Mon, 11 May 2009 16:30:56 -0700
6484957: G1: parallel concurrent refinement
6826318: G1: remove traversal-based refinement code
Summary: Removed traversal-based refinement code as it's no longer used. Made the concurrent refinement (queue-based) parallel.
Reviewed-by: tonyp
ysr@777 | 1 | /* |
ysr@777 | 2 | * Copyright 2001-2007 Sun Microsystems, Inc. All Rights Reserved. |
ysr@777 | 3 | * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. |
ysr@777 | 4 | * |
ysr@777 | 5 | * This code is free software; you can redistribute it and/or modify it |
ysr@777 | 6 | * under the terms of the GNU General Public License version 2 only, as |
ysr@777 | 7 | * published by the Free Software Foundation. |
ysr@777 | 8 | * |
ysr@777 | 9 | * This code is distributed in the hope that it will be useful, but WITHOUT |
ysr@777 | 10 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or |
ysr@777 | 11 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License |
ysr@777 | 12 | * version 2 for more details (a copy is included in the LICENSE file that |
ysr@777 | 13 | * accompanied this code). |
ysr@777 | 14 | * |
ysr@777 | 15 | * You should have received a copy of the GNU General Public License version |
ysr@777 | 16 | * 2 along with this work; if not, write to the Free Software Foundation, |
ysr@777 | 17 | * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. |
ysr@777 | 18 | * |
ysr@777 | 19 | * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, |
ysr@777 | 20 | * CA 95054 USA or visit www.sun.com if you need additional information or |
ysr@777 | 21 | * have any questions. |
ysr@777 | 22 | * |
ysr@777 | 23 | */ |
ysr@777 | 24 | |
ysr@777 | 25 | #include "incls/_precompiled.incl" |
ysr@777 | 26 | #include "incls/_concurrentG1Refine.cpp.incl" |
ysr@777 | 27 | |
ysr@777 | 28 | ConcurrentG1Refine::ConcurrentG1Refine() : |
ysr@777 | 29 | _card_counts(NULL), _cur_card_count_histo(NULL), _cum_card_count_histo(NULL), |
ysr@777 | 30 | _hot_cache(NULL), |
ysr@777 | 31 | _def_use_cache(false), _use_cache(false), |
iveresov@1229 | 32 | _n_periods(0), _total_cards(0), _total_travs(0), |
iveresov@1229 | 33 | _threads(NULL), _n_threads(0) |
ysr@777 | 34 | { |
ysr@777 | 35 | if (G1ConcRefine) { |
iveresov@1229 | 36 | _n_threads = (G1ParallelRSetThreads > 0) ? G1ParallelRSetThreads : ParallelGCThreads; |
iveresov@1229 | 37 | if (_n_threads > 0) { |
iveresov@1229 | 38 | _threads = NEW_C_HEAP_ARRAY(ConcurrentG1RefineThread*, _n_threads); |
iveresov@1229 | 39 | ConcurrentG1RefineThread *next = NULL; |
iveresov@1229 | 40 | for (int i = _n_threads - 1; i >= 0; i--) { |
iveresov@1229 | 41 | ConcurrentG1RefineThread* t = new ConcurrentG1RefineThread(this, next, i); |
iveresov@1229 | 42 | assert(t != NULL, "Conc refine should have been created"); |
iveresov@1229 | 43 | assert(t->cg1r() == this, "Conc refine thread should refer to this"); |
