Thread::current()->resource_area()); |
| 93 | |
| 94 | // Outer loop: must compute local live-in sets and push into predecessors. |
| 95 | for (uint j = _cfg.number_of_blocks(); j > 0; j--) { |
| 96 | Block* block = _cfg.get_block(j - 1); |
| 97 | |
| 98 | // Compute the local live-in set. Start with any new live-out bits. |
| 99 | IndexSet* use = getset(block); |
| 100 | IndexSet* def = &_defs[block->_pre_order-1]; |
| 101 | DEBUG_ONLY(IndexSet *def_outside = getfreeset();) |
| 102 | uint i; |
| 103 | for (i = block->number_of_nodes(); i > 1; i--) { |
| 104 | Node* n = block->get_node(i-1); |
| 105 | if (n->is_Phi()) { |
| 106 | break; |
| 107 | } |
| 108 | |
| 109 | uint r = _names.at(n->_idx); |
| 110 | assert(!def_outside->member(r), "Use of external LRG overlaps the same LRG defined in this block" ); |
| 111 | def->insert( r ); |
| 112 | use->remove( r ); |
| 113 | uint cnt = n->req(); |
| 114 | for (uint k = 1; k < cnt; k++) { |
| 115 | Node *nk = n->in(k); |
| 116 | uint nkidx = nk->_idx; |
| 117 | if (_cfg.get_block_for_node(nk) != block) { |
| 118 | uint u = _names.at(nkidx); |
| 119 | use->insert(u); |
| 120 | DEBUG_ONLY(def_outside->insert(u);) |
| 121 | } |
| 122 | } |
| 123 | } |
| 124 | #ifdef ASSERT |
| 125 | def_outside->set_next(_free_IndexSet); |
| 126 | _free_IndexSet = def_outside; // Drop onto free list |
| 127 | #endif |
| 128 | // Remove anything defined by Phis and the block start instruction |
| 129 | for (uint k = i; k > 0; k--) { |
| 130 | uint r = _names.at(block->get_node(k - 1)->_idx); |
| 131 | def->insert(r); |
| 132 | use->remove(r); |
| 133 | } |
| 134 | |
| 135 | // Push these live-in things to predecessors |
| 136 | for (uint l = 1; l < block->num_preds(); l++) { |
| 137 | Block* p = _cfg.get_block_for_node(block->pred(l)); |
| 138 | add_liveout(p, use, first_pass); |
| 139 | |
| 140 | // PhiNode uses go in the live-out set of prior blocks. |
| 141 | for (uint k = i; k > 0; k--) { |
| 142 | Node *phi = block->get_node(k - 1); |
| 143 | if (l < phi->req()) { |
| 144 | add_liveout(p, _names.at(phi->in(l)->_idx), first_pass); |
| 145 | } |
| 146 | } |
| 147 | } |
| 148 | freeset(block); |
| 149 | first_pass.set(block->_pre_order); |
| 150 | |
| 151 | // Inner loop: blocks that picked up new live-out values to be propagated |
| 152 | while (_worklist->size()) { |
| 153 | Block* block = _worklist->pop(); |
| 154 | IndexSet *delta = getset(block); |
| 155 | assert( delta->count(), "missing delta set" ); |
| 156 | |
| 157 | // Add new-live-in to predecessors live-out sets |
| 158 | for (uint l = 1; l < block->num_preds(); l++) { |
| 159 | Block* predecessor = _cfg.get_block_for_node(block->pred(l)); |
| 160 | add_liveout(predecessor, delta, first_pass); |
| 161 | } |
| 162 | |
| 163 | freeset(block); |
| 164 | } // End of while-worklist-not-empty |
| 165 | |
| 166 | } // End of for-all-blocks-outer-loop |
| 167 | |
| 168 | // We explicitly clear all of the IndexSets which we are about to release. |
| 169 | // This allows us to recycle their internal memory into IndexSet's free list. |
| 170 | |
