Add AArch64 SVE2 Vector Length Agnostic builders #317
Merged
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
Contributor
Author
|
The first 2 commits of this are #311. Please review only the last commit here. @maxim-kuvyrkov please confirm that this is exactly what was requested. |
asb
reviewed
Nov 20, 2024
DavidSpickett
force-pushed
the
sve2-builders
branch
from
682c0c8 to
bc0a174
Compare
Contributor
Author
|
I've also confirmed that at least right now, these configurations are green. They will be silent to begin with anyway, so that we can work out any capacity issues. |
maxim-kuvyrkov
approved these changes
Nov 21, 2024
Contributor
maxim-kuvyrkov
left a comment
maxim-kuvyrkov
left a comment
There was a problem hiding this comment.
Other than the parallelism comment -- looks great. Thanks!
Contributor
|
This looks good to me overall. just one minor nit inline. |
omjavaid
approved these changes
Nov 29, 2024
These generate code that uses the SVE2 extension, as opposed to our existing builders that only use SVE. Therefore they will run on our Graviton 4 workers only, which are Neoverse v2 cores with SVE and SVE2. The optimisation guide: https://github.com/aws/aws-graviton-getting-started/blob/main/c-c++.md Suggests -mcpu=neoverse-v2 for best performance, but here I'm using it just so we get the SVE2 feature enabled. If we used the balanced -mcpu=neoverse-512tvb, we wouldn't get SVE2 without fiddly extra flags. ``` $ ~/clang+llvm-18.1.8-aarch64-linux-gnu/bin/clang /tmp/test.c -o /dev/null -mcpu=neoverse-v2 -### <...> "-target-feature" "+sve" "-target-feature" "+sve2-bitperm" "-target-feature" "+sve2" <...> ``` I have not added vector length settings as these builders will generate scalable ("length agnostic" - LA) code. Otherwise, the builders have identical setups to the existing SVE VLA builders.
DavidSpickett
force-pushed
the
sve2-builders
branch
from
26e7ee1 to
f1a9b25
Compare
DavidSpickett
force-pushed
the
sve2-builders
branch
from
f1a9b25 to
8be6daa
Compare
Contributor
Author
|
CI failure is due to an unrelated change, merging. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 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.
These generate code that uses the SVE2 extension, as opposed to our existing builders that only use SVE. Therefore they will run on our Graviton 4 workers only, which are Neoverse v2 cores with SVE and SVE2.
The optimisation guide:
https://github.com/aws/aws-graviton-getting-started/blob/main/c-c++.md
Suggests -mcpu=neoverse-v2 for best performance, but here I'm using it just so we get the SVE2 feature enabled. If we used the balanced -mcpu=neoverse-512tvb, we wouldn't get SVE2 without fiddly extra flags.
I have not added vector length settings as these builders will generate scalable ("length agnostic" - LA) code.
Otherwise, the builders have identical setups to the existing SVE VLA builders.