Mercurial > hg > graal-jvmci-8

comparison src/share/vm/opto/ifg.cpp @ 12071:adb9a7d94cb5

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
8023003: Cleanup the public interface to PhaseCFG Summary: public methods that don't need to be public should be private. Reviewed-by: kvn, twisti
author adlertz
date Fri, 16 Aug 2013 10:23:55 +0200
parents d1034bd8cefc
children 650868c062a9
comparison
equal deleted inserted replaced
12070:afbe18ae0905 12071:adb9a7d94cb5
35 #include "opto/indexSet.hpp" 35 #include "opto/indexSet.hpp"
36 #include "opto/machnode.hpp" 36 #include "opto/machnode.hpp"
37 #include "opto/memnode.hpp" 37 #include "opto/memnode.hpp"
38 #include "opto/opcodes.hpp" 38 #include "opto/opcodes.hpp"
39 39
40 //=============================================================================
41 //------------------------------IFG--------------------------------------------
42 PhaseIFG::PhaseIFG( Arena *arena ) : Phase(Interference_Graph), _arena(arena) { 40 PhaseIFG::PhaseIFG( Arena *arena ) : Phase(Interference_Graph), _arena(arena) {
43 } 41 }
44 42
45 //------------------------------init-------------------------------------------
46 void PhaseIFG::init( uint maxlrg ) { 43 void PhaseIFG::init( uint maxlrg ) {
47 _maxlrg = maxlrg; 44 _maxlrg = maxlrg;
48 _yanked = new (_arena) VectorSet(_arena); 45 _yanked = new (_arena) VectorSet(_arena);
49 _is_square = false; 46 _is_square = false;
50 // Make uninitialized adjacency lists 47 // Make uninitialized adjacency lists
57 _adjs[i].initialize(maxlrg); 54 _adjs[i].initialize(maxlrg);
58 _lrgs[i].Set_All(); 55 _lrgs[i].Set_All();
59 } 56 }
60 } 57 }
61 58
62 //------------------------------add--------------------------------------------
63 // Add edge between vertices a & b. These are sorted (triangular matrix), 59 // Add edge between vertices a & b. These are sorted (triangular matrix),
64 // then the smaller number is inserted in the larger numbered array. 60 // then the smaller number is inserted in the larger numbered array.
65 int PhaseIFG::add_edge( uint a, uint b ) { 61 int PhaseIFG::add_edge( uint a, uint b ) {
66 lrgs(a).invalid_degree(); 62 lrgs(a).invalid_degree();
67 lrgs(b).invalid_degree(); 63 lrgs(b).invalid_degree();
69 assert( !_is_square, "only on triangular" ); 65 assert( !_is_square, "only on triangular" );
70 if( a < b ) { uint tmp = a; a = b; b = tmp; } 66 if( a < b ) { uint tmp = a; a = b; b = tmp; }
71 return _adjs[a].insert( b ); 67 return _adjs[a].insert( b );
72 } 68 }
73 69
74 //------------------------------add_vector-------------------------------------
75 // Add an edge between 'a' and everything in the vector. 70 // Add an edge between 'a' and everything in the vector.
76 void PhaseIFG::add_vector( uint a, IndexSet *vec ) { 71 void PhaseIFG::add_vector( uint a, IndexSet *vec ) {
77 // IFG is triangular, so do the inserts where 'a' < 'b'. 72 // IFG is triangular, so do the inserts where 'a' < 'b'.
78 assert( !_is_square, "only on triangular" ); 73 assert( !_is_square, "only on triangular" );
79 IndexSet *adjs_a = &_adjs[a]; 74 IndexSet *adjs_a = &_adjs[a];
84 while ((neighbor = elements.next()) != 0) { 79 while ((neighbor = elements.next()) != 0) {
85 add_edge( a, neighbor ); 80 add_edge( a, neighbor );
86 } 81 }
87 } 82 }
88 83
89 //------------------------------test-------------------------------------------
90 // Is there an edge between a and b? 84 // Is there an edge between a and b?
91 int PhaseIFG::test_edge( uint a, uint b ) const { 85 int PhaseIFG::test_edge( uint a, uint b ) const {
92 // Sort a and b, so that a is larger 86 // Sort a and b, so that a is larger
93 assert( !_is_square, "only on triangular" ); 87 assert( !_is_square, "only on triangular" );
94 if( a < b ) { uint tmp = a; a = b; b = tmp; } 88 if( a < b ) { uint tmp = a; a = b; b = tmp; }
95 return _adjs[a].member(b); 89 return _adjs[a].member(b);
96 } 90 }
97 91
98 //------------------------------SquareUp---------------------------------------
99 // Convert triangular matrix to square matrix 92 // Convert triangular matrix to square matrix
100 void PhaseIFG::SquareUp() { 93 void PhaseIFG::SquareUp() {
101 assert( !_is_square, "only on triangular" ); 94 assert( !_is_square, "only on triangular" );
102 95
103 // Simple transpose 96 // Simple transpose
109 } 102 }
110 } 103 }
111 _is_square = true; 104 _is_square = true;
112 } 105 }
113 106
114 //------------------------------Compute_Effective_Degree-----------------------
115 // Compute effective degree in bulk 107 // Compute effective degree in bulk
116 void PhaseIFG::Compute_Effective_Degree() { 108 void PhaseIFG::Compute_Effective_Degree() {
117 assert( _is_square, "only on square" ); 109 assert( _is_square, "only on square" );
118 110
119 for( uint i = 0; i < _maxlrg; i++ ) 111 for( uint i = 0; i < _maxlrg; i++ )
120 lrgs(i).set_degree(effective_degree(i)); 112 lrgs(i).set_degree(effective_degree(i));
121 } 113 }
122 114
123 //------------------------------test_edge_sq-----------------------------------
124 int PhaseIFG::test_edge_sq( uint a, uint b ) const { 115 int PhaseIFG::test_edge_sq( uint a, uint b ) const {
125 assert( _is_square, "only on square" ); 116 assert( _is_square, "only on square" );
126 // Swap, so that 'a' has the lesser count. Then binary search is on 117 // Swap, so that 'a' has the lesser count. Then binary search is on
127 // the smaller of a's list and b's list. 118 // the smaller of a's list and b's list.
