[petsc-users] error while allocating dense distributed matrix

[Michael Povolotskyi]

Thank you, Barry.
What I did I changed the call to

ierr = MatCreateMPIDense(comm, m, num_cols ,PETSC_DECIDE,PETSC_DECIDE,data,petsc_matrix)

and it works. Do you approve this change?
Michael.



On 06/06/2013 07:30 PM, Barry Smith wrote:
>    The first "num_cols" argument refers to the number of "local" columns on that process the second one the total number of columns. With size > 1 they cannot be equal. The sum of all the "local" columns across all processes must equal the total number of columns.
>
> This is not explained well in the MatCreateDense manual page, sorry. Look at the manual page http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/Mat/MatCreateAIJ.html and see the definition of m and n which explains it much better. The "local" row and column sizes are define by the "left" and "right" parallel vector layouts for vectors compatible with the matrix.
>
>     Barry
>
> On Jun 6, 2013, at 4:55 PM, Michael Povolotskyi <mpovolot at purdue.edu> wrote:
>
>> Dear Petsc developers,
>> petsc gives error with the error code equal to 60 when I'm doing the following:
>>
>> ierr = MatCreateMPIDense(comm, m, num_cols ,num_rows,num_cols,data,petsc_matrix)
>>
>> on rank = 0 the gebugger gives:
>> (gdb) print m
>> $1 = 240
>> (gdb) print num_cols
>> $2 = 18
>> (gdb) print num_rows
>> $3 = 360
>>
>> on rank = 1 the gebugger gives:
>> ierr = MatCreateMPIDense(comm, m, num_cols ,num_rows,num_cols,data,petsc_matrix);
>> (gdb) print m
>> $1 = 120
>> (gdb) print num_cols
>> $2 = 18
>> (gdb) print num_rows
>> $3 = 360
>> (gdb)
>>
>> Could you, please, tell me what is wrong here?
>> Thank you,
>> Michael.