| 171 | for (i = 0; i < _cfg.number_of_blocks(); i++) { |
| 172 | _defs[i].clear(); |
| 173 | if (_deltas[i]) { |
| 174 | // Is this always true? |
| 175 | _deltas[i]->clear(); |
| 176 | } |
| 177 | } |
| 178 | IndexSet *free = _free_IndexSet; |
| 179 | while (free != NULL) { |
| 180 | IndexSet *temp = free; |
| 181 | free = free->next(); |
| 182 | temp->clear(); |
| 183 | } |
| 184 | |
| 185 | } |
| 186 | |
| 187 | #ifndef PRODUCT |
| 188 | void PhaseLive::stats(uint iters) const { |
| 189 | } |
| 190 | #endif |
| 191 | |
| 192 | // Get an IndexSet for a block. Return existing one, if any. Make a new |
| 193 | // empty one if a prior one does not exist. |
| 194 | IndexSet *PhaseLive::getset( Block *p ) { |
| 195 | IndexSet *delta = _deltas[p->_pre_order-1]; |
| 196 | if( !delta ) // Not on worklist? |
| 197 | // Get a free set; flag as being on worklist |
| 198 | delta = _deltas[p->_pre_order-1] = getfreeset(); |
| 199 | return delta; // Return set of new live-out items |
| 200 | } |
| 201 | |
| 202 | // Pull from free list, or allocate. Internal allocation on the returned set |
| 203 | // is always from thread local storage. |
| 204 | IndexSet *PhaseLive::getfreeset( ) { |
| 205 | IndexSet *f = _free_IndexSet; |
| 206 | if( !f ) { |
| 207 | f = new IndexSet; |
| 208 | // f->set_arena(Thread::current()->resource_area()); |
| 209 | f->initialize(_maxlrg, Thread::current()->resource_area()); |
| 210 | } else { |
| 211 | // Pull from free list |
| 212 | _free_IndexSet = f->next(); |
| 213 | //f->_cnt = 0; // Reset to empty |
| 214 | // f->set_arena(Thread::current()->resource_area()); |
| 215 | f->initialize(_maxlrg, Thread::current()->resource_area()); |
| 216 | } |
| 217 | return f; |
| 218 | } |
| 219 | |
| 220 | // Free an IndexSet from a block. |
| 221 | void PhaseLive::freeset( Block *p ) { |
| 222 | IndexSet *f = _deltas[p->_pre_order-1]; |
| 223 | if ( _keep_deltas ) { |
| 224 | add_livein(p, f); |
| 225 | } |
| 226 | f->set_next(_free_IndexSet); |
| 227 | _free_IndexSet = f; // Drop onto free list |
| 228 | _deltas[p->_pre_order-1] = NULL; |
| 229 | } |
| 230 | |
| 231 | // Add a live-out value to a given blocks live-out set. If it is new, then |
| 232 | // also add it to the delta set and stick the block on the worklist. |
| 233 | void PhaseLive::add_liveout( Block *p, uint r, VectorSet &first_pass ) { |
| 234 | IndexSet *live = &_live[p->_pre_order-1]; |
| 235 | if( live->insert(r) ) { // If actually inserted... |
| 236 | // We extended the live-out set. See if the value is generated locally. |
| 237 | // If it is not, then we must extend the live-in set. |
| 238 | if( !_defs[p->_pre_order-1].member( r ) ) { |
| 239 | if( !_deltas[p->_pre_order-1] && // Not on worklist? |
| 240 | first_pass.test(p->_pre_order) ) |
| 241 | _worklist->push(p); // Actually go on worklist if already 1st pass |
| 242 | getset(p)->insert(r); |
| 243 | } |
| 244 | } |
| 245 | } |
| 246 | |
| 247 | // Add a vector of live-out values to a given blocks live-out set. |
| 248 | void PhaseLive::add_liveout( Block *p, IndexSet *lo, VectorSet &first_pass ) { |