128 if( neighbor_cnt(a) > neighbor_cnt(b) ) { uint tmp = a; a = b; b = tmp; } 119 if( neighbor_cnt(a) > neighbor_cnt(b) ) { uint tmp = a; a = b; b = tmp; }
129 //return _adjs[a].unordered_member(b); 120 //return _adjs[a].unordered_member(b);
130 return _adjs[a].member(b); 121 return _adjs[a].member(b);
131 } 122 }
132 123
133 //------------------------------Union------------------------------------------
134 // Union edges of B into A 124 // Union edges of B into A
135 void PhaseIFG::Union( uint a, uint b ) { 125 void PhaseIFG::Union( uint a, uint b ) {
136 assert( _is_square, "only on square" ); 126 assert( _is_square, "only on square" );
137 IndexSet *A = &_adjs[a]; 127 IndexSet *A = &_adjs[a];
138 IndexSetIterator b_elements(&_adjs[b]); 128 IndexSetIterator b_elements(&_adjs[b]);
144 lrgs(datum).invalid_degree(); 134 lrgs(datum).invalid_degree();
145 } 135 }
146 } 136 }
147 } 137 }
148 138
149 //------------------------------remove_node------------------------------------
150 // Yank a Node and all connected edges from the IFG. Return a 139 // Yank a Node and all connected edges from the IFG. Return a
151 // list of neighbors (edges) yanked. 140 // list of neighbors (edges) yanked.
152 IndexSet *PhaseIFG::remove_node( uint a ) { 141 IndexSet *PhaseIFG::remove_node( uint a ) {
153 assert( _is_square, "only on square" ); 142 assert( _is_square, "only on square" );
154 assert( !_yanked->test(a), "" ); 143 assert( !_yanked->test(a), "" );
163 lrgs(datum).inc_degree( -lrg_a.compute_degree(lrgs(datum)) ); 152 lrgs(datum).inc_degree( -lrg_a.compute_degree(lrgs(datum)) );
164 } 153 }
165 return neighbors(a); 154 return neighbors(a);
166 } 155 }
167 156
168 //------------------------------re_insert--------------------------------------
169 // Re-insert a yanked Node. 157 // Re-insert a yanked Node.
170 void PhaseIFG::re_insert( uint a ) { 158 void PhaseIFG::re_insert( uint a ) {
171 assert( _is_square, "only on square" ); 159 assert( _is_square, "only on square" );
172 assert( _yanked->test(a), "" ); 160 assert( _yanked->test(a), "" );
173 (*_yanked) >>= a; 161 (*_yanked) >>= a;
178 _adjs[datum].insert(a); 166 _adjs[datum].insert(a);
179 lrgs(datum).invalid_degree(); 167 lrgs(datum).invalid_degree();
180 } 168 }
181 } 169 }
182 170
183 //------------------------------compute_degree---------------------------------
184 // Compute the degree between 2 live ranges. If both live ranges are 171 // Compute the degree between 2 live ranges. If both live ranges are
185 // aligned-adjacent powers-of-2 then we use the MAX size. If either is 172 // aligned-adjacent powers-of-2 then we use the MAX size. If either is
186 // mis-aligned (or for Fat-Projections, not-adjacent) then we have to 173 // mis-aligned (or for Fat-Projections, not-adjacent) then we have to
187 // MULTIPLY the sizes. Inspect Brigg's thesis on register pairs to see why 174 // MULTIPLY the sizes. Inspect Brigg's thesis on register pairs to see why
188 // this is so. 175 // this is so.
194 ? (num_regs * nregs) // then use product 181 ? (num_regs * nregs) // then use product
195 : MAX2(num_regs,nregs); // else use max 182 : MAX2(num_regs,nregs); // else use max
196 return tmp; 183 return tmp;
197 } 184 }
198 185
199 //------------------------------effective_degree-------------------------------
200 // Compute effective degree for this live range. If both live ranges are 186 // Compute effective degree for this live range. If both live ranges are
201 // aligned-adjacent powers-of-2 then we use the MAX size. If either is 187 // aligned-adjacent powers-of-2 then we use the MAX size. If either is
202 // mis-aligned (or for Fat-Projections, not-adjacent) then we have to 188 // mis-aligned (or for Fat-Projections, not-adjacent) then we have to
203 // MULTIPLY the sizes. Inspect Brigg's thesis on register pairs to see why 189 // MULTIPLY the sizes. Inspect Brigg's thesis on register pairs to see why
204 // this is so. 190 // this is so.
