February 2022 - hotspot-gc-dev

RFR: 8281254: GCs with survivor spaces report Init > Max through MXBeans
by Aleksey Shipilev 23 Jun '22

23 Jun '22

See the discussion in the bug. This realistically affects only Serial and Parallel. My first attempt was to push through `initial_size` to generations properly, but it is too intrusive to my liking. The current patch "just" caps the initial to max. Additional testing: - [x] New regression test now passes on all GCs - [x] Linux x86_64 `hotspot:tier1` ------------- Commit messages: - Revert extensive changes - Fix Parallel too - Better fix - Fix Changes: https://git.openjdk.java.net/jdk/pull/7341/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=7341&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8281254 Stats: 95 lines in 2 files changed: 94 ins; 0 del; 1 mod Patch: https://git.openjdk.java.net/jdk/pull/7341.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/7341/head:pull/7341 PR: https://git.openjdk.java.net/jdk/pull/7341

2 4

RFR: 8256265: G1: Improve parallelism in regions that failed evacuation
by Hamlin Li 07 Jun '22

07 Jun '22

Currently G1 assigns a thread per failed evacuated region. This can in effect serialize the whole process as often (particularly with region pinning) there is only one region to fix up. This patch tries to improve parallelism when walking over the regions in chunks Latest implementation scans regions in chunks to bring parallelism, it's based on JDK-8278917 which changes to uses prev bitmap to mark evacuation failure objs. ------------- Commit messages: - clean vm options - sync region preparation before iterate through chunks in a region - use const G1CMBitMap - adapt to update_bot_if_crossing_boundary changes - Merge branch 'master' into parallelize-evac-failure-in-bm - Merge from master - Initial commit Changes: https://git.openjdk.java.net/jdk/pull/7047/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=7047&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8256265 Stats: 504 lines in 17 files changed: 461 ins; 17 del; 26 mod Patch: https://git.openjdk.java.net/jdk/pull/7047.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/7047/head:pull/7047 PR: https://git.openjdk.java.net/jdk/pull/7047

4 59

RFR: 8280106: G1: Make reclaiming memory functionality more general in G1SegmentedArrayFreePool and G1SegmentedArrayFreeMemoryTask
by Hamlin Li 07 Apr '22

07 Apr '22

We have planned to make memory reclaimation more general, and reuse it in evacuation failure processing after JDK-8254739. Some of the work has already been done in JDK-8276670 and JDK-8277542. But after JDK-8278917, there is no need to return the memory used by G1SegmentedArray in evacuation failure processing, as it uses another way to do it. Currently, G1SegmentedArrayFreePool and G1SegmentedArrayFreeMemoryTask is half-bound to card set, and half-general, it should be made more general for the readability of the code and future possible reuse of this functionality returning memory (used by G1SegmentedArray) to OS. ------------- Commit messages: - Make G1SegmentedArrayFreePool public for tests - Move G1SegmentedArrayMemoryStats nested in G1SegmentedArrayFreePool - Initial commit Changes: https://git.openjdk.java.net/jdk/pull/7121/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=7121&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8280106 Stats: 178 lines in 14 files changed: 56 ins; 39 del; 83 mod Patch: https://git.openjdk.java.net/jdk/pull/7121.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/7121/head:pull/7121 PR: https://git.openjdk.java.net/jdk/pull/7121

3 8

RFR: 8280579: Shenandoah: Skip regions in the back of sorted array when choosing cset
by Yude Lin 25 Mar '22

25 Mar '22

Can I have review on this small change that skips some unnecessary work in cset choosing? When choosing regions to add to cset, we sort the regions from most garbage to least garbage. We then iterate the sorted array. We can break early from the loop if we find a region with (garbage <= garbage_threshold). Because we know the regions left won't have enough garbage and won't be added anyway. ------------- Commit messages: - 8280579: Shenandoah: Skip regions in the back of sorted array when choosing cset Changes: https://git.openjdk.java.net/jdk/pull/7211/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=7211&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8280579 Stats: 4 lines in 1 file changed: 4 ins; 0 del; 0 mod Patch: https://git.openjdk.java.net/jdk/pull/7211.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/7211/head:pull/7211 PR: https://git.openjdk.java.net/jdk/pull/7211

3 7

RFR: 8281879: Serial: Merge CardGeneration into TenuredGeneration
by Albert Mingkun Yang 21 Mar '22

21 Mar '22

Mostly mechanical changes; `git diff --color-moved=zebra` probably helps to see the actual (small) change, such as dropping the `virtual` modifier, introducing small helper methods to avoid naming collision, etc. Test: tier1-6 ------------- Commit messages: - flatten-cardgen Changes: https://git.openjdk.java.net/jdk/pull/7482/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=7482&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8281879 Stats: 893 lines in 14 files changed: 350 ins; 526 del; 17 mod Patch: https://git.openjdk.java.net/jdk/pull/7482.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/7482/head:pull/7482 PR: https://git.openjdk.java.net/jdk/pull/7482

3 5

RFR: 8271195: Use largest available large page size smaller than LargePageSizeInBytes when available
by Swati Sharma 16 Mar '22

