Mercurial > hg > graal-compiler
view src/share/vm/oops/markOop.hpp @ 2149:7e37af9d69ef
7011379: G1: overly long concurrent marking cycles
Summary: This changeset introduces filtering of SATB buffers at the point when they are about to be enqueued. If this filtering clears enough entries on each buffer, the buffer can then be re-used and not enqueued. This cuts down the number of SATB buffers that need to be processed by the concurrent marking threads.
Reviewed-by: johnc, ysr
author | tonyp |
---|---|
date | Wed, 19 Jan 2011 09:35:17 -0500 |
parents | f95d63e2154a |
children | 3582bf76420e |
line wrap: on
line source
/* * Copyright (c) 1997, 2010, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it * under the terms of the GNU General Public License version 2 only, as * published by the Free Software Foundation. * * This code is distributed in the hope that it will be useful, but WITHOUT * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License * version 2 for more details (a copy is included in the LICENSE file that * accompanied this code). * * You should have received a copy of the GNU General Public License version * 2 along with this work; if not, write to the Free Software Foundation, * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. * * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA * or visit www.oracle.com if you need additional information or have any * questions. * */ #ifndef SHARE_VM_OOPS_MARKOOP_HPP #define SHARE_VM_OOPS_MARKOOP_HPP #include "oops/oop.hpp" // The markOop describes the header of an object. // // Note that the mark is not a real oop but just a word. // It is placed in the oop hierarchy for historical reasons. // // Bit-format of an object header (most significant first, big endian layout below): // // 32 bits: // -------- // hash:25 ------------>| age:4 biased_lock:1 lock:2 (normal object) // JavaThread*:23 epoch:2 age:4 biased_lock:1 lock:2 (biased object) // size:32 ------------------------------------------>| (CMS free block) // PromotedObject*:29 ---------->| promo_bits:3 ----->| (CMS promoted object) // // 64 bits: // -------- // unused:25 hash:31 -->| unused:1 age:4 biased_lock:1 lock:2 (normal object) // JavaThread*:54 epoch:2 unused:1 age:4 biased_lock:1 lock:2 (biased object) // PromotedObject*:61 --------------------->| promo_bits:3 ----->| (CMS promoted object) // size:64 ----------------------------------------------------->| (CMS free block) // // unused:25 hash:31 -->| cms_free:1 age:4 biased_lock:1 lock:2 (COOPs && normal object) // JavaThread*:54 epoch:2 cms_free:1 age:4 biased_lock:1 lock:2 (COOPs && biased object) // narrowOop:32 unused:24 cms_free:1 unused:4 promo_bits:3 ----->| (COOPs && CMS promoted object) // unused:21 size:35 -->| cms_free:1 unused:7 ------------------>| (COOPs && CMS free block) // // - hash contains the identity hash value: largest value is // 31 bits, see os::random(). Also, 64-bit vm's require // a hash value no bigger than 32 bits because they will not // properly generate a mask larger than that: see library_call.cpp // and c1_CodePatterns_sparc.cpp. // // - the biased lock pattern is used to bias a lock toward a given // thread. When this pattern is set in the low three bits, the lock // is either biased toward a given thread or "anonymously" biased, // indicating that it is possible for it to be biased. When the // lock is biased toward a given thread, locking and unlocking can // be performed by that thread without using atomic operations. // When a lock's bias is revoked, it reverts back to the normal // locking scheme described below. // // Note that we are overloading the meaning of the "unlocked" state // of the header. Because we steal a bit from the age we can // guarantee that the bias pattern will never be seen for a truly // unlocked object. // // Note also that the biased state contains the age bits normally // contained in the object header. Large increases in scavenge // times were seen when these bits were absent and an arbitrary age // assigned to all biased objects, because they tended to consume a // significant fraction of the eden semispaces and were not // promoted promptly, causing an increase in the amount of copying // performed. The