[petsc-dev] -pc_asm_blocks

[Jed Brown]

On Mon, 18 Jan 2010 07:58:46 -0600, Matthew Knepley <knepley at gmail.com> wrote:
> Okay, I just tried it and I understand the problem: This never really
> worked. RASM was not doing the proper thing.

So what happens in release-3.0.0?  Are you saying that what it was doing
for multiple subdomains per process was some other Schwarz variant, but
not RASM?  That might explain this:

  $ mpirun -n 7 ./ex2 -pc_type asm -pc_asm_blocks 7
  Norm of error 0.000458307 iterations 7
  $ mpirun -n 1 ./ex2 -pc_type asm -pc_asm_blocks 7
  Norm of error 0.000290453 iterations 11

Is there an argument against (optionally) solving the subdomains
multiplicatively within each process?

> Now it does, but we need to correct the code to produce is_local
> automatically with this argument.

Great.

Jed