8390907: Genshen: Improve mark loop performance by earthling-amzn · Pull Request #32488 · openjdk/jdk

earthling-amzn · 2026-08-21T22:09:56Z

There are several small changes here that add up to a 10% to 12% reduction in mark times on specjbb2015.

Gate the re-remembering work on a template parameter so that only the remembered set scan closure needs to do it
Bias the affiliation byte array to eliminate a subtraction for every oop
Loosen the check on heap inclusion for decoded oops
Inline the card marking function
Inline age census add function
Test for two pointers being in the same region before testing if old points to young

I confirm that I make this contribution in accordance with the OpenJDK Interim AI Policy.

Progress

Change must not contain extraneous whitespace
Commit message must refer to an issue
Change must be properly reviewed (2 reviews required, with at least 1 Reviewer, 1 Author)

Issue

JDK-8390907: Genshen: Improve mark loop performance (Enhancement - P4)

Reviewers

Kelvin Nilsen (@kdnilsen - Reviewer)
Aleksey Shipilev (@shipilev - Reviewer)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/32488/head:pull/32488
$ git checkout pull/32488

Update a local copy of the PR:
$ git checkout pull/32488
$ git pull https://git.openjdk.org/jdk.git pull/32488/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 32488

View PR using the GUI difftool:
$ git pr show -t 32488

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/32488.diff

Using Webrev

Link to Webrev Comment

…ring-mark

bridgekeeper · 2026-08-21T22:11:28Z

👋 Welcome back wkemper! A progress list of the required criteria for merging this PR into master will be added to the body of your pull request. There are additional pull request commands available for use with this pull request.

openjdk · 2026-08-21T22:12:06Z

@earthling-amzn This change now passes all automated pre-integration checks.

ℹ️ This project also has non-automated pre-integration requirements. Please see the file CONTRIBUTING.md for details.

After integration, the commit message for the final commit will be:

8390907: Genshen: Improve mark loop performance

Reviewed-by: kdnilsen, shade

You can use pull request commands such as /contributor and /issue to adjust it as needed.

At the time when this comment was updated there had been 7 new commits pushed to the master branch:

c436633: 8391044: [zgc] unused parameter "node" in z_color and z_uncolor
f86752b: 8389938: Follow up after removing UseObjectMonitorTable flag
22933b7: 8358889: C2 hits assert in backend due to malformed uncommon trap
... and 4 more: https://git.openjdk.org/jdk/compare/f40a2c3625484087cfdb41d34b360414ccf0ebd0...master

As there are no conflicts, your changes will automatically be rebased on top of these commits when integrating. If you prefer to avoid this automatic rebasing, please check the documentation for the /integrate command for further details.

➡️ To integrate this PR with the above commit message to the master branch, type /integrate in a new comment.

openjdk · 2026-08-21T22:12:52Z

@earthling-amzn The following labels will be automatically applied to this pull request:

hotspot-gc
shenandoah

When this pull request is ready to be reviewed, an "RFR" email will be sent to the corresponding mailing lists. If you would like to change these labels, use the /label pull request command.

openjdk · 2026-08-21T22:12:55Z

The total number of required reviews for this PR has been set to 2 based on the presence of this label: hotspot-gc. This can be overridden with the /reviewers command.

shipilev · 2026-08-24T14:51:31Z

+      // may visit class metadata that lives outside the heap so we cannot
+      // assume (or assert) that `p` is in old.
+      heap->old_generation()->mark_card_as_dirty(p);
+    } else if (GENERATION == GLOBAL && heap->has_affiliation(p, OLD_GENERATION) && heap->has_affiliation(obj, YOUNG_GENERATION)) {


Should this be also predicated with DIRTY, or am I missing something?

REDIRTY is specifically for the remembered set scan. I have added more asserts to make this more clear. I also though about calling this parameter REREMEMBER, but it looked too weird.

shipilev · 2026-08-24T14:52:14Z

  inline bool is_write_card_dirty(HeapWord* p) const;
-  inline void mark_card_as_dirty(HeapWord* p);
+
+  void mark_card_as_dirty(HeapWord* p) const {


Functions like these really belong in .inline.hpp. Putting anything non-trivial in the .hpp risks circular dependencies that are hard to untangle.

shipilev · 2026-08-24T14:53:07Z


 public:
-  template<class T, ShenandoahGenerationType GENERATION>
+  template<class T, ShenandoahGenerationType GENERATION, bool REDIRTY = false>


We usually do not do default values for template parameters to force callers to decide whether they want it or not. Are there many instances that would require fixing if you drop = false?

Only one caller would set it true. The instantiations aren't spread all over the code base or anything, I was just trying to minimize changes in the PR.

openjdk · 2026-08-24T18:10:44Z

@earthling-amzn this pull request can not be integrated into master due to one or more merge conflicts. To resolve these merge conflicts and update this pull request you can run the following commands in the local repository for your personal fork:

git checkout simplify-reremember-during-mark
git fetch https://git.openjdk.org/jdk.git master
git merge FETCH_HEAD
# resolve conflicts and follow the instructions given by git merge
git commit -m "Merge master"
git push

shipilev · 2026-08-24T18:12:57Z

In my runs, I also see ShenandoahAgeCensus::add not inlined. Have you seen it in yours?

…ring-mark

shipilev · 2026-08-25T07:48:10Z

+void ShenandoahDirectCardMarkRememberedSet::mark_card_as_dirty(HeapWord* p) const {
+  size_t index = card_index_for_addr(p);
+  CardValue* bp = &(_card_table->write_byte_map())[index];
+  bp[0] = CardTable::dirty_card_val();


Another question: does this write need to be conditional? We default to +UseCondCardMark to avoid contention on marks that are already dirty. Is the path we are optimizing benefits from the same optimization?

For a young GC, the read card table will have been cleaned and then swapped with the write table, so the write table is clean here (which is why the rset closure redirties the cards at it scans the read table). During a global GC in generational mode, the rset is not scanned, but the tables are still swapped, so cards should also be clean here.

This function is also called during operations on the discovered list for reference processing, but I expect the write volume would not be high there as the card is only marked when an old reference is linked to a young reference.

…ring-mark

mlbridge · 2026-08-25T21:58:37Z

Webrevs

00: Full (38039654)

kdnilsen

Thanks. Nice performance improvements.

shipilev

Looks all right with a few nits.

shipilev · 2026-08-26T09:10:47Z

 * Copyright (c) 2015, 2020, Red Hat, Inc. All rights reserved.
 * Copyright Amazon.com Inc. or its affiliates. All Rights Reserved.
- * Copyright (c) 2025, 2026, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2025, Oracle and/or its affiliates. All rights reserved.


This moves dates backwards, so it does not look right.

shipilev · 2026-08-26T09:25:54Z

+  if (ShenandoahHeapRegion::is_in_same_region(maybe_old, maybe_young)) {
+    return false;
+  }


Just for my understanding. This is a narrower check than we can do? I would have tried to check that region affiliations are different. But that performs access to affiliation bitmap, so that's why we do this address-only check, right?

earthling-amzn added 11 commits August 12, 2026 00:20

Have only rset scan 're-remember' the interesting pointers

875c383

Fix C++ error

97f8f87

Can't assume rset scan closure won't encounter a young pointer

6ed1c27

Merge remote-tracking branch 'jdk/master' into simplify-reremember-du…

6086fa7

…ring-mark

Bias affiliation array

6bf7711

Inline card dirtying

fb2d4f4

Skip in-heap check when we have decoded the oop already

b2475c2

Optimize new rset scan mark closure

be3f3fb

Use lighterweight is_in_reserved when old marking finds a young pointer

b39e172

Use AtomicLoad for biased_affiliations (same access as _affiliations)

1b1a738

Add REDIRTY template parameter to reduce duplication of mark logic

ce6c6b3

earthling-amzn marked this pull request as draft August 21, 2026 22:11

openjdk Bot added hotspot-gc hotspot-gc-dev@openjdk.org shenandoah shenandoah-dev@openjdk.org labels Aug 21, 2026

shipilev reviewed Aug 24, 2026

View reviewed changes

earthling-amzn added 7 commits August 24, 2026 09:32

Update copyright years

c38843b

Strengthen assertions for marking invariants

fb3380f

Document requirements for oop overload of has_affiliation

53a7092

Have existing is_in_young, is_in_old delegate to has_affiliation

e87261d

Fix typo in comment

ebd85b3

Initial new member to nullptr

0d9da14

Fix idiosyncratic whitespace

35ffaa0

openjdk Bot added the merge-conflict Pull request has merge conflict with target branch label Aug 24, 2026

Remove default value for REDIRTY template parameter

061c038

earthling-amzn added 5 commits August 24, 2026 13:50

Move mark_card_as_dirty into inline header (and fix all the fallout)

3e42246

Inline ShAgeCensus::age

44f0cf8

Update copyrights harder

0798c1e

Test if field and value are in same region before checking affiliation

b22d7ea

Merge remote-tracking branch 'jdk/master' into simplify-reremember-du…

57973e5

…ring-mark

openjdk Bot removed the merge-conflict Pull request has merge conflict with target branch label Aug 24, 2026

shipilev reviewed Aug 25, 2026

View reviewed changes

earthling-amzn added 2 commits August 25, 2026 19:07

Test for a nullptr before looking up region

ca7c146

Merge remote-tracking branch 'jdk/master' into simplify-reremember-du…

3803965

…ring-mark

earthling-amzn marked this pull request as ready for review August 25, 2026 21:53

openjdk Bot added the rfr Pull request is ready for review label Aug 25, 2026

kdnilsen approved these changes Aug 26, 2026

View reviewed changes

shipilev approved these changes Aug 26, 2026

View reviewed changes

openjdk Bot added the ready Pull request is ready to be integrated label Aug 26, 2026

Uh oh!

Conversation

earthling-amzn commented Aug 21, 2026 • edited by openjdk Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Progress

Issue

Reviewers

Reviewing

Uh oh!

bridgekeeper Bot commented Aug 21, 2026

Uh oh!

openjdk Bot commented Aug 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

openjdk Bot commented Aug 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

openjdk Bot commented Aug 21, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

openjdk Bot commented Aug 24, 2026

Uh oh!

shipilev commented Aug 24, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

earthling-amzn Aug 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mlbridge Bot commented Aug 25, 2026

Webrevs

Uh oh!

kdnilsen left a comment

Choose a reason for hiding this comment

Uh oh!

shipilev left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

3 participants

earthling-amzn commented Aug 21, 2026 •

edited by openjdk Bot

Loading

openjdk Bot commented Aug 21, 2026 •

edited

Loading

openjdk Bot commented Aug 21, 2026 •

edited

Loading

earthling-amzn Aug 25, 2026 •

edited

Loading

shipilev left a comment •

edited

Loading