[CIG-LONG] problem with gale on a cluster

js at cp.dias.ie js at cp.dias.ie
Sun Dec 16 16:33:13 PST 2007


> js at cp.dias.ie wrote:
>> Hello Walter,
>> thank you for the input file.
>>
>> Unfortunately I am unable to compile mumps with petsc-2.3.2.
>>
>> Currently I have mpich2-1.0.4p1 installed which was
>> compiled with gcc/g77/gfortran for mpicc/mpif77/mpif90
>> respectively.
>>
>> the problem I encounter is as follows,
>>
>> compiling petsc with blacs requires mpif77. By default
>> petsc chooses mpif90 and this will not compile blacs.
>
> That is odd.  What is the error?


I think it is related to this,
http://www.beowulf.org/archive/2001-March/002549.html



I will recompile everything with intel and see how it goes.

Thanks again.

john.

>
>> This I can change with a flag, however then mumps will
>> not compile as it wants mpif90.
>>
>> I can compile mumps independently with mpif90 and build
>> petsc with mpif77, but then I get complaints about unknown
>> functions beginning with _gfortran_ and build fails.
>>
>> I was considering using the intel compiler, but I am unsure
>> if that will have any unwelcome side-effects.
>
> I have built it with the intel compiler, so it should work.  You may
> have to recompile mpich with it, but otherwise everything should work.
>
> Cheers,
> Walter Landry
> walter at geodynamics.org
>




More information about the CIG-LONG mailing list