8296419: [REDO] JDK-8295319: pending_cards_at_gc_start doesn't include cards in thread buffers #11053

kimbarrett · 2022-11-09T04:57:38Z

Let's try this again. This is the original JDK-8295319 change, plus a couple
of small additions to fix problems with that original change.

The description of the original change was:

Please review this change to G1 to include the per-thread buffers in the
number of pending cards at the start of a young GC.

DCQS::concatenate_logs has been renamed to concatenate_logs_and_stats, and now
also merges the per-thread refinement stats during the thread walk to flush
buffers. That replaces the separate thread walk to merge and record these
stats earlier in the GC. The merged stats and related info don't seem to be
needed until after the buffer flushing.

Also, when abandoning dirty card buffers and stats because of a full GC, fixed
to also abandon any buffers in the paused buffers lists.

The problem was that by moving log concatenation earlier, it no longer
followed the call to retire_tlabs(). Among other things, that function
calls flush_deferred_card_mark_barrier() on each thread, possibly adding cards
to the thread's dirty card queue. Doing that after the log concatenation may
lead to unprocessed cards in thread queues, with chaos ensuing.

To fix this, the call to retire_tlabs has also been moved, so that it once
again precedes the log concatenation. Also added (debug-only) verification
that all thread dirty card queues are empty at the end of
pre_evacuate_collection_set.

A possible followup is to refactor those two operations, which each do a
(single-threaded) walk of all threads. We could combine that into a single
walk. We could also parallelize it if that seems warranted, though the
per-thread work is usually pretty small, so might not be worth parallelizing.

There might be an opportunity to do some similar refactoring for fullgc and
log abandonment, though there the two operations are done far away from each
other.

Testing:
mach5 tier1-6

Progress

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

Issue

JDK-8296419: [REDO] JDK-8295319: pending_cards_at_gc_start doesn't include cards in thread buffers

Reviewers

Thomas Schatzl (@tschatzl - Reviewer) ⚠️ Review applies to 576e799c
Stefan Johansson (@kstefanj - Reviewer) ⚠️ Review applies to 576e799c

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk pull/11053/head:pull/11053
$ git checkout pull/11053

Update a local copy of the PR:
$ git checkout pull/11053
$ git pull https://git.openjdk.org/jdk pull/11053/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 11053

View PR using the GUI difftool:
$ git pr show -t 11053

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/11053.diff

…sn't include cards in thread buffers" This reverts commit b847fb6.

bridgekeeper · 2022-11-09T04:58:46Z

👋 Welcome back kbarrett! 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 · 2022-11-09T05:01:41Z

@kimbarrett The following label will be automatically applied to this pull request:

hotspot-gc

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

mlbridge · 2022-11-09T05:06:05Z

Webrevs

tschatzl · 2022-11-09T08:59:55Z

src/hotspot/share/gc/g1/g1YoungCollector.cpp

+    // Flush early, so later phases don't need to account for per-thread stuff.
+    // Flushes deferred card marks, so must precede concatenting logs.
+    retire_tlabs();
+


I was wondering whether we should do something about all the methods (retire_tlabs, concatenate_dirty_card_logs..., calculate_collection_set) methods here that are part of the pre_evacuate_collection_set phase but are now located outside of that method.

Looking at the per_thread_states parameter passed to pre_evacuate_collection_set, it's only used for some verification inside pre_evacuate_collection_set, i.e. we could move that verification and initialization of the PSSS :) afterpre_evacuate_collection_set instead.

Since the change touches that code, I think it is appropriate (and better) to consolidate right away.

I agree about that deferring combining and parallelizing the various phases of pre_evacuate_collection_set can be done extra; the reason why we want parallelization of the TLAB retiring is that there can be tens of thousands threads, and even little work per thread adds up (there is a CR for that already actually, JDK-8211104).

Other than that it looks good to me.

