Learning Forum

[parsaary]

hello,

i would like to run a two-way fluent-rocky simulation on my uni's hpc, on 2 or more nodes. i can run these simulations fine on 1 hpc node, but i need to speed things up. i know rocky can only handle 1 node while fluent can do distributed parallel, but i can't figure out how to do this.

i have asked this question on here before but haven't gotten a conclusive answer. there's no documentation with regards to running these types of sims on hpc, i have looked at all the tutorials, manuals, etc, there's nothing i can find for this issue.

if you are an ansys employee, may i please schedule a short zoom session with you to show you the issue? i have been stuck for months and need some help, my university hpc admins don't have ansys specific expertise and i have not been able to solve the issue on my own. thanks for your help.

[lucas.hoenicke]

Hello, Parsaary.

Below you can see an image of the 2-Way Fluent menu in Ansys Rocky. In the "Fluent" tab, you will find options related to the processing of the simulation.

As mentioned, you are looking for distributed parallel execution.
You can set the Execution Mode for Ansys Fluent as "Distributed Parallel", which will allow to choose different hosts for processing.

I believe there is valuable information regarding parallel processing in the Ansys Fluent documentation.
Please, check the Fluent User's Guide, Chapter III - Solution Mode, Section 44 - Parallel Processing to see if there is any content that may help you further.

Best regards,
Lucas

[parsaary]

Hi Lucas,

Thanks for the reply, I am familiar with the "distributed parallel" option under the Fluent tab on the Rocky GUI, and that's the option I select when setting up the simulation. However, I can't use the GUI after I transfer the files onto an HPC, since that's where I want to launch it on.

To summarize, I'm attempting to launch Rocky with Fluent integration. I need to specify multiple nodes for Fluent Distributed Parallel mode. My system is using Slurm to schedule compute nodes, so for each run the nodes may be different. Is there a way to set these nodes dynamically on launch from a command line either by an argument or via a configuration file?

My current command line launch command is as follows:

Rocky --simulate "RProcky.rocky" --resume=1 --use-gpu=1 --gpu-num=0 --gpu-num=1 --gpu-num=2 --gpu-num=3 --ncpus=4