Make initialization in OnceNonZeroUsize::get_or_try_init #[cold].
#273
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
In typical use, the Some branch is going to be taken extremely frequently but the None branch is only going to be taken once (or a very small number of times) at startup. If `get_or_try_init` is in a performance-sensitive segment of code, then it is important that `get_or_try_init` be inlined and that the compiler understands that the Some branch is (much) more likely than the None branch. When this happens, the call site basically becomes a load followed by a conditional jump that is basically never taken; which is ideal. When `get_or_try_init` is used in many places in the user's code, it is important to avoid inlining any of the None branch into the call sites. Unfortunately, the Rust compiler is sometimes not good at recognizing that code that calls a #[cold] function unconditionally must be cold itself. So, sometimes it isn't enough to mark our f as `#[cold] #[inline(never)]`. Move the entire body of the None branch into a function that is marked because some post-inlining optimization passes in the compiler seem to not understand `#[cold]`, and because we don't want any part of that branch to be in the calling code.
matklad
approved these changes
Feb 6, 2025
|
Thanks! |
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Outdated suggestions cannot be applied.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.
In typical use, the Some branch is going to be taken extremely frequently but the None branch is only going to be taken once (or a very small number of times) at startup. If
get_or_try_initis in a performance-sensitive segment of code, then it is important thatget_or_try_initbe inlined and that the compiler understands that the Some branch is (much) more likely than the None branch. When this happens, the call site basically becomes a load followed by a conditional jump that is basically never taken; which is ideal.When
get_or_try_initis used in many places in the user's code, it is important to avoid inlining any of the None branch into the call sites.Unfortunately, the Rust compiler is sometimes not good at recognizing that code that calls a #[cold] function unconditionally must be cold itself. So, sometimes it isn't enough to mark our f as
#[cold] #[inline(never)].Move the entire body of the None branch into a function that is marked because some post-inlining optimization passes in the compiler seem to not understand
#[cold], and because we don't want any part of that branch to be in the calling code.