Skip to content

Conversation

@pramods-cad
Copy link
Contributor

@pramods-cad pramods-cad commented Oct 13, 2025

Optimization in 8x16 lstm layer for HiFi by using xa_nn_matXvec_v2_sym8sxsym16s_sym16s with appropriate conditions for FC.

bug=fixes #3231

…m8sxsym16s_sym16s with appropriate conditions for FC.
@pramods-cad pramods-cad removed their assignment Oct 15, 2025
@rameshkunasi
Copy link
Contributor

@pramods-cad Thank you for your PR. Can you please create a issue and add the issue in the PR description?
Please refer to PR description.

Example:
bug=fixes #123456

@pramods-cad
Copy link
Contributor Author

@rameshkunasi - I have added #3231 in PR description now.

@rameshkunasi
Copy link
Contributor

@veblush - Looks Good to Me

@ddavis-2015 ddavis-2015 requested a review from veblush October 15, 2025 18:24
@veblush veblush added the ci:run label Oct 17, 2025
@veblush veblush enabled auto-merge (squash) October 17, 2025 17:03
@TFLM-bot TFLM-bot removed the ci:run label Oct 17, 2025
@veblush veblush merged commit cee9550 into tensorflow:main Oct 17, 2025
52 of 53 checks passed
@pramods-cad pramods-cad deleted the lstm_opt branch October 27, 2025 07:14
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.

Performance improvement for 8x16 LSTM on HiFi for single batch case.

5 participants