Learning Forum

[Atul Singh]

Hi,

When i dont have a udf in my example, resources online have indicated that fluent on the cluster (slurm based) can be done using the -mpi=intel -pinfiniband flags. (See for instance this link from university of south florida, which suggest parallel jobs require the above two flags,
https://wiki.rc.usf.edu/index.php/ANSYS#Documentation

Putting the image as well, incase you deny opening the link,

However, when I do have the udf in my input file towards a batch submission, this seemed to have thrown an error.

So I have two questions,

1) Does using parallel necessary require giving -mpi=intel -pinfiniband flags????
2) If yes, what other commands/things i could look into to compile and load a udf as well, to this. Currently, if I have a udf , I simply remove these two flags.
3) What recommendations can you give to compile the udf using the mpi and pib flags?

[Rob]

The command script required will depend on how the cluster is configured. On the Ansys cluster I launch Fluent at a queue and don't specify any parallel options other than the number of cores. 

What suggestions have local IT made?