Instead of using solve, dot etc we may speed things up by simplifying some of the rout

They were added in <a class="issue-link js-issue-link" data-error-text="Failed to load

This has actually been partially implemented in <a class="commit-link" data-hovercard-

Create fast lapack functions about sisl HOT 4 CLOSED

zerothi commented on July 23, 2024

Create fast lapack functions

from sisl.

zerothi commented on July 23, 2024

This, sadly requires some additional work.

Currently the routines zhegst does not exist in scipy which makes generalized eigenvalue problems difficult to solve.

from sisl.

zerothi commented on July 23, 2024

They were added in scipy/scipy#7927 but will first be available in 1.1.0

from sisl.

zerothi commented on July 23, 2024

Currently, the solve method is fast.
I will add more later.

from sisl.

zerothi commented on July 23, 2024

This has actually been partially implemented in 1cdec31 by moving calls to direct LAPACK functions.

It is thus enabled using a simple routine which quickly fetches the corresponding LAPACK routine with very little overhead.

Note that typically this is only required if a method is called many times! Like for self-energies with small matrix sizes.

I'll consider this closed since the tools is there.

from sisl.