<div dir="ltr"><div><div><br>Thanks a lot for your answer!<br></div>Regards, <br></div>Natacha <br><div><div><br></div></div></div><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Oct 9, 2014 at 10:19 AM, Barry Smith <span dir="ltr"><<a href="mailto:bsmith@mcs.anl.gov" target="_blank">bsmith@mcs.anl.gov</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class=""><br>
On Oct 9, 2014, at 3:12 AM, Natacha BEREUX <<a href="mailto:natacha.bereux@gmail.com">natacha.bereux@gmail.com</a>> wrote:<br>
<br>
> Dear PETSc users,<br>
><br>
> I am a bit confused about blocksizes in MPI matrices.<br>
><br>
> I define a MPI matrix A with a certain blocksize, let say bs = 3.<br>
> As far as I understand, the sparse pattern of the matrix A is made of square blocks of size 3x3, regardless of the actual values of the terms of A.<br>
> Am I right ?<br>
<br>
</span> Only if you use the BAIJ or SBAIJ matrix format. If you use AIJ then it still stores the values without regard to the block size but it carries knowledge of the block size around and that gets used in some places such as the GAMG solver.<br>
<span class="">><br>
> The matrix A is distributed among several processors: does the local number of rows of A on each processor have to be a multiple of bs ?<br>
<br>
</span> Correct<br>
<span class="">><br>
> I want to use a multigrid preconditionner to solve the linear system of matrix A : A x = b<br>
> Is it mandatory to define the right hand side b as a vector with the same blocksize bs=3 ?<br>
<br>
</span> It probably doesn’t matter, but if you know something has an associated block size it is generally best to capture it as soon as you can in the problem.<br>
<span class="HOEnZb"><font color="#888888"><br>
Barry<br>
</font></span><div class="HOEnZb"><div class="h5"><br>
><br>
> Thank you very much of your help,<br>
> Best regards,<br>
> Natacha<br>
<br>
</div></div></blockquote></div><br></div>