Running Polyflow across multiple cluster nodes

  • Last Post 25 May 2020
allenpvar posted this 22 May 2020


I have access to a cluster that has 10x NVIDIA Tesla P100. The configuration is setup so that there are 5 GPU node servers with 2 Tesla GPU's connected to each.

1. Is there a way I can run polyflow in batch mode, across all the server nodes thereby utilizing all 10 GPU's

2. Is there a way I can combine running polyflow in CPU (to increase available memory) and GPU (to maximize computing)


Currently I am able to run polyflow on a single node utilizing a maximum of 2 GPU's

Current approach:

I have been attempting to run Polyflow in batch mode using the command,

polyflow -ES full -acc nvidia -na 2 -s polyflow.p3rc -MGR -default_output -th 32 < polyflow.dat > polyflow.lst


Screenshot from polyflow.lst:


Please advice,




bducoeur posted this 25 May 2020


Polyflow runs using shared memory parallel only, not distributed memory. That means, it cannot use cores on more than one compute node.

Best Regards,