On 21.04.2011, at 20:28, Derek Gaston wrote:
> On Apr 21, 2011, at 10:32 AM, John Peterson wrote:
>
>> So, looks like you didn't download BLAS/Lapack through PETSc, and
>> therefore it must be finding and using some other BLAS on your system,
>> possibly C BLAS since you specified --with-fc=0?
>
On Apr 21, 2011, at 12:34 PM, Kirk, Benjamin (JSC-EG311) wrote:
> FWIW, I use --with-blas-lapack-lib="-framework vecLib" when I compile PETSc
> on OSX to get their BLAS/LAPACK support. Works great.
Aha - that sounds right. Have you done any benchmarking with that? When I
tried... I couldn't s
> This is definitely the issue. OSX ships with a version of BLAS/Lapack... it's
> called Veclib and it's one of the "Frameworks" on OSX. It can be used by
> Petsc (and libMesh)... but last time I tried I seem to remember that I had to
> specify extra libraries to be linked into Petsc so that they
On Apr 21, 2011, at 10:32 AM, John Peterson wrote:
> So, looks like you didn't download BLAS/Lapack through PETSc, and
> therefore it must be finding and using some other BLAS on your system,
> possibly C BLAS since you specified --with-fc=0?
>
> Can you look at the link line being used with libm
> If you haven't recompiled Petsc since you switched compilers that could
> easily be a source of this problem. I would suggest that you make sure that
> your HPC compilers are completely removed and that all of your various search
> paths have been sanitized of those paths before reconfigurin
On Thu, Apr 21, 2011 at 10:16 AM, Anna-Lena Gerner
wrote:
>
>> It looks like you have some issues with your underlying BLAS and Lapack
>> libraries. How are you configuring Petsc and what compilers are you using?
>> Are you using the Apple supplied compilers and MPICH wrappers or did you
>> b
On Apr 21, 2011, at 10:16 AM, Anna-Lena Gerner wrote:
>
>> It looks like you have some issues with your underlying BLAS and Lapack
>> libraries. How are you configuring Petsc and what compilers are you using?
>> Are you using the Apple supplied compilers and MPICH wrappers or did you
>> bui
> It looks like you have some issues with your underlying BLAS and Lapack
> libraries. How are you configuring Petsc and what compilers are you using?
> Are you using the Apple supplied compilers and MPICH wrappers or did you
> build any of that on your own?
>
Yes, I use the Apple compilers
Looks like you are working on a Mac. As a troubleshooting step you might
consider configuring libMesh with "--disable-shared" and remaking the library.
Then you should receive a compile error when building ex16, ex17 and perhaps
even the libMesh library itself. Those errors are usually a litt
Yes, all PETSc and SLEPc tests succeed, and also all the other libMesh examples
(apart from 16, 17) work fine.
My own applications using SLEPc fail, too.
-Anna
On 21.04.2011, at 11:44, Roy Stogner wrote:
>
> On Thu, 21 Apr 2011, Anna-Lena Gerner wrote:
>
>> All of a sudden, I receive the fo
On Thu, 21 Apr 2011, Anna-Lena Gerner wrote:
> All of a sudden, I receive the following runtime-error whenever an
> application uses slepc:
>
> dyld: lazy symbol binding failed: Symbol not found: _dhseqr
Looks like a BLAS function... weird that it's being dynamically
linked. Do the PETSc and S
11 matches
Mail list logo