16 Mar '22

Hi Team, In this patch I have fixed two issues related to large pages, following is the summary of changes :- 1. Patch fixes existing large page allocation functionality where if a commit over 1GB pages fails allocation should happen over next small page size i.e. 2M where as currently its happening over 4kb pages resulting into significant TLB miss penalty. Patch includes new JTREG Test case covering various scenarios for checking the correct explicit page allocation according to the 1G, 2M, 4K priority. 2. While attempting commit over larger pages we first try to reserve requested bytes over the virtual address space, in case commit to large page fails we should be un reserving entire reservation to avoid leaving any leaks in virtual address space. Please find below the performance data with and without patch for the JMH benchmark included with the patch. ![image](https://user-images.githubusercontent.com/96874289/152189587-4822a4ca-f5e2-4621-b405-0da941485143.png) Please review and provide your valuable comments. Thanks, Swati Sharma Runtime Software Development Engineer Intel ------------- Commit messages: - 8271195: Resolving jcheck failure. - 8271195: Use largest available large page size smaller than LargePageSizeInBytes when available Changes: https://git.openjdk.java.net/jdk/pull/7326/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=7326&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8271195 Stats: 352 lines in 5 files changed: 323 ins; 21 del; 8 mod Patch: https://git.openjdk.java.net/jdk/pull/7326.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/7326/head:pull/7326 PR: https://git.openjdk.java.net/jdk/pull/7326

8 64

RFR: 8256265 G1: Improve parallelism in regions that failed evacuation
by Hamlin Li 11 Mar '22

11 Mar '22

Currently G1 assigns a thread per failed evacuated region. This can in effect serialize the whole process as often (particularly with region pinning) there is only one region to fix up. Try to improve the parallelsim when walking over the regions by - first, split a region into tasks; - then, process these task in parallel and load balance among GC threads; - last, necessary cleanup The perf test based on lastest implementation + JDK-8277736 shows that: - when ParallelGCThreads=32, when G1EvacuationFailureALotCSetPercent <= 50, the parallelism bring more benefit than regression; - when ParallelGCThreads=128, whatever G1EvacuationFailureALotCSetPercent is, the parallelism bring more benefit than regression; For detailed perf test result, please check: - https://bugs.openjdk.java.net/secure/attachment/97227/parallel.evac.failure… - https://bugs.openjdk.java.net/secure/attachment/97228/parallel.evac.failure… NOTE: load balance part of code is almost same as G1ParScanThreadState, if necessary and feasible, consider to refactor this part into a shared code base. ------------- Commit messages: - Adjust worker cost by a factor; initialize task queues set and terminator threads by active workers - Fix wrong merge - Merge with master - Remove and merge code of G1ParRemoveSelfForwardPtrsTask into RemoveSelfForwardPtrsTask - Fix crashes in ~G1GCParPhaseTimesTracker(), G1PreRemoveSelfForwardClosure::do_heap_region, G1CollectedHeap::par_iterate_regions_array()=>~StubRoutines::atomic entry points; Refine comments - Fix inconsistent length between task queues and terminator - Fix crash when heap verification; Fix compilation error; Refine comments - Initial commit Changes: https://git.openjdk.java.net/jdk/pull/6627/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=6627&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8256265 Stats: 1055 lines in 22 files changed: 698 ins; 280 del; 77 mod Patch: https://git.openjdk.java.net/jdk/pull/6627.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/6627/head:pull/6627 PR: https://git.openjdk.java.net/jdk/pull/6627

5 20

RFR: 8276887: G1: Move precleaning to Concurrent Mark From Roots subphase
by Albert Mingkun Yang 04 Mar '22

04 Mar '22

Simple change of moving the "Preclean" subphase into "Mark from roots" subphase so that precleaning becomes parallel automatically. Evaluation using a contrived java program shows that multiple GC threads do precleaning. More detailed testing results are available in the JBS ticket. Test: hotspot_gc ------------- Commit messages: - preclean Changes: https://git.openjdk.java.net/jdk/pull/6327/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=6327&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8276887 Stats: 189 lines in 6 files changed: 38 ins; 128 del; 23 mod Patch: https://git.openjdk.java.net/jdk/pull/6327.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/6327/head:pull/6327 PR: https://git.openjdk.java.net/jdk/pull/6327

5 17

RFR: 8282381: Parallel: Remove unnecessary PCReferenceProcessor
by Albert Mingkun Yang 02 Mar '22

02 Mar '22

Simple change of removing a redundant class. Test: tier1-5 ------------- Commit messages: - ps-ref-discover Changes: https://git.openjdk.java.net/jdk/pull/7618/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=7618&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8282381 Stats: 32 lines in 1 file changed: 3 ins; 27 del; 2 mod Patch: https://git.openjdk.java.net/jdk/pull/7618.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/7618/head:pull/7618 PR: https://git.openjdk.java.net/jdk/pull/7618

3 6

RFR: 8267834: Refactor G1CardSetAllocator and BufferNode::Allocator to use a common base class
by Ivan Walulya 01 Mar '22

01 Mar '22

Hi all, Please review this change to combine free list allocation used in G1CardSetAllocator and BufferNode::Allocator into a single class. Testing: Tier 1-5 Thanks Ivan ------------- Commit messages: - 8267834: Refactor G1CardSetAllocator and BufferNode::Allocator to use a common base class Changes: https://git.openjdk.java.net/jdk/pull/7615/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=7615&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8267834 Stats: 717 lines in 10 files changed: 390 ins; 281 del; 46 mod Patch: https://git.openjdk.java.net/jdk/pull/7615.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/7615/head:pull/7615 PR: https://git.openjdk.java.net/jdk/pull/7615

3 7