-
Notifications
You must be signed in to change notification settings - Fork 6.1k
8370325: G1: Disallow GC for TLAB allocation #27932
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
8370325: G1: Disallow GC for TLAB allocation #27932
Conversation
Hi all, please review this change to the allocation path to only allow one path doing GCs during that time. Currently, when allocating an object, G1 (or actually any collector) first allocates from TLABs, then allocates from outside TLABs. When G1 can't find space for a TLAB in the current region, it passes on control to `G1CollectedHeap::attempt_allocation[_slow]` that first tries to allocate a new region, or do a GC. Allocation from outside TLABs does the same. G1's behavior to do a GC during failed TLAB allocation is problematic for the `UseGCOverheadLimit` functionality I'm also working on: if the GC overhead limit triggers in a particular operation, returning null for that TLAB allocation. This will will cause another GC right away trying to allocate outside TLAB. Since the garbage collections for TLAB allocations did free some memory (but because of exceeded gc overhead we returned null for the TLAB allocation), that allocation/collection for this outside-tlab allocation will succeed, effectively swallowing the attempt to tell the mutator that GC overhead has been exceeded. Testing: tier1-5, performance neutral after some tests Hth, Thomas
|
👋 Welcome back tschatzl! A progress list of the required criteria for merging this PR into |
|
@tschatzl 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: 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 56 new commits pushed to the
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 |
Webrevs
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Maybe the title can be revised a bit -- sth like "G1: Disallow GC for TLAB allocation".
Much better, thanks! |
|
Thanks @albertnetymk @walulyai for your reviews |
|
Going to push as commit 027aea9.
Your commit was automatically rebased without conflicts. |
Hi all,
please review this change to the object allocation path to only allow direct allocation after failing TLAB allocation doing GCs.
Currently, when allocating an object, G1 (or actually any collector) first allocates from TLABs, then allocates from outside TLABs.
When G1 can't find space for a TLAB in the current region, it passes on control to
G1CollectedHeap::attempt_allocation[_slow]that first tries to allocate that memory using a new region, or if that fails, do a GC. Potential allocation outside TLABs that would follow TLAB-allocation does the same.G1's behavior to do a GC during failed TLAB allocation is problematic for the
UseGCOverheadLimitfunctionality (https://bugs.openjdk.org/browse/JDK-8212084): if the GC overhead limit triggers in a TLAB allocation, it returnsnullfor that TLAB allocation.Control will be passed to outside-tlab allocation as described above. Since the garbage collections for TLAB allocations did free some memory (but because of exceeded gc overhead we returned null for the TLAB allocation), that allocation/collection for this outside-tlab allocation will succeed and return non-
nullto the mutator, effectively swallowing the information that GC overhead has been exceeded.I split this out from the
UseGCOverheadLimitchange because I thought it is somewhat standalone, but I can merge it.Testing: tier1-5, performance neutral after some tests
Hth,
Thomas
Progress
Issue
Reviewers
Reviewing
Using
gitCheckout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/27932/head:pull/27932$ git checkout pull/27932Update a local copy of the PR:
$ git checkout pull/27932$ git pull https://git.openjdk.org/jdk.git pull/27932/headUsing Skara CLI tools
Checkout this PR locally:
$ git pr checkout 27932View PR using the GUI difftool:
$ git pr show -t 27932Using diff file
Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/27932.diff
Using Webrev
Link to Webrev Comment