iveresov@1229 | 44 | _threads[i] = t; |
iveresov@1229 | 45 | next = t; |
iveresov@1229 | 46 | } |
iveresov@1229 | 47 | } |
ysr@777 | 48 | } |
ysr@777 | 49 | } |
ysr@777 | 50 | |
ysr@777 | 51 | void ConcurrentG1Refine::init() { |
ysr@777 | 52 | if (G1ConcRSLogCacheSize > 0 || G1ConcRSCountTraversals) { |
ysr@777 | 53 | G1CollectedHeap* g1h = G1CollectedHeap::heap(); |
ysr@777 | 54 | _n_card_counts = |
ysr@777 | 55 | (unsigned) (g1h->g1_reserved_obj_bytes() >> CardTableModRefBS::card_shift); |
ysr@777 | 56 | _card_counts = NEW_C_HEAP_ARRAY(unsigned char, _n_card_counts); |
ysr@777 | 57 | for (size_t i = 0; i < _n_card_counts; i++) _card_counts[i] = 0; |
ysr@777 | 58 | ModRefBarrierSet* bs = g1h->mr_bs(); |
ysr@777 | 59 | guarantee(bs->is_a(BarrierSet::CardTableModRef), "Precondition"); |
ysr@777 | 60 | CardTableModRefBS* ctbs = (CardTableModRefBS*)bs; |
ysr@777 | 61 | _ct_bot = ctbs->byte_for_const(g1h->reserved_region().start()); |
ysr@777 | 62 | if (G1ConcRSCountTraversals) { |
ysr@777 | 63 | _cur_card_count_histo = NEW_C_HEAP_ARRAY(unsigned, 256); |
ysr@777 | 64 | _cum_card_count_histo = NEW_C_HEAP_ARRAY(unsigned, 256); |
ysr@777 | 65 | for (int i = 0; i < 256; i++) { |
ysr@777 | 66 | _cur_card_count_histo[i] = 0; |
ysr@777 | 67 | _cum_card_count_histo[i] = 0; |
ysr@777 | 68 | } |
ysr@777 | 69 | } |
ysr@777 | 70 | } |
ysr@777 | 71 | if (G1ConcRSLogCacheSize > 0) { |
ysr@777 | 72 | _def_use_cache = true; |
ysr@777 | 73 | _use_cache = true; |
ysr@777 | 74 | _hot_cache_size = (1 << G1ConcRSLogCacheSize); |
ysr@777 | 75 | _hot_cache = NEW_C_HEAP_ARRAY(jbyte*, _hot_cache_size); |
ysr@777 | 76 | _n_hot = 0; |
ysr@777 | 77 | _hot_cache_idx = 0; |
ysr@777 | 78 | } |
ysr@777 | 79 | } |
ysr@777 | 80 | |
iveresov@1229 | 81 | void ConcurrentG1Refine::stop() { |
iveresov@1229 | 82 | if (_threads != NULL) { |
iveresov@1229 | 83 | for (int i = 0; i < _n_threads; i++) { |
iveresov@1229 | 84 | _threads[i]->stop(); |
iveresov@1229 | 85 | } |
iveresov@1229 | 86 | } |
iveresov@1229 | 87 | } |
iveresov@1229 | 88 | |
ysr@777 | 89 | ConcurrentG1Refine::~ConcurrentG1Refine() { |
ysr@777 | 90 | if (G1ConcRSLogCacheSize > 0 || G1ConcRSCountTraversals) { |
ysr@777 | 91 | assert(_card_counts != NULL, "Logic"); |
ysr@777 | 92 | FREE_C_HEAP_ARRAY(unsigned char, _card_counts); |
ysr@777 | 93 | assert(_cur_card_count_histo != NULL, "Logic"); |
ysr@777 | 94 | FREE_C_HEAP_ARRAY(unsigned, _cur_card_count_histo); |
ysr@777 | 95 | assert(_cum_card_count_histo != NULL, "Logic"); |
ysr@777 | 96 | FREE_C_HEAP_ARRAY(unsigned, _cum_card_count_histo); |
ysr@777 | 97 | } |
ysr@777 | 98 | if (G1ConcRSLogCacheSize > 0) { |
ysr@777 | 99 | assert(_hot_cache != NULL, "Logic"); |
ysr@777 | 100 | FREE_C_HEAP_ARRAY(jbyte*, _hot_cache); |
ysr@777 | 101 | } |
