I was wondering what is required just to run a CUDA application? Does the CUDA nvidia driver & toolkit still have to be installed?
I’m trying to implement CUDA in an application I am writing but don’t want people to have to install all the extra CUDA drivers/toolkit, so is it possible just to supply some dlls with the program to get it to work? If so what are the dlls required?
OK thats great, so standard drivers are ok? Why does the CUDA Website say that a “cuda” driver is required - or is supplying the libraries/dlls with the executable basically doing the same thing?
It has been a while since I’ve tested CUDA, and haven’t used version 2.0 yet, but IIRC, the driver and the toolkit are necessary as the toolkit has all the RTLs needed for the CUDA calls, but the SDK is necessary to compile code. I do remember testing VMD and having to have the toolkit installed, otherwise the software would complain of missing libraries.
I think Nvidia allows to supply the libraries independently from the toolkit (you may want to confirm that if you’re working on something commercially) so if it’s only the problem of libraries, the developer could just ship them with the executable, no?
Not sure. I’d have to look through the license agreement on that. I’m more of a hardware guy, not a software guy. I’m just saying from experience from testing other people apps on our system.