note this is still suboptimal.
for example, if you run a job with two MPI tasks with two OpenMP threads
each on the same node,
then it is likely the OpenMP runtime will bind both thread 0 on core 0,
and both thread 1 on core 1, which one more time means time sharing.
Cheers,
Gilles
On 6/8/2016 3:30 PM, Maxim Reshetnyak wrote:
thank you!
mpirun --bind-to none ...
gives what I need:
echo " run 1 " ; export OMP_NUM_THREADS=1 ; time mpirun -np 1
--bind-to none a.out ; echo " run 2 " ; export OMP_NUM_THREADS=2 ;
time mpirun -np 1 --bind-to none a.out
run 1
0 0
0 0
real 0m43.593s
user 0m43.282s
sys 0m0.187s
run 2
0 0
0 1
real 0m21.970s
user 0m43.202s
sys 0m0.148s
_______________________________________________
users mailing list
us...@open-mpi.org
Subscription: https://www.open-mpi.org/mailman/listinfo.cgi/users
Link to this post:
http://www.open-mpi.org/community/lists/users/2016/06/29402.php
