about gmshfem
Dear @ARoyer and @geuzaine and all other contributors,
Congratulations for this project, it looks, to say the least, truly amazing. I have just discovered it and literally spent the whole teleworking day compiling/exploring the demos.
Here is a little feedback about my wanderings:
- I tried to compile gmshfem against the gmsh sdk (ubuntu) and couldn't figure out the linking conflict due to the old (10.2) petsc shipped with the gmsh sdk and my own petsc (this was solved using my own gmsh).
- I add to disable slepc for now, another linking issue.
- The demos are running smoothly - the openmp parallel assembly does scales nicely! Two minor issues:
- in
utils/basisFunction
:LagrangeDual
does not seem to exist... - in the demos, command line arguments are seemingly passed correctly but runs end with a misleading warning claiming the opposite, e.g. in
demos/maxwell3d
running./demo -order 1
triggers first order but ends withWARNING! There are options you set that were not used!
etc...
- in
Finally, from a more prospective point of view and pure interest, two features could lead us to gradually migrate from getdp to gmshfem:
- Bloch boundary conditions (getdp's
LinkCplx
) - an eigensolver (e.g. slepc ;-) )
Actually most of non-periodic 'direct' Maxwell 2D/3D problems could be adapted right away! If I can be of any help with these last points, I would be glad to spend time on it.
Salutations, Guillaume
Edited by Guillaume Demesy