[petsc-dev] refactoring petsccusp.h needed

Paul Mullowney paulm at txcorp.com
Fri Mar 15 17:43:25 CDT 2013


Barry,

Is March 21 still the target release date? I will look into this asap if 
that's the case but I've been trying to fix some other PETSc GPU 
problems in GMRES and BCGS algorithms.

For GMRES, the current performance of VecMDot_SeqCUSP sucks. I have an 
solution, but I haven't tested all cases yet.
For BCGS, some part of the algorithm is broken but I don't know what it 
is. By broken, I mean that CPU and GPU residuals diverge fairly quickly.

-Paul
>    Paul,
>
>      We are thinking of a PETSc release around March 21. Before that time petsccusp.h
>
> #if !defined(__PETSCCUSP_H)
> #define __PETSCCUSP_H
> /*
>      This should only be included in user code that uses CUSP directly and hence the file name ends with .cu
> */
> #include<../src/vec/vec/impls/dvecimpl.h>
> #include<../src/vec/vec/impls/seq/seqcusp/cuspvecimpl.h>
> #endif
>
> so that ONLY the public interface of the stuff (what needs to be known to user code) is available and the rest is kept in the cuspvecimpl.h file. Maybe pets ccusp.h could be empty, I don't know).
>
>     Do you think you can make that change and anything else that needs doing before the release?
>
>     Thanks
>      Barry
>




More information about the petsc-dev mailing list