Skip to content

Conversation

@shalinib-ibm
Copy link
Contributor

This patch gets rid of the redundant vec_perm insns by re-ordering the matrix multiplication in kernel

Make sure to read the contributing guidelines before submitting a PR

This patch gets rid of the redundant vec_perm insns
by re-ordering the matrix multiplication in kernel

Signed-off-by: Shalini Salomi Bodapati <[email protected]>
@shalinib-ibm
Copy link
Contributor Author

image

llama_file sgemm computes C = A^T * B in column major
For MMA , this equates to C^T = A * B^T in row major
Now A * B^T ( in row major) = (B * A^T)^T where B and A^T should be row major, but we have B and A^T in column major order only
So, this approach is not working

@github-actions github-actions bot added the ggml changes relating to the ggml tensor library for machine learning label May 29, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ggml changes relating to the ggml tensor library for machine learning

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant