[petsc-dev] Petsc Pthreads Capability

[Dave Nystrom]

Shrirang Abhyankar writes:
 > Dave, 
 >  
 > I doubt if you would see a significant performance improvement in linear
 > solves using pthreads since currently we have support for threaded seq and
 > mpi vectors only. The pthread work is an ongoing work so it might be also
 > unstable at times. I suggest experimenting with the linear solvers and
 > preconditioners available with PETSc first and then move to using pthreads
 > if you feel necessary.
 >  
 > To use Petsc with pthreads
 > 1) configure with --with-pthreadclasses=1
 > 2) pthread runtime options 
 >    A) -thread_max <max_threads>
 >    B) -use_thread_pool <main, true, chain, tree>
 >    C) -vec_type <pthread, seqpthread, mpipthread>
 >    D) -mat_type <seqaijpthread>
 >  
 > Note that the matrix class seqiajpthread only has threaded MatMult
 > operation, all others are sequential.

Hi Shri,

Thanks for your reply.  Maybe I should hold off on evaluating the petsc
pthreads work.  I'd be happy to give it a try when you think it is more
mature development wise and has the potential to offer some meaningful
performance gain.

Thanks,

Dave

 > Shri
 > 
 > On Nov 29, 2011, at 11:47 PM, Dave Nystrom <dnystrom1 at comcast.net> wrote:
 > 
 > > I have a 2d resistive mhd code which spends over 95 percent of its run time
 > > doing linear solves and I have interfaced these linear solves to petsc.  I'm
 > > interested in optimizing the single node performance of this code right now
 > > and was wondering if there could be any benefit to trying out the petsc
 > > pthreads capability that is in petsc-dev.  Is this work at a stage yet where
 > > I might be able to benefit from trying it with my code when running on a
 > > multi-core cpu?
 > > 
 > > Thanks,
 > > 
 > > Dave
 > > 
 > > -- 
 > > Dave Nystrom
 > > 
 > > phone: 505-661-9943 (home office)
 > >       505-662-6893 (home)
 > > skype: dave.nystrom76
 > > email: dnystrom1 at comcast.net
 > > smail: 219 Loma del Escolar
 > >       Los Alamos, NM 87544