PETSc does NOT use a 2D partition. You need to get that concept out of your head.<br><br><div class="gmail_quote">On Wed, Aug 8, 2012 at 3:09 PM, Jinquan Zhong <span dir="ltr"><<a href="mailto:jzhong@scsolutions.com" target="_blank">jzhong@scsolutions.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Satish,<br>
<br>
I looked at<br>
<a href="http://www.mcs.anl.gov/petsc/petsc-current/src/snes/examples/tutorials/ex5s.c.html" target="_blank">http://www.mcs.anl.gov/petsc/petsc-current/src/snes/examples/tutorials/ex5s.c.html</a><br>
where<br>
MatCreateAIJ(PETSC_COMM_WORLD,rend-rstart,rend-rstart,N,N,5,0,0,0,&J);<br>
<br>
was used to create J (N,N) with local J dimensioned as (rend-rstart,rend-rstart).<br>
<br>
For my application, I have LDA=LDB=3 and N=9 in the following<br>
<br>
ierr = MatCreateAIJ(PETSC_COMM_WORLD,LDA,LDB,N,N,LDA,0,0,0,&A); CHKERRQ(ierr);<br>
<br>
I got the following error message:<br>
<br>
[0]PETSC ERROR: --------------------- Error Message ------------------------------------<br>
[0]PETSC ERROR: Nonconforming object sizes!<br>
[0]PETSC ERROR: Sum of local lengths 27 does not equal global length 9, my local length 3<br>
likely a call to VecSetSizes() or MatSetSizes() is wrong.<br>
<br>
<br>
It does NOT appear PETSc had the 9x9 matrix. It tried to build 27*27 since I am using 9 procs. Is there a way to resolve this local size and global size issue?<br>
<br>
Thanks,<br>
<br>
Jinquan<br>
<div class="im HOEnZb"><br>
<br>
<br>
-----Original Message-----<br>
From: <a href="mailto:petsc-users-bounces@mcs.anl.gov">petsc-users-bounces@mcs.anl.gov</a> [mailto:<a href="mailto:petsc-users-bounces@mcs.anl.gov">petsc-users-bounces@mcs.anl.gov</a>] On Behalf Of Satish Balay<br>
Sent: Wednesday, August 08, 2012 1:45 PM<br>
To: PETSc users list<br>
Subject: Re: [petsc-users] Customizeing MatSetValuesBlocked(...)<br>
<br>
</div><div class="HOEnZb"><div class="h5">On Wed, 8 Aug 2012, Jed Brown wrote:<br>
<br>
> On Wed, Aug 8, 2012 at 2:25 PM, Jinquan Zhong <<a href="mailto:jzhong@scsolutions.com">jzhong@scsolutions.com</a>>wrote:<br>
><br>
> > If I understand you correctly, ScaLAPACK blocks don't have anything<br>
> > to do with the sparse matrix structure.****<br>
> ><br>
> > ** **<br>
> ><br>
> ><br>
> > ********************************************************************<br>
> > ********************************************************************<br>
> > **<br>
> > ****<br>
> ><br>
> > ** **<br>
> ><br>
> > You are correct. What I meant was how to define the diagonal and<br>
> > off-diagonal parts of each submatrix matrix A (LDA, LDB). For<br>
> > example, in the following matrix,****<br>
> ><br>
> > ** **<br>
> ><br>
> > *Proc0* *Proc1* *Proc2*****<br>
> ><br>
> > 1 2 0 | 0 3 0 | 0 4****<br>
> ><br>
> > *Proc0* 0 5 6 | 7 0 0 | 8 0****<br>
> ><br>
> > 9 0 10 | 11 0 0 | 12 0****<br>
> ><br>
> > -------------------------------------****<br>
> ><br>
> > 13 0 14 | 15 16 17 | 0 0****<br>
> ><br>
> > *Proc3* 0 18 0 | 19 20 21 | 0 0 <=== owned by Proc 5****<br>
> ><br>
> > 0 0 0 | 22 23 0 | 24 0****<br>
> ><br>
> > -------------------------------------****<br>
> ><br>
> > *Proc6* 25 26 27 | 0 0 28 | 29 0****<br>
> ><br>
> > 30 0 0 | 31 32 33 | 0 34****<br>
> ><br>
> > ** **<br>
> ><br>
> > I am not sure how to fill out the values for d_nz, d_nnz,o_nz, o_nnz<br>
> > peoperly for the subblock (0 0 ;0 0; 24 0) owned by Proc 5 since it<br>
> > was based on diagonal and off-diagonal parts.<br>
> ><br>
><br>
> Throw your 2D block cyclic nonsense out the window. These are sparse<br>
> matrices and that layout would be terrible. Logically permute your<br>
> matrices all you want, then define a global ordering and chunk it into<br>
> contiguous blocks of rows (no partition of columns). Work this out<br>
> with a pencil and paper. You should have a function that translates<br>
> row/column pairs from your ordering to our ordering. Now compute the<br>
> sparsity pattern in the new ordering. (Usually you can figure this out<br>
> on paper as well.) Then preallocate and call MatSetValues() with the new (row,column) locations.<br>
><br>
<br>
also check the 'example usage' at:<br>
<br>
<a href="http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/Mat/MatCreateAIJ.html" target="_blank">http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/Mat/MatCreateAIJ.html</a><br>
<br>
Satish<br>
</div></div></blockquote></div><br>