8253037: G1: Improve check for string dedup #981

kimbarrett · 2020-11-01T07:04:29Z

Please review this change to G1 evacuation's checking for string
deduplication. The old code would go out of line to the G1StringDedup
support code for every object when deduplication is enabled, only to usually
discover the object isn't a string. Instead we now have a simpl inline
test for the combination of dedup enabled and object is a string, and only
call out to the dedup support code when that's true. This eliminates some
work for every non-string (non-array) object when dedup is enabled.

The performance impact seems to be pretty small and hard to measure, since
enabling deduplication has other costs that seem to overwhelm the cost
here. I'm hoping to improve that with JDK-8254598.

Testing:
tier1 on Oracle supported platforms.
Performance testing with deduplication enabled.

Progress

Change must not contain extraneous whitespace
Commit message must refer to an issue
Change must be properly reviewed

Issue

JDK-8253037: G1: Improve check for string dedup

Reviewers

Albert Mingkun Yang (@albertnetymk - Author) ⚠️ Review applies to cc583d7
Thomas Schatzl (@tschatzl - Reviewer) ⚠️ Review applies to 50fb6ad

Download

$ git fetch https://git.openjdk.java.net/jdk pull/981/head:pull/981
$ git checkout pull/981

bridgekeeper · 2020-11-01T07:05:27Z

👋 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 · 2020-11-01T07:07:26Z

@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.

kimbarrett · 2020-11-01T07:09:02Z

/summary Combine dedup enabled and is_string into a single test, using the already in-hand klass of the object.

openjdk · 2020-11-01T07:10:15Z

@kimbarrett Setting summary to Combine dedup enabled and is_string into a single test, using the already in-hand klass of the object.

mlbridge · 2020-11-01T07:12:59Z

Webrevs

albertnetymk

Nit: maybe mentioning java_string in the method name could make the precondition more explicit, sth like enqueue_java_string_from_evacuation.

kimbarrett · 2020-11-01T11:38:56Z

Nit: maybe mentioning java_string in the method name could make the precondition more explicit, sth like enqueue_java_string_from_evacuation.

That seems redundant, since this function is in StringDedup.

albertnetymk · 2020-11-01T12:18:51Z

That seems redundant, since this function is in StringDedup.

OK, then the fact that its neighbor, enqueue_from_mark, doesn't have such precondition (must be java_string), but shares the same naming schema could be a bit surprising. Or, maybe it's better to add the same precondition for enqueue_from_mark as well?

tschatzl · 2020-11-02T08:41:18Z

src/hotspot/share/gc/g1/g1ParScanThreadState.cpp

    _num_optional_regions(optional_cset_length),
    _numa(g1h->numa()),
    _obj_alloc_stat(NULL)
 {
+  // Verify klass comparison with _string_klass_or_null is sufficient


Not sure if this is the correct place to check this property: it needs not only hold during evacuation but basically anywhere. Is it possible to change finality of a java class?

What would be the impact of java.lang.String not passing this check? From what I understand only that these subclasses of j.l.String are not deduplicated. This does not seem to be a breaking (but surprising) problem, only that dedup for all subclasses of j.l.String does not work.

I would probably move that check into initialization, printing a warning.

I don't believe finality can be changed in a relevant way. (Not counting changing the source code and recompiling, or perhaps patching the jar file.) If some project were to try to change String, I think they'd very quickly trip this assert and know something needed to be done. Depending on what the change was, it could effectively disable deduplication (for example, if String were changed to be abstract).

I don't see a good way to check this during initialization that doesn't put the check far away from the code that cares, which makes it much less helpful. I thought about putting the assert at the point of use, but that seemd a little excessive.

tschatzl · 2020-11-02T08:41:18Z

src/hotspot/share/gc/g1/g1ParScanThreadState.hpp

@@ -83,6 +84,8 @@ class G1ParScanThreadState : public CHeapObj<mtGC> {
  // Size (in elements) of a partial objArray task chunk.
  int _partial_objarray_chunk_size;
  PartialArrayTaskStepper _partial_array_stepper;
+  // Used to check whether string dedup should be applied to an object.
+  Klass* _string_klass_or_null;


Maybe add a note that we replicate java_lang_String::is_instance_inlined() here for performance reasons (skipping the NULL check), so that when it changes, maybe a grep will find this here?

I think such a comment is better placed at the point of use, which is what I intend to do.

tschatzl · 2020-11-02T08:41:18Z

src/hotspot/share/gc/g1/g1StringDedup.hpp

@@ -75,7 +75,8 @@ class G1StringDedup : public StringDedup {
  // Enqueues a deduplication candidate for later processing by the deduplication
  // thread. Before enqueuing, these functions apply the appropriate candidate
  // selection policy to filters out non-candidates.
-  static void enqueue_from_mark(oop java_string, uint worker_id);
+  static void enqueue_from_mark(oop obj, uint worker_id);


As Albert mentioned, it would be nice if the caller of this would also get this optimization as it seems straightforward to implement there as well for consistency.

While the caller does not get the Klass* directly, it already reads the mark oop, so getting the Klass there should be practically free too.

