Configuring your development environment
It is fundamental to have NVIDIA's proprietary graphic driver installed. This is an important item because the default driver that works with NVIDIA GPUs in Linux installations is the nouveau open source driver, which can execute graphics but is not capable of running CUDA code.We are going to learn about two possible configurations. We will follow an order that is less intrusive to our machines; you will have an opportunity to experiment with CUDA before committing to installing it directly. In the first configuration, we will discuss execution through Docker, and in the second configuration, we will see the installation of everything directly on our system. Each configuration has pros and cons, which we will also discuss.Just to make our strategy clearer, let's summarize it:
- Installing the CUDA proprietary device driver is mandatory, regardless of the CUDA Toolkit – so we are going to deal with that...
