[petsc-users] command line option for superlu_dist

[Hong Zhang]

I got the matrix and start testing it:

- sequential mumps fails during numerical factorization
[0]PETSC ERROR: [0] MatFactorNumeric_MUMPS line 314 src/mat/impls/aij/ 
mpi/mumps/mumps.c
[0]PETSC ERROR: [0] MatLUFactorNumeric line 2378 src/mat/interface/ 
matrix.c

- sequential superlu seems hang ...

Something is wrong with your matrix. I'll further check it and let you  
know
what I get.

How did you get mumps work? Sequential or parallel (num of processors?)?

Hong

On Jan 5, 2010, at 1:59 PM, Barry Smith wrote:

>
>   Ok, we got the file. For "security reasons" ls is disabled in that  
> directory.
>
>   Barry
>
> On Jan 5, 2010, at 1:54 PM, hxie at umn.edu wrote:
>
>>
>> I use ftp to upload the file matrix.gz to incoming. But I cannot  
>> see it using 'ls'.
>>
>> Hui
>>
>> -------------
>> Likely SuperLU_Dist is failing on the provided matrix. Could you   
>> use -ksp_view_binary on the run and copy the resulting file   
>> binaryoutput to the anonymous ftp site ftp.mcs.anl.gov in the   
>> directory incoming and tell us when the matrix is there.  Sparse   
>> parallel factorizations are a bit fragile and can fail.
>>
>>
>>   Barry
>