219 return eff; 205 return eff;
220 } 206 }
221 207
222 208
223 #ifndef PRODUCT 209 #ifndef PRODUCT
224 //------------------------------dump-------------------------------------------
225 void PhaseIFG::dump() const { 210 void PhaseIFG::dump() const {
226 tty->print_cr("-- Interference Graph --%s--", 211 tty->print_cr("-- Interference Graph --%s--",
227 _is_square ? "square" : "triangular" ); 212 _is_square ? "square" : "triangular" );
228 if( _is_square ) { 213 if( _is_square ) {
229 for( uint i = 0; i < _maxlrg; i++ ) { 214 for( uint i = 0; i < _maxlrg; i++ ) {
258 tty->print("}\n"); 243 tty->print("}\n");
259 } 244 }
260 tty->print("\n"); 245 tty->print("\n");
261 } 246 }
262 247
263 //------------------------------stats------------------------------------------
264 void PhaseIFG::stats() const { 248 void PhaseIFG::stats() const {
265 ResourceMark rm; 249 ResourceMark rm;
266 int *h_cnt = NEW_RESOURCE_ARRAY(int,_maxlrg*2); 250 int *h_cnt = NEW_RESOURCE_ARRAY(int,_maxlrg*2);
267 memset( h_cnt, 0, sizeof(int)*_maxlrg*2 ); 251 memset( h_cnt, 0, sizeof(int)*_maxlrg*2 );
268 uint i; 252 uint i;
274 if( h_cnt[i] ) 258 if( h_cnt[i] )
275 tty->print("%d/%d ",i,h_cnt[i]); 259 tty->print("%d/%d ",i,h_cnt[i]);
276 tty->print_cr(""); 260 tty->print_cr("");
277 } 261 }
278 262
279 //------------------------------verify-----------------------------------------
280 void PhaseIFG::verify( const PhaseChaitin *pc ) const { 263 void PhaseIFG::verify( const PhaseChaitin *pc ) const {
281 // IFG is square, sorted and no need for Find 264 // IFG is square, sorted and no need for Find
282 for( uint i = 0; i < _maxlrg; i++ ) { 265 for( uint i = 0; i < _maxlrg; i++ ) {
283 assert(!((*_yanked)[i]) || !neighbor_cnt(i), "Is removed completely" ); 266 assert(!((*_yanked)[i]) || !neighbor_cnt(i), "Is removed completely" );
284 IndexSet *set = &_adjs[i]; 267 IndexSet *set = &_adjs[i];
296 assert(!lrgs(i)._degree_valid || effective_degree(i) == lrgs(i).degree(), "degree is valid but wrong"); 279 assert(!lrgs(i)._degree_valid || effective_degree(i) == lrgs(i).degree(), "degree is valid but wrong");
297 } 280 }
298 } 281 }
299 #endif 282 #endif
300 283
301 //------------------------------interfere_with_live----------------------------
302 // Interfere this register with everything currently live. Use the RegMasks 284 // Interfere this register with everything currently live. Use the RegMasks
303 // to trim the set of possible interferences. Return a count of register-only 285 // to trim the set of possible interferences. Return a count of register-only
304 // interferences as an estimate of register pressure. 286 // interferences as an estimate of register pressure.
305 void PhaseChaitin::interfere_with_live( uint r, IndexSet *liveout ) { 287 void PhaseChaitin::interfere_with_live( uint r, IndexSet *liveout ) {
306 uint retval = 0; 288 uint retval = 0;
313 while( (l = elements.next()) != 0 ) 295 while( (l = elements.next()) != 0 )
314 if( rm.overlap( lrgs(l).mask() ) ) 296 if( rm.overlap( lrgs(l).mask() ) )
315 _ifg->add_edge( r, l ); 297 _ifg->add_edge( r, l );
316 } 298 }
317 299
318 //------------------------------build_ifg_virtual------------------------------
319 // Actually build the interference graph. Uses virtual registers only, no 300 // Actually build the interference graph. Uses virtual registers only, no
320 // physical register masks. This allows me to be very aggressive when 301 // physical register masks. This allows me to be very aggressive when
321 // coalescing copies. Some of this aggressiveness will have to be undone 302 // coalescing copies. Some of this aggressiveness will have to be undone
322 // later, but I'd rather get all the copies I can now (since unremoved copies 303 // later, but I'd rather get all the copies I can now (since unremoved copies
323 // at this point can end up in bad places). Copies I re-insert later I have 304 // at this point can end up in bad places). Copies I re-insert later I have
324 // more opportunity to insert them in low-frequency locations. 305 // more opportunity to insert them in low-frequency locations.
325 void PhaseChaitin::build_ifg_virtual( ) { 306 void PhaseChaitin::build_ifg_virtual( ) {
326 307
327 // For all blocks (in any order) do... 308 // For all blocks (in any order) do...
328 for( uint i=0; i<_cfg._num_blocks; i++ ) { 309 for (uint i = 0; i < _cfg.number_of_blocks(); i++) {
329 Block *b = _cfg._blocks[i]; 310 Block* block = _cfg.get_block(i);
330 IndexSet *liveout = _live->live(b); 311 IndexSet* liveout = _live->live(block);
331 312
332 // The IFG is built by a single reverse pass over each basic block. 313 // The IFG is built by a single reverse pass over each basic block.
333 // Starting with the known live-out set, we remove things that get 314 // Starting with the known live-out set, we remove things that get
334 // defined and add things that become live (essentially executing one 315 // defined and add things that become live (essentially executing one
335 // pass of a standard LIVE analysis). Just before a Node defines a value 316 // pass of a standard LIVE analysis). Just before a Node defines a value
336 // (and removes it from the live-ness set) that value is certainly live. 317 // (and removes it from the live-ness set) that value is certainly live.