I made a prototype for this suggestion: 587660c that seems to at least compile. No testing. One may also move rdcqs and preserved_marks_set completely into the G1ParScanThreadStateSet as neither is used elsewhere.

I also added that suggested refactoring here: https://github.com/openjdk/jdk/compare/pr/11053...tschatzl:jdk:pull/11053-suggestion?expand=1

kimbarrett · 2022-11-22T03:58:23Z

I've made some adjustments based on @tschatzl suggestion. (The suggestion didn't work as-is, but it was in the right direction, and just needed a bit of cleaning up.)

tschatzl

Lgtm, and thanks for the changes.

Apologies for introducing a few bugs in my suggestion.

tschatzl · 2022-11-22T09:18:58Z

src/hotspot/share/gc/g1/g1YoungCollector.cpp

@@ -501,13 +501,22 @@ void G1YoungCollector::verify_empty_dirty_card_logs() const {
 }
 #endif // ASSERT

-void G1YoungCollector::pre_evacuate_collection_set(G1EvacInfo* evacuation_info, G1ParScanThreadStateSet* per_thread_states) {
+void G1YoungCollector::pre_evacuate_collection_set(G1EvacInfo* evacuation_info) {
+  // Flush early, so later phases don't need to account for per-thread stuff.


I think this one is duplicate from the comment to concatenate_dirty_card_logs.

It was intentional, as it's useful to deal with TLABs early for the same reason as the DCQs. The 2nd line of the comment is about the respective ordering of these two early "flushes". (The plan of merging these into a single pass over the threads will render the duplicate comment question moot.)

openjdk · 2022-11-22T09:22:54Z

@kimbarrett 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:

8296419: [REDO] JDK-8295319: pending_cards_at_gc_start doesn't include cards in thread buffers

Reviewed-by: tschatzl, sjohanss

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

At the time when this comment was updated there had been no new commits pushed to the master branch. If another commit should be pushed before you perform the /integrate command, your PR will be automatically rebased. If you prefer to avoid any potential 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.

kstefanj

Looks good.

tschatzl

One more typo to fix before pushing.

tschatzl · 2022-11-24T10:38:58Z

src/hotspot/share/gc/g1/g1YoungCollector.cpp

+
+void G1YoungCollector::pre_evacuate_collection_set(G1EvacInfo* evacuation_info) {
+  // Flush early, so later phases don't need to account for per-thread stuff.
+  // Flushes deferred card marks, so must precede concatenting logs.


Suggested change

// Flushes deferred card marks, so must precede concatenting logs.

// Flushes deferred card marks, so must precede concatenating logs.

Fixed. Thanks.

kimbarrett · 2022-11-25T01:59:49Z

Thanks for reviews @tschatzl and @kstefanj .

kimbarrett · 2022-11-25T02:00:00Z

/integrate

openjdk · 2022-11-25T02:03:05Z

Going to push as commit 1fb9dad.

openjdk · 2022-11-25T02:03:22Z

@kimbarrett Pushed as commit 1fb9dad.

💡 You may see a message that your pull request was closed with unmerged commits. This can be safely ignored.

kimbarrett added 3 commits November 9, 2022 04:53

Revert "8296414: [BACKOUT] JDK-8295319: pending_cards_at_gc_start doe…

bd27ae2

…sn't include cards in thread buffers" This reverts commit b847fb6.

verify empty dirty card logs

218a139

retire tlabs before processing thread dcq

928b68b

openjdk bot added the rfr Pull request is ready for review label Nov 9, 2022

openjdk bot added the hotspot-gc hotspot-gc-dev@openjdk.org label Nov 9, 2022

tschatzl suggested changes Nov 9, 2022

View reviewed changes

kimbarrett added 3 commits November 21, 2022 07:54

Merge branch 'master' into cards-in-thread-buffers

c094464

rearrange pre-evacuation

49f5d7b

fix usage of _preserved_marks_set

576e799

tschatzl approved these changes Nov 22, 2022

View reviewed changes

openjdk bot added the ready Pull request is ready to be integrated label Nov 22, 2022

