Mercurial > hg > graal-jvmci-8

comparison src/share/vm/runtime/advancedThresholdPolicy.cpp @ 22296:1b16dd28160c

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
removed forward looking statements from comment
author Doug Simon <doug.simon@oracle.com>
date Wed, 22 Jul 2015 15:11:03 +0200
parents be896a1983c0
children 569c82ebb96e
comparison
equal deleted inserted replaced
22295:11c97c1d824a 22296:1b16dd28160c
373 if (common(p, method, CompLevel_full_profile, disable_feedback) == CompLevel_full_optimization) { 373 if (common(p, method, CompLevel_full_profile, disable_feedback) == CompLevel_full_optimization) {
374 next_level = CompLevel_full_optimization; 374 next_level = CompLevel_full_optimization;
375 } else if ((this->*p)(i, b, cur_level)) { 375 } else if ((this->*p)(i, b, cur_level)) {
376 #ifdef COMPILERJVMCI 376 #ifdef COMPILERJVMCI
377 // Since JVMCI takes a while to warm up, its queue inevitably backs up during 377 // Since JVMCI takes a while to warm up, its queue inevitably backs up during
378 // early VM execution. As of 2014-06-13, JVMCI's inliner assumes that the root 378 // early VM execution.
379 // compilation method and all potential inlinees have mature profiles (which
380 // includes type profiling). If it sees immature profiles, JVMCI's inliner
381 // can perform pathologically bad (e.g., causing OutOfMemoryErrors due to
382 // exploring/inlining too many graphs). Since a rewrite of the inliner is
383 // in progress, we simply disable the dialing back heuristic for now and will
384 // revisit this decision once the new inliner is completed.
385 next_level = CompLevel_full_profile; 379 next_level = CompLevel_full_profile;
386 #else 380 #else
387 // C1-generated fully profiled code is about 30% slower than the limited profile 381 // C1-generated fully profiled code is about 30% slower than the limited profile
388 // code that has only invocation and backedge counters. The observation is that 382 // code that has only invocation and backedge counters. The observation is that
389 // if C2 queue is large enough we can spend too much time in the fully profiled code 383 // if C2 queue is large enough we can spend too much time in the fully profiled code