Support VO bit #158
Merged
Support VO bit #158
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Performance comparison of using the VO bit (inlined in the fastpath allocation): |
|
|
…expose functions from mmtk-core
qinsoon
reviewed
Jul 8, 2024
mmtk/src/api.rs
Outdated
| @@ -226,6 +226,16 @@ pub extern "C" fn mmtk_post_alloc( | |||
| bytes: usize, | |||
| semantics: AllocationSemantics, | |||
| ) { | |||
| #[cfg(all(feature = "object_pinning", not(feature = "non_moving")))] | |||
There was a problem hiding this comment.
Did you include this code when you ran benchmarks in #158 (comment)?
This should be fixed in mmtk-core by bulk clearing the pinning bits. We don't want to clear the bits here for every object.
There was a problem hiding this comment.
Yes, I did include it. Is it fixed in the mmtk-core version we're currently using?
I'll remove that code then and maybe try to re-run the benchmarks in the meantime.
There was a problem hiding this comment.
I don't think we have fixed it in mmtk-core. But we should fix it in mmtk-core.
There was a problem hiding this comment.
You also only clear the bits in this function, not in the JIT'd code. So for objects allocated in the JIT'd code, the pinning bits are not cleared properly.
There was a problem hiding this comment.
You also only clear the bits in this function, not in the JIT'd code. So for objects allocated in the JIT'd code, the pinning bits are not cleared properly.
I think I'm clearing them in the slowpath, not sure if they need to be cleared in the fastpath as well (since they will be bulk cleared for the memory chunk I get in the slowpath).
There was a problem hiding this comment.
There is no slowpath or fastpath for post_alloc. We either inline them, or not inline them. Both versions should be the same.
There was a problem hiding this comment.
What I meant was that in the allocation slowpath, you could clear the pin bits by calling mmtk_post_alloc, and it would be cleared for the chunk of memory you get between the new cursor and limit. But I checked the Julia PR and I wasn't calling mmtk_post_alloc there either, since I'm now inlining the code to set the VO bit. So, in fact, that code to bulk zero the pinning bits was not being executed in the results above.
There was a problem hiding this comment.
I see you removed the code for the pin bits. Just make sure that we are on the same page. With this PR, the pin bits are not cleared, right? It sounds fine to me. But it is an issue that we will need to fix.
There was a problem hiding this comment.
That's right. It should be fixed inside mmtk-core. I'm currently looking into it and will open a PR there instead.
Merged
qinsoon
added a commit
that referenced
this pull request
Dec 4, 2024
This PR ports #158 to `dev`, and updates Julia to mmtk/julia#76.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR introduces the
is_mmtk_objectfeature supporting a valid object (VO) bit for conservative stack scanning. It also sets this feature as default.NB: merge with mmtk/julia#59.
NB2: it requires a change inmmtk-coreto expose an api function to bulk set the VO bit. (see mmtk/mmtk-core#1157)