[Getdp] getdp Digest, Vol 91, Issue 4

Ruth V. Sabariego r.sabariego at ulg.ac.be
Tue Oct 12 08:56:31 CEST 2010


  Weird...
Mumps is min 10 times faster in my computer (even running serial).

The main difference between my configuration and yours is that I am 
using openmpi for compiling petsc. These are my options (on MacOSX):

export PETSC_DIR=$PWD
export PETSC_ARCH=darwin-complex

./configure \
--CC=/opt/local/bin/openmpicc \
--CXX=/opt/local/bin/openmpicxx \
--FC=/opt/local/bin/openmpif90 \
--with-debugging=0 \
--with-clanguage=cxx \
--with-shared=0 \
--with-x=0 \
--download-mumps=ifneeded \
--download-parmetis=ifneeded \
--download-scalapack=ifneeded \
--download-blacs=ifneeded \
--download-umfpack=ifneeded \
--with-scalar-type=complex

Nothing particular with regard to GetDP. What version are you using? 
Lately there have been some improvements concerning the solver.

Best regards,
Ruth

On 11/10/10 10:08, Zacharie De Grève wrote:
> Hello Ruth,
>
> It works now, with the two solutions you proposed (umfpack and mumps). Thanks
> for your help.
>
> I have tried it on a much bigger mesh (more or less 40000 dofs), with the
> command 'getdp<pro-file>  -msh<msh-file>' etc. But umfpack still performs
> better than mumps (approximately 20% faster).
>
> Maybe it is due to a wrong step/missing option during the compilations of
> Petsc and GetDP ? I employed the following commands on a Kubuntu 10.04
> distribution:
>
> * Petsc :
> 1. ./config/configure.py --with-debugging=0 --with-clanguage=cxx --with-shared=0
> --with-x=0 --download-mumps=ifneeded --download-parmetis=ifneeded --download-
> scalapack=ifneeded --download-blacs=ifneeded --with-scalar-type=complex --
> download-openmpi=ifneeded
> 2. make PETSC_DIR=..._ARCH=linux-gnu-cxx-opt all test
>
> * GetDP
> ./configure --with-petsc-prefix=... --with-petsc-arch=...
> make
>
> Thank you again.
>
> Zacharie
>
>
>
>
>
> Le vendredi 8 octobre 2010 12:00:01, getdp-request at ace20.montefiore.ulg.ac.be a
> écrit :
>> Send getdp mailing list submissions to
>> 	getdp at geuz.org
>>
>> To subscribe or unsubscribe via the World Wide Web, visit
>> 	http://www.geuz.org/mailman/listinfo/getdp
>> or, via email, send a message with subject or body 'help' to
>> 	getdp-request at geuz.org
>>
>> You can reach the person managing the list at
>> 	getdp-owner at geuz.org
>>
>> When replying, please edit your Subject line so it is more specific
>> than "Re: Contents of getdp digest..."
>>
>>
>> Today's Topics:
>>
>>     1. Re: "Zero Pivot in LU factorization" error after upgrading
>>        (Ruth V. Sabariego)
>>
>>
>> ----------------------------------------------------------------------
>>
>> Message: 1
>> Date: Fri, 08 Oct 2010 09:16:38 +0200
>> From: "Ruth V. Sabariego"<r.sabariego at ulg.ac.be>
>> Subject: Re: [Getdp] "Zero Pivot in LU factorization" error after
>> 	upgrading
>> To: getdp at geuz.org
>> Message-ID:<4CAEC556.3010404 at ulg.ac.be>
>> Content-Type: text/plain; charset="us-ascii"
>>
>> An HTML attachment was scrubbed...
>> URL:
>> <http://www.geuz.org/pipermail/getdp/attachments/20101008/f24b0aad/attachm
>> ent.html>
>>
>> ------------------------------
>>
>> _______________________________________________
>> getdp mailing list
>> getdp at geuz.org
>> http://www.geuz.org/mailman/listinfo/getdp
>>
>>
>> End of getdp Digest, Vol 91, Issue 4
>> ************************************

-- 
Dr. Ir. Ruth V. Sabariego
University of Liege, Dept. of Electrical Engineering&  Computer Science,
Applied&  Computational Electromagnetics (ACE),
phone: +32-4-3663737 - fax: +32-4-3662910 - http://ace.montefiore.ulg.ac.be/