Hello all,

I'm using a 16 processor linux SMP machine, and the calculation seems to hang every time I use the host_to_node_real(array,size) function. I suspect that the problem is in the blocking send/recieve pair called by this macro, as it is impossible to get Fluent to do anything besides pop up dialogue boxes saying that it is busy until I kill one of the processes named "fluent_smpi.6.2". At this point, the main Cortex terminal displays something like "Fluent process could not be started", and it is possible to exit fluent normally.

Should I used a different version of the MPI standard for this system? I've tried the options available in the FLUENT menus, but with the same result each time.

Alternatively, is it possible, or recomended, to point all of the slave processes at some chunk of shared memory that is written by the host process?

Many thanks, Tom

You don't believe but yesterday I had the same problem. Before calling host_to_node(array,szie) send the size variable: host_to_node_int1(size). It works! Luca

Thanks for the idea, but unfortunately, my system hangs when I use the host_to_node_int_1() macro as well.

The problem may be with how our SMP machine is configured.