Re: [Boost-users] Boost graph parallel issue

On Thu, Jul 16, 2015 at 6:48 AM Mayank Bakshi wrote:

Hi Marcin,

Thanks for the reply!

You're welcome. :)

I used a different graph for timings which has 15,606 nodes and 45,878 edges, link: http://people.sc.fsu.edu/~jburkardt/data/metis_graph/4elt.graph. Can you suggest any graph which gives boost mpi performance with increasing processes?

You could use one of the generators. Your timing shows 40.80ms for one process. I don't think that there is much that MPI can improve on this result. At such a short run time, multiple MPI processes will be just a pure overhead.

I am running boost tests for connected components, betweenness centrality and page rank in directory $(BOOST_ROOT)/libs/graph_parallel/test, can I use as input a .gr file rather than random graph which gets generated there for input. Please suggest.

I am not sure I understand the question. You can use a .gr file if you have one.

Also can I know how to convert this .gr file to matrix market file format(.mtx) or vice versa?

Hmm, I am not sure if there is a ready made converter.

Thanks & Regards Mayank Bakshi Tata Consultancy Services Limited Cell:- 7083955990 Mailto: mayank.bakshi@tcs.com Website: http://www.tcs.com ____________________________________________ Experience certainty. IT Services Business Solutions Consulting ____________________________________________

-----"Boost-users" wrote: ----- To: boost-users@lists.boost.org From: Marcin Zalewski Sent by: "Boost-users" Date: 07/16/2015 05:56AM Subject: Re: [Boost-users] Boost graph parallel issue

Mayank,

Which graph do you use as input? Are you using https://github.com/boostorg/graph/blob/master/test/weighted_graph.gr ? This graph is probably just too small to benefit from additional processes. Could you try some input that runs for at least few seconds on a single process?

-m

On Tue, Jul 14, 2015 at 9:36 AM Mayank Bakshi wrote:

...

I want to utilize boost MPI to increase performance of graph code. When I run standard examples in libs/graph_parallel/examples/, bfs does not give better performance with more than 1 process taken in mpirun (using openmpi 1.8.3). The timing is as follows:

No of Processes Time (ms)

1 40.80 2 82.19 4 92.67 8 76.86 16 91.61

Any suggestions to improve the performance with processes as graph parallel should use more ranks for better performance.

Thanks & Regards Mayank Bakshi Tata Consultancy Services Limited Cell:- 7083955990 Mailto: mayank.bakshi@tcs.com Website: http://www.tcs.com ____________________________________________ Experience certainty. IT Services Business Solutions Consulting ____________________________________________

=====-----=====-----===== Notice: The information contained in this e-mail message and/or attachments to it may contain confidential or privileged information. If you are not the intended recipient, any dissemination, use, review, distribution, printing or copying of the information contained in this e-mail message and/or attachments to it are strictly prohibited. If you have received this communication in error, please notify us by reply e-mail or telephone and immediately and permanently delete the message and any attachments. Thank you

_______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

_______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users _______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users