perf: Use one query per crate for lang items, not one per lang item #21149
Conversation
This file contains hidden or 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
Via separate methods. This is both more clear, shorter, and will be required for the next commit.
rustbot
added
the
S-waiting-on-review
ChayimFriedman2
commented
Nov 28, 2025
| } | ||
| "#, | ||
| expect
ChayimFriedman2
force-pushed
the
lang-items
branch
2 times, most recently
from
24c68f5 to
d3d0845
Compare
Lang items rarely change, so putting a query for each doesn't give us anything. On the other hand, putting them behind only one query not only saves memory, it also has a giant benefit: we can store the struct with all lang items in the interner, making access to them very cheap. That basically means that anything in the hot path can avoid a *very common* query, and exchange it for a simple field access.
ChayimFriedman2
force-pushed
the
lang-items
branch
from
d3d0845 to
d9bee82
Compare
Contributor
Author
|
So I checked and this saves 7mb. Cute. Speed impact is more real. (However thinking about it it's amusing: those 7mb were likely caused, almost all of them, only by query dependencies!) |
Veykril
reviewed
Nov 29, 2025
| /// Salsa query. This will look for lang items in a specific crate. | ||
| #[salsa_macros::tracked(returns(ref))] | ||
| pub fn crate_lang_items(db: &dyn DefDatabase, krate: Crate) -> Option<Box<LangItems>> { | ||
| pub fn crate_lang_items(db: &dyn DefDatabase, krate: Crate) -> LangItems { |
Member
There was a problem hiding this comment.
I think we should keep the Option<Box<_>> here, 99.9% of all crates will have this by completely empty after all, unlike the lang_items query
Contributor
Author
There was a problem hiding this comment.
It doesn't really matter and makes the code a tad more complicated, having few extra kilobytes per crate won't matter for memory usage.
Veykril
approved these changes
Nov 29, 2025
rustbot
removed
the
S-waiting-on-review
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Lang items rarely change, so putting a query for each doesn't give us anything. On the other hand, putting them behind only one query not only saves memory, it also has a giant benefit: we can store the struct with all lang items in the interner, making access to them very cheap. That basically means that anything in the hot path can avoid a very common query, and exchange it for a simple field access.
This saves 3 seconds on analysis-stats on self. I can't measure the impact on memory as it fluctuates wildly due to some recent changes, but I believe the impact will be positive, and definitely not negative.
Reviewing commit-by-commit is recommended. The first commit is more technical and could use a skim review.