[parsec-users] Setting affinity to parsec threads

Swarnendu Biswas biswas.38 at buckeyemail.osu.edu
Wed Aug 31 09:09:21 EDT 2016


Hi,

You could instrument thread begin, and they try setting 
pthreads_setaffinity_np() from within the instrumentation. I haven't 
tried it myself, just an idea.

Regards,
Swarnendu Biswas.

On 08/31/2016 09:03 AM, hergys rexha wrote:
> Hello Pierre,
> Thanks for the reply. Actually i am using pthreads and i want to lock 
> the threads of blackscholes into different cores. If somebody has some 
> experience?
> Thanks.
>
> On Wed, Aug 31, 2016 at 3:42 PM, Pierre-Yves Péneau 
> <pierre-yves.peneau at lirmm.fr <mailto:pierre-yves.peneau at lirmm.fr>> wrote:
>
>     Hi,
>
>     If you are using the OpenMP version of blackscholes (before OpenMP
>     3.1),
>     you can use this environment variable:
>     GOMP_CPU_AFFINITY=list-of-cpu-id [1]
>
>     With OpenMP greater or equal to 3.1, it seems you can only specify if
>     the runtime could move threads during the execution or not, by using
>     OMP_PROC_BIND=true/false [2]
>
>
>     Bests,
>
>
>     [1]
>     https://gcc.gnu.org/onlinedocs/libgomp/GOMP_005fCPU_005fAFFINITY.html
>     <https://gcc.gnu.org/onlinedocs/libgomp/GOMP_005fCPU_005fAFFINITY.html>
>     [2] http://www.openmp.org/mp-documents/OpenMP3.1.pdf
>     <http://www.openmp.org/mp-documents/OpenMP3.1.pdf> (see section 4.4)
>
>     On 31/08/2016 14:19, hergys rexha wrote:
>     > Hello,
>     > I am trying to make benchmarking with Parsec blackscholes and i
>     want to
>     > lock threads to different cpu cores. I am using a linux system
>     and dont
>     > have so much experience with the benchmark.
>     > Any help will be appreciated.
>     > Thank you,
>     > Hergys
>     >
>     >
>     >
>     > _______________________________________________
>     > parsec-users mailing list
>     > parsec-users at lists.cs.princeton.edu
>     <mailto:parsec-users at lists.cs.princeton.edu>
>     > https://lists.cs.princeton.edu/mailman/listinfo/parsec-users
>     <https://lists.cs.princeton.edu/mailman/listinfo/parsec-users>
>     >
>
>     --
>     +-------------------------------------------------------------+
>     | Pierre-Yves Péneau - PhD student |  first.last at lirmm.fr
>     <http://lirmm.fr>  |
>     | LIRMM / CNRS - SYSMIC team       | + 33 4 67 41 86 33
>     <tel:%2B%2033%204%2067%2041%2086%2033>    |
>     | Building 4 Office H2.2           | http://walafc0.org   |
>     +-------------------------------------------------------------+
>
>
>     _______________________________________________
>     parsec-users mailing list
>     parsec-users at lists.cs.princeton.edu
>     <mailto:parsec-users at lists.cs.princeton.edu>
>     https://lists.cs.princeton.edu/mailman/listinfo/parsec-users
>     <https://lists.cs.princeton.edu/mailman/listinfo/parsec-users>
>
>
>
>
> _______________________________________________
> parsec-users mailing list
> parsec-users at lists.cs.princeton.edu
> https://lists.cs.princeton.edu/mailman/listinfo/parsec-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/parsec-users/attachments/20160831/8dc61e53/attachment-0001.html>


More information about the parsec-users mailing list