[Coin-ipopt] Paradiso

[Olaf.Schenk at unibas.ch]

Paul Harper Hines <phines at cmu.edu> wrote:

> Has anyone tried to use the paradiso linear solver instead of the HSL
> routines that are used by IPOPT?  Any thoughts on how they would compare in
> terms of run-time on non-multi threaded systems?
>
>
Hi Paul,

The run time of both solvers strongly depends on the size and the structure of
the sparse systems. The C++ version of IPOPT has an interface to PARDISO which
should work in principle, but we have not tested the interface for large
problems so far. Anyway, our plan is to test the solver extensively with IPOPT
in September and we will work on modifications that might be necessary in both
packages PARDISO and IPOPT.

Regards,
Olaf

----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.