Emulating in CUDA with NO CUDA enabled graphics card can you do this?

Someone on this Forum yesterday said that I should install NVIDIA-linux-x86-180.22.-pkg1.run to avoid some not implemented errors when i run certain projects that come with 2.1 CUDA. I get a lot of runtime errors that go back to cutilCheckMsg. I was told that to get rid of those I must install NVIDIA-linux-x86-180.22.-pkg1.run. As I said before, I am only emulating for now. I am doing this because I want to learn CUDA and at present I do not have a CUDA enabled graphics card - so I emulate.

But to install NVIDIA-linux-x86-180.22.-pkg1.run would to avoid the cutilCheckMsg errors, i must shutdown X windows and run NVIDIA-linux-x86-180.22.-pkg1.run as root. It will state again that this software will ignore the graphics card. I guess this is good - since all I want is some ancillary subprograms to avoid getting cutilCheckMsg related errors. i do not want tit to install any new drivers for my non-CUDA enabled graphics card.

So how can i install NVIDIA-linux-x86-180.22.-pkg1.run to get it to install, but to not have anything to do with the current graphics card which as I said is not CUDA enabled. If it did have any thing to do with my graphics card things would get mess up quick. I know that once I run native, this problem will no longer exist.



Run the installer with --advanced-options and you will notice that amongst the cornucopia of options is one to not install the kernel driver, and several to control the installation of the NVIDIA X and GL libraries in a non standard path. You probably want to do both, and then customize the SDK makefile to look for the libraries in the non standard path you chose. This means there will be no conflict with either the drivers nor the X and GL libraries of your current video card and X setup.