Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

src: cpu: aarch64: Re-enable matmul and convolution static quantisation through ACL. #2308

Open
wants to merge 2 commits into
base: main
Choose a base branch
from

Conversation

renato-arantes
Copy link
Contributor

Description

This is to re-enable matmul and conv static quantization operations through ACL. Currently, the supported data type combinations are s8:s8:s8 and u8:s8:u8. Neoverse-N1 is skipped until failing tests are fixed.

General

  • Do all unit and benchdnn tests (make test and make test_benchdnn_*) pass locally for each commit?
  • Have you formatted the code using clang-format?

@renato-arantes
Copy link
Contributor Author

Hi @mgouicem, it's the same as before but more restrictive with Neoverse-N1 until the failing tests are fixed.

if (!ok) return status::unimplemented;

if (weights_md_.ndims != 4) return status::unimplemented;

// reject in case the op is running in a Neoverse-N1 with mixed sign quantization.
if (!arm_compute::CPUInfo::get().has_i8mm() && src_t == dnnl_u8
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This if statement is the only difference b/w the previous version, right?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, that is true.

@github-actions github-actions bot added the platform:cpu-aarch64 Codeowner: @oneapi-src/onednn-cpu-aarch64 label Dec 23, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
platform:cpu-aarch64 Codeowner: @oneapi-src/onednn-cpu-aarch64
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants