Graphics Processing Units (GPUs) were originally developed for computer gaming and other graphical tasks, but for many years have been exploited for general purpose computing in a number of areas. They offer advantages over traditional CPUs because they have greater computational capability, and use high-bandwidth memory systems (memory bandwidth is the main bottleneck for many scientific applications).
This introductory course will describe GPUs, and the advantages they offer.
It will teach participants how to start to program GPUs, which cannot be used in isolation, but are usually used in conjunction with CPUs.
Important issues affecting performance will be covered.
The course focuses on NVIDIA GPUs, and the CUDA programming language (an extension to C/C++ or Fortran). Please note the course is aimed at application programmers; it does not consider machine learning or any of the packages available in the machine learning arena.
Hands-on practical sessions are included.
You will require your laptop, and your institutional credentials to connect to eduroam. The training will be run on a web-based system so all you will need is a relatively recent web browser (Firefox, Chrome and Safari are known to work).
10:20 GPU Concepts/Architectures
11:20 CUDA Programming
12:00 First CUDA Programming Exercise
14:30 CUDA Optimisation Practical Exercise
15:20 CUDA Constant and Shared memory
16:00 Practical exercise
10:10 OpenCL and OpenCL exercise
11:30 NVIDIA Volta (Alan Gray NVIDIA)
14:00 OpenACC directives and exercise
15:20 The future? Performance portability and Kokkos