Paul, Sometime back I had reported an issue where using BGCS in parallel led to different results on CPU and GPU. Attaching that mail and other mails in that conversation. I'm not sure if it is the same issue that you've observed.
Chetan > -----Original Message----- > From: petsc-dev-bounces at mcs.anl.gov [mailto:petsc-dev-bounces at > mcs.anl.gov] On Behalf Of Paul Mullowney > Sent: Friday, March 15, 2013 3:43 PM > To: Barry Smith > Cc: For users of the development version of PETSc > Subject: Re: [petsc-dev] refactoring petsccusp.h needed > > 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 > > -------------- next part -------------- An embedded message was scrubbed... From: "Chetan Jhurani" <[email protected]> Subject: discrepancy with bcgs, mpi, cusp, and no preconditioning Date: Thu, 10 May 2012 12:45:30 -0700 Size: 6181 URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20130315/5e25279d/attachment-0003.mht> -------------- next part -------------- An embedded message was scrubbed... From: "Matthew Knepley" <[email protected]> Subject: Re: [petsc-dev] discrepancy with bcgs, mpi, cusp, and no preconditioning Date: Thu, 10 May 2012 12:49:55 -0700 Size: 6805 URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20130315/5e25279d/attachment-0004.mht> -------------- next part -------------- An embedded message was scrubbed... From: "Chetan Jhurani" <[email protected]> Subject: RE: [petsc-dev] discrepancy with bcgs, mpi, cusp, and no preconditioning Date: Thu, 10 May 2012 12:56:47 -0700 Size: 426500 URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20130315/5e25279d/attachment-0005.mht>