iveresov@1229 | 102 | if (_threads != NULL) { |
iveresov@1229 | 103 | for (int i = 0; i < _n_threads; i++) { |
iveresov@1229 | 104 | delete _threads[i]; |
iveresov@1229 | 105 | } |
iveresov@1229 | 106 | FREE_C_HEAP_ARRAY(ConcurrentG1RefineThread*, _n_threads); |
ysr@777 | 107 | } |
ysr@777 | 108 | } |
ysr@777 | 109 | |
iveresov@1229 | 110 | void ConcurrentG1Refine::threads_do(ThreadClosure *tc) { |
iveresov@1229 | 111 | if (_threads != NULL) { |
iveresov@1229 | 112 | for (int i = 0; i < _n_threads; i++) { |
iveresov@1229 | 113 | tc->do_thread(_threads[i]); |
iveresov@1229 | 114 | } |
ysr@777 | 115 | } |
ysr@777 | 116 | } |
ysr@777 | 117 | |
ysr@777 | 118 | |
ysr@777 | 119 | int ConcurrentG1Refine::add_card_count(jbyte* card_ptr) { |
ysr@777 | 120 | size_t card_num = (card_ptr - _ct_bot); |
ysr@777 | 121 | guarantee(0 <= card_num && card_num < _n_card_counts, "Bounds"); |
ysr@777 | 122 | unsigned char cnt = _card_counts[card_num]; |
ysr@777 | 123 | if (cnt < 255) _card_counts[card_num]++; |
ysr@777 | 124 | return cnt; |
ysr@777 | 125 | _total_travs++; |
ysr@777 | 126 | } |
ysr@777 | 127 | |
ysr@777 | 128 | jbyte* ConcurrentG1Refine::cache_insert(jbyte* card_ptr) { |
ysr@777 | 129 | int count = add_card_count(card_ptr); |
ysr@777 | 130 | // Count previously unvisited cards. |
ysr@777 | 131 | if (count == 0) _total_cards++; |
ysr@777 | 132 | // We'll assume a traversal unless we store it in the cache. |
ysr@777 | 133 | if (count < G1ConcRSHotCardLimit) { |
ysr@777 | 134 | _total_travs++; |
ysr@777 | 135 | return card_ptr; |
ysr@777 | 136 | } |
ysr@777 | 137 | // Otherwise, it's hot. |
ysr@777 | 138 | jbyte* res = NULL; |
ysr@777 | 139 | MutexLockerEx x(HotCardCache_lock, Mutex::_no_safepoint_check_flag); |
ysr@777 | 140 | if (_n_hot == _hot_cache_size) { |
ysr@777 | 141 | _total_travs++; |
ysr@777 | 142 | res = _hot_cache[_hot_cache_idx]; |
ysr@777 | 143 | _n_hot--; |
ysr@777 | 144 | } |
ysr@777 | 145 | // Now _n_hot < _hot_cache_size, and we can insert at _hot_cache_idx. |
ysr@777 | 146 | _hot_cache[_hot_cache_idx] = card_ptr; |
ysr@777 | 147 | _hot_cache_idx++; |
ysr@777 | 148 | if (_hot_cache_idx == _hot_cache_size) _hot_cache_idx = 0; |
ysr@777 | 149 | _n_hot++; |
ysr@777 | 150 | return res; |
ysr@777 | 151 | } |
ysr@777 | 152 | |
ysr@777 | 153 | |
ysr@777 | 154 | void ConcurrentG1Refine::clean_up_cache(int worker_i, G1RemSet* g1rs) { |
ysr@777 | 155 | assert(!use_cache(), "cache should be disabled"); |
ysr@777 | 156 | int start_ind = _hot_cache_idx-1; |
ysr@777 | 157 | for (int i = 0; i < _n_hot; i++) { |
ysr@777 | 158 | int ind = start_ind - i; |
ysr@777 | 159 | if (ind < 0) ind = ind + _hot_cache_size; |
ysr@777 | 160 | jbyte* entry = _hot_cache[ind]; |
ysr@777 | 161 | if (entry != NULL) { |
ysr@777 | 162 | g1rs->concurrentRefineOneCard(entry, worker_i); |
