wl@eimb_ru
Member
I am trying to run the fully coupled CCSM3 (version 3.0) on the Cray XD1.
Chips - AMD Opterons, OS - Cray's Linux, compiler is PGI 6.1.4
Since I am limited in a number of simultaneous MPI tasks,
I tried to set more than 1 OMP threads for CAM and CLM in order to speed up integration.
However, there was no obvious increase in speed.
Unfortunately I cannot connect with a shell to a compute node on the Cray and inspect processes it runs with the 'ps' command.
I can easily determine from logs, how many MPI processes started, model components write about that.
However, its not obvious for me how to determine the number of OMP threads actually started.
Chips - AMD Opterons, OS - Cray's Linux, compiler is PGI 6.1.4
Since I am limited in a number of simultaneous MPI tasks,
I tried to set more than 1 OMP threads for CAM and CLM in order to speed up integration.
However, there was no obvious increase in speed.
Unfortunately I cannot connect with a shell to a compute node on the Cray and inspect processes it runs with the 'ps' command.
I can easily determine from logs, how many MPI processes started, model components write about that.
However, its not obvious for me how to determine the number of OMP threads actually started.