With the advent of GPGPU cards, all computers are becoming massively
parallel systems that are very difficult to program efficiently. Indeed,
the latest generation of NVIDIA cards provide more than 3000 cores grouped into multi-processors of 32 SIMD cores. This means that in order to efficiently execute an existing algorithm on a single PC with a hexacore CPU and a top notch GPGPU card, one should decompose the algorithm into 6 major tasks and 1024 minor ones which, 32 by 32, should execute the same instruction at the same time!

Because of necessary synchronizations between cores and data exchanges, exploiting efficiently such a machine (a standard PC) is virtually impossible with standard algorithms.

Fortunately, Complex Systems produce results that emerge from the multi-level interaction of many independent entities that can be directly implemented in a very efficient way on multi-level massively parallel machines such as the ones described above.

This tutorial will show how the EASEA-CLOUD massively parallel evolutionary platform implements evolutionary algorithms (that can optimize nearly any kind of continous, discrete, combinatorial, mixed problems) as a kind of Complex System, where entities are individuals that interact through genetic operators, not only on one machine, but on computing eco-systems such as clusters or grids of GPGPU machines, or a cloud of computers.

Authors

Pierre Collet After a PhD in 1997 in virtual reality, Pierre Collet joined INRIA from 1998 to 2000, where he first developed the EAsy Specification of Evolutionary Algorithms (EASEA) language in 1999. He then went to the French Ecole Polytechnique, as a researcher for the European DREAM project, that used EASEA as a programming language. After several years as associate professor at Université du Littoral, he was appointed full professor at Université de Strasbourg in 2007, where he co-leads the CSTB (Complex Systems and Translational Bioinformatics) team of the ICUBE laboratory. He was Head of the department of Computer Science of the Strasbourg University between 2011 and 2015. Pierre Collet is also co-coordinator of the CS-DC-UNESCO UniTwin since 2014.

Tutorials

Photos by : Derek K. Miller