[petsc-users] Using PETSc for performance evaluation, tuning on multicore machines

[Matthew Knepley]

On Wed, May 23, 2018 at 6:10 AM, Najeeb Ahmad <nahmad16 at ku.edu.tr> wrote:

> Hi All,
>
> I am a PhD student in parallel computing working on unstructured meshes
> optimizations on multicore architectures. For this purpose, I want to use a
> simple block Jacobi iterative solver using unstructured meshes to see how I
> can improve performance of the solver on a multicore machine (KNL for
> example). I wanted to know if PETSc can be good choice for this purpose?
> Will it allow me to experiment with data layout, compiler optimizations
> (e.g. AVX512) etc.?
>

It is intended to be. For example, there are matrix classes specialized for
MKL (
http://www.mcs.anl.gov/petsc/petsc-master/docs/manualpages/Mat/MATBAIJMKL.html
),
and many parts of the code, e.g. VecScatter, which have specific KNL
optimizations.

  Thanks,

     Matt


> I thank you for your assistance and useful feedback.
>
> --
> *Najeeb Ahmad*
>
>
> *Research and Teaching Assistant*
> *PARallel and MultiCORE Computing Laboratory (ParCoreLab) *
>
> *Computer Science and Engineering*
> *Koç University, Istanbul, Turkey*
>
>


-- 
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which their
experiments lead.
-- Norbert Wiener

https://www.cse.buffalo.edu/~knepley/ <http://www.caam.rice.edu/~mk51/>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20180523/bf229b13/attachment-0001.html>