archive/jdk
1
0
Fork 0
mirror of https://github.com/openjdk/jdk.git synced 2025-08-26 22:34:27 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

7121496: G1: do the per-region evacuation failure handling work in parallel

Browse source 
Parallelize the removal of self forwarding pointers etc. by wrapping in a HeapRegion closure, which is then wrapped inside an AbstractGangTask.

Reviewed-by: tonyp, iveresov
This commit is contained in:
John Cuthbertson 2011-12-23 11:14:18 -08:00
parent c18293c806
commit 39d61f89a2
5 changed files with 262 additions and 168 deletions
Download patch file Download diff file Expand all files Collapse all files

20
hotspot/src/share/vm/gc_implementation/g1/concurrentMark.cpp
View file

@ -1,5 +1,5 @@
/*
* Copyright (c) 2001, 2011, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2001, 2012, 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
@ -3080,19 +3080,6 @@ public:
}
};
class SetClaimValuesInCSetHRClosure: public HeapRegionClosure {
jint _claim_value;
public:
SetClaimValuesInCSetHRClosure(jint claim_value) :
_claim_value(claim_value) { }
bool doHeapRegion(HeapRegion* hr) {
hr->set_claim_value(_claim_value);
return false;
}
};
class G1ParCompleteMarkInCSetTask: public AbstractGangTask {
protected:
G1CollectedHeap* _g1h;
@ -3135,9 +3122,8 @@ void ConcurrentMark::complete_marking_in_collection_set() {
assert(g1h->check_cset_heap_region_claim_values(HeapRegion::CompleteMarkCSetClaimValue), "sanity");
// Now reset the claim values in the regions in the collection set.
SetClaimValuesInCSetHRClosure set_cv_cl(HeapRegion::InitialClaimValue);
g1h->collection_set_iterate(&set_cv_cl);
// Reset the claim values in the regions in the collection set.
g1h->reset_cset_heap_region_claim_values();
assert(g1h->check_cset_heap_region_claim_values(HeapRegion::InitialClaimValue), "sanity");

