Problems with installing GPU-supported Gromacs on a cluster and can't find cuda root

User discussions
,
1

GROMACS version: 2022.4

Dear gromacs users:

Recently I’m trying to install a GPU-supported version on the login node of a cluster which doesn’t have a GPU hardware. And I don’t have sudo access.

From the topic discussed in Gromacs 2020.4 compilation with GPU-support on non-GPU nodes, I understand that compling a GPU version does not necessarily need a GPU hardware, but I didn’t know the details in installation.

In my case, I loaded cuda from module on the node and however I couldn’t find the cuda root. It’s not in the defalut root: /usr/local/cuda. So is there any suggestions to find the cuda root which could help me install, or should I install another cuda toolkit for myself in my own directory?

2

Try:
module show <cuda_module>

3

Thanks for your suggestions, I tried this command before, but I can’t find the indicated path:

(base) xia.m@hebe:~$ module load cuda
        CUDA runtime library path

(base) xia.m@hebe:~$ module show cuda
-------------------------------------------------------------------
/hosts/hebe/programs/modules/modulefiles/cuda:

module-whatis   {Sets runtime library path for CUDA executables}
prepend-path    LD_LIBRARY_PATH /usr/local/cuda/lib64
-------------------------------------------------------------------

(base) xia.m@hebe:~$ ls /usr/local/cuda
ls: cannot access '/usr/local/cuda': No such file or directory
4

Then it’s probably the wrong module. There’s got to be a module called “cuda_toolkit” or something like that. Use module spider cuda to find it, and after loading it, run which nvcc to get the path. “nvidia” might also be another keyword to search for.

5

Thanks a lot for your help! I’ll try these methods soon.