337 // The defined value interferes with everything currently live. The 318 // The defined value interferes with everything currently live. The
338 // value is then removed from the live-ness set and it's inputs are 319 // value is then removed from the live-ness set and it's inputs are
339 // added to the live-ness set. 320 // added to the live-ness set.
340 for( uint j = b->end_idx() + 1; j > 1; j-- ) { 321 for (uint j = block->end_idx() + 1; j > 1; j--) {
341 Node *n = b->_nodes[j-1]; 322 Node* n = block->_nodes[j - 1];
342 323
343 // Get value being defined 324 // Get value being defined
344 uint r = _lrg_map.live_range_id(n); 325 uint r = _lrg_map.live_range_id(n);
345 326
346 // Some special values do not allocate 327 // Some special values do not allocate
406 } 387 }
407 } // End of forall instructions in block 388 } // End of forall instructions in block
408 } // End of forall blocks 389 } // End of forall blocks
409 } 390 }
410 391
411 //------------------------------count_int_pressure-----------------------------
412 uint PhaseChaitin::count_int_pressure( IndexSet *liveout ) { 392 uint PhaseChaitin::count_int_pressure( IndexSet *liveout ) {
413 IndexSetIterator elements(liveout); 393 IndexSetIterator elements(liveout);
414 uint lidx; 394 uint lidx;
415 uint cnt = 0; 395 uint cnt = 0;
416 while ((lidx = elements.next()) != 0) { 396 while ((lidx = elements.next()) != 0) {
422 cnt += lrgs(lidx).reg_pressure(); 402 cnt += lrgs(lidx).reg_pressure();
423 } 403 }
424 return cnt; 404 return cnt;
425 } 405 }
426 406
427 //------------------------------count_float_pressure---------------------------
428 uint PhaseChaitin::count_float_pressure( IndexSet *liveout ) { 407 uint PhaseChaitin::count_float_pressure( IndexSet *liveout ) {
429 IndexSetIterator elements(liveout); 408 IndexSetIterator elements(liveout);
430 uint lidx; 409 uint lidx;
431 uint cnt = 0; 410 uint cnt = 0;
432 while ((lidx = elements.next()) != 0) { 411 while ((lidx = elements.next()) != 0) {
436 cnt += lrgs(lidx).reg_pressure(); 415 cnt += lrgs(lidx).reg_pressure();
437 } 416 }
438 return cnt; 417 return cnt;
439 } 418 }
440 419
441 //------------------------------lower_pressure---------------------------------
442 // Adjust register pressure down by 1. Capture last hi-to-low transition, 420 // Adjust register pressure down by 1. Capture last hi-to-low transition,
443 static void lower_pressure( LRG *lrg, uint where, Block *b, uint *pressure, uint *hrp_index ) { 421 static void lower_pressure( LRG *lrg, uint where, Block *b, uint *pressure, uint *hrp_index ) {
444 if (lrg->mask().is_UP() && lrg->mask_size()) { 422 if (lrg->mask().is_UP() && lrg->mask_size()) {
445 if (lrg->_is_float || lrg->_is_vector) { 423 if (lrg->_is_float || lrg->_is_vector) {
446 pressure[1] -= lrg->reg_pressure(); 424 pressure[1] -= lrg->reg_pressure();
458 } 436 }
459 } 437 }
460 } 438 }
461 } 439 }
462 440
463 //------------------------------build_ifg_physical-----------------------------
464 // Build the interference graph using physical registers when available. 441 // Build the interference graph using physical registers when available.
465 // That is, if 2 live ranges are simultaneously alive but in their acceptable 442 // That is, if 2 live ranges are simultaneously alive but in their acceptable
466 // register sets do not overlap, then they do not interfere. 443 // register sets do not overlap, then they do not interfere.
467 uint PhaseChaitin::build_ifg_physical( ResourceArea *a ) { 444 uint PhaseChaitin::build_ifg_physical( ResourceArea *a ) {
468 NOT_PRODUCT( Compile::TracePhase t3("buildIFG", &_t_buildIFGphysical, TimeCompiler); ) 445 NOT_PRODUCT( Compile::TracePhase t3("buildIFG", &_t_buildIFGphysical, TimeCompiler); )
469 446
470 uint spill_reg = LRG::SPILL_REG;
471 uint must_spill = 0; 447 uint must_spill = 0;
472 448
473 // For all blocks (in any order) do... 449 // For all blocks (in any order) do...
474 for( uint i = 0; i < _cfg._num_blocks; i++ ) { 450 for (uint i = 0; i < _cfg.number_of_blocks(); i++) {
475 Block *b = _cfg._blocks[i]; 451 Block* block = _cfg.get_block(i);
476 // Clone (rather than smash in place) the liveout info, so it is alive 452 // Clone (rather than smash in place) the liveout info, so it is alive
477 // for the "collect_gc_info" phase later. 453 // for the "collect_gc_info" phase later.
