[Bug 3620] [IcedTea8] Shenandoah broken on ppc64
bugzilla-daemon at icedtea.classpath.org
bugzilla-daemon at icedtea.classpath.org
Sat Nov 24 02:45:49 UTC 2018
https://icedtea.classpath.org/bugzilla/show_bug.cgi?id=3620
--- Comment #5 from hg commits <mercurial at icedtea.classpath.org> ---
details:
http://icedtea.classpath.org//hg/icedtea8?cmd=changeset;node=966b5fcf9312
author: Andrew John Hughes <gnu_andrew at member.fsf.org>
date: Sat Nov 03 14:32:58 2018 +0000
Bump shenandoah to
aarch64-shenandoah-jdk8u181-b15-shenandoah-merge-2018-09-19.
PR3634: Shenandoah still broken on s390 with
aarch64-shenandoah-jdk8u181-b16
Upstream changes:
- [backport] Adaptive CSet selection overshoots max-CSet
- [backport] Adaptive CSet selection selects excessively when memory is
tight
- [backport] Adaptive/Traversal heuristics rewrite for allocation rate
- [backport] Added logging for the number of workers used for GC cycles
- [backport] Add task termination and enhanced task queue state tracking
+ weakrefs
- [backport] Allocation path should not touch GC barriers for metadata
- [backport] Allocation tracker should really report bytes
- [backport] AlwaysPreTouch fails with non-default ConcGCThreads
- [backport] Application pacing precision fixes
- [backport] Apply ShenandoahEvacOOMScope only for evac-taking paths in
ShenandoahBarrierSet
- [backport] Assembler write barriers should consistently check for
forwarded objects
- [backport] Avoid indirection to next-mark-context
- [backport] Avoid using uintx in ShenandoahHeapRegion
- [backport] C1 shenandoah_wb expects obj in a register
- [backport] Check and ensure that Shenandoah-enabled compilations
succeed
- [backport] Check heap stability in C1 WBs
- [backport] ClassUnloadingWithConcurrentMark should be opt-in with
Shenandoah
- [backport] clean up obsolete c2 code
- [backport] Cleanup: remove unused root processor's sub tasks
- [backport] Cleanup UseShenandoahOWST blocks
- [backport] CollectedHeap::max_tlab_size is measured in words
- [backport] Complete liveness for recently allocated regions outside the
allocation path
- [backport] Comprehensible GC trigger logging
- [backport] Concurrent uncommit should be recorded as GC event
- [backport] Constify ShHeapRegionSet and ShCollectionSet
- [backport] Convert magic value to ShenandoahPacingSurcharge
- [backport] Default to ShenandoahCodeRootsStyle = 2
- [backport] Degenerated evacuation
- [backport] Disable UseFastJNIAccessors for Shenandoah
- [backport] Elastic TLABs support for Shenandoah
- [backport] Enable ShenandoahEvacReserveOverflow by default
- [backport] Evac assist should touch marked objects only
- [backport] Evac reserve: make sure GC has untouchable space to move the
objects into
- [backport] EvilSyncBug test is too slow
- [backport] Explicit GC should actually uncommit the heap
- [backport] Exponential backoff with pacing
- [backport] Fix aarch64 CAS predicates
- [backport] Fix CAS-obj predicates and add expected-null-versions for
cmpxchg-narrow-oop
- [backport] Fixed SA due to code refactoring and merging
- [backport] Fix Minimal and Zero builds
- [backport] Fix ShHeap::notify_alloc usages: it accepts words, not bytes
- [backport] Fix TestCommonGCLoads test
- [backport] Fix TestFullGCALot test failure
- [backport] Fix TestGCThreadGroups test
- [backport] Forceful SATB buffer flushes should be time-periodic, not
traffic-dependent
- [backport] Full GC always comes with liveness data
- [backport] Full GC should not always update references
- [backport] GCLAB slowpath allocations should fit the object into GCLAB
- [backport] GC trace messages have to be immortal
- [backport] Handle missing ShenandoahWriteBarrierRB case
- [backport] Heap region count selection should only consider max heap
size
- [backport] Hook up GCLABs to Elastic LAB support
- [backport] Improve scheduling and interleaving of SATB processing in
mark loop
- [backport] Including metaspace info when reporting heap info
- [backport] Incorrect label for static heuristics
- [backport] Make in-cset checks use signed bytes to match C2 better
- [backport] Micro-optimize AArch64 assembly write-barriers
- [backport] Missing Shenandoah entry in GCNameHelper::to_string
- [backport] More detailed pacing histogram
- [backport] More verbose profiling for phase 4 in mark-compact
- [backport] Move heuristics from ShCollectorPolicy to ShHeap
- [backport] Move (Java)Thread::_gc_state to lower offset to optimize
barrier fast-path encoding
- [backport] Move ParallelCodeIterator to ShenandoahCodeRoots
- [backport] Move periodic GC decision making to GC heuristics base class
- [backport] Move Shenandoah stress tests to tier3
- [backport] Non-cancellable mark loops should have sensible stride
- [backport] Off-by-one error in degen progress calculation
- [backport] Only Java and GC worker threads should get GCLABs
- [backport] Out-of-cycle Degenerated GC should process references and
unload classes
- [backport] Pacer for evacuation should print "Avail" to capture
discounting
- [backport] Pacer should account actual size for elastic TLABs
- [backport] Parallel +AlwaysPreTouch should run with max workers
- [backport] Passive heuristics should enter degen GC, not full GC
- [backport] Perform gc-state checks with LoadB to fit C2 matchers
