CUDA Tutorial

Introduction

Welcome to this tutorial. I will guid your from the installation of CUDA Toolkit to some coding samples and implementation.

Brievely:

• What is CUDA ?

CUDA is a parallel computing platform and programming model invented by NVIDIA. It enables increases in computing performance by using the power of the graphics processing unit. You can code with a specific language called CUDA (.cu) and improve your program in C/C++ or python.

• What is CUDA Toolkit ?

CUDA Toolkit provides an environnement that allows you to compile CUDA code

• What is CUDA for ?

You can use it for AI and Data Science, High Performance Computing, Design and Simulation, Data Center and Cloud Computing and many more. There is the NGC catalog where you can find a very large collection of software developped by Nvidia: click here

Installation of CUDA Toolkit

You can install CUDA on Windows and Linux but my suggestion is to install the linux version and this why:

• Linux installation is faster, take less space on your hard drive and code performance are always better with linux.
• Developping CUDA code with Linux is more convenient: you can chose your IDLE, create project more faster and don't have tons of depencies like Visual Studio.

Special mention for WSL: I didn't succeed to install CUDA Toolkit on WSL. I'm struggling with my CG detection and it make the installation impossible. If you have a solution I will be pleased to know about it.

You can find the tutorials in the docs/ folder.

Let's code

Compilation

The CUDA toolkit allows you to compile CUDA code. The file extension look like this: .cu and the syntax is pretty much like C++ language. You can find some examples in the src folder. Within the toolkit, a compiler has been installed: nvcc. Let's try it !

We're going to compile the mult.cu file:

nvcc src/mult.cu -o mult.o

You now have an executable file: mult.o and you can execute it:

./mult.o

Deep learning softwares and pre-trained models

More software for DL

NGC Catalog

Ressources

• CUDA toolkit installation: https://docs.nvidia.com/cuda/cuda-quick-start-guide/contents.html

• CUDA C++ Programming Guide: https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#

• CUDA python tutorial: https://www.vincent-lunot.com/post/an-introduction-to-cuda-in-python-part-1/