478 IndexSet liveout(_live->live(b)); 454 IndexSet liveout(_live->live(block));
479 uint last_inst = b->end_idx(); 455 uint last_inst = block->end_idx();
480 // Compute first nonphi node index 456 // Compute first nonphi node index
481 uint first_inst; 457 uint first_inst;
482 for( first_inst = 1; first_inst < last_inst; first_inst++ ) 458 for (first_inst = 1; first_inst < last_inst; first_inst++) {
483 if( !b->_nodes[first_inst]->is_Phi() ) 459 if (!block->_nodes[first_inst]->is_Phi()) {
484 break; 460 break;
461 }
462 }
485 463
486 // Spills could be inserted before CreateEx node which should be 464 // Spills could be inserted before CreateEx node which should be
487 // first instruction in block after Phis. Move CreateEx up. 465 // first instruction in block after Phis. Move CreateEx up.
488 for( uint insidx = first_inst; insidx < last_inst; insidx++ ) { 466 for (uint insidx = first_inst; insidx < last_inst; insidx++) {
489 Node *ex = b->_nodes[insidx]; 467 Node *ex = block->_nodes[insidx];
490 if( ex->is_SpillCopy() ) continue; 468 if (ex->is_SpillCopy()) {
491 if( insidx > first_inst && ex->is_Mach() && 469 continue;
492 ex->as_Mach()->ideal_Opcode() == Op_CreateEx ) { 470 }
471 if (insidx > first_inst && ex->is_Mach() && ex->as_Mach()->ideal_Opcode() == Op_CreateEx) {
493 // If the CreateEx isn't above all the MachSpillCopies 472 // If the CreateEx isn't above all the MachSpillCopies
494 // then move it to the top. 473 // then move it to the top.
495 b->_nodes.remove(insidx); 474 block->_nodes.remove(insidx);
496 b->_nodes.insert(first_inst, ex); 475 block->_nodes.insert(first_inst, ex);
497 } 476 }
498 // Stop once a CreateEx or any other node is found 477 // Stop once a CreateEx or any other node is found
499 break; 478 break;
500 } 479 }
501 480
502 // Reset block's register pressure values for each ifg construction 481 // Reset block's register pressure values for each ifg construction
503 uint pressure[2], hrp_index[2]; 482 uint pressure[2], hrp_index[2];
504 pressure[0] = pressure[1] = 0; 483 pressure[0] = pressure[1] = 0;
505 hrp_index[0] = hrp_index[1] = last_inst+1; 484 hrp_index[0] = hrp_index[1] = last_inst+1;
506 b->_reg_pressure = b->_freg_pressure = 0; 485 block->_reg_pressure = block->_freg_pressure = 0;
507 // Liveout things are presumed live for the whole block. We accumulate 486 // Liveout things are presumed live for the whole block. We accumulate
508 // 'area' accordingly. If they get killed in the block, we'll subtract 487 // 'area' accordingly. If they get killed in the block, we'll subtract
509 // the unused part of the block from the area. 488 // the unused part of the block from the area.
510 int inst_count = last_inst - first_inst; 489 int inst_count = last_inst - first_inst;
511 double cost = (inst_count <= 0) ? 0.0 : b->_freq * double(inst_count); 490 double cost = (inst_count <= 0) ? 0.0 : block->_freq * double(inst_count);
512 assert(!(cost < 0.0), "negative spill cost" ); 491 assert(!(cost < 0.0), "negative spill cost" );
513 IndexSetIterator elements(&liveout); 492 IndexSetIterator elements(&liveout);
514 uint lidx; 493 uint lidx;
515 while ((lidx = elements.next()) != 0) { 494 while ((lidx = elements.next()) != 0) {
516 LRG &lrg = lrgs(lidx); 495 LRG &lrg = lrgs(lidx);
517 lrg._area += cost; 496 lrg._area += cost;
518 // Compute initial register pressure 497 // Compute initial register pressure
519 if (lrg.mask().is_UP() && lrg.mask_size()) { 498 if (lrg.mask().is_UP() && lrg.mask_size()) {
520 if (lrg._is_float || lrg._is_vector) { // Count float pressure 499 if (lrg._is_float || lrg._is_vector) { // Count float pressure
521 pressure[1] += lrg.reg_pressure(); 500 pressure[1] += lrg.reg_pressure();
522 if( pressure[1] > b->_freg_pressure ) 501 if (pressure[1] > block->_freg_pressure) {
523 b->_freg_pressure = pressure[1]; 502 block->_freg_pressure = pressure[1];
503 }
524 // Count int pressure, but do not count the SP, flags 504 // Count int pressure, but do not count the SP, flags
525 } else if( lrgs(lidx).mask().overlap(*Matcher::idealreg2regmask[Op_RegI]) ) { 505 } else if(lrgs(lidx).mask().overlap(*Matcher::idealreg2regmask[Op_RegI])) {
526 pressure[0] += lrg.reg_pressure(); 506 pressure[0] += lrg.reg_pressure();
527 if( pressure[0] > b->_reg_pressure ) 507 if (pressure[0] > block->_reg_pressure) {
528 b->_reg_pressure = pressure[0]; 508 block->_reg_pressure = pressure[0];
509 }
529 } 510 }
530 } 511 }
531 } 512 }
532 assert( pressure[0] == count_int_pressure (&liveout), "" ); 513 assert( pressure[0] == count_int_pressure (&liveout), "" );
533 assert( pressure[1] == count_float_pressure(&liveout), "" ); 514 assert( pressure[1] == count_float_pressure(&liveout), "" );
539 // (and removes it from the live-ness set) that value is certainly live. 520 // (and removes it from the live-ness set) that value is certainly live.
540 // The defined value interferes with everything currently live. The 521 // The defined value interferes with everything currently live. The
541 // value is then removed from the live-ness set and it's inputs are added 522 // value is then removed from the live-ness set and it's inputs are added
542 // to the live-ness set. 523 // to the live-ness set.
