[petsc-users] MatSetNearNullSpace() or PCSetCoordinate for GAMG ?

[Mark F. Adams]

On Nov 8, 2012, at 3:54 PM, A.L. Siahaan <als88 at cam.ac.uk> wrote:

> Referring to the following thread about setting nullspace and coordinate for GAMG : http://lists.mcs.anl.gov/pipermail/petsc-users/2012-April/012790.html
> 
> Is there any difference of memory requirement (to set nullspace) between MatSetNearNullSpace() and PCSetCoordinate ?

No,  you can delete the data that your provide and the semantics of GAMG are exactly the same.

> 
> For PCSetCoordinate, is the order of x-y-z coordinate in thevector [0.x, 0.y, 0.z, 1.x, ..., N-1.z] with respect to the ordering of solution vector [0.u_x, 0.u_y, 0.u_z, 1.u_z, ..., N-1.u_z] ?

This seems obvious, maybe I'm missing something.

> 
> I tried to use PCSetCoordinate for GAMG, but then I get an error :
> ============================================================
> [0]PETSC ERROR: Floating point exception!
> [0]PETSC ERROR: Infinite or not-a-number generated in mdot, entry 0!
> ========================================================
> 
> I am wondering if it is due to the ordering of coordinate.
> 

I doubt it.  Try commenting out your setCoordinates line and see if that helps.

> Regards,
> Antony
> 
>