Parallel Computing technologies brought dramatic changes to mainstream computing. This trend is accelerating as the end of the development of hardware following Moore's law looms on the horizon. The majority of standard PC's and even notebooks today incorporate multiprocessor chips with up to four processors. This number is expected to soon reach eight and more.
These standard components, COTS (Components Off The Shelf), are increasingly combined with powerful parallel processors originally designed for high-speed graphics processing, GPU's (Graphics Processing Units), and FPGA's (Free Programmable Gate Arrays) to build heterogeneous parallel computers that offer a wide spectrum of high speed processing functions. The number of processors incorporated in such systems are today of the order of up to 104 to 106. This vast number of processors allows the construction of high speed computers in the petascale range, and even the exascale range, at a reasonable cost. The limiting factor for constructing more powerful hardware is the energy consumption and thermal control of such systems. Many research efforts concentrate on reducing the overall energy consumed.
In addition to the hardware design and build limitations, the practical use of petascale or exascale machines is hampered by the difficulties of developing software that efficiently and effectively run on such architectures. This holds for system as well as application software. The ParCo conference aimed at addressing many of these issues through contributed papers as well as the various mini-symposia presentations.
In this book, which includes selected and refereed papers presented at the international Parallel Computing conference (ParCo2009) held from 1–4 September 2009 at ENS (École Normale Supérieure), Lyon, France, problems associated with the development of high speed parallel systems using new hardware concepts and the associated software development issues are considered. The papers were presented as keynote papers, in regular sessions, an industrial session and various mini-symposia covering specialised topics. Overall these give a snapshot of the state-of-the-art of parallel computing technologies, both in hardware as well as application and software development.
This year's highlight is no doubt the increasing number of papers addressing the programming of general-purpose graphics processing units. Considering the main track of the conference, as well as the mini-symposium dedicated to GPU's, ParCo2009 turned into one of the main scientific venues covering this important research topic in 2009.
The editors wish to express their sincere gratitude for all persons who supported this venture and lastly made it feasible. In particular we wish to thank the many reviewers who, as members of the international Program Committee, not only assessed papers, but also acted as session chairmen during the conference.
Sincere thanks is due to the members of the Organising Committee, and in particular to Laurent Lefèvre, Eddy Caron and Jean-Christophe Mignot, who spent many hours assisting in organising a very successful event. We are also very grateful for work done by Virginie Mahdi from Genci, Paris in attracting a considerable number of sponsors as well as participants in the Industrial Session.
Please note that versions of papers with colour images and diagrams are available in the electronic version of the book on http://www.booksonline.iospress.nl/ under Advances in Parallel Computing.
Barbara Chapman, University of Houston, USA
Frédéric Desprