Adavanced-Options setting abs. and rel. paths

I am confused. I want to install the NVIDA X drivers and GL libraries. I already have a non-CUDA enabled graphics card working in the system. I do not want to mess with that.

When I run the installer with --advanced-options I can see my first option should be

–no-kernel-module so everything about the CUDA enabled driver is installed except the kernel, which I do not want to install. I want the libraries, but not the kernel. I already have a working graphics kernel and I cannot have two working kernels in one computer. I leave the new one out and keep the old one in.

When running the installe r the second options should be:


which is installed to put these CUDA enabled NVIDIA X libraries into a new directory and path (so as to be used only on compiling) so as not to interfere with the working libraries on the current graphics card. I also must put this path into the makefile so that during compilation the makefile can find these new paths. How do I designate a path in Linux? The current one in the makefiles is:
include …/…/common/

So I guess it must be something like that for X-LIBRARY-PATH, but of course not that exactly.

Now in setting the path for the NVIDIA GL libraries I need to use, I think:


and I must again specify a path (OPENGL-LIBDIR) and again use it in the makefiles for compling. This time if I read it correctly it must be a relative path. So what do I specify. Also, did I choose the right option?

The advanced option use the terms: libraries, modules and components interchangeably so I I am not sure sometimes what they are speaking about. I care about libraries here.

In the third option why change to a relative path suddenly? Is this a typo. It must be.

I know from DOS/Windows what an absolute and a relative path are, but do they apply here in LINUX? See the make path include …/…/common/

I am using Ubuntu 8.04.

I know this sound a bit confusing, but I am lost here. Any help appreciated. I believe my first two options are correct; I am not sure about the third one with the OPENGL libraries.



I can’t think of a polite way to put this, so I am not going to try. You need to spend some time and familiarize yourself thoroughly with the operating system and the development toolchain before you try and go any further. Using the SDK requires a elementary degree of competency with the development environment that you obviously lack. Engineering what amounts to your own customized installation of the SDK and support libraries and using it in a non-standard manner is clearly not something you are presently capable of, and I suspect nothing anyone posts in reply to your questions is going to help you with your present level of understanding.

The cheapest video cards which support CUDA now cost less than $50, and their thermal and power requirements are so modest that any PC with the requisite PCIe x16 slot should work without further expense or modification. If you goal is to get a working SDK you can develop with, then my advice is to procure a cheap supported card and install the drivers, CUDA and SDK over a supported Linux distribution exactly as described in the NVIDIA instructions. Your chances of success will be several orders of magnitude better than they are with your current approach.