Åsmund Ødegård wrote: > > > On Dec 1, 2007 9:41 PM, Matthew Knepley <[EMAIL PROTECTED] > <mailto:[EMAIL PROTECTED]>> wrote: > > On Dec 1, 2007 2:34 PM, Åsmund Ødegård <[EMAIL PROTECTED] > <mailto:[EMAIL PROTECTED]>> wrote: > > > > > > On Nov 30, 2007 6:08 PM, Garth N. Wells < [EMAIL PROTECTED] > <mailto:[EMAIL PROTECTED]>> wrote: > > > To get the flags right for MPI, the test in petsc.py should use > mpicc > > > (or mpicxx) if PETSc has been compiled using MPI, instead of g++. > > > > > > > You're right - I haven't tought about that. Unfortunately, > pkg-config which > > we use quite heavily does not really supports the concept of > telling about > > compilers, so I have to find some clever way to deal with this > issue :-) > > You can read this info out of PETSc: > > 1) make getincludedirs, make getlinklibs > > or > > 2) Loading up the pickled Python dict and reading the flags out of > the MPI in particular > > > Sure - getting the stuff out of petsc is easy; I can just read, say > ${PCC} and ${PCC-LINKER} with make. The hard part is to persist this in > the pkg-config file I create for petsc. As Garth mention, there is > always the option of using --variable=compiler. But I have to consider > this "a hack" and really want something cleaner if possible... > >
We used --variable=compiler originally to deal with users using different compilers, e.g. gcc or Intel, which require different compiler flags. We wanted DOLFIN executables to be compiled with the same compiler as the DOLFIN library and couldn't come up with any other solution when using pkg-config. It seems to me a real weakness that pkg-config doesn't provide simple support for different compilers. Garth > > Matt > > > Thanks for testing! > > > > > > > > Garth > > > > > > > > > > > > > > > Garth N. Wells wrote: > > > > > > > > Åsmund Ødegård wrote: > > > >> > > > >> On Nov 30, 2007 5:32 PM, Garth N. Wells < [EMAIL PROTECTED] > <mailto:[EMAIL PROTECTED]> > > > >> <mailto: [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>>> wrote: > > > >> > > > >> [snip] > > > >> > > > > >> > Usually, you should set what you used as prefix when > installing > > > >> umfpack. > > > >> > > > > >> > > > >> I'm using Ubuntu 7.10 with the libsuitespare-dev > package. The > > header > > > >> files are located under > > > >> > > > >> /usr/include/suitesparse > > > >> > > > >> "suitesparse" needs to be added to > > > >> > > > >> ./simula-scons/scons/pkgconfiggenerator/umfpack.py > > > >> > > > >> > > > >> Thanks Garth. I thought that was already in there, but > obviously it > > > >> wasn't. Added upstream now. > > > >> > > > > > > > > I'm also having a problem with petsc.py. I'm still trying to > figure it > > > > out, but it appears to be related to the location of the MPI > header > > files. > > > > > > > > Garth > > > > > > > > > > > >> -- > > > >> [simula.research laboratory] > > > >> Åsmund Ødegård > > > >> IT-Manager > > > >> phone: +4790069915 > > > >> http://www.simula.no/ < http://www.simula.no/> > > > > > > > > > > > > > > > _______________________________________________ > > > > DOLFIN-dev mailing list > > > > [email protected] <mailto:[email protected]> > > > > > > > http://www.fenics.org/mailman/listinfo/dolfin-dev > <http://www.fenics.org/mailman/listinfo/dolfin-dev> > > > > > > > > > > -- > > > > > > > > > > > -- > > [simula.research laboratory] > > Åsmund Ødegård > > IT-Manager > > phone: +4790069915 > > http://www.simula.no/ > > _______________________________________________ > > DOLFIN-dev mailing list > > [email protected] <mailto:[email protected]> > > http://www.fenics.org/mailman/listinfo/dolfin-dev > > > > > > > > -- > What most experimenters take for granted before they begin their > experiments is infinitely more interesting than any results to which > their experiments lead. > -- Norbert Wiener > > > > > -- > [simula.research laboratory] > Åsmund Ødegård > IT-Manager > phone: +4790069915 > http://www.simula.no/ _______________________________________________ DOLFIN-dev mailing list [email protected] http://www.fenics.org/mailman/listinfo/dolfin-dev
