Re: Working on bvp4c

[Bill Greene]

I have been using Sundials and (indirectly) idaDlsBandDQJac for the past several months and it seems to be

working fine. If a band-matrix representation is acceptable in your bvp4c implementation, leveraging Sundials

might be a good idea.

For the next phase of my work, I need a FD jacobian routine for general sparse matrices and,

unfortunately, Sundials doesn't have this. So I've been looking further at the sparse jacobian

paper and implementation by Coleman, et. al. Since my last post, I discovered that this approach

and software are being used in PETSc, http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/MatFD/index.html.

If I find out more, I'll pass it along.

Bill

On Sun, Aug 7, 2016 at 2:28 PM, lakerluke <address@hidden> wrote:

Thank you for the summary and information Bill. I have been looking into
Sundials via the link you mentioned. I couldn't actually find a reference to
idaDlsBandDQJac (presumably either in IDA or IDAS) in the following user
guides linked to from here:

http://computation.llnl.gov/projects/sundials-suite-nonlinear-differential-algebraic-equation-solvers/sundials-software

However I did find the code here:

http://sundials.sourcearchive.com/documentation/2.4.0/idas__direct_8c-source.html

Given your background research do you think the best way to write numjac is
via the Sundials approach? Did you get anywhere with this?

Luke



--
View this message in context: http://octave.1599824.n4.nabble.com/Working-on-bvp4c-tp4677540p4679058.html

Sent from the Octave - Maintainers mailing list archive at Nabble.com.