| 249 | IndexSet *live = &_live[p->_pre_order-1]; |
| 250 | IndexSet *defs = &_defs[p->_pre_order-1]; |
| 251 | IndexSet *on_worklist = _deltas[p->_pre_order-1]; |
| 252 | IndexSet *delta = on_worklist ? on_worklist : getfreeset(); |
| 253 | |
| 254 | IndexSetIterator elements(lo); |
| 255 | uint r; |
| 256 | while ((r = elements.next()) != 0) { |
| 257 | if( live->insert(r) && // If actually inserted... |
| 258 | !defs->member( r ) ) // and not defined locally |
| 259 | delta->insert(r); // Then add to live-in set |
| 260 | } |
| 261 | |
| 262 | if( delta->count() ) { // If actually added things |
| 263 | _deltas[p->_pre_order-1] = delta; // Flag as on worklist now |
| 264 | if( !on_worklist && // Not on worklist? |
| 265 | first_pass.test(p->_pre_order) ) |
| 266 | _worklist->push(p); // Actually go on worklist if already 1st pass |
| 267 | } else { // Nothing there; just free it |
| 268 | delta->set_next(_free_IndexSet); |
| 269 | _free_IndexSet = delta; // Drop onto free list |
| 270 | } |
| 271 | } |
| 272 | |
| 273 | // Add a vector of live-in values to a given blocks live-in set. |
| 274 | void PhaseLive::add_livein(Block *p, IndexSet *lo) { |
| 275 | IndexSet *livein = &_livein[p->_pre_order-1]; |
| 276 | IndexSetIterator elements(lo); |
| 277 | uint r; |
| 278 | while ((r = elements.next()) != 0) { |
| 279 | livein->insert(r); // Then add to live-in set |
| 280 | } |
| 281 | } |
| 282 | |
| 283 | #ifndef PRODUCT |
| 284 | // Dump the live-out set for a block |
| 285 | void PhaseLive::dump( const Block *b ) const { |
| 286 | tty->print("Block %d: " ,b->_pre_order); |
| 287 | if ( _keep_deltas ) { |
| 288 | tty->print("LiveIn: " ); _livein[b->_pre_order-1].dump(); |
| 289 | } |
| 290 | tty->print("LiveOut: " ); _live[b->_pre_order-1].dump(); |
| 291 | uint cnt = b->number_of_nodes(); |
| 292 | for( uint i=0; i<cnt; i++ ) { |
| 293 | tty->print("L%d/" , _names.at(b->get_node(i)->_idx)); |
| 294 | b->get_node(i)->dump(); |
| 295 | } |
| 296 | tty->print("\n" ); |
| 297 | } |
| 298 | |
| 299 | // Verify that base pointers and derived pointers are still sane. |
| 300 | void PhaseChaitin::verify_base_ptrs( ResourceArea *a ) const { |
| 301 | #ifdef ASSERT |
| 302 | Unique_Node_List worklist(a); |
| 303 | for (uint i = 0; i < _cfg.number_of_blocks(); i++) { |
| 304 | Block* block = _cfg.get_block(i); |
| 305 | for (uint j = block->end_idx() + 1; j > 1; j--) { |
| 306 | Node* n = block->get_node(j-1); |
| 307 | if (n->is_Phi()) { |
| 308 | break; |
| 309 | } |
| 310 | // Found a safepoint? |
| 311 | if (n->is_MachSafePoint()) { |
| 312 | MachSafePointNode *sfpt = n->as_MachSafePoint(); |
| 313 | JVMState* jvms = sfpt->jvms(); |
| 314 | if (jvms != NULL) { |
| 315 | // Now scan for a live derived pointer |
| 316 | if (jvms->oopoff() < sfpt->req()) { |
| 317 | // Check each derived/base pair |
| 318 | for (uint idx = jvms->oopoff(); idx < sfpt->req(); idx++) { |
| 319 | Node *check = sfpt->in(idx); |
| 320 | bool is_derived = ((idx - jvms->oopoff()) & 1) == 0; |
| 321 | // search upwards through spills and spill phis for AddP |
