Yes. I tested it, it does not work. Like PETSC, they have --with-batch so
that we can run the test manually by typing : mpirun -np 1 ./conftest
Is there a way to let libmesh do this?
Thank you .
On 5/25/07, John Peterson <peterson@...> wrote:
> LibMesh has not worked with PGI compilers in the past. I'm unaware if
> it now does.
> You can set various configure options on the command line, e.g.
> CXX=mpiCC CC=mpicc ./configure
> but I don't know of a way to have configure tests run using mpirun.
> Bin writes:
> > Hello,
> > I am now trying to install libmesh on a batch system with mpi compiler
> > compilers). But I got configuration errors saying that:
> > checking whether the C++ compiler works... configure: error: cannot
> > C++ compiled programs
> > Does libmesh have the configuration options which allow me to run the
> > code by mpirun? Like PETSC, they have --with-batch.
> > Thank you
> > --
> > Best regards,
> > Bin
From: Roy Stogner <roystgnr@ic...> - 2007-05-25 21:26:01
On Fri, 25 May 2007, Bin wrote:
> Yes. I tested it, it does not work. Like PETSC, they have --with-batch so
> that we can run the test manually by typing : mpirun -np 1 ./conftest
> Is there a way to let libmesh do this?
I don't suppose there are any experts on autoconf reading this mailing
PETSc uses their own configure.py script; we're not going to rewrite
our build system, so libMesh is pretty much limited to whatever
facilities autoconf gives us, and I don't know how to tell autoconf to
prepend a program name like mpirun to the test executables it runs.
Autoconf should have a way to handle systems which can't run their own
newly created executables (after all, lots of GNU developers use cross
compilers), but I don't know what that way is.
As a workaround, you may be able to disable some of those tests (e.g.
by commenting out AC_LANG_CPLUSPLUS in configure.in and regenerating
configure?), but I don't know how far doing that will get you.