Skip to content

Add HiFi-optimized Xtensa Conv kernels - #3726

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

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

Conversation

@unmeshna017

Copy link
Copy Markdown
Contributor

Optimize the Xtensa conv kernels using the HiFi NNLib: dilated/group conv, float32, and Int4 filter dispatch, plus new reference fallbacks (conv_float32_reference.cc, conv_int8_int16_float32.cc). Registers the new helper sources in xtensa.inc.

Depends on the 09_03_2026 NNLib release (PR#3703) for the v2 per-channel conv APIs

@cad-audio @joshih-cad

Optimize the Xtensa conv kernels using the HiFi NNLib: dilated/group conv, float32, and Int4 filter dispatch, plus new reference fallbacks (conv_float32_reference.cc, conv_int8_int16_float32.cc). Registers the new helper sources in xtensa.inc. Depends on the 09_03_2026 NNLib release (PR#3703) for the v2 per-channel conv APIs (e.g. xa_nn_conv2d_std_v2_per_chan_sym4sxasym8s).
@unmeshna017
unmeshna017 requested a review from a team as a code owner September 17, 2026 07:05
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.

This branch has not been deployed

No deployments
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