Dynamic scheduling and memory locality

jimdempseyatthecove
Total Points: 13304
Status Points: 13304
Black Belt
July 2, 2008 9:23 AM PDT
Rate
 
#3

Here is a suggestion:

Zone the matricies work into more zones than you have cores, and where the initial "number of cores" number of zones contains the majority of the work. Then for the remainder of zoned work on first come first serve.

$OMP PARALLEL
{work on zone=OpenMPteamMemberNumber}
$OMP DO SCHEDULE(YourChoice)
{do remainder work as parallel do}
$OMP END DO
$OMP END PARALLEL

You _may_ also find it advantageous to set thread affinity to a given processor.

Jim Dempsey

 

Forum Statistics

4488 users have contributed to 24040 threads and 70114 posts to date.
In the past 24 hours, we have 22 new thread(s) 111 new posts(s), and 143 new user(s).

In the past 3 days, the most popular thread for everyone has been IVF 11.0 and VS2008 The most posts were made to IVF 11.0 and VS2008 The post with the most views is Leigh, Have you built this fo

Please welcome our newest member janet.p.simpsonnasa.gov