runtime system aligns all JavaThread* pointers to // a very large value (currently 128 bytes (32bVM) or 256 bytes (64bVM)) // to make room for the age bits & the epoch bits (used in support of // biased locking), and for the CMS "freeness" bit in the 64bVM (+COOPs). // // [JavaThread* | epoch | age | 1 | 01] lock is biased toward given thread // [0 | epoch | age | 1 | 01] lock is anonymously biased // // - the two lock bits are used to describe three states: locked/unlocked and monitor. // // [ptr | 00] locked ptr points to real header on stack // [header | 0 | 01] unlocked regular object header // [ptr | 10] monitor inflated lock (header is wapped out) // [ptr | 11] marked used by markSweep to mark an object // not valid at any other time // // We assume that stack/thread pointers have the lowest two bits cleared. class BasicLock; class ObjectMonitor; class JavaThread; class markOopDesc: public oopDesc { private: // Conversion uintptr_t value() const { return (uintptr_t) this; } public: // Constants enum { age_bits = 4, lock_bits = 2, biased_lock_bits = 1, max_hash_bits = BitsPerWord - age_bits - lock_bits - biased_lock_bits, hash_bits = max_hash_bits > 31 ? 31 : max_hash_bits, cms_bits = LP64_ONLY(1) NOT_LP64(0), epoch_bits = 2 }; // The biased locking code currently requires that the age bits be // contiguous to the lock bits. Class data sharing would prefer the // hash bits to be lower down to provide more random hash codes for // shared read-only symbolOop objects, because these objects' mark // words are set to their own address with marked_value in the lock // bit, and using lower bits would make their identity hash values // more random. However, the performance decision was made in favor // of the biased locking code. enum { lock_shift = 0, biased_lock_shift = lock_bits, age_shift = lock_bits + biased_lock_bits, cms_shift = age_shift + age_bits, hash_shift = cms_shift + cms_bits, epoch_shift = hash_shift }; enum { lock_mask = right_n_bits(lock_bits), lock_mask_in_place = lock_mask << lock_shift, biased_lock_mask = right_n_bits(lock_bits + biased_lock_bits), biased_lock_mask_in_place= biased_lock_mask << lock_shift, biased_lock_bit_in_place = 1 << biased_lock_shift, age_mask = right_n_bits(age_bits), age_mask_in_place = age_mask << age_shift, epoch_mask = right_n_bits(epoch_bits), epoch_mask_in_place = epoch_mask << epoch_shift, cms_mask = right_n_bits(cms_bits), cms_mask_in_place = cms_mask << cms_shift #ifndef _WIN64 ,hash_mask = right_n_bits(hash_bits), hash_mask_in_place = (address_word)hash_mask << hash_shift #endif }; // Alignment of JavaThread pointers encoded in object header required by biased locking enum { biased_lock_alignment = 2 << (epoch_shift + epoch_bits) }; #ifdef _WIN64 // These values are too big for Win64 const static uintptr_t hash_mask = right_n_bits(hash_bits); const static uintptr_t hash_mask_in_place = (address_word)hash_mask << hash_shift; #endif enum { locked_value = 0, unlocked_value = 1, monitor_value = 2, marked_value = 3, biased_lock_pattern = 5 }; enum { no_hash = 0 }; // no hash value assigned enum { no_hash_in_place = (address_word)no_hash << hash_shift, no_lock_in_place = unlocked_value }; enum { max_age = age_mask }; enum { max_bias_epoch = epoch_mask }; // Biased Locking accessors. // These must be checked by all code which calls into the // ObjectSynchronizer and other code. The biasing is not understood // by the lower-level CAS-based locking code, although the runtime // fixes up biased locks to be compatible with it when a bias is // revoked. bool has_bias_pattern() const { return (mask_bits(value(), biased_lock_mask_in_place) == biased_lock_pattern); } JavaThread* biased_locker() const { assert(has_bias_pattern(), "should not call this otherwise"); return (JavaThread*) ((intptr_t) (mask_bits(value(), ~(biased_lock_mask_in_place | age_mask_in_place | epoch_mask_in_place)))); } // Indicates that the mark has the bias bit set but that it has not // yet been biased toward a particular thread bool is_biased_anonymously() const { return (has_bias_pattern() && (biased_locker() == NULL)); } // Indicates epoch in which this bias was acquired. If the epoch // changes due to too many bias revocations occurring, the biases // from the previous epochs are all considered invalid. int bias_epoch() const { assert(has_bias_pattern(), "should not call this otherwise"); return (mask_bits(value(), epoch_mask_in_place) >> epoch_shift); } markOop set_bias_epoch(int epoch) { assert(has_bias_pattern(), "should not call this otherwise"); assert((epoch & (~epoch_mask)) == 0, "epoch overflow"); return markOop(mask_bits(value(), ~epoch_mask_in_place) | (epoch << epoch_shift)); } markOop incr_bias_epoch() { return set_bias_epoch((1 + bias_epoch()) & epoch_mask); } // Prototype mark for initialization static markOop biased_locking_prototype() { return markOop( biased_lock_pattern ); } // lock accessors (note that these assume lock_shift == 0) bool is_locked() const { return (mask_bits(value(), lock_mask_in_place) != unlocked_value); } bool is_unlocked() const { return (mask_bits(value(), biased_lock_mask_in_place) == unlocked_value); } bool is_marked() const { return (mask_bits(value(), lock_mask_in_place) == marked_value); } bool is_neutral() const { return (mask_bits(value(), biased_lock_mask_in_place) == unlocked_value); } // Special temporary state of the markOop while being inflated. // Code that looks at mark outside a lock need to take this into account. bool is_being_inflated() const { return (value() == 0); } // Distinguished markword value - used when inflating over // an existing stacklock. 0 indicates the markword is "BUSY". // Lockword mutators that use a LD...CAS idiom should always // check for and avoid overwriting a 0 value installed by some // other thread. (They should spin or block instead. The 0 value // is transient and *should* be short-lived). static markOop INFLATING() { return (markOop) 0; } // inflate-in-progress // Should this header be preserved during GC? inline bool must_be_preserved(oop obj_containing_mark) const; inline bool must_be_preserved_with_bias(oop obj_containing_mark) const; // Should this header (including its age bits) be preserved in the // case of a promotion failure during scavenge? // Note that we special case this situation. We want to avoid // calling BiasedLocking::preserve_marks()/restore_marks() (which // decrease the number of mark words that need to be preserved // during GC) during each scavenge. During scavenges in which there // is no promotion failure, we actually don't need to call the above // routines at all, since we don't mutate and re-initialize the // marks of promoted objects using init_mark(). However, during // scavenges which result in promotion failure, we do re-initialize // the mark words of objects, meaning that we should have called // these mark word preservation routines. Currently there's no good // place in which to call them in any of the scavengers (although // guarded by appropriate locks we could make one), but the // observation is that promotion failures are quite rare and // reducing the number of mark words preserved during them isn't a // high priority. inline bool must_be_preserved_for_promotion_failure(oop obj_containing_mark) const; inline bool must_be_preserved_with_bias_for_promotion_failure(oop obj_containing_mark) const; // Should this header be preserved during a scavenge where CMS is // the old generation? // (This is basically the same body as must_be_preserved_for_promotion_failure(), // but takes the klassOop as argument instead) inline bool must_be_preserved_for_cms_scavenge(klassOop klass_of_obj_containing_mark) const; inline bool must_be_preserved_with_bias_for_cms_scavenge(klassOop klass_of_obj_containing_mark) const; // WARNING: The following routines are used EXCLUSIVELY by // synchronization functions. They are not really gc safe. // They must get updated if markOop layout get changed. markOop set_unlocked() const { return markOop(value() | unlocked_value); } bool has_locker() const { return ((value() & lock_mask_in_place) == locked_value); } BasicLock* locker() const { assert(has_locker(), "check"); return (BasicLock*) value(); } bool has_monitor() const { return ((value() & monitor_value) != 0); } ObjectMonitor* monitor() const { assert(has_monitor(), "check"); // Use xor instead of &~ to provide one extra tag-bit check. return (ObjectMonitor*) (value() ^ monitor_value); } bool has_displaced_mark_helper() const { return ((value() & unlocked_value) == 0); } markOop displaced_mark_helper() const { assert(has_displaced_mark_helper(), "check"); intptr_t ptr = (value() & ~monitor_value); return *(markOop*)ptr; } void set_displaced_mark_helper(markOop m) const { assert(has_displaced_mark_helper(), "check"); intptr_t ptr = (value() & ~monitor_value); *(markOop*)ptr = m; } markOop copy_set_hash(intptr_t hash) const { intptr_t tmp = value() & (~hash_mask_in_place); tmp |= ((hash & hash_mask) << hash_shift); return (markOop)tmp; } // it is only used to be stored into BasicLock as the // indicator that the lock is using heavyweight monitor static markOop unused_mark() { return (markOop) marked_value; } // the following two functions create the markOop to be // stored into object header, it encodes monitor info static markOop encode(BasicLock* lock) { return (markOop) lock; } static markOop encode(ObjectMonitor* monitor) { intptr_t tmp = (intptr_t) monitor; return (markOop) (tmp | monitor_value); } static markOop encode(JavaThread* thread, int age, int bias_epoch) { intptr_t tmp = (intptr_t) thread; assert(UseBiasedLocking && ((tmp & (epoch_mask_in_place | age_mask_in_place | biased_lock_mask_in_place)) == 0), "misaligned JavaThread pointer"); assert(age <= max_age, "age too large"); assert(bias_epoch <= max_bias_epoch, "bias epoch too large"); return (markOop) (tmp | (bias_epoch << epoch_shift) | (age << age_shift) | biased_lock_pattern); } // used to encode pointers during GC markOop clear_lock_bits() { return markOop(value() & ~lock_mask_in_place); } // age operations markOop set_marked() { return markOop((value() & ~lock_mask_in_place) | marked_value); } int age() const { return mask_bits(value() >> age_shift, age_mask); } markOop set_age(int v) const { assert((v & ~age_mask) == 0, "shouldn't overflow age field"); return markOop((value() & ~age_mask_in_place) | (((intptr_t)v & age_mask) << age_shift)); } markOop incr_age() const { return age() == max_age ? markOop(this) : set_age(age() + 1); } // hash operations intptr_t hash() const { return mask_bits(value() >> hash_shift, hash_mask); } bool has_no_hash() const { return hash() == no_hash; } // Prototype mark for initialization static markOop prototype() { return markOop( no_hash_in_place | no_lock_in_place ); } // Helper function for restoration of unmarked mark oops during GC static inline markOop prototype_for_object(oop obj); // Debugging void print_on(outputStream* st) const; // Prepare address of oop for placement into mark inline static markOop encode_pointer_as_mark(void* p) { return markOop(p)->set_marked(); } // Recover address of oop from encoded form used in mark inline void* decode_pointer() { if (UseBiasedLocking && has_bias_pattern()) return NULL; return clear_lock_bits(); } // see the definition in markOop.cpp for the gory details bool should_not_be_cached() const; // These markOops indicate cms free chunk blocks and not objects. // In 64 bit, the markOop is set to distinguish them from oops. // These are defined in 32 bit mode for vmStructs. const static uintptr_t cms_free_chunk_pattern = 0x1; // Constants for the size field. enum { size_shift = cms_shift + cms_bits, size_bits = 35 // need for compressed oops 32G }; // These values are too big for Win64 const static uintptr_t size_mask = LP64_ONLY(right_n_bits(size_bits)) NOT_LP64(0); const static uintptr_t size_mask_in_place = (address_word)size_mask << size_shift; #ifdef _LP64 static markOop cms_free_prototype() { return markOop(((intptr_t)prototype() & ~cms_mask_in_place) | ((cms_free_chunk_pattern & cms_mask) << cms_shift)); } uintptr_t cms_encoding() const { return mask_bits(value() >> cms_shift, cms_mask); } bool is_cms_free_chunk() const { return is_neutral() && (cms_encoding() & cms_free_chunk_pattern) == cms_free_chunk_pattern; } size_t get_size() const { return (size_t)(value() >> size_shift); } static markOop set_size_and_free(size_t size) { assert((size & ~size_mask) == 0, "shouldn't overflow size field"); return markOop(((intptr_t)cms_free_prototype() & ~size_mask_in_place) | (((intptr_t)size & size_mask) << size_shift)); } #endif // _LP64 }; #endif // SHARE_VM_OOPS_MARKOOP_HPP