Copying to libmesh-users; this is getting into questions of general
interest that I'd like a search engine to see.

On Thu, 19 Nov 2015, Brian Cohen wrote:

> ./configure --disable-strict-lgpl --enable-everything 
> --with-nlopt-lib=/usr/local/lib/
>
> --enable-parmesh

Be warned - this makes the default Mesh into a ParallelMesh, with
elements distributed between processors, so no one MPI rank can even
see the entire mesh.  That's the most scalable way to do things, and
it's what I use, but if you have to do anything sophisticated then
writing the code for it becomes harder.  You can still construct a
SerialMesh manually if you find the need though.

> --enable-complex

You don't want this on unless you're solving complex-valued problems.
For various reasons (primarily the fact that we can only link to one
PETSc at a time) complex-valued Number support mostly *replaces*, not
just augments, real-valued Number support, and solving real-valued
PDEs is more efficient with real-valued matrices.

> --enable-tecplot --enable-curl --enable-blocked-storage --enable-gzstreams
> 
> Still don't have gzstream

Ah!  It looks like we have gzstream in contrib, but it depends on
finding libz libs and headers installed on your system.  I've never
used a system without libz-dev installed so I've never noticed.  :-)

> or hdf5 (not sure if that is needed).

It's not, unless you want to use HDF5-based files.

> Also I installed cppunit but not sure how to point to it.

--with-cppunit-prefix=/where/ever

Assuming that /where/ever/bin/cppunit-config exists.

> Also do would it be advisable to get tbb and vtk??

VTK only if you need to output those files.  TBB no; I believe if you
don't have it then we fall back on openmp which may work better, if
you do want to use threads or hybrid parallelism rather than just
pure MPI.
---
Roy

------------------------------------------------------------------------------
_______________________________________________
Libmesh-users mailing list
Libmesh-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libmesh-users

Reply via email to