Because of compressed klasses, getting the klass is never completely trivial. However, since effectively the first thing done by enqueue_from_mark is to test for the obj being a String, hoisting that out to fullgc's mark_object seems reasonable. And that gives the consistency between the enqueue functions that Albert asked for. I'll add that to the PR after re-running tests.

tschatzl

(Not necessarily requesting changes, but feedback to my comments)

openjdk · 2020-11-02T12:31:19Z

@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:

8253037: G1: Improve check for string dedup

Combine dedup enabled and is_string into a single test, using the already in-hand klass of the object.

Reviewed-by: ayang, tschatzl

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.

albertnetymk · 2020-11-02T13:11:41Z

Thank you for the change, Kim.

kimbarrett · 2020-11-04T05:00:01Z

/integrate

openjdk · 2020-11-04T05:06:06Z

@kimbarrett Pushed as commit 4b88119.

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

Combine dedup enabled and is_string into a single test, using the already in-hand klass of the object. Reviewed-by: ayang, tschatzl

kimbarrett · 2020-11-04T05:07:23Z

Thanks for reviewing @albertnetymk and @tschatzl

new check

cc583d7

kimbarrett changed the title ~~new check~~ 8253037: G1: Improve check for string dedupnew check Nov 1, 2020

kimbarrett changed the title ~~8253037: G1: Improve check for string dedupnew check~~ 8253037: G1: Improve check for string dedup Nov 1, 2020

openjdk bot added the hotspot-gc label Nov 1, 2020

kimbarrett marked this pull request as ready for review Nov 1, 2020

openjdk bot added the rfr label Nov 1, 2020

albertnetymk approved these changes Nov 1, 2020

View changes

tschatzl reviewed Nov 2, 2020

View changes

tschatzl suggested changes Nov 2, 2020

View changes

kimbarrett added 2 commits Nov 2, 2020

move string check from is_candidate_from_mark to fullgc caller

a0e505c

comment the combined is_enabled + is_string test

50fb6ad

tschatzl approved these changes Nov 2, 2020

View changes

openjdk bot added the ready label Nov 2, 2020

Merge branch 'master' into string_dedup_check

701f628

openjdk bot closed this Nov 4, 2020

openjdk bot added integrated and removed ready rfr labels Nov 4, 2020

kimbarrett deleted the string_dedup_check branch Nov 4, 2020

openjdk-notifier bot referenced this issue Nov 4, 2020

8253037: G1: Improve check for string dedup

4b88119

Combine dedup enabled and is_string into a single test, using the already in-hand klass of the object. Reviewed-by: ayang, tschatzl

8253037: G1: Improve check for string dedup #981

8253037: G1: Improve check for string dedup #981

kimbarrett commented Nov 1, 2020 •

edited by openjdk[bot] bot

bridgekeeper bot commented Nov 1, 2020

openjdk bot commented Nov 1, 2020

kimbarrett commented Nov 1, 2020

openjdk bot commented Nov 1, 2020

mlbridge bot commented Nov 1, 2020 •

edited

albertnetymk left a comment

kimbarrett commented Nov 1, 2020

albertnetymk commented Nov 1, 2020

tschatzl Nov 2, 2020

kimbarrett Nov 2, 2020

tschatzl Nov 2, 2020

kimbarrett Nov 2, 2020

tschatzl Nov 2, 2020 •

edited

kimbarrett Nov 2, 2020

tschatzl left a comment

openjdk bot commented Nov 2, 2020 •

edited

albertnetymk commented Nov 2, 2020

kimbarrett commented Nov 4, 2020

openjdk bot commented Nov 4, 2020

kimbarrett commented Nov 4, 2020

8253037: G1: Improve check for string dedup #981

8253037: G1: Improve check for string dedup #981

Conversation

kimbarrett commented Nov 1, 2020 • edited by openjdk[bot] bot

Progress

Issue

Reviewers

Download

bridgekeeper bot commented Nov 1, 2020

openjdk bot commented Nov 1, 2020

kimbarrett commented Nov 1, 2020

openjdk bot commented Nov 1, 2020

mlbridge bot commented Nov 1, 2020 • edited

Webrevs

albertnetymk left a comment

kimbarrett commented Nov 1, 2020

albertnetymk commented Nov 1, 2020

tschatzl Nov 2, 2020

Choose a reason for hiding this comment

kimbarrett Nov 2, 2020

Choose a reason for hiding this comment

tschatzl Nov 2, 2020

Choose a reason for hiding this comment

kimbarrett Nov 2, 2020

Choose a reason for hiding this comment

tschatzl Nov 2, 2020 • edited

Choose a reason for hiding this comment

kimbarrett Nov 2, 2020

Choose a reason for hiding this comment

tschatzl left a comment

openjdk bot commented Nov 2, 2020 • edited

albertnetymk commented Nov 2, 2020

kimbarrett commented Nov 4, 2020

openjdk bot commented Nov 4, 2020

kimbarrett commented Nov 4, 2020

kimbarrett commented Nov 1, 2020 •

edited by openjdk[bot] bot

mlbridge bot commented Nov 1, 2020 •

edited

tschatzl Nov 2, 2020 •

edited

openjdk bot commented Nov 2, 2020 •

edited