[petsc-users] Why the code needs much longer time when I use MPI

[Patrick Sanan]

Is the number of iterations to convergence also increasing with the
number of processors?

A couple of possibly relevant FAQs:
http://www.mcs.anl.gov/petsc/documentation/faq.html#differentiterations
http://www.mcs.anl.gov/petsc/documentation/faq.html#slowerparallel

On Thu, Sep 29, 2016 at 10:56 AM, Ji Zhang <gotofd at gmail.com> wrote:
> Dear all,
> I'm using KSP included in PETSc to solve some linear equations. The code
> runs well if I do not use mpirun. However, the running duration since
> increase linearly with the number of CUPs.
>
> The solve method is gmres and without precondition method. For a same case,
> it need about 4.9s with 'mpirun -n 1', which is about 7.9s if I use 'mpirun
> -n 2', and 12.5s when I use 'mpirun -n 4'. It looks like that I need double
> time when I use double CUPs. Is there any one could give me some suggestion?
> Thanks.
>
> Best,
> Regards,
> Zhang Ji, PhD student
> Beijing Computational Science Research Center
> Zhongguancun Software Park II, No. 10 Dongbeiwang West Road, Haidian
> District, Beijing 100193, China