Skip to content

Add HiFi-optimized Xtensa LSTM kernels - #3724

Open
unmeshna017 wants to merge 2 commits into
tensorflow:mainfrom
unmeshna017:xtensa-lstm-hifi
Open

unmeshna017 wants to merge 2 commits into
tensorflow:mainfrom
unmeshna017:xtensa-lstm-hifi

Conversation

@unmeshna017

Copy link
Copy Markdown
Contributor

Optimize the Xtensa unidirectional_sequence_lstm / lstm_eval kernels using the HiFi NNLib.

Depends on the 09_03_2026 NNLib release (PR#3703).

@cad-audio @joshih-cad

Optimize the Xtensa unidirectional_sequence_lstm / lstm_eval kernels using the HiFi NNLib. lstm_eval.cc uses xa_nn_matmul_sym8sxasym8s_sym16s for the int8xint8->int16 gate matmuls. Depends on the 09_03_2026 NNLib release (PR#3703).
@unmeshna017
unmeshna017 requested a review from a team as a code owner September 16, 2026 11:54
Add HIFI_IQ to the include guard so xa_nnlib_api.h and xa_nnlib_standards.h are pulled in for the hifi_iq build.
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.

1 participant