173
hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.cpp
View file

@ -1,5 +1,5 @@
/*
* Copyright (c) 2001, 2011, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2001, 2012, 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
@ -32,6 +32,7 @@
#include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
#include "gc_implementation/g1/g1CollectorPolicy.hpp"
#include "gc_implementation/g1/g1ErgoVerbose.hpp"
#include "gc_implementation/g1/g1EvacFailure.hpp"
#include "gc_implementation/g1/g1MarkSweep.hpp"
#include "gc_implementation/g1/g1OopClosures.inline.hpp"
#include "gc_implementation/g1/g1RemSet.inline.hpp"
@ -2618,12 +2619,16 @@ public:
}
};
void
G1CollectedHeap::reset_heap_region_claim_values() {
void G1CollectedHeap::reset_heap_region_claim_values() {
ResetClaimValuesClosure blk;
heap_region_iterate(&blk);
}
void G1CollectedHeap::reset_cset_heap_region_claim_values() {
ResetClaimValuesClosure blk;
collection_set_iterate(&blk);
}
#ifdef ASSERT
// This checks whether all regions in the heap have the correct claim
// value. I also piggy-backed on this a check to ensure that the
@ -3998,157 +4003,26 @@ void G1CollectedHeap::finalize_for_evac_failure() {
_evac_failure_scan_stack = NULL;
}
class UpdateRSetDeferred : public OopsInHeapRegionClosure {
private:
G1CollectedHeap* _g1;
DirtyCardQueue *_dcq;
CardTableModRefBS* _ct_bs;
public:
UpdateRSetDeferred(G1CollectedHeap* g1, DirtyCardQueue* dcq) :
_g1(g1), _ct_bs((CardTableModRefBS*)_g1->barrier_set()), _dcq(dcq) {}
virtual void do_oop(narrowOop* p) { do_oop_work(p); }
virtual void do_oop( oop* p) { do_oop_work(p); }
template <class T> void do_oop_work(T* p) {
assert(_from->is_in_reserved(p), "paranoia");
if (!_from->is_in_reserved(oopDesc::load_decode_heap_oop(p)) &&
!_from->is_survivor()) {
size_t card_index = _ct_bs->index_for(p);
if (_ct_bs->mark_card_deferred(card_index)) {
_dcq->enqueue((jbyte*)_ct_bs->byte_for_index(card_index));
}
}
}
};
class RemoveSelfPointerClosure: public ObjectClosure {
private:
G1CollectedHeap* _g1;
ConcurrentMark* _cm;
HeapRegion* _hr;
size_t _prev_marked_bytes;
size_t _next_marked_bytes;
OopsInHeapRegionClosure *_cl;
public:
RemoveSelfPointerClosure(G1CollectedHeap* g1, HeapRegion* hr,
OopsInHeapRegionClosure* cl) :
_g1(g1), _hr(hr), _cm(_g1->concurrent_mark()), _prev_marked_bytes(0),
_next_marked_bytes(0), _cl(cl) {}
size_t prev_marked_bytes() { return _prev_marked_bytes; }
size_t next_marked_bytes() { return _next_marked_bytes; }
// <original comment>
// The original idea here was to coalesce evacuated and dead objects.
// However that caused complications with the block offset table (BOT).
// In particular if there were two TLABs, one of them partially refined.
// |----- TLAB_1--------|----TLAB_2-~~~(partially refined part)~~~|
// The BOT entries of the unrefined part of TLAB_2 point to the start
// of TLAB_2. If the last object of the TLAB_1 and the first object
// of TLAB_2 are coalesced, then the cards of the unrefined part
// would point into middle of the filler object.
// The current approach is to not coalesce and leave the BOT contents intact.
// </original comment>
//
// We now reset the BOT when we start the object iteration over the
// region and refine its entries for every object we come across. So
// the above comment is not really relevant and we should be able
// to coalesce dead objects if we want to.
void do_object(oop obj) {
HeapWord* obj_addr = (HeapWord*) obj;
assert(_hr->is_in(obj_addr), "sanity");
size_t obj_size = obj->size();
_hr->update_bot_for_object(obj_addr, obj_size);
if (obj->is_forwarded() && obj->forwardee() == obj) {
// The object failed to move.
assert(!_g1->is_obj_dead(obj), "We should not be preserving dead objs.");
_cm->markPrev(obj);
assert(_cm->isPrevMarked(obj), "Should be marked!");
_prev_marked_bytes += (obj_size * HeapWordSize);
if (_g1->mark_in_progress() && !_g1->is_obj_ill(obj)) {
_cm->markAndGrayObjectIfNecessary(obj);
}
obj->set_mark(markOopDesc::prototype());
// While we were processing RSet buffers during the
// collection, we actually didn't scan any cards on the
// collection set, since we didn't want to update remebered
// sets with entries that point into the collection set, given
// that live objects fromthe collection set are about to move
// and such entries will be stale very soon. This change also
// dealt with a reliability issue which involved scanning a
// card in the collection set and coming across an array that
// was being chunked and looking malformed. The problem is
// that, if evacuation fails, we might have remembered set
// entries missing given that we skipped cards on the
// collection set. So, we'll recreate such entries now.
obj->oop_iterate(_cl);
assert(_cm->isPrevMarked(obj), "Should be marked!");
} else {
// The object has been either evacuated or is dead. Fill it with a
// dummy object.
MemRegion mr((HeapWord*)obj, obj_size);
CollectedHeap::fill_with_object(mr);
_cm->clearRangeBothMaps(mr);
}
}
};
void G1CollectedHeap::remove_self_forwarding_pointers() {
UpdateRSetImmediate immediate_update(_g1h->g1_rem_set());
DirtyCardQueue dcq(&_g1h->dirty_card_queue_set());
UpdateRSetDeferred deferred_update(_g1h, &dcq);
OopsInHeapRegionClosure *cl;
if (G1DeferredRSUpdate) {
cl = &deferred_update;
} else {
cl = &immediate_update;
}
HeapRegion* cur = g1_policy()->collection_set();
while (cur != NULL) {
assert(check_cset_heap_region_claim_values(HeapRegion::InitialClaimValue), "sanity");
assert(g1_policy()->assertMarkedBytesDataOK(), "Should be!");
assert(!cur->isHumongous(), "sanity");
if (cur->evacuation_failed()) {
assert(cur->in_collection_set(), "bad CS");
RemoveSelfPointerClosure rspc(_g1h, cur, cl);
G1ParRemoveSelfForwardPtrsTask rsfp_task(this);
// In the common case we make sure that this is done when the
// region is freed so that it is "ready-to-go" when it's
// re-allocated. However, when evacuation failure happens, a
// region will remain in the heap and might ultimately be added
// to a CSet in the future. So we have to be careful here and
// make sure the region's RSet is ready for parallel iteration
// whenever this might be required in the future.
cur->rem_set()->reset_for_par_iteration();
cur->reset_bot();
cl->set_region(cur);
cur->object_iterate(&rspc);
if (G1CollectedHeap::use_parallel_gc_threads()) {
set_par_threads();
workers()->run_task(&rsfp_task);
set_par_threads(0);
} else {
rsfp_task.work(0);
}
// A number of manipulations to make the TAMS be the current top,
// and the marked bytes be the ones observed in the iteration.
if (_g1h->concurrent_mark()->at_least_one_mark_complete()) {
// The comments below are the postconditions achieved by the
// calls. Note especially the last such condition, which says that
// the count of marked bytes has been properly restored.
cur->note_start_of_marking(false);
// _next_top_at_mark_start == top, _next_marked_bytes == 0
cur->add_to_marked_bytes(rspc.prev_marked_bytes());
// _next_marked_bytes == prev_marked_bytes.
cur->note_end_of_marking();
// _prev_top_at_mark_start == top(),
// _prev_marked_bytes == prev_marked_bytes
}
// If there is no mark in progress, we modified the _next variables
// above needlessly, but harmlessly.
if (_g1h->mark_in_progress()) {
cur->note_start_of_marking(false);
// _next_top_at_mark_start == top, _next_marked_bytes == 0
// _next_marked_bytes == next_marked_bytes.
}
}
cur = cur->next_in_collection_set();
}
assert(check_cset_heap_region_claim_values(HeapRegion::ParEvacFailureClaimValue), "sanity");
// Reset the claim values in the regions in the collection set.
reset_cset_heap_region_claim_values();
assert(check_cset_heap_region_claim_values(HeapRegion::InitialClaimValue), "sanity");
assert(g1_policy()->assertMarkedBytesDataOK(), "Should be!");
// Now restore saved marks, if any.
@ -4161,6 +4035,7 @@ void G1CollectedHeap::remove_self_forwarding_pointers() {
markOop m = _preserved_marks_of_objs->at(i);
obj->set_mark(m);
}
// Delete the preserved marks growable arrays (allocated on the C heap).
delete _objs_with_preserved_marks;
delete _preserved_marks_of_objs;

