Skip to content

Conversation

@darshanp4
Copy link

@darshanp4 darshanp4 commented Dec 19, 2023

For issue dgemm performance degradation on ARM NEOVERSEV1 with lower P*Q #4323 updated the SGEMM and DGEMM PQ param as per the configuration of NEOVERSEV1 cache size. Not keeping too big but enough to get better performance and close enough to half of the cache size.
Below is OpenBLAS/gemm.c benchmark on single core ( also seen on multicore) of graviton3 machine. Performance for SGEMM improved by ~ 2-5% and DGEMM improved by ~2-12%

image

@martin-frbg please help to get it review and merge.

Copy link
Contributor

@Mousius Mousius left a comment

Choose a reason for hiding this comment

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

These seem sensible to me! Thanks @darshanp4 😸

@darshanp4
Copy link
Author

@Mousius thank you for review , much appreciated! 👍

@martin-frbg martin-frbg added this to the 0.3.26 milestone Dec 19, 2023
@martin-frbg
Copy link
Collaborator

Great, thank you very much.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

dgemm performance degradation on ARM NEOVERSEV1 with lower P*Q

3 participants