So there have been quite a few changes since the topics on this were posted, various hacks have been suggested, but wondering what the best way to do it would be.
In a cluster environment, how do I enable Double Precision if there is no X environment available?
Scenario: I submit a job via ssh which lands on a compute node and is executed.
I can get which card(s) I am assigned via $CUDA_VISIBLE_DEVICES, but I seem unable to enable double precision:
$ nvidia-settings -a :0[gpu:$CUDA_VISIBLE_DEVICES]/GPUDoublePrecisionBoostImmediate=1
ERROR: The control display is undefined; please run
nvidia-settings --help for usage information.
Is there a tool that will set that, bypassing the check for an X display?
Naval Postgraduate School