[Getdp] Problem when compiling GetDP with sequential MUMPS

Christophe Geuzaine cgeuzaine at ulg.ac.be
Mon Sep 26 11:06:47 CEST 2011


On 21 Sep 2011, at 07:59, John Sorker wrote:

> Hi Christophe,
>  
> sorry for bothering you, but it would be very helpful to give me some
> hints regarding how you've compiled GetDP with Petsc together with sequential
> Mumps for the binaries on the GetDP webpage. Or maybe you can send me the
> modified source/configure files.
> Your help is highly appreciated!

Hi John,

I had to slightly modify petsc to do it. Here's the procedure for petsc-3.1-p8:

* compile Metis 4.0
* compile sequential version of MUMPS
* copy libmetis.a and libseq/libmpiseq.a into mumps' lib dir
* replace config/PETSc/packages/MUMPS.py with the attached MUMPS_SEQUENTIAL.py
* edit src/sys/mpiuni/mpi.c and #define MPIUNI_AVOID_MPI_NAMESPACE
* on win64 edit include/mpiuni/mpi.h and define MPIUNI_INTPTR as "long long"
* configure petsc with "--with-mumps-dir=/path/to/your/mumps"
* on win32/win64, edit PETSC_ARCH/include/petscconf.h and #undef HAVE_SLEEP, #undef HAVE_PAGESIZE

Hope this helps,

Christophe

-------------- next part --------------
A non-text attachment was scrubbed...
Name: MUMPS_SEQUENTIAL.py
Type: text/x-python-script
Size: 865 bytes
Desc: not available
URL: <http://www.geuz.org/pipermail/getdp/attachments/20110926/ba3295fb/attachment.py>
-------------- next part --------------



>  
> Regards
> John
>  
>  
> 
> ----- Weitergeleitete Message -----
> Von: John Sorker <john.sorker at yahoo.com>
> An: "getdp at geuz.org" <getdp at geuz.org>
> Gesendet: 13:49 Mittwoch, 17.August 2011 
> Betreff: Problem when compiling GetDP with sequential MUMPS
> 
>  
> Hi Christophe,
>  
> Dave told me that it was tricky to compile the Mumps sequential version and Petsc with it.
> Could you please tell me how to do it? It would be very helpful!
>  
> Best regards,
> John
>  
>  
> 
> Von: David Colignon <David.Colignon at ulg.ac.be>
> An: John Sorker <john.sorker at yahoo.com>
> Cc: "getdp at geuz.org" <getdp at geuz.org>
> Gesendet: 9:10 Mittwoch, 20.Juli 2011 
> Betreff: Re: [Getdp] Problem when compiling GetDP with sequential MUMPS
> 
> Hi Hohn,
> 
> Christophe is on holiday, but I remember he had a hard time to compile these sequential binaries. He had to first 
> compile the Mumps sequential version independently from Petsc, then hack the Petsc configure procedure and other source 
> files to be able to compile it with this sequential version of Mumps...
> 
> Regards,
> 
> Dave
> 
> --
> David Colignon, Ph.D.
> Collaborateur Logistique du F.R.S.-FNRS
> CÉCI - Consortium des Équipements de Calcul Intensif
> ACE - Applied & Computational Electromagnetics
> Institut Montefiore B28
> Université de Liège
> 4000 Liège - BELGIQUE
> Tél: +32 (0)4 366 37 32
> Fax: +32 (0)4 366 29 10
> WWW:    http://hpc.montefiore.ulg.ac.be/
> Agenda: http://www.google.com/calendar/embed?src=david.colignon%40gmail.com
> 
> On 19/07/11 22:12, John Sorker wrote:
> > Hi all!
> >
> > I'm trying to compile GetDP with PETSc and sequential MUMPS without MPI
> > but unfortunately I fail with compiling PETSc (Ubuntu 11.04 32 bit)
> >
> > I used the following configure flags:
> > ./configure --CC=gcc --CXX=g++ --FC=gfortran --download-f-blas-lapack=ifneeded --with-debugging=0 --with-clanguage=cxx
> > --with-shared=0 --with-x=0 --with-mpi=0 --download-mumps=ifneeded --with-scalar-type=real --useThreads=0
> >
> > Then I get this configure error:
> > Did not find package PARMETIS needed by MUMPS.
> >
> > As I do not want to build a MPI version PARMETIS cannot be used because it depends obviously on MPI.
> >
> >
> > I would be really happy if someone could tell me how to compile GetDP with PETSc and MUMPS
> > without MPI as it is done for the binaries.
> >
> > Best regards,
> > John
> >
> >
> >
> >
> > _______________________________________________
> > getdp mailing list
> > getdp at geuz.org
> > http://www.geuz.org/mailman/listinfo/getdp
> 
> 
> 
> 
> 
> _______________________________________________
> getdp mailing list
> getdp at geuz.org
> http://www.geuz.org/mailman/listinfo/getdp

-- 
Prof. Christophe Geuzaine
University of Liege, Electrical Engineering and Computer Science 
http://www.montefiore.ulg.ac.be/~geuzaine