[coolcrasher]

I am running my openFoam cases in a cluster.

Previously I used to run them on single node. Following is the code used to run ( from user guide )

mpirun -pinexpr S0:0-9@S1:0-9 -np 20 simpleFoam -parallel

Each node has two sockets. Each socket has 10 physical processors. One node has 20 physical processors indeed ( 40 procs if SMT also taken into account )

S0,S1 are the two sockets and 0-9 represents 10 physical processors in the socket ( pinning is done using the LIKWID tool )

However I want to run simulation with finer mesh and one node is taking lot time. Hence I want to run it on more than one node.

Following are my questions:

1) How to pin the threads if I am running in more than two nodes. ( say 40 physical processsors in 2 nodes )
2) If I dont do pinning and I just run on one node ( 20 cores ) using following

mpirun -np 20 simpleFoam -parallel

Will the openFoam run them on 20 physical cores or it may run on some physical cores and some on artificial ( SMT )