Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Implement "sandwich product" matrix operations #798

Draft
wants to merge 1 commit into
base: master
Choose a base branch
from
Draft

Conversation

devinamatthews
Copy link
Member

This PR adds new BLAS-like level 3 operations which include a diagonal matrix "sandwiched" between two other matrices:

  • bli_gemdm: C = alpha*A*D*B + beta*C
  • bli_gemdmt: C = alpha*tri[ul](A*D*B) + beta*C
  • bli_syrkd: C = alpha*A*D*A^T + beta*C
  • bli_herkd: C = alpha*A*D*A^H + beta*C
  • bli_syr2kd: C = alpha*(A*D*B^T + B*D*A^T) + beta*C
  • bli_her2kd: C = alpha*A*D*B^H + conj(alpha)*B*conj(D)*A^H + beta*C

Currently, there are the following things to do:

  • Implement remaining kernels needed for mixed-precision/mixed-domain and 1m.
  • Integrate into the BLIS testsuite.
  • herkd: D has to be real (not checked or implemented in packing)
  • her2kd: need to use conj(D) in second product

@fgvanzee fgvanzee changed the title Implement "sandich product" matrix operations Implement "sandwich product" matrix operations Feb 28, 2024
…*D*A^{T,H} (syrk2d/her2kd) operations (and gemdmt for good measure). Some rough edges still:

- Complex herk2 and her2kd will not work.
- No mixed-type/mixed-domain or 1m.
- Not integrated into testsuite yet.
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