kstefanj approved these changes Nov 24, 2022

View reviewed changes

tschatzl approved these changes Nov 24, 2022

View reviewed changes

kimbarrett added 2 commits November 25, 2022 01:43

fix typo

81154e3

Merge branch 'master' into cards-in-thread-buffers

70dfb13

openjdk bot added the integrated Pull request has been integrated label Nov 25, 2022

openjdk bot closed this Nov 25, 2022

openjdk bot removed ready Pull request is ready to be integrated rfr Pull request is ready for review labels Nov 25, 2022

kimbarrett deleted the cards-in-thread-buffers branch November 25, 2022 06:52

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

8296419: [REDO] JDK-8295319: pending_cards_at_gc_start doesn't include cards in thread buffers #11053

8296419: [REDO] JDK-8295319: pending_cards_at_gc_start doesn't include cards in thread buffers #11053

kimbarrett commented Nov 9, 2022 •

edited by openjdk bot

Loading

bridgekeeper bot commented Nov 9, 2022

openjdk bot commented Nov 9, 2022

mlbridge bot commented Nov 9, 2022 •

edited

Loading

tschatzl Nov 9, 2022 •

edited

Loading

tschatzl Nov 15, 2022 •

edited

Loading

tschatzl Nov 15, 2022

kimbarrett commented Nov 22, 2022

tschatzl left a comment

tschatzl Nov 22, 2022

kimbarrett Nov 22, 2022

openjdk bot commented Nov 22, 2022 •

edited

Loading

kstefanj left a comment

tschatzl left a comment

tschatzl Nov 24, 2022

kimbarrett Nov 25, 2022

kimbarrett commented Nov 25, 2022

kimbarrett commented Nov 25, 2022

openjdk bot commented Nov 25, 2022

openjdk bot commented Nov 25, 2022

	// Flushes deferred card marks, so must precede concatenting logs.
	// Flushes deferred card marks, so must precede concatenating logs.

8296419: [REDO] JDK-8295319: pending_cards_at_gc_start doesn't include cards in thread buffers #11053

8296419: [REDO] JDK-8295319: pending_cards_at_gc_start doesn't include cards in thread buffers #11053

Conversation

kimbarrett commented Nov 9, 2022 • edited by openjdk bot Loading

Progress

Issue

Reviewers

Reviewing

bridgekeeper bot commented Nov 9, 2022

openjdk bot commented Nov 9, 2022

mlbridge bot commented Nov 9, 2022 • edited Loading

Webrevs

tschatzl Nov 9, 2022 • edited Loading

Choose a reason for hiding this comment

tschatzl Nov 15, 2022 • edited Loading

Choose a reason for hiding this comment

tschatzl Nov 15, 2022

Choose a reason for hiding this comment

kimbarrett commented Nov 22, 2022

tschatzl left a comment

Choose a reason for hiding this comment

tschatzl Nov 22, 2022

Choose a reason for hiding this comment

kimbarrett Nov 22, 2022

Choose a reason for hiding this comment

openjdk bot commented Nov 22, 2022 • edited Loading

kstefanj left a comment

Choose a reason for hiding this comment

tschatzl left a comment

Choose a reason for hiding this comment

tschatzl Nov 24, 2022

Choose a reason for hiding this comment

kimbarrett Nov 25, 2022

Choose a reason for hiding this comment

kimbarrett commented Nov 25, 2022

kimbarrett commented Nov 25, 2022

openjdk bot commented Nov 25, 2022

openjdk bot commented Nov 25, 2022

kimbarrett commented Nov 9, 2022 •

edited by openjdk bot

Loading

mlbridge bot commented Nov 9, 2022 •

edited

Loading

tschatzl Nov 9, 2022 •

edited

Loading

tschatzl Nov 15, 2022 •

edited

Loading

openjdk bot commented Nov 22, 2022 •

edited

Loading