Disable jiterp safepoints in single-threaded WASM #99558
Merged
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
|
Tagging subscribers to this area: @BrzVlad, @kotlarmilos |
|
Tagging subscribers to 'arch-wasm': @lewing |
vargaz
approved these changes
Mar 12, 2024
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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Outdated suggestions cannot be applied.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.
May make #99554 partially or completely go away.
The introduction of the imm branch opcodes is an improvement for native interp since safepoints are pretty cheap there, but in jiterp safepoints are somewhat expensive (theoretically) because they bloat traces, add more branches to push things out of the branch predictor table, and potentially cause trace compilation to end earlier. Before these superinstructions were added, the relevant branch would have been a non-safepoint branch.
Vlad pointed out the obvious: safepoints don't do anything in single threaded wasm, so the jiterp can just not generate any code for them. This should hopefully make the affected vector code fast again by shrinking traces, and potentially speed up other traces too.