- [backport] Pre-filter oops before enqueing them in SBS slowpaths
- [backport] Print task queue statistics at the end of GC cycle
- [backport] Process remaining SATB buffers in final mark/traverse loop
instead of separate phase
- [backport] Proper units for allocation failure messages
- [backport] Prune undefined and unused methods
- [backport] Purge partial heuristics and connection matrix
infrastructure
- [backport] Purge support for ShenandoahConcurrentEvacCodeRoots and
ShenandoahBarriersForConst
- [backport] Rearrange Shenandoah tests into 3 tiers
- [backport] Reclaim immediate garbage after mark-compact marking
- [backport] Recycle the regions only once
- [backport] Refactor alive-closures to deal better with new marking
contexts
- [backport] Refactor allocation path to accept ShenandoahAllocRequest
tuple
- [backport] Refactor and improve ShenandoahCodeRoots strategies
- [backport] Refactor FreeSet logging: support evac-reserve, denser
printouts
- [backport] Refactor gc+init logging
- [backport] Refactoring ShenandoahStrDedupStress test to reduce test
time
- [backport] Refactor to group marking bitmap and TAMS structure in one
class ShenandoahMarkingContext
- [backport] Remove C2 write-barrier from .ad files
- [backport] Removed racy assertion
- [backport] Remove NMethodSizeLimit adjustment for Shenandoah
- [backport] Remove obsolete/unused logging usages
- [backport] Remove safe_equals()
- [backport] Remove ShHeuristics::print_threshold
- [backport] Rename and move ShenandoahPrepareForMarkClosure
- [backport] Rename "cancel_concgc" to "cancel_gc"
- [backport] Replace custom asserts with shenandoah_assert_*
- [backport] Replace risky SBS::need_update_refs_barrier with
straightforward check
- [backport] Replace ShBarrierSet* casts with accessor
- [backport] Report actual free size in non-verbose FreeSet status
- [backport] Report heap region stats in proper units
- [backport] Resettable iterators to avoid dealing with
copying/assignment compilation differences
- [backport] Reshuffle tests: verify STW GC is working first, then verify
under aggressive, then the rest
- [backport] Rework ClassUnloading* flags handling
- [backport] Rework GC degradation on allocation failure
- [backport] SATB buffer filtering/compaction hides unmarked objects
until final-mark
- [backport] shenandoah_assert_correct should check object/forwardee
klasses
- [backport] shenandoah_assert_correct should verify classes before
claiming _safe_oop
- [backport] Shenandoah changes to allow enabling -Wreorder
- [backport] Shenandoah/PPC barrier stubs
- [Backport] Shenandoah string deduplication
- [backport] SH::make_(tlabs)_parsable() should work correctly
with/without TLABs
- [backport] Shortcut regions that are known not to be alive
- [backport] Should cleanup previous/bad versions of redefined classes
during full gc
- [backport] Skip RESOLVE in SATBBufferClosure if no forwarded objects
are in heap
- [backport] Soft-refs policy needs reliable heap usage data after the GC
cycle
- [backport] Soft refs should be purged reliably on allocation failure,
or with compact heuristics
- [backport] Some trivial-ish cleanups
- [backport] Split write barrier paths for mutator and GC workers
- [backport] StringInternCleanup times out
- [backport] TestHeapDump runs much faster with small heap
- [backport] Tests should use -XX:+ShenandoahVerify in some OOM-evac
configurations
- [backport] TLAB sizing policy should converge faster with Shenandoah
- [backport] Trace and report total allocation latency and sizes
- [backport] Traversal should resize TLABs
- [backport] Trivial enhancement to avoid costly deletion array element
- [backport] Uncommit should relinquish the heap lock regularly
- [backport] Unreachable assert in ShenandoahCodeRoots::acquire_lock
- [backport] Verifier should dump raw memory around the problematic oops
- [backport] Verify global and local gc-state status
- [backport] VSC++ requires space(s) in between two string literals
- [backport] WB slowpath should assist with evacuation of adjacent
objects
- [backport] Wiring GC events to JFR + Restore heap occupancy in GC logs
after JFR changes
- [backport] Wiring heap and metaspace info to JFR
- [backport] Wrap worker id in thread local worker session
- [backport] -XX:-UseTLAB should disable GCLABs too
- Cleanup undeclared methods in barrier stubs
- Disable evac assist by default until bugfixes arrive
- Fix build failure: signedness mismatch in assert
- Fix MacOS/Clang build failure
- Fix x86_32 build
- JDK8u: Silence compilation warnings on implicit type conversion
- Move JNI Weak References workaround to Shenandoah-specific root
processor
2018-10-29 Andrew John Hughes <gnu_andrew at member.fsf.org>
Bump shenandoah to
aarch64-shenandoah-jdk8u181-b15-shenandoah-merge-2018-09-19.
* patches/hotspot/shenandoah/pr3619.patch,
* patches/hotspot/shenandoah/pr3620.patch:
Removed as applied upstream.
* Makefile.am:
(ICEDTEA_PATCHES): Remove PR3619 and PR3620
patches. Add PR3634 patch.
* NEWS: Updated.
* hotspot.map.in: Bump shenandoah to
aarch64-shenandoah-jdk8u181-b15-shenandoah-merge-2018-09-19.
--
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/distro-pkg-dev/attachments/20181124/c689efcc/attachment-0001.html>
More information about the distro-pkg-dev
mailing list