ysr@777 | 163 | } |
ysr@777 | 164 | } |
ysr@777 | 165 | _n_hot = 0; |
ysr@777 | 166 | _hot_cache_idx = 0; |
ysr@777 | 167 | } |
ysr@777 | 168 | |
ysr@777 | 169 | void ConcurrentG1Refine::clear_and_record_card_counts() { |
ysr@777 | 170 | if (G1ConcRSLogCacheSize == 0 && !G1ConcRSCountTraversals) return; |
ysr@777 | 171 | _n_periods++; |
ysr@777 | 172 | if (G1ConcRSCountTraversals) { |
ysr@777 | 173 | for (size_t i = 0; i < _n_card_counts; i++) { |
ysr@777 | 174 | unsigned char bucket = _card_counts[i]; |
ysr@777 | 175 | _cur_card_count_histo[bucket]++; |
ysr@777 | 176 | _card_counts[i] = 0; |
ysr@777 | 177 | } |
ysr@777 | 178 | gclog_or_tty->print_cr("Card counts:"); |
ysr@777 | 179 | for (int i = 0; i < 256; i++) { |
ysr@777 | 180 | if (_cur_card_count_histo[i] > 0) { |
ysr@777 | 181 | gclog_or_tty->print_cr(" %3d: %9d", i, _cur_card_count_histo[i]); |
ysr@777 | 182 | _cum_card_count_histo[i] += _cur_card_count_histo[i]; |
ysr@777 | 183 | _cur_card_count_histo[i] = 0; |
ysr@777 | 184 | } |
ysr@777 | 185 | } |
ysr@777 | 186 | } else { |
ysr@777 | 187 | assert(G1ConcRSLogCacheSize > 0, "Logic"); |
ysr@777 | 188 | Copy::fill_to_words((HeapWord*)(&_card_counts[0]), |
ysr@777 | 189 | _n_card_counts / HeapWordSize); |
ysr@777 | 190 | } |
ysr@777 | 191 | } |
ysr@777 | 192 | |
ysr@777 | 193 | void |
ysr@777 | 194 | ConcurrentG1Refine:: |
ysr@777 | 195 | print_card_count_histo_range(unsigned* histo, int from, int to, |
ysr@777 | 196 | float& cum_card_pct, |
ysr@777 | 197 | float& cum_travs_pct) { |
ysr@777 | 198 | unsigned cards = 0; |
ysr@777 | 199 | unsigned travs = 0; |
ysr@777 | 200 | guarantee(to <= 256, "Precondition"); |
ysr@777 | 201 | for (int i = from; i < to-1; i++) { |
ysr@777 | 202 | cards += histo[i]; |
ysr@777 | 203 | travs += histo[i] * i; |
ysr@777 | 204 | } |
ysr@777 | 205 | if (to == 256) { |
ysr@777 | 206 | unsigned histo_card_sum = 0; |
ysr@777 | 207 | unsigned histo_trav_sum = 0; |
ysr@777 | 208 | for (int i = 1; i < 255; i++) { |
ysr@777 | 209 | histo_trav_sum += histo[i] * i; |
ysr@777 | 210 | } |
ysr@777 | 211 | cards += histo[255]; |
ysr@777 | 212 | // correct traversals for the last one. |
ysr@777 | 213 | unsigned travs_255 = (unsigned) (_total_travs - histo_trav_sum); |
ysr@777 | 214 | travs += travs_255; |
ysr@777 | 215 | |
ysr@777 | 216 | } else { |
ysr@777 | 217 | cards += histo[to-1]; |
ysr@777 | 218 | travs += histo[to-1] * (to-1); |
ysr@777 | 219 | } |
ysr@777 | 220 | float fperiods = (float)_n_periods; |
ysr@777 | 221 | float f_tot_cards = (float)_total_cards/fperiods; |
ysr@777 | 222 | float f_tot_travs = (float)_total_travs/fperiods; |
ysr@777 | 223 | if (cards > 0) { |
ysr@777 | 224 | float fcards = (float)cards/fperiods; |
ysr@777 | 225 | float ftravs = (float)travs/fperiods; |
ysr@777 | 226 | if (to == 256) { |
