Open bjodah opened 9 years ago
Hi Björn,
yes, unfortunately we have currently these types hard-coded into the implicit steppers. This is still very bad and indeed bothers us as well since quite some time now. We have made plans to attack this problem and scheduled some code sprint in December to finally start improving this situation. For now I can only ask you to be patient, but we might come back to you if we have something ready so you can try out some early versions if you are interested.
Perfectly understandable. I can definitely provide some feedback on early drafts if wanted. Thanks!
I was trying to figure out if it is possible to use
ublas::banded_matrix
with odeint when defining the jacobian. But it seems that the use ofublas::matrix
is hard-coded at the moment?I have a system where the jacobian is banded (large matrix with narrow band) and the computation of a full dense LU decomposition is prohibitively expensive.
I could of course use another ODE solver but I want a reference solution calculated with multiprecision values which is real killer feature of odeint in my oppinion.