ggml-cpu: drop support for nnpa intrinsics #15821
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Aaron Teo <[email protected]>
github-actions
bot
added
documentation
slaren
reviewed
Sep 5, 2025
Signed-off-by: Aaron Teo <[email protected]>
slaren
approved these changes
Sep 5, 2025
walidbr
pushed a commit
to walidbr/llama.cpp
that referenced
this pull request
Sep 7, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Closes #15721
Supersedes #15739
This Pull Request drops support for the NNPA Vector Intrinsics as the maintenance cost required does not justify the performance improvements for FP32 ↔ FP16 conversion.
Tested with both
-fa offand-fa onand ensured that the inference is correct on both modes.For future reference to IBMers that want to bring this acceleration back,
-fa on, turned on by default) somehow causes tensor data to be invalid i.e.,-infandnan. Make sure to check that the data is clean before determining if the conversion implementation is correct. See: ggml-cpu: fixes instability in NNPA Vector Intrinsics #15739 (comment)ggml_compute_forward_dup_f32.