543 uint j; 524 uint j;
544 for( j = last_inst + 1; j > 1; j-- ) { 525 for (j = last_inst + 1; j > 1; j--) {
545 Node *n = b->_nodes[j - 1]; 526 Node* n = block->_nodes[j - 1];
546 527
547 // Get value being defined 528 // Get value being defined
548 uint r = _lrg_map.live_range_id(n); 529 uint r = _lrg_map.live_range_id(n);
549 530
550 // Some special values do not allocate 531 // Some special values do not allocate
551 if(r) { 532 if(r) {
552 // A DEF normally costs block frequency; rematerialized values are 533 // A DEF normally costs block frequency; rematerialized values are
553 // removed from the DEF sight, so LOWER costs here. 534 // removed from the DEF sight, so LOWER costs here.
554 lrgs(r)._cost += n->rematerialize() ? 0 : b->_freq; 535 lrgs(r)._cost += n->rematerialize() ? 0 : block->_freq;
555 536
556 // If it is not live, then this instruction is dead. Probably caused 537 // If it is not live, then this instruction is dead. Probably caused
557 // by spilling and rematerialization. Who cares why, yank this baby. 538 // by spilling and rematerialization. Who cares why, yank this baby.
558 if( !liveout.member(r) && n->Opcode() != Op_SafePoint ) { 539 if( !liveout.member(r) && n->Opcode() != Op_SafePoint ) {
559 Node *def = n->in(0); 540 Node *def = n->in(0);
560 if( !n->is_Proj() || 541 if( !n->is_Proj() ||
561 // Could also be a flags-projection of a dead ADD or such. 542 // Could also be a flags-projection of a dead ADD or such.
562 (_lrg_map.live_range_id(def) && !liveout.member(_lrg_map.live_range_id(def)))) { 543 (_lrg_map.live_range_id(def) && !liveout.member(_lrg_map.live_range_id(def)))) {
563 b->_nodes.remove(j - 1); 544 block->_nodes.remove(j - 1);
564 if (lrgs(r)._def == n) { 545 if (lrgs(r)._def == n) {
565 lrgs(r)._def = 0; 546 lrgs(r)._def = 0;
566 } 547 }
567 n->disconnect_inputs(NULL, C); 548 n->disconnect_inputs(NULL, C);
568 _cfg.unmap_node_from_block(n); 549 _cfg.unmap_node_from_block(n);
578 if (lrgs(r)._fat_proj) { 559 if (lrgs(r)._fat_proj) {
579 // Count the int-only registers 560 // Count the int-only registers
580 RegMask itmp = lrgs(r).mask(); 561 RegMask itmp = lrgs(r).mask();
581 itmp.AND(*Matcher::idealreg2regmask[Op_RegI]); 562 itmp.AND(*Matcher::idealreg2regmask[Op_RegI]);
582 int iregs = itmp.Size(); 563 int iregs = itmp.Size();
583 if( pressure[0]+iregs > b->_reg_pressure ) 564 if (pressure[0]+iregs > block->_reg_pressure) {
584 b->_reg_pressure = pressure[0]+iregs; 565 block->_reg_pressure = pressure[0] + iregs;
585 if( pressure[0] <= (uint)INTPRESSURE && 566 }
586 pressure[0]+iregs > (uint)INTPRESSURE ) { 567 if (pressure[0] <= (uint)INTPRESSURE && pressure[0] + iregs > (uint)INTPRESSURE) {
587 hrp_index[0] = j-1; 568 hrp_index[0] = j - 1;
588 } 569 }
589 // Count the float-only registers 570 // Count the float-only registers
590 RegMask ftmp = lrgs(r).mask(); 571 RegMask ftmp = lrgs(r).mask();
591 ftmp.AND(*Matcher::idealreg2regmask[Op_RegD]); 572 ftmp.AND(*Matcher::idealreg2regmask[Op_RegD]);
592 int fregs = ftmp.Size(); 573 int fregs = ftmp.Size();
593 if( pressure[1]+fregs > b->_freg_pressure ) 574 if (pressure[1] + fregs > block->_freg_pressure) {
594 b->_freg_pressure = pressure[1]+fregs; 575 block->_freg_pressure = pressure[1] + fregs;
595 if( pressure[1] <= (uint)FLOATPRESSURE && 576 }
596 pressure[1]+fregs > (uint)FLOATPRESSURE ) { 577 if(pressure[1] <= (uint)FLOATPRESSURE && pressure[1]+fregs > (uint)FLOATPRESSURE) {
597 hrp_index[1] = j-1; 578 hrp_index[1] = j - 1;
598 } 579 }
599 } 580 }
600 581
601 } else { // Else it is live 582 } else { // Else it is live
602 // A DEF also ends 'area' partway through the block. 583 // A DEF also ends 'area' partway through the block.
605 586
606 // Insure high score for immediate-use spill copies so they get a color 587 // Insure high score for immediate-use spill copies so they get a color
607 if( n->is_SpillCopy() 588 if( n->is_SpillCopy()
608 && lrgs(r).is_singledef() // MultiDef live range can still split 589 && lrgs(r).is_singledef() // MultiDef live range can still split
609 && n->outcnt() == 1 // and use must be in this block 590 && n->outcnt() == 1 // and use must be in this block
610 && _cfg.get_block_for_node(n->unique_out()) == b ) { 591 && _cfg.get_block_for_node(n->unique_out()) == block) {
611 // All single-use MachSpillCopy(s) that immediately precede their 592 // All single-use MachSpillCopy(s) that immediately precede their
612 // use must color early. If a longer live range steals their 593 // use must color early. If a longer live range steals their
613 // color, the spill copy will split and may push another spill copy 594 // color, the spill copy will split and may push another spill copy
614 // further away resulting in an infinite spill-split-retry cycle. 595 // further away resulting in an infinite spill-split-retry cycle.
