[Getdp] large problems with getdp

Guillaume Demesy gdemesy at physics.utoronto.ca
Fri Dec 17 16:20:52 CET 2010


Thanks for the feedback on the compilation on Mac OS!

Have fun!

Guillaume


On 17/12/2010 10:15, Helmut Müller wrote:
> Hi all,
>
> just to keep you updated:
>
> after some hours of compiling getdp and gmesh are now full functional on a MacBook Pro. Mesh generation and calculation up to 1.6 million DOF can be done with 8GB RAM. Mesh generation for larger meshes is also possible but lasts really long.
>
> With large problems it works now quite impressive,no more memory allocation errors.
>
> The main problem for compilation on the mac was, that there are 4.0 and 4.2 apple compilers installed and also mpi compilers based on 4.2, but without fortran support.
> openmpi (installed as suggested via macport) messed it further up, as some of the openmpi compilers pointed to 4.2 apple compiler while the fortran ones pointed to the as a dependency installed 4.3.5 GNU compiler. Linking was (no surprise) impossible with this setup.
>
> With a proper setup it was possible to compile, link and run the apps. Just arpack in getdp produced an error, so I had to disable it.
>
> So, thanks again, especially to Guillaume.
>
> Just as an idea, I think on the homepage of gmsh and getdp there should be a section which describes what sizes of  problems can be handled with an optimal installation. At least I didn´t knew
> what should be solvable with my hardware, so I was unsure wether it would make sense to recompile the whole bunch of src or not.
>
> Finally thanks a lot to all the people responsible for creating and maintaining such a great piece of software!
>
> Helmut Müller
>
>
> _______________________________________________
> getdp mailing list
> getdp at geuz.org
> http://www.geuz.org/mailman/listinfo/getdp