On Tue, 2014-01-28 at 14:21 -0500, Guanglei Cui wrote:
> Hi Jan-Philip and Ross,
>
> Thanks for the response. I suspect the performance drop I saw is most
> likely due to running 2 jobs on the same card because I didn't specify
> CUDA_VISIBLE_DEVICES in the first place. And, the same CUDA device ID was
> reported in the output files.
>
> However, setting CUDA_VISIBLE_DEVICES in the job script doesn't seem to
> affect the device ID reported in mdout. With the variable set to 0 and 1
> individually, CUDA DEVICE ID 0 is reported in both mdouts, even though
> nvidia-smi seems to suggest both GPUs are being used.
Unfortunately this is expected behavior. The device number is numbered
from 0 to N of the N GPUs specified by CUDA_VISIBLE_DEVICES. It may
make sense for Amber to start printing out the value of the
CUDA_VISIBLE_DEVICES environment variable when it prints out device
number, too.
This is part of the CUDA API itself, so Amber has no control.
One thing to be careful of is that the device numbering in nvidia-smi
does _not_ necessarily match the device numbering that one gets from the
CUDA API. If you want to get the mapping that the CUDA API returns, you
need to use the "deviceQuery" program included with the CUDA SDK
instead.
HTH,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Jan 28 2014 - 12:00:03 PST