6
hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.hpp
View file

@ -1,5 +1,5 @@
/*
* Copyright (c) 2001, 2011, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2001, 2012, 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
@ -1313,6 +1313,10 @@ public:
// It resets all the region claim values to the default.
void reset_heap_region_claim_values();
// Resets the claim values of regions in the current
// collection set to the default.
void reset_cset_heap_region_claim_values();
#ifdef ASSERT
bool check_heap_region_claim_values(jint claim_value);

228
hotspot/src/share/vm/gc_implementation/g1/g1EvacFailure.hpp Normal file
View file

@ -0,0 +1,228 @@
/*
* Copyright (c) 2012, 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_GC_IMPLEMENTATION_G1_G1EVACFAILURE_HPP
#define SHARE_VM_GC_IMPLEMENTATION_G1_G1EVACFAILURE_HPP
#include "gc_implementation/g1/concurrentMark.inline.hpp"
#include "gc_implementation/g1/dirtyCardQueue.hpp"
#include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
#include "gc_implementation/g1/g1_globals.hpp"
#include "gc_implementation/g1/g1OopClosures.inline.hpp"
#include "gc_implementation/g1/heapRegion.hpp"
#include "gc_implementation/g1/heapRegionRemSet.hpp"
#include "utilities/workgroup.hpp"
// Closures and tasks associated with any self-forwarding pointers
// installed as a result of an evacuation failure.
class UpdateRSetDeferred : public OopsInHeapRegionClosure {
private:
G1CollectedHeap* _g1;
DirtyCardQueue *_dcq;
CardTableModRefBS* _ct_bs;
public:
UpdateRSetDeferred(G1CollectedHeap* g1, DirtyCardQueue* dcq) :
_g1(g1), _ct_bs((CardTableModRefBS*)_g1->barrier_set()), _dcq(dcq) {}
virtual void do_oop(narrowOop* p) { do_oop_work(p); }
virtual void do_oop( oop* p) { do_oop_work(p); }
template <class T> void do_oop_work(T* p) {
assert(_from->is_in_reserved(p), "paranoia");
if (!_from->is_in_reserved(oopDesc::load_decode_heap_oop(p)) &&
!_from->is_survivor()) {
size_t card_index = _ct_bs->index_for(p);
if (_ct_bs->mark_card_deferred(card_index)) {
_dcq->enqueue((jbyte*)_ct_bs->byte_for_index(card_index));
}
}
}
};
class RemoveSelfForwardPtrObjClosure: public ObjectClosure {
private:
G1CollectedHeap* _g1;
ConcurrentMark* _cm;
HeapRegion* _hr;
size_t _prev_marked_bytes;
size_t _next_marked_bytes;
OopsInHeapRegionClosure *_update_rset_cl;
public:
RemoveSelfForwardPtrObjClosure(G1CollectedHeap* g1, ConcurrentMark* cm,
HeapRegion* hr,
OopsInHeapRegionClosure* update_rset_cl) :
_g1(g1), _cm(cm), _hr(hr),
_update_rset_cl(update_rset_cl),
_prev_marked_bytes(0), _next_marked_bytes(0) {}
size_t prev_marked_bytes() { return _prev_marked_bytes; }
size_t next_marked_bytes() { return _next_marked_bytes; }
// <original comment>
// The original idea here was to coalesce evacuated and dead objects.
// However that caused complications with the block offset table (BOT).
// In particular if there were two TLABs, one of them partially refined.
// |----- TLAB_1--------|----TLAB_2-~~~(partially refined part)~~~|
// The BOT entries of the unrefined part of TLAB_2 point to the start
// of TLAB_2. If the last object of the TLAB_1 and the first object
// of TLAB_2 are coalesced, then the cards of the unrefined part
// would point into middle of the filler object.
// The current approach is to not coalesce and leave the BOT contents intact.
// </original comment>
//
// We now reset the BOT when we start the object iteration over the
// region and refine its entries for every object we come across. So
// the above comment is not really relevant and we should be able
// to coalesce dead objects if we want to.
void do_object(oop obj) {
HeapWord* obj_addr = (HeapWord*) obj;
assert(_hr->is_in(obj_addr), "sanity");
size_t obj_size = obj->size();
_hr->update_bot_for_object(obj_addr, obj_size);
if (obj->is_forwarded() && obj->forwardee() == obj) {
// The object failed to move.
assert(!_g1->is_obj_dead(obj), "We should not be preserving dead objs.");
_cm->markPrev(obj);
assert(_cm->isPrevMarked(obj), "Should be marked!");
_prev_marked_bytes += (obj_size * HeapWordSize);
if (_g1->mark_in_progress() && !_g1->is_obj_ill(obj)) {
_cm->markAndGrayObjectIfNecessary(obj);
}
obj->set_mark(markOopDesc::prototype());
// While we were processing RSet buffers during the collection,
// we actually didn't scan any cards on the collection set,
