Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Allow separate directories for cuda and cudnn #590

Closed
dsfenn opened this issue Dec 5, 2020 · 1 comment
Closed

Allow separate directories for cuda and cudnn #590

dsfenn opened this issue Dec 5, 2020 · 1 comment
Labels
installation CUDA is easy to install, right?

Comments

@dsfenn
Copy link

dsfenn commented Dec 5, 2020

I'm working on a cluster that has cuda and cudnn installed in different locations. I don't have write permission to these locations, so I can't change how they're laid out. CUDA.jl successfully finds the cuda location, but it looks in the same place for cudnn and thinks it's not installed because it can't find it there. I can't figure out how to change that.

If there is a way to specify a separate cudnn directory, I haven't been able to find it. If there isn't, it would be helpful to have.

@maleadt maleadt added the installation CUDA is easy to install, right? label Dec 7, 2020
@maleadt
Copy link
Member

maleadt commented Dec 7, 2020

Make sure those directories are part of the library search path (LD_LIBRARY_PATH), which your cluster's module management system probably ought to do. That should do the job. CUDA_HOME etc is a common shortcut, but generally libraries and binaries you want to use should just be available on standard paths, and CUDA.jl shouldn't be looking all across the system for them.

@maleadt maleadt closed this as completed Dec 7, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
installation CUDA is easy to install, right?
Projects
None yet
Development

No branches or pull requests

2 participants