Skip to content

Conversation

@ddavis-2015
Copy link
Member

@ddavis-2015 ddavis-2015 commented Sep 1, 2025

@tensorflow/micro

Add optimized xtensa implementation of these decompression types to DECODE operator:

  • LUT
  • Pruning
  • Huffman

Makefile changes.

bug=fixes #3189

@tensorflow/micro

Add initial support for DECODE operator.
Add reference implementation.
Add LUT decompression support.
Update op resolvers.
Update Makefiles and Bazel BUILD files.
Add kernel unit test.

bug=fixes tensorflow#3131
@tensorflow/micro

Additional support for DECODE operator.
Add Xtensa optimizations for LUT decompression.
Move all Xtensa kernel source references to the Xtensa target makefile.

bug=fixes tensorflow#3150
@tensorflow/micro

Reorganize Xtensa makefiles such that all references to optimized kernel sources are moved to the Xtensa target makefile.

Move hifimini kernel sources to the parent directory, and rename them so they do not interfere with the target overlay mechanism of the root makefile.

bug=fixes tensorflow#3153
Fix code style errors.
@tensorflow/micro

Add reference implementation of pruning to DECODE operator.
Makefile and Bazel BUILD file changes.
Additional unit tests.

bug=fixes tensorflow#3161
@tensorflow/micro

Add optimized xtensa implementation of pruning to DECODE operator.
Makefile changes.
Additional unit tests.

bug=fixes tensorflow#3171
@tensorflow/micro

Add reference implementation of Huffman decompression to DECODE operator.
Makefile and Bazel BUILD file changes.
Additional unit tests.

bug=fixes tensorflow#3187
@tensorflow/micro

Add optimized xtensa implementation of Huffman decompression to DECODE operator.
Makefile changes.

bug=fixes tensorflow#3189
@ddavis-2015 ddavis-2015 marked this pull request as ready for review November 1, 2025 01:47
@ddavis-2015 ddavis-2015 requested a review from a team as a code owner November 1, 2025 01:47
@ddavis-2015 ddavis-2015 requested a review from veblush November 1, 2025 01:48
Copy link
Collaborator

@veblush veblush left a comment

Choose a reason for hiding this comment

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

I've reviewed the non-optimization parts of this PR (very small) and LGTM.

@veblush veblush self-requested a review November 5, 2025 17:59
Copy link
Collaborator

@veblush veblush left a comment

Choose a reason for hiding this comment

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

Approving. Relying on CI and the unit tests for the XTensa-specific implementation.

@veblush veblush merged commit 14b8584 into tensorflow:main Nov 6, 2025
95 of 96 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Support for DECODE operator

3 participants