[release/9.0-staging] Backport #113095: Better ARM intrinsics implementation for dn_simdhash #113234
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
Restrict neon implementation to arm64 because arm32 neon doesn't have wide vector registers Add missing vreinterpret arm64 build fixes Ampere ARM64 optimizations Cleanup simdhash arch header and remove redundant assert Better define name
ghost
added
the
needs-area-label
kg
changed the title
jeffschwMSFT
approved these changes
Mar 6, 2025
3 tasks
|
I suggest to wait until #113241 is merged and then backport the final changes. Also, let's standup the build measurements in the perf lab to validate the improvements. |
jeffschwMSFT
removed
the
Servicing-consider
|
Closing in favor of #113316 |
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Outdated suggestions cannot be applied.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.
See #113074.
Customer impact
Multiple customers have reported seeing significantly slower iOS/MacCatalyst AOT compilation times in .NET 9 compared to .NET 8. We found the slowdown was specific to ARM macs and was due to ARM specific inefficiencies with a hashtable we introduced early on in the .NET 9 cycle.
Both internal and external customers are affected by this regression. This PR delivers performance on par with what customers experienced in .NET 8.
Regression
Yes, 9.0 is significantly slower than 8.0 for this scenario
Testing
We tested on Mac arm64/x64 and Linux arm64 for completeness. Performance was acceptable on all three.
Risk
Low as the change is fairly narrow in scope.