Fedora 21 install on VMware

Hello,

I am trying to install CUDA 7.5 to fedora 21 on VMware…

I follow instructions from Linux install document. After installation nvcc works…

However after a reboot, fedora crashes and gives “oh no! Something has gone wrong” , even for rescue kernel.

Anyone experienced same…? could you get around…?

Thanks in advance for all suggestions…

If you intend to use that setup with a GPU in the VMware instance/VM, then you’ll need to be a lot more specific about your setup and how you intend to achieve that.

If you intend to use that setup without a GPU (so, just for compiling, I guess) then probably the solution is to install Fedora 21 in your VM, then install the CUDA toolkit but not the driver.

Using the CUDA toolkit runfile installer method, simply select “no” when prompted to install the driver.

Thank you…

This is CUDA equipped device (GeForce)… I installed CUDA for VS 2013 and could use it.

I am also trying to use it in linux environment…

If you are using a GeForce device from within a VMware VM, that is a non-trivial setup. You would need to give more specifics about how you have set it up. My guess is that if/when you were using CUDA with VS2013 on it, that was not taking place from within a VMWare VM.

Thank you…

Main OS is Windows and VS2013 runs on it (no other OS). VMware runs on windows. Installed Fedora 21 from original ISO file.

Before CUDA everyrhing is fine. After installation CUDA works but after a reboot F21 crashes and does not boot up…

Let’s go back to my first comment then.

If you’re intending this Fedora setup just for compiling (not running) codes, then proceed, but when you install CUDA, use the runfile installer method and decline the option to install the driver. That should work fine but you won’t be able to run any CUDA codes there.

If you’re intending this Fedora setup to also run CUDA codes, it won’t work without additional configuration steps to provide a (CUDA) GPU in the VM instance. This is a non-trivial thing to do, and probably won’t work with a GeForce GPU.

The reason things worked for you in the Windows/VS case is because that OS was not running in a VM, and so had direct access to the CUDA GPU.

I thank for the reply which helped me to save a lot of time…