Implement faster pointer rebasing. #1696
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
Motivation: I recently discovered that UnsafeRawBufferPointer.init(rebasing:) is surprisingly expensive, with 7 traps and 11 branches. A simple replacement can make it a lot cheaper, down to two traps and four branches. This ends up having pretty drastic effects on ByteBuffer-heavy NIO code, which often outlines the call to that initializer and loses the ability to make a bunch of site-local optimisations. While this has been potentially fixed upstream with apple/swift#34879, there is no good reason to wait until Swift 5.4 for this improvement. Due to the niche use-case, I didn't bother doing this for _every_ rebasing in the program. In particular, there is at least one UnsafeBufferPointer(rebasing:) that I didn't do this with, and there are uses in both NIOTLS and NIOHTTP1 that I didn't change. While we can fix those if we really need to, it would be nice to avoid this helper proliferating too far through our codebase. Modifications: - Replaced the use of URBP.init(rebasing:) with a custom hand-rolled version that avoids Slice.count. Result: Cheaper code. One NIOHTTP2 benchmark sees a 2.9% speedup from this change alone.
Lukasa
added
the
patch-version-bump-only
glbrntt
approved these changes
Nov 27, 2020
Lukasa
added a commit
to Lukasa/swift-nio
that referenced
this pull request
Jan 7, 2021
This reverts commit 4bf379b.
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Outdated suggestions cannot be applied.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.
Motivation:
I recently discovered that UnsafeRawBufferPointer.init(rebasing:) is
surprisingly expensive, with 7 traps and 11 branches. A simple
replacement can make it a lot cheaper, down to two traps and four
branches. This ends up having pretty drastic effects on
ByteBuffer-heavy NIO code, which often outlines the call to that
initializer and loses the ability to make a bunch of site-local
optimisations.
While this has been potentially fixed upstream with
apple/swift#34879, there is no good reason to
wait until Swift 5.4 for this improvement.
Due to the niche use-case, I didn't bother doing this for every
rebasing in the program. In particular, there is at least one
UnsafeBufferPointer(rebasing:) that I didn't do this with, and there are
uses in both NIOTLS and NIOHTTP1 that I didn't change. While we can fix
those if we really need to, it would be nice to avoid this helper
proliferating too far through our codebase.
Modifications:
version that avoids Slice.count.
Result:
Cheaper code. One NIOHTTP2 benchmark sees a 2.9% speedup from this
change alone.