// since we didn't want to update remembered sets with entries
// that point into the collection set, given that live objects
// from the collection set are about to move and such entries
// will be stale very soon.
// This change also dealt with a reliability issue which
// involved scanning a card in the collection set and coming
// across an array that was being chunked and looking malformed.
// The problem is that, if evacuation fails, we might have
// remembered set entries missing given that we skipped cards on
// the collection set. So, we'll recreate such entries now.
obj->oop_iterate(_update_rset_cl);
assert(_cm->isPrevMarked(obj), "Should be marked!");
} else {
// The object has been either evacuated or is dead. Fill it with a
// dummy object.
MemRegion mr((HeapWord*)obj, obj_size);
CollectedHeap::fill_with_object(mr);
_cm->clearRangeBothMaps(mr);
}
}
};
class RemoveSelfForwardPtrHRClosure: public HeapRegionClosure {
G1CollectedHeap* _g1h;
ConcurrentMark* _cm;
OopsInHeapRegionClosure *_update_rset_cl;
public:
RemoveSelfForwardPtrHRClosure(G1CollectedHeap* g1h,
OopsInHeapRegionClosure* update_rset_cl) :
_g1h(g1h), _update_rset_cl(update_rset_cl),
_cm(_g1h->concurrent_mark()) { }
bool doHeapRegion(HeapRegion *hr) {
assert(!hr->isHumongous(), "sanity");
assert(hr->in_collection_set(), "bad CS");
if (hr->claimHeapRegion(HeapRegion::ParEvacFailureClaimValue)) {
if (hr->evacuation_failed()) {
RemoveSelfForwardPtrObjClosure rspc(_g1h, _cm, hr, _update_rset_cl);
// In the common case (i.e. when there is no evacuation
// failure) we make sure that the following is done when
// the region is freed so that it is "ready-to-go" when it's
// re-allocated. However, when evacuation failure happens, a
// region will remain in the heap and might ultimately be added
// to a CSet in the future. So we have to be careful here and
// make sure the region's RSet is ready for parallel iteration
// whenever this might be required in the future.
hr->rem_set()->reset_for_par_iteration();
hr->reset_bot();
_update_rset_cl->set_region(hr);
hr->object_iterate(&rspc);
// A number of manipulations to make the TAMS for this region
// be the current top, and the marked bytes be the ones observed
// in the iteration.
if (_cm->at_least_one_mark_complete()) {
// The comments below are the postconditions achieved by the
// calls. Note especially the last such condition, which says that
// the count of marked bytes has been properly restored.
hr->note_start_of_marking(false);
// _next_top_at_mark_start == top, _next_marked_bytes == 0
hr->add_to_marked_bytes(rspc.prev_marked_bytes());
// _next_marked_bytes == prev_marked_bytes.
hr->note_end_of_marking();
// _prev_top_at_mark_start == top(),
// _prev_marked_bytes == prev_marked_bytes
}
// If there is no mark in progress, we modified the _next variables
// above needlessly, but harmlessly.
if (_g1h->mark_in_progress()) {
hr->note_start_of_marking(false);
// _next_top_at_mark_start == top, _next_marked_bytes == 0
// _next_marked_bytes == next_marked_bytes.
}
}
}
return false;
}
};
class G1ParRemoveSelfForwardPtrsTask: public AbstractGangTask {
protected:
G1CollectedHeap* _g1h;
public:
G1ParRemoveSelfForwardPtrsTask(G1CollectedHeap* g1h) :
AbstractGangTask("G1 Remove Self-forwarding Pointers"),
_g1h(g1h) { }
void work(uint worker_id) {
UpdateRSetImmediate immediate_update(_g1h->g1_rem_set());
DirtyCardQueue dcq(&_g1h->dirty_card_queue_set());
UpdateRSetDeferred deferred_update(_g1h, &dcq);
OopsInHeapRegionClosure *update_rset_cl = &deferred_update;
if (!G1DeferredRSUpdate) {
update_rset_cl = &immediate_update;
}
RemoveSelfForwardPtrHRClosure rsfp_cl(_g1h, update_rset_cl);
HeapRegion* hr = _g1h->start_cset_region_for_worker(worker_id);
_g1h->collection_set_iterate_from(hr, &rsfp_cl);
}
};
#endif // SHARE_VM_GC_IMPLEMENTATION_G1_G1EVACFAILURE_HPP

5
hotspot/src/share/vm/gc_implementation/g1/heapRegion.hpp
View file

@ -1,5 +1,5 @@
/*
* Copyright (c) 2001, 2011, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2001, 2012, 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
@ -373,7 +373,8 @@ class HeapRegion: public G1OffsetTableContigSpace {
ScrubRemSetClaimValue = 3,
ParVerifyClaimValue = 4,
RebuildRSClaimValue = 5,
CompleteMarkCSetClaimValue = 6
CompleteMarkCSetClaimValue = 6,
ParEvacFailureClaimValue = 7
};
inline HeapWord* par_allocate_no_bot_updates(size_t word_size) {