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

heev: document and check for square MPI grid and lower triangular matrix #204

Merged
merged 3 commits into from
Dec 28, 2024

Conversation

mgates3
Copy link
Collaborator

@mgates3 mgates3 commented Dec 28, 2024

  • heev function now asserts that the MPI process grid is square, and that the matrix is lower triangular. Note that slate_assert throws an exception and is not disabled by -DNDEBUG, whereas C++ assert exits and is disabled by -DNDEBUG.
  • Documentation is updated with comments about both conditions.
  • test_heev checks that heev throws the appropriate exceptions, rather than just skipping the tests.
  • Also updated doxygen layout to the latest, since doxygen renamed "Modules" to "Topics", which we rename to "Routines".
  • Also fixed a minor tester bug that caused a compiler warning (using pointer after deleting it).
  • Use "2-stage", meaning 2-stage reduction to bi- or tri-diagonal as the more common terminology (cf. ELPA 2), instead of "3-stage", the 3rd stage being reduction to diagonal.

Copy link
Collaborator Author

@mgates3 mgates3 left a comment

Choose a reason for hiding this comment

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

Reviewed.

@mgates3 mgates3 merged commit 95a481c into icl-utk-edu:master Dec 28, 2024
8 checks passed
@hfp
Copy link

hfp commented Jan 8, 2025

The restrictions for (P)HEEV(D) are:

  1. Matrix extents need to be an exact multiple of the blocksize.
  2. Lower triangular storage Hermitian matrix.
  3. Square MPI process grid ($p \times p$).

What are the chances to relax the limitations?

I am currently hitting 1) with my application (via ScaLAPACK interface), but likely all limitations making Slate not applicable.

@mgates3
Copy link
Collaborator Author

mgates3 commented Jan 8, 2025

I don't think (0.) is true. A SLATE matrix can be any size, unrelated to the block size.

Due to limited funding, the other limitations are unlikely to be addressed soon.

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