Ask Your Question

Revision history [back]

Create a Legacy Run Environment (Kernel, Compiler, etc) for CUDA

I'm new to Fedora but not Linux, and I recently installed and updated Fedora 27. I'm going through the process of setting up TensorFlow with GPU Support using this pdf guide for CUDA installation. Reports from NVidia Developer forums have led me to believe the actual distro doesn't matter so much as the requirements for Kernel and C Compiler versions in Section 1.1 of the aforelinked pdf.

My question is this: would it be possible for me to setup a working area that exactly meets those requirements without downgrading my entire box? Just some directory with a shell that would have the correct versions? If not the kernel, at least the compiler?