Skip to content

Add experimental CMSIS-NN float16/float32 support#222

Merged
mansnils merged 2 commits into
ARM-software:mainfrom
FabKlein:cmsis_nn_float
Jun 15, 2026
Merged

Add experimental CMSIS-NN float16/float32 support#222
mansnils merged 2 commits into
ARM-software:mainfrom
FabKlein:cmsis_nn_float

Conversation

@FabKlein

Copy link
Copy Markdown
Contributor
- Introduce float16/float32 operator APIs aligned with quantized CMSIS-NN, using channel-last layouts only.

- Make float support fully opt-in (`f16`, `f32`, or both) to avoid extra code size for integer-only builds.

- Add optional optimized kernels for convolution, depthwise convolution, and pooling.

- Expand float unit test coverage.

- Add GitHub Actions for FVP tests, host tests, and code-size tracking.

Fabien Klein and others added 2 commits June 15, 2026 16:13
    - Introduce float16/float32 operator APIs aligned with quantized CMSIS-NN,
    using channel-last layouts only.

    - Make float support fully opt-in (`f16`, `f32`, or both) to avoid extra
    code size for integer-only builds.

    - Add optional optimized kernels for convolution, depthwise convolution,
    and pooling.

    - Expand float unit test coverage.

    - Add GitHub Actions for FVP tests, host tests, and code-size tracking.
Pass --agree-embedded-license to cpackget installs used by the CI workflows so fresh runners do not stop at an embedded license prompt when the pack cache is empty.

Also apply clang-format spacing fixes reported by the formatter check.

Signed-off-by: Fabien Klein <fabien.klein@arm.com>
@mansnils mansnils merged commit 91f84c8 into ARM-software:main Jun 15, 2026
11 checks passed
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.

2 participants