615 // Assigning a zero area results in a high score() and a good 596 // Assigning a zero area results in a high score() and a good
616 // location in the simplify list. 597 // location in the simplify list.
617 // 598 //
618 599
619 Node *single_use = n->unique_out(); 600 Node *single_use = n->unique_out();
620 assert( b->find_node(single_use) >= j, "Use must be later in block"); 601 assert(block->find_node(single_use) >= j, "Use must be later in block");
621 // Use can be earlier in block if it is a Phi, but then I should be a MultiDef 602 // Use can be earlier in block if it is a Phi, but then I should be a MultiDef
622 603
623 // Find first non SpillCopy 'm' that follows the current instruction 604 // Find first non SpillCopy 'm' that follows the current instruction
624 // (j - 1) is index for current instruction 'n' 605 // (j - 1) is index for current instruction 'n'
625 Node *m = n; 606 Node *m = n;
626 for( uint i = j; i <= last_inst && m->is_SpillCopy(); ++i ) { m = b->_nodes[i]; } 607 for (uint i = j; i <= last_inst && m->is_SpillCopy(); ++i) {
627 if( m == single_use ) { 608 m = block->_nodes[i];
609 }
610 if (m == single_use) {
628 lrgs(r)._area = 0.0; 611 lrgs(r)._area = 0.0;
629 } 612 }
630 } 613 }
631 614
632 // Remove from live-out set 615 // Remove from live-out set
633 if( liveout.remove(r) ) { 616 if( liveout.remove(r) ) {
634 // Adjust register pressure. 617 // Adjust register pressure.
635 // Capture last hi-to-lo pressure transition 618 // Capture last hi-to-lo pressure transition
636 lower_pressure( &lrgs(r), j-1, b, pressure, hrp_index ); 619 lower_pressure(&lrgs(r), j - 1, block, pressure, hrp_index);
637 assert( pressure[0] == count_int_pressure (&liveout), "" ); 620 assert( pressure[0] == count_int_pressure (&liveout), "" );
638 assert( pressure[1] == count_float_pressure(&liveout), "" ); 621 assert( pressure[1] == count_float_pressure(&liveout), "" );
639 } 622 }
640 623
641 // Copies do not define a new value and so do not interfere. 624 // Copies do not define a new value and so do not interfere.
644 if (idx) { 627 if (idx) {
645 uint x = _lrg_map.live_range_id(n->in(idx)); 628 uint x = _lrg_map.live_range_id(n->in(idx));
646 if (liveout.remove(x)) { 629 if (liveout.remove(x)) {
647 lrgs(x)._area -= cost; 630 lrgs(x)._area -= cost;
648 // Adjust register pressure. 631 // Adjust register pressure.
649 lower_pressure(&lrgs(x), j-1, b, pressure, hrp_index); 632 lower_pressure(&lrgs(x), j - 1, block, pressure, hrp_index);
650 assert( pressure[0] == count_int_pressure (&liveout), "" ); 633 assert( pressure[0] == count_int_pressure (&liveout), "" );
651 assert( pressure[1] == count_float_pressure(&liveout), "" ); 634 assert( pressure[1] == count_float_pressure(&liveout), "" );
652 } 635 }
653 } 636 }
654 } // End of if live or not 637 } // End of if live or not
716 699
717 } // End of if normal register-allocated value 700 } // End of if normal register-allocated value
718 701
719 // Area remaining in the block 702 // Area remaining in the block
720 inst_count--; 703 inst_count--;
721 cost = (inst_count <= 0) ? 0.0 : b->_freq * double(inst_count); 704 cost = (inst_count <= 0) ? 0.0 : block->_freq * double(inst_count);
722 705
723 // Make all inputs live 706 // Make all inputs live
724 if( !n->is_Phi() ) { // Phi function uses come from prior block 707 if( !n->is_Phi() ) { // Phi function uses come from prior block
725 JVMState* jvms = n->jvms(); 708 JVMState* jvms = n->jvms();
726 uint debug_start = jvms ? jvms->debug_start() : 999999; 709 uint debug_start = jvms ? jvms->debug_start() : 999999;
741 LRG &lrg = lrgs(x); 724 LRG &lrg = lrgs(x);
742 // No use-side cost for spilling debug info 725 // No use-side cost for spilling debug info
743 if (k < debug_start) { 726 if (k < debug_start) {
744 // A USE costs twice block frequency (once for the Load, once 727 // A USE costs twice block frequency (once for the Load, once
745 // for a Load-delay). Rematerialized uses only cost once. 728 // for a Load-delay). Rematerialized uses only cost once.
746 lrg._cost += (def->rematerialize() ? b->_freq : (b->_freq + b->_freq)); 729 lrg._cost += (def->rematerialize() ? block->_freq : (block->_freq + block->_freq));
747 } 730 }
748 // It is live now 731 // It is live now
749 if (liveout.insert(x)) { 732 if (liveout.insert(x)) {
750 // Newly live things assumed live from here to top of block 733 // Newly live things assumed live from here to top of block
751 lrg._area += cost; 734 lrg._area += cost;
752 // Adjust register pressure 735 // Adjust register pressure
753 if (lrg.mask().is_UP() && lrg.mask_size()) { 736 if (lrg.mask().is_UP() && lrg.mask_size()) {
754 if (lrg._is_float || lrg._is_vector) { 737 if (lrg._is_float || lrg._is_vector) {
755 pressure[1] += lrg.reg_pressure(); 738 pressure[1] += lrg.reg_pressure();
756 if( pressure[1] > b->_freg_pressure ) 739 if (pressure[1] > block->_freg_pressure) {
757 b->_freg_pressure = pressure[1]; 740 block->_freg_pressure = pressure[1];
741 }
758 } else if( lrg.mask().overlap(*Matcher::idealreg2regmask[Op_RegI]) ) { 742 } else if( lrg.mask().overlap(*Matcher::idealreg2regmask[Op_RegI]) ) {
759 pressure[0] += lrg.reg_pressure(); 743 pressure[0] += lrg.reg_pressure();
760 if( pressure[0] > b->_reg_pressure ) 744 if (pressure[0] > block->_reg_pressure) {
761 b->_reg_pressure = pressure[0]; 745 block->_reg_pressure = pressure[0];
746 }
762 } 747 }
763 } 748 }
764 assert( pressure[0] == count_int_pressure (&liveout), "" ); 749 assert( pressure[0] == count_int_pressure (&liveout), "" );
765 assert( pressure[1] == count_float_pressure(&liveout), "" ); 750 assert( pressure[1] == count_float_pressure(&liveout), "" );
766 } 751 }
770 } // End of reverse pass over all instructions in block 755 } // End of reverse pass over all instructions in block
771 756
772 // If we run off the top of the block with high pressure and 757 // If we run off the top of the block with high pressure and
773 // never see a hi-to-low pressure transition, just record that 758 // never see a hi-to-low pressure transition, just record that
774 // the whole block is high pressure. 759 // the whole block is high pressure.
775 if( pressure[0] > (uint)INTPRESSURE ) { 760 if (pressure[0] > (uint)INTPRESSURE) {
776 hrp_index[0] = 0; 761 hrp_index[0] = 0;
777 if( pressure[0] > b->_reg_pressure ) 762 if (pressure[0] > block->_reg_pressure) {
778 b->_reg_pressure = pressure[0]; 763 block->_reg_pressure = pressure[0];
779 } 764 }
780 if( pressure[1] > (uint)FLOATPRESSURE ) { 765 }
766 if (pressure[1] > (uint)FLOATPRESSURE) {
781 hrp_index[1] = 0; 767 hrp_index[1] = 0;
782 if( pressure[1] > b->_freg_pressure ) 768 if (pressure[1] > block->_freg_pressure) {
783 b->_freg_pressure = pressure[1]; 769 block->_freg_pressure = pressure[1];
770 }
784 } 771 }
785 772
786 // Compute high pressure indice; avoid landing in the middle of projnodes 773 // Compute high pressure indice; avoid landing in the middle of projnodes
787 j = hrp_index[0]; 774 j = hrp_index[0];
788 if( j < b->_nodes.size() && j < b->end_idx()+1 ) { 775 if (j < block->_nodes.size() && j < block->end_idx() + 1) {
789 Node *cur = b->_nodes[j]; 776 Node* cur = block->_nodes[j];
790 while( cur->is_Proj() || (cur->is_MachNullCheck()) || cur->is_Catch() ) { 777 while (cur->is_Proj() || (cur->is_MachNullCheck()) || cur->is_Catch()) {
791 j--; 778 j--;
792 cur = b->_nodes[j]; 779 cur = block->_nodes[j];
793 } 780 }
794 } 781 }
795 b->_ihrp_index = j; 782 block->_ihrp_index = j;
796 j = hrp_index[1]; 783 j = hrp_index[1];
797 if( j < b->_nodes.size() && j < b->end_idx()+1 ) { 784 if (j < block->_nodes.size() && j < block->end_idx() + 1) {
798 Node *cur = b->_nodes[j]; 785 Node* cur = block->_nodes[j];
799 while( cur->is_Proj() || (cur->is_MachNullCheck()) || cur->is_Catch() ) { 786 while (cur->is_Proj() || (cur->is_MachNullCheck()) || cur->is_Catch()) {
800 j--; 787 j--;
801 cur = b->_nodes[j]; 788 cur = block->_nodes[j];
802 } 789 }
803 } 790 }
804 b->_fhrp_index = j; 791 block->_fhrp_index = j;
805 792
806 #ifndef PRODUCT 793 #ifndef PRODUCT
807 // Gather Register Pressure Statistics 794 // Gather Register Pressure Statistics
808 if( PrintOptoStatistics ) { 795 if( PrintOptoStatistics ) {
809 if( b->_reg_pressure > (uint)INTPRESSURE || b->_freg_pressure > (uint)FLOATPRESSURE ) 796 if (block->_reg_pressure > (uint)INTPRESSURE || block->_freg_pressure > (uint)FLOATPRESSURE) {
810 _high_pressure++; 797 _high_pressure++;
811 else 798 } else {
812 _low_pressure++; 799 _low_pressure++;
800 }
813 } 801 }
814 #endif 802 #endif
815 } // End of for all blocks 803 } // End of for all blocks
816 804
817 return must_spill; 805 return must_spill;