After SC, I need to investigate why this inconsistency happens. This applies for competitions, not for the main/10 node list.
Example from two different systems
Output from IOR 1:
IOR-3.3.0+dev: MPI Coordinated Test of Parallel I/O
...
task offset : 1
nodes : 1
tasks : 40
clients per node : 40
Output from IOR 2:
IOR-3.3.0+dev: MPI Coordinated Test of Parallel I/O
...
task offset : 1
tasks : 56
clients per node : 56
repetitions : 1
We used in our script the variable nodes, we need to change to tasks/clients per node, but I am wondering why this happens.