Introduction to Performance Engineering

Performance engineering is a complex activity, which is based on existing, fundamental knowledge, but also requires active investigation.

The goal typically is to reduce the consumption of resources (be it time, compute time or energy - this is getting more and more important) for a given computational task.

This brief introductory courses aims to:

  • give an overview of the knowledge required to do performance engineering work, in the spirit of reducing the unknown unknowns for beginners, with a focus on HPC and scientific computing;

  • discuss the tools to analyze application behaviour in the most common programming languages used in HPC in the field of scientific research (profiling, tracing);

  • give elementary notions of computer architecture, including CPU, memory systems, GPU and Filesystem, which sets the maximum performance that can be achieved;

  • give a hands-on demonstration of the performance optimization loop.

The choice of topics is based on the sources of inspiration (see below) and on the personal experiences of the maintainers of this lesson, with a slight bias - which must be acknowledged and managed - towards interesting problems.

Episodes

Credits

The current course material is inspired by and based on:

Supplementary Material