ysr@777 | 227 | gclog_or_tty->print(" %4d- %10.2f%10.2f", from, fcards, ftravs); |
ysr@777 | 228 | } else { |
ysr@777 | 229 | gclog_or_tty->print(" %4d-%4d %10.2f%10.2f", from, to-1, fcards, ftravs); |
ysr@777 | 230 | } |
ysr@777 | 231 | float pct_cards = fcards*100.0/f_tot_cards; |
ysr@777 | 232 | cum_card_pct += pct_cards; |
ysr@777 | 233 | float pct_travs = ftravs*100.0/f_tot_travs; |
ysr@777 | 234 | cum_travs_pct += pct_travs; |
ysr@777 | 235 | gclog_or_tty->print_cr("%10.2f%10.2f%10.2f%10.2f", |
ysr@777 | 236 | pct_cards, cum_card_pct, |
ysr@777 | 237 | pct_travs, cum_travs_pct); |
ysr@777 | 238 | } |
ysr@777 | 239 | } |
ysr@777 | 240 | |
ysr@777 | 241 | void ConcurrentG1Refine::print_final_card_counts() { |
ysr@777 | 242 | if (!G1ConcRSCountTraversals) return; |
ysr@777 | 243 | |
ysr@777 | 244 | gclog_or_tty->print_cr("Did %d total traversals of %d distinct cards.", |
ysr@777 | 245 | _total_travs, _total_cards); |
ysr@777 | 246 | float fperiods = (float)_n_periods; |
ysr@777 | 247 | gclog_or_tty->print_cr(" This is an average of %8.2f traversals, %8.2f cards, " |
ysr@777 | 248 | "per collection.", (float)_total_travs/fperiods, |
ysr@777 | 249 | (float)_total_cards/fperiods); |
ysr@777 | 250 | gclog_or_tty->print_cr(" This is an average of %8.2f traversals/distinct " |
ysr@777 | 251 | "dirty card.\n", |
ysr@777 | 252 | _total_cards > 0 ? |
ysr@777 | 253 | (float)_total_travs/(float)_total_cards : 0.0); |
ysr@777 | 254 | |
ysr@777 | 255 | |
ysr@777 | 256 | gclog_or_tty->print_cr("Histogram:\n\n%10s %10s%10s%10s%10s%10s%10s", |
ysr@777 | 257 | "range", "# cards", "# travs", "% cards", "(cum)", |
ysr@777 | 258 | "% travs", "(cum)"); |
ysr@777 | 259 | gclog_or_tty->print_cr("------------------------------------------------------------" |
ysr@777 | 260 | "-------------"); |
ysr@777 | 261 | float cum_cards_pct = 0.0; |
ysr@777 | 262 | float cum_travs_pct = 0.0; |
ysr@777 | 263 | for (int i = 1; i < 10; i++) { |
ysr@777 | 264 | print_card_count_histo_range(_cum_card_count_histo, i, i+1, |
ysr@777 | 265 | cum_cards_pct, cum_travs_pct); |
ysr@777 | 266 | } |
ysr@777 | 267 | for (int i = 10; i < 100; i += 10) { |
ysr@777 | 268 | print_card_count_histo_range(_cum_card_count_histo, i, i+10, |
ysr@777 | 269 | cum_cards_pct, cum_travs_pct); |
ysr@777 | 270 | } |
ysr@777 | 271 | print_card_count_histo_range(_cum_card_count_histo, 100, 150, |
ysr@777 | 272 | cum_cards_pct, cum_travs_pct); |
ysr@777 | 273 | print_card_count_histo_range(_cum_card_count_histo, 150, 200, |
ysr@777 | 274 | cum_cards_pct, cum_travs_pct); |
ysr@777 | 275 | print_card_count_histo_range(_cum_card_count_histo, 150, 255, |
ysr@777 | 276 | cum_cards_pct, cum_travs_pct); |
ysr@777 | 277 | print_card_count_histo_range(_cum_card_count_histo, 255, 256, |
ysr@777 | 278 | cum_cards_pct, cum_travs_pct); |
ysr@777 | 279 | } |