[petsc-dev] GMRES with one processor crashes

[Alexander Grayver]

Ok, I finally figured out reason. When I configure with option:

--with-blas-lapack-dir=/opt/intel/Compiler/11.1/072/mkl/lib/em64t

PETSc doesn't find blas/lapack libraries there because actual names are 
libmkl_blas95_lp64.a/libmkl_lapack95_lp64.a and (just my suggestion) 
either automatically substitutes libraries from somewhere else or uses 
just stubs without actual subroutines (as Matt correctly suggested).

Since 1 year I use petsc I've never come acrross this problem because I 
always built with blacs/scalapack and used more than one process in 
which case routines from blacs/scalapack libraries were used.

-- 
Regards,
Alexander

On 13.01.2012 11:41, Alexander Grayver wrote:
> I tested various combinations, for the error.
>
> It appears only in this combination:
>
> * debugging configuration
> * MKL blas/lapack (but not blacs and scalapack!)
> * development branch
>
> The release configuration, reference blas/lapack installation or 
> petsc-3.2-p5 do not have the problem.
>
> Regards,
> Alexander
>
> On 11.01.2012 19:55, Matthew Knepley wrote:
>> On Wed, Jan 11, 2012 at 11:10 AM, Alexander Grayver 
>> <agrayver at gfz-potsdam.de <mailto:agrayver at gfz-potsdam.de>> wrote:
>>
>>     The error still happens if I configure and make with the
>>     following line:
>>
>>     ./configure --with-petsc-arch=openmpi-intel-complex-debug-c
>>     --with-mpi-dir=/opt/mpi/intel/openmpi-1.4.2
>>     --with-scalar-type=complex
>>     --with-blas-lapack-dir=/opt/intel/Compiler/11.1/072/mkl/lib/em64t
>>     --with-precision=double
>>
>>
>> This sounds like a problem with the MKL install. Are the complex 
>> routines there and not just stubs? Is that
>> the correct set of libraries for this machine?
>>
>>    Matt
>>
>>     And disappears with the following one:
>>
>>     ./configure --with-petsc-arch=openmpi-intel-complex-debug-c
>>     --with-mpi-dir=/opt/mpi/intel/openmpi-1.4.2
>>     --with-scalar-type=complex --download-blas-lapack
>>     --with-precision=double
>>
>>     What could that mean?
>>
>>     Regards,
>>     Alexander
>>
>>     On 10.01.2012 20:29, Matthew Knepley wrote:
>>>     On Tue, Jan 10, 2012 at 4:27 AM, Alexander Grayver
>>>     <agrayver at gfz-potsdam.de <mailto:agrayver at gfz-potsdam.de>> wrote:
>>>
>>>         Can I do something about that I should I wait for patch?
>>>
>>>
>>>     I was not being direct enough. I want you to repeatedly
>>>     reconfigure and run until you find the
>>>     offending package. We still cannot get this to fail.
>>>
>>>        Matt
>>>
>>
>>
>>
>> -- 
>> 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
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20120113/c07a439d/attachment.html>