Dear Jason,
As you said, that is simply CUDA tool kit problem.
I resolved it by making machine name file.
Thank you for usual support.
Kurisaki IKuo
-----Original Message-----
From: Jason Swails [mailto:jason.swails.gmail.com]
Sent: Friday, November 30, 2012 10:03 PM
To: AMBER Mailing List
Subject: Re: [AMBER] GPU compile TEST does not work.
On Fri, Nov 30, 2012 at 2:06 AM, kurisaki < kurisaki.ncube.human.nagoya-u.ac.jp>
wrote:
> Dear Amber developers and users,
>
> Thank you for kind support.
>
> I would resolve the following error by Adding $CUDA_HOME/* to
> $LD_LIBRARY_PATH.
>
> However, in the TEST step,
> There still error message,
> " cudaGetDeviceCount failed no CUDA-capable device is detected"
>
> I am most grateful if you give me an advice to treat it.
>
This is not actually an Amber problem. The CUDA API cannot find any GPU
accelerator, which means the driver may not be properly installed (or it may not
be the right version). If you haven't installed the CUDA SDK yet, I would
suggest doing that. The "nvidia-smi" program is helpful in determining what
GPUs are available (and their current load, if that model supports those
functions).
Good luck,
Jason
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Nov 30 2012 - 06:30:03 PST
