Mikkel Paltorp

Hi again, Thanks for the quick answers! @askhamwhat Ahh perfect, I tried looking into the code but I did not realize that calling `hfmm3d_` also included the vectorized version (the...

Okay, I can see that the price to pay for using the interface is that the data is not as closely available. I guess you could technically pass user defined...

I guess technically you can do (1), but it would just be wrong / have some undefined behaviour. Good to know. Related to (2), then I guess the same could...

Yes, its only C/ C++ where this is giving some issues. In Rust and Julia you can at least get around the issue fairly easy as you have access to...

If it is an issue for you with Pardiso I can also try to make a PR with a fix.

> Part of the problem is that Pardiso wants to perform iterative refinement as part of its solution procedure, and this clashes somewhat with the iterative refinement we already do...

> as you suggest and the matrix would be shared between the LDL solver object and the one that contains it. In that case there would be no duplication of...

Hmm, would this be fixed by providing Pardiso with someting like `KKT - shift*S` (in order to just retrieve $K$) or is it a little more complicated than that to...

Perfect. I will run with `static_regularization_enable = false` for now then. Thanks a lot for the many insights. They have definitely saved me from many hours of debugging!

An update: I figured out that in it is possible to completely disable the internal iterative refinement in Panua-Pardiso by setting iparm(8)=-99. Doing so and having Clarabel take care of...