[petsc-users] Matrix assembly error in PETSc

Barry Smith bsmith at mcs.anl.gov
Thu Aug 1 13:02:05 CDT 2013


   Please always send the ENTIRE error message, it makes it much easier for us to deduce what is going on. 

   Error code 63 is PETSC_ERR_ARG_OUTOFRANGE which presumably is generated in  MatSetValues_MPIAIJ() which means a row or column index is out of range. But since this is called within the MatAssemblyEnd_MPIAIJ() it should never be out of range. The most likely cause is data corruption on values passed between processes with MPI.  It is possible the error is due to bugs in the MPI implementation or due to memory corruption elsewhere. I would first recommend running the code with valgrind (and enormously powerful tool) to eliminate the chance of memory corruption http://www.mcs.anl.gov/petsc/documentation/faq.html#valgrind  

   Let us know what happens,

    Barry

MPI 2.0 vs MPI 3.0 is likely not the issue.


On Aug 1, 2013, at 10:08 AM, Wadud.Miah at awe.co.uk wrote:

> Hello,
>  
> I am running an application code which works with 4, 8 and 18 processes but crashes with 16 processes. I have used MPICH2 and MVAPICH2 (both adhered to the MPI 3.0 standard) and both cause the same problem. I get the following error message:
>  
> [12]PETSC ERROR: MatSetValues_MPIAIJ() line 564 in src/mat/impls/aij/mpi/mpiaij.c
> [12]PETSC ERROR: MatAssemblyEnd_MPIAIJ() line 680 in src/mat/impls/aij/mpi/mpiaij.c
> [12]PETSC ERROR: MatAssemblyEnd() line 4879 in src/mat/interface/matrix.c
>  
> [12]     -->  Error in "MatAssemblyEnd()".
> [12]     -->  Code: 63
>  
> However, I do not get this using the Intel MPI (which adheres to the MPI 2.0 standard) library. Any help will be greatly appreciated.
>  
> Regards,
>  
> --------------------------
> Wadud Miah
> HPC, Design and Theoretical Physics
> Direct: 0118 98 56220 
> AWE, Aldermaston, Reading, RG7 4PR
>  
>  
> ___________________________________________________ ____________________________ The information in this email and in any attachment(s) is commercial in confidence. If you are not the named addressee(s) or if you receive this email in error then any distribution, copying or use of this communication or the information in it is strictly prohibited. Please notify us immediately by email at admin.internet(at)awe.co.uk, and then delete this message from your computer. While attachments are virus checked, AWE plc does not accept any liability in respect of any virus which is not detected. AWE Plc Registered in England and Wales Registration No 02763902 AWE, Aldermaston, Reading, RG7 4PR
> 



More information about the petsc-users mailing list