[Ipopt] Metis

[Jonathan Hogg]

Unless they've gone back to compatibility with the old version the API 
has changed, and the best way forwards is probably to write some form of 
wrapper that takes calls from the old version and translates to the new 
version.

I think the only routine used by IPOPT (or rather the linear solvers 
involved) is Metis_nodeND which as an extra argument and a changed 
options structure.

Our advice to users of the HSL codes (incl MA57) is just to grab the 
v4.0 tarball while we try and figure out the nicest way to call v5 which 
is fairly C-centric (eg requires passing of NULL pointers for missing 
arguments, not so easy from Fortran without requiring a f2003 compliant 
compiler).

The old version is available from the following webpage:
http://glaros.dtc.umn.edu/gkhome/fsroot/sw/metis/OLD

Jonathan.

On 05/08/11 05:05, Giacomo Nannicini wrote:
> Hi all,
> it seems to me that the code in ThirdParty/Metis for getting Metis
> does not work anymore: version 4.0 is no longer available from what I
> gather, after the release of 5.0. This is a very recent change (new
> version was released today, I believe).
>
> Not sure if it's the Ipopt people that should have a look at
> ThirdParty software, but Metis is a dependency of Ipopt in any case.
>
> You may want to have a look. The modifications to do seem pretty
> straightforward, only a couple of directory names have changed (most
> notably, /Lib is now /libmetis).
>
> Giacomo
>
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/ipopt

-- 
Scanned by iCritical.