Barry, Satish,<br><br>Thank you very much for these precious informations!<br><br>Nicolas<br><br><div><span class="gmail_quote">2008/3/21, Barry Smith <<a href="mailto:bsmith@mcs.anl.gov">bsmith@mcs.anl.gov</a>>:</span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<br> On Mar 21, 2008, at 8:03 AM, Nicolas Tardieu wrote:<br> <br> > Dear Barry,<br> ><br> <br>> I have a few questions about inodes :<br> > - can the user specify the inodes sizes (it seems to me this is an<br>
> automatic PETSc feature to gain performance, right?)?<br> <br> <br> No, they must be inodes, they are not arbitary<br> <br>><br> > - can the inodes have different sizes?<br> <br> <br> Yes, they are what ever they are.<br>
<br>><br> ><br> > Otherwise I plan to define a new matrix type MATSEQVBAIJ (VB for<br> > variable block) based on MATSEQAIJ with an additional attribute<br> > defining the block sizes.<br> <br> <br> This is overkill, you will end up doing a lot more work then you<br>
need to!<br> The locations for storing the diagonal blocks is already in the<br> MATSEQAIJ data structures, if you<br> write code to replace the MatRelax_Inode (that uses that same diagonal<br> space) you are done.<br> <br>
<br> Barry<br> <br><br> <br> ><br> ><br> > Thanks,<br> ><br> > Nicolas<br> ><br> > 2008/3/20, Barry Smith <<a href="mailto:bsmith@mcs.anl.gov">bsmith@mcs.anl.gov</a>>:<br> > DON"T USE GREP! Use etags<br>
> # To access the tags in EMACS, type M-x visit-tags-table and specify<br> > # the file petsc/TAGS.<br> > # 1) To move to where a PETSc function is defined, enter M-. and the<br> > # function name.<br> > # 2) To search for a string and move to the first occurrence,<br>
> # use M-x tags-search and the string.<br> > # To locate later occurrences, use M-,<br> ><br> > It may not be in petsc-2.3.3 you may need to use petsc-dev<br> > <a href="http://www-unix.mcs.anl.gov/petsc/petsc-as/developers/index.html">http://www-unix.mcs.anl.gov/petsc/petsc-as/developers/index.html</a><br>
><br> > it is in src/mat/impls/aij/seq/inode.c<br> ><br> ><br> ><br> > Barry<br> ><br> ><br> ><br> > On Mar 20, 2008, at 1:52 PM, Nicolas Tardieu wrote:<br> ><br> > > Dear Barry,<br>
> ><br> > > Thank you very much for your quick answer.<br> > > I am searching for MatRelax_Inode in PETSc src directory using grep.<br> > > But I just can't find it! Please note I am using 2.3.3p6<br>
> > Could you please tell me where this routine is?<br> > ><br> > > Nicoals<br> > ><br> > > 2008/3/20, Barry Smith <<a href="mailto:bsmith@mcs.anl.gov">bsmith@mcs.anl.gov</a>>:<br> > > Nicolas,<br>
> ><br> > > Actually we are very close to this. Check out the routine<br> > > MatRelax_Inode() which is for AIJ matrices.<br> > > For the block size it uses the inode size. If your matrix has inodes<br>
> > that match the blocks you want<br> > > then everything is there already just use the PCSOR.<br> > ><br> > > If you do not have inodes then you can mimic the<br> > MatRelax_Inode()<br> > > code to make a new<br>
> > routine MatPBRelax_SeqAIJ(). If you do implement this we'd love to<br> > add<br> > > it to PETSc.<br> > ><br> > ><br> > > Please let us know if you have additional questions.<br>
> ><br> > ><br> > > Barry<br> > ><br> > ><br> > ><br> > ><br> > > On Mar 20, 2008, at 1:07 PM, Nicolas Tardieu wrote:<br> > ><br> > > > Dear PETSc experts,<br>
> > ><br> > > > I am currently designing a multigrid PC that needs to operate on<br> > > > variable block size matrices. I know PETSc does not support this<br> > > > feature.<br> > > > So, I will do it "by hand". Neverthless I would like to be able<br>
> to<br> > > > use some smoothers like Jacobi or SOR in their block version.<br> > > ><br> > > > What is the best solution that would work both in sequential and<br> > > > parallel?<br>
> > ><br> > > > - define some MATSHELL that handle variable block size<br> > > > -> will I be able to use in some way a block Jacobi or SOR?<br> > > > -> would it be possible to use CSR inside the MATSHELL and<br>
> > > manage the variable blocks.<br> > > ><br> > > > - use a AIJ matrix and manage the blocks using some indirection<br> > > > mechanism and reprogram Jacobi or SOR<br> > > ><br>
> > ><br> > > ><br> > > > Thank you,<br> > > ><br> > > > Nicolas<br> > > ><br> > ><br> > ><br> ><br> ><br> <br> </blockquote></div><br>