| 322 | worklist.clear(); |
| 323 | worklist.push(check); |
| 324 | uint k = 0; |
| 325 | while( k < worklist.size() ) { |
| 326 | check = worklist.at(k); |
| 327 | assert(check,"Bad base or derived pointer" ); |
| 328 | // See PhaseChaitin::find_base_for_derived() for all cases. |
| 329 | int isc = check->is_Copy(); |
| 330 | if( isc ) { |
| 331 | worklist.push(check->in(isc)); |
| 332 | } else if( check->is_Phi() ) { |
| 333 | for (uint m = 1; m < check->req(); m++) |
| 334 | worklist.push(check->in(m)); |
| 335 | } else if( check->is_Con() ) { |
| 336 | if (is_derived) { |
| 337 | // Derived is NULL+offset |
| 338 | assert(!is_derived || check->bottom_type()->is_ptr()->ptr() == TypePtr::Null,"Bad derived pointer" ); |
| 339 | } else { |
| 340 | assert(check->bottom_type()->is_ptr()->_offset == 0,"Bad base pointer" ); |
| 341 | // Base either ConP(NULL) or loadConP |
| 342 | if (check->is_Mach()) { |
| 343 | assert(check->as_Mach()->ideal_Opcode() == Op_ConP,"Bad base pointer" ); |
| 344 | } else { |
| 345 | assert(check->Opcode() == Op_ConP && |
| 346 | check->bottom_type()->is_ptr()->ptr() == TypePtr::Null,"Bad base pointer" ); |
| 347 | } |
| 348 | } |
| 349 | } else if( check->bottom_type()->is_ptr()->_offset == 0 ) { |
| 350 | if(check->is_Proj() || (check->is_Mach() && |
| 351 | (check->as_Mach()->ideal_Opcode() == Op_CreateEx || |
| 352 | check->as_Mach()->ideal_Opcode() == Op_ThreadLocal || |
| 353 | check->as_Mach()->ideal_Opcode() == Op_CMoveP || |
| 354 | check->as_Mach()->ideal_Opcode() == Op_CheckCastPP || |
| 355 | #ifdef _LP64 |
| 356 | (UseCompressedOops && check->as_Mach()->ideal_Opcode() == Op_CastPP) || |
| 357 | (UseCompressedOops && check->as_Mach()->ideal_Opcode() == Op_DecodeN) || |
| 358 | (UseCompressedClassPointers && check->as_Mach()->ideal_Opcode() == Op_DecodeNKlass) || |
| 359 | #endif |
| 360 | check->as_Mach()->ideal_Opcode() == Op_LoadP || |
| 361 | check->as_Mach()->ideal_Opcode() == Op_LoadKlass))) { |
| 362 | // Valid nodes |
| 363 | } else { |
| 364 | check->dump(); |
| 365 | assert(false,"Bad base or derived pointer" ); |
| 366 | } |
| 367 | } else { |
| 368 | assert(is_derived,"Bad base pointer" ); |
| 369 | assert(check->is_Mach() && check->as_Mach()->ideal_Opcode() == Op_AddP,"Bad derived pointer" ); |
| 370 | } |
| 371 | k++; |
| 372 | assert(k < 100000,"Derived pointer checking in infinite loop" ); |
| 373 | } // End while |
| 374 | } |
| 375 | } // End of check for derived pointers |
| 376 | } // End of Kcheck for debug info |
| 377 | } // End of if found a safepoint |
| 378 | } // End of forall instructions in block |
| 379 | } // End of forall blocks |
| 380 | #endif |
| 381 | } |
| 382 | |
| 383 | // Verify that graphs and base pointers are still sane. |
| 384 | void PhaseChaitin::verify( ResourceArea *a, bool verify_ifg ) const { |
| 385 | #ifdef ASSERT |
| 386 | if( VerifyOpto || VerifyRegisterAllocator ) { |
| 387 | _cfg.verify(); |
| 388 | verify_base_ptrs(a); |
| 389 | if(verify_ifg) |
| 390 | _ifg->verify(this); |
| 391 | } |
| 392 | #endif |
| 393 | } |
| 394 | |
| 395 | #endif |
| 396 | |