This is a training course in GPU-programming with OpenACC including an overview of the hardware architectures (i.e. CPU vs GPU). The course involves hands-on and is physical. It took place in Bergen 29/03/2022.
The course contains slides, exercises and solutions. The content of the slides is split into two parts as described below:
0. Part 0
- State-of-the-art supercomputers
I. Part I
-
CPU and GPU architectures (CPU vs GPU)
-
GPU-programming models and compiler supports
-
Functionality of OpenACC
II. Part II
-
Synchronous OpenACC
-
Combining OpenACC with other programming models
-
Asynchronous OpenACC
-
Code profiling
III. Hands-on + discussion