Hi -- I have a fork that caches symbolic factorization in between calls to PardisoSolver::PreProcess.
I think this relies on the fact that my model has no contacts, but basically gives me a 1.85x speedup. I don't know how general this will be for other people.
My change is here, it's AI-written, and I've read it but not reviewed it as I would review production code. (And I'm not qualified to assess the linear algebra claims).
Do you want me to actually review this and submit it as a real PR? Or not relevant right now.
Hi -- I have a fork that caches symbolic factorization in between calls to PardisoSolver::PreProcess.
I think this relies on the fact that my model has no contacts, but basically gives me a 1.85x speedup. I don't know how general this will be for other people.
My change is here, it's AI-written, and I've read it but not reviewed it as I would review production code. (And I'm not qualified to assess the linear algebra claims).
Do you want me to actually review this and submit it as a real PR? Or not relevant right now.