Introduction to CUDA C/C++

Training is for ORNL Badged Staff only.

In the CUDA programming model, a program is separated into serial regions that are run on CPUs and parallel regions that are performed (by a hierarchy of thread groups) on GPUs. This workshop will introduce you to the basic concepts, syntax, and CUDA API functions needed to transfer data to and from GPUs, write GPU functions (kernels), and manage thread groups. We will include a presentation of these basic concepts as well as hands-on portions of the workshop where you will run/alter example problems and use what you have learned to change a simple C program into a CUDA program that offloads work to a GPU.

Requirements

• You must have a laptop with ssh client to log into remote machine.
• You must have basic experience with Linux/UNIX command line and text editor (e.g. emacs, vim, nano, etc.)

Slides:  Introduction to CUDA C/C++