Removed the HeapRegion::_top_at_conc_mark_count field. It is no longer needed as a result of the changes for 6888336 and 7127706. Refactored the closures that finalize and verify the liveness counting data so that common functionality was placed into a base class.
Reviewed-by: brutisso, tonyp
Cleanup of the CSet chooser class: standardize on uints for region num and indexes (instead of int, jint, etc.), make the method / field naming style more consistent, remove a lot of dead code.
Reviewed-by: johnc, brutisso
Change the type of fields / variables / etc. that represent region counts and indeces from size_t to uint.
Reviewed-by: iveresov, brutisso, jmasa, jwilhelm
Added log levels "fine", "finer" and "finest". Let PrintGC map to "fine" and PrintGCDetails map to "finer". Separated out the per worker information in the G1 logging to the "finest" level.
Reviewed-by: stefank, jwilhelm, tonyp, johnc
Tiered compilation has increased the number of nmethods in the code cache. This has, in turn, significantly increased the number of marked nmethods processed during the StrongRootsScope destructor. Create a specialized version of CodeBlobToOopClosure for G1 which places only those nmethods that contain pointers into the collection set on to the marked nmethods list.
Reviewed-by: iveresov, tonyp
Make two G1 cmd line flags available in product builds: G1HeapWastePercent (previously called: G1OldReclaimableThresholdPercent) and G1MixedGCCountTarget (previous called: G1MaxMixedGCNum). Also changed the default of the former from 1% to 5% and the default for G1OldCSetRegionLiveThresholdPercent to 90%.
Reviewed-by: azeemj, jwilhelm, johnc
Revamp of the mechanism that chooses old regions for inclusion in the CSet. It simplifies the code and introduces min and max bounds on the number of old regions added to the CSet at each mixed GC to avoid pathological cases. It also ensures that when we do a mixed GC we'll always find old regions to add to the CSet (i.e., it eliminates the case where a mixed GC will collect no old regions which can happen today).
Reviewed-by: johnc, brutisso
Re-enable survivors during the initial-mark pause. Afterwards, the concurrent marking threads have to scan them and mark everything reachable from them. The next GC will have to wait for the survivors to be scanned.
Reviewed-by: brutisso, johnc
This change simplifies the interaction between GC and concurrent marking. By disabling survivor spaces during the initial-mark pause we don't need to propagate marks of objects we copy during each GC (since we never need to copy an explicitly marked object).
Reviewed-by: johnc, brutisso
Change variables representing the number of GC workers to uint from int and size_t. Change the parameter in work(int i) to work(uint worker_id).
Reviewed-by: brutisso, tonyp
Renamed partially-young GCs as mixed and fully-young GCs as young. Change all external output that includes those terms (GC log and GC ergo log) as well as any comments, fields, methods, etc. The changeset also includes very minor code tidying up (added some curly brackets).
Reviewed-by: johnc, brutisso
Parallelize the serial code that was used to mark objects reachable from survivor objects in the collection set. Some minor improvments in the timers used to track the freeing of the collection set along with some tweaks to PrintGCDetails.
Reviewed-by: tonyp, brutisso
Race between the VM thread reading G1CollectedHeap::_mark_in_progress and it being set by the concurrent mark thread when concurrent marking is aborted by a full GC. Have the concurrent mark thread join the SuspendibleThreadSet before changing the marking state.
Reviewed-by: tonyp, brutisso
Major cleanup of the G1CollectorPolicy class. It removes a lot of unused fields and methods and also consolidates replicated information (mainly various ways of counting the number of CSet regions) into one copy.
Reviewed-by: johnc, brutisso
7102445: G1: Unnecessary Resource allocations during RSet scanning
Add a new per-worker thread line in the PrintGCDetails output. GC Worker Other is the difference between the elapsed time for the parallel phase of the evacuation pause and the sum of the times of the sub-phases (external root scanning, mark stack scanning, RSet updating, RSet scanning, object copying, and termination) for that worker. During RSet scanning, stack allocate DirtyCardToOopClosure objects; allocating these in a resource area was causing abnormally high GC Worker Other times while the worker thread freed ResourceArea chunks.
Reviewed-by: tonyp, jwilhelm, brutisso
Removed unused fields and methods, removed the G1CollectoryPolicy_BestRegionsFirst class and folded its functionality into the G1CollectorPolicy class.
Reviewed-by: ysr, brutisso, jcoomes
Some objects may not be marked in the event of an evacuation failure in a partially young GC, during a marking cycle. Avoid this situation by not allowing partially young GCs during a marking cycle.
Reviewed-by: tonyp, ysr, brutisso
G1 now uses two reference processors - one is used by concurrent marking and the other is used by STW GCs (both full and incremental evacuation pauses). In an evacuation pause, the reference processor is embedded into the closures used to scan objects. Doing so causes causes reference objects to be 'discovered' by the reference processor. At the end of the evacuation pause, these discovered reference objects are processed - preserving (and copying) referent objects (and their reachable graphs) as appropriate.
Reviewed-by: ysr, jwilhelm, brutisso, stefank, tonyp
The desired min / max heap sizes are miscalculated at initialization when NewRatio is used. The changeset also includes an additional small change to turn a print statement into a warning.
Reviewed-by: johnc, jmasa, ysr, brutisso
When refactoring the code for a previous fix, a condition was not correctly negated which prevents the G1 policy from adding the correct number of old regions to the CSet when the young gen size is fixed. The changeset also fixes a small syntactical issue in g1ErgoVerbose.hpp which is causing compiler warnings.
Reviewed-by: brutisso, ysr
Fixed inconsistencies and mistakes in the young list target length calculations so that a) the calculated target length is optimal (before, it was not), b) other parameters like max survivor size and max gc locker eden expansion are always consistent with the calculated target length (before, they were not always), and c) the resulting target length was always bound by desired min and max values (before, it was not).
Reviewed-by: brutisso, johnc
It introduces ergonomic decision logging in G1 for the following heuristics: heap sizing, collection set construction, concurrent cycle initiation, and partially-young GC start/end. The code has a bit of refactoring in a few places to make the decision logging possible. It also replaces alternative ad-hoc logging that we have under different parameters and switches (G1_DEBUG, G1PolicyVerbose).
Reviewed-by: johnc, ysr
Refactor the allocation code during GC to use the G1AllocRegion abstraction. Use separate subclasses of G1AllocRegion for survivor and old regions. Avoid BOT updates and dirty survivor cards incrementally for the former.
Reviewed-by: brutisso, johnc, ysr
In root processing, move the scanning of the reference processor's discovered lists to before RSet updating and scanning. When scanning the reference processor's discovered lists, use a buffering closure so that the time spent copying any reference object is correctly attributed. Also removed a couple of unused and irrelevant timers.
Reviewed-by: ysr, jmasa
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
This changeset extends the logging information generated by +PrintGCDetails to also print out separate size transitions for the eden, survivors, and old regions.
Reviewed-by: ysr, brutisso
Changed default value of WorkStealingYieldsBeforeSleep from 1000 to 5000. Added more information to G1 pause logging.
Reviewed-by: jwilhelm, tonyp, jmasa
7007446: G1: expand the heap with a single step, not one region at a time
Changed G1CollectedHeap::expand() to expand the committed space by calling VirtualSpace::expand_by() once rather than for every region in the expansion amount. This allows the success or failure of the expansion to be determined before creating any heap regions. Introduced a develop flag G1ExitOnExpansionFailure (false by default) that, when true, will exit the VM if the expansion of the committed space fails. Finally G1CollectedHeap::expand() returns a status back to it's caller so that the caller knows whether to attempt the allocation.
Reviewed-by: brutisso, tonyp
This changeset removes the zero-filling thread from G1 and collapses the two free region lists we had before (the "free" and "unclean" lists) into one. The new free list uses the new heap region sets / lists abstractions that we'll ultimately use it to keep track of all regions in the heap. A heap region set was also introduced for the humongous regions. Finally, this change increases the concurrency between the thread that completes freeing regions (after a cleanup pause) and the rest of the system (before we'd have to wait for said thread to complete before allocating a new region). The changest also includes a lot of refactoring and code simplification.
Reviewed-by: jcoomes, johnc
This change revamps the slow allocation path of G1. Improvements include the following: a) Allocations directly to old regions are now totally banned. G1 now only allows allocations out of young regions (with the only exception being humongous regions). b) The thread that allocates a new region (which is now guaranteed to be young) does not dirty all its cards. Each thread that successfully allocates out of a young region is now responsible for dirtying the cards that corresponding to the "block" that just got allocated. c) allocate_new_tlab() and mem_allocate() are now implemented differently and TLAB allocations are only done by allocate_new_tlab(). d) If a thread schedules an evacuation pause in order to satisfy an allocation request, it will perform the allocation at the end of the safepoint so that the thread that initiated the GC also gets "first pick" of any space made available by the GC. e) If a thread is unable to allocate a humongous object it will schedule an evacuation pause in case it reclaims enough regions so that the humongous allocation can be satisfied aftewards. f) The G1 policy is more careful to set the young list target length to be the survivor number +1. g) Lots of code tidy up, removal, refactoring to make future changes easier.
Reviewed-by: johnc, ysr
