[petsc-users] Tuning the parallel performance of a 3D FEM CFD code

[Barry Smith]

On May 15, 2011, at 11:47 AM, Jed Brown wrote:

> On Sun, May 15, 2011 at 18:41, Barry Smith <bsmith at mcs.anl.gov> wrote:
> This may help a bit but I'm not sure how dramatically. What we could really use is a point block ILU for AIJ matrices with inodes. That is each inode set of rows is eliminated simultaneously (by factoring the little diagonal block associated with that inode).
> 
> But, pivoting aside, what is the difference between point and point-block ILU when there is no discarded fill within the block?

  You are right, since the little diagonal blocks are dense by definition of inodes there is no difference. 

   So you could be right, using an interlaced ordering (if it is not currently being used) could help the convergence, possibly a good deal. Worth trying.

    Barry