Energy efficiency has been the dominant criterion guiding computer design for the past 20 years. By abruptly halting the increase in the clock frequency of computers, the energy constraint has led manufacturers to produce machines whose power results from the assembly of an ever-increasing number of calculation units. The acceleration of computer calculations is therefore mainly based on the division of work into an ever-increasing number of elementary tasks executed in parallel. Because of this energy constraint, parallel computing architectures are also increasingly specialised (CPU, GPU TPU, DSP, FPGA) and their programming is based on adapted paradigms and languages.

Mobile phone, laptop, desktop or supercomputer: parallelism is omnipresent. Mastering the programming of parallel machines is therefore no longer reserved for laboratories, but is part of the necessary background for all developers concerned with the performance of their software.

This training aims to expose the major issues of programming and parallel machines and to guide developers towards the tools best suited to their needs. In addition, the course will provide a methodology for designing energy-efficient software. This knowledge will also be useful for IT managers in establishing their equipment strategy.