I think that if you need highest performance, you can probably find libraries that are faster in shared memory than PBGL (e.g., Ligra, Galois). Some of these systems are experimental and somewhat harder to use than PBGL, so you should investigate the performance tradeoff for the particular use you plan.

-m

On Wed, Mar 2, 2016 at 4:40 PM Ярослав <yaroslavkishchenko@gmail.com> wrote:

Hi guys, I need to write some graph parallel algorithms
and I decided to use boost graph library for this. There
are also PBGL. So my question is, is PBGL works great in
shared memory environment. My algorithms will be used in a
competition, so I know that it will be launched on a
single machine, but with quite high number of cores. Is it
worth to use sequential BGL with OpenMP or something, PBGL
can handle shared memory and prevent overhead from copying
data from one process space to another (as I get it is
using MPI)?
_______________________________________________
Boost-users mailing list
Boost-users@lists.boost.org
http://lists.boost.org/mailman/listinfo.cgi/boost-users