[Openmm-betatesters] CUDA tests are failing
Peter Eastman
peastman at stanford.edu
Thu Jan 3 11:35:04 PST 2013
Hi Nabil,
You can specify the path to nvcc with the OPENMM_CUDA_COMPILER environment variable. So in your case,
export OPENMM_CUDA_COMPILER=/opt/cuda/bin/nvcc
See chapter 11 of the manual for more details.
Peter
On Jan 2, 2013, at 10:56 PM, Nabil Faruk <nffaruk at gmail.com> wrote:
> Hi,
> I have installed OpenMM 5.0 beta from source on our linux CentOS 5.6 machine. All of the CUDA tests fail after running make test. When running an individual test I received this error:
>
> exception: Error launching CUDA compiler: 32512
> sh: /usr/local/cuda/bin/nvcc: No such file or directory
>
> I know that in my case the path should be opt/cuda/bin/nvcc and cmake seems to recognize it since in the CMakeCache.txt I have:
>
> //Path to a program.
> CUDA_NVCC_EXECUTABLE:FILEPATH=/opt/cuda/bin/nvcc
>
> So why don't the tests realize this? Is there some environment variable that I should set? I didn't have this issue with OpenMM 4.1.1.
>
> Thanks,
>
> --
> Nabil
> _______________________________________________
> Openmm-betatesters mailing list
> Openmm-betatesters at simtk.org
> https://simtk.org/mailman/listinfo/openmm-betatesters
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://simtk.org/pipermail/openmm-betatesters/attachments/20130103/4833ec43/attachment.html
More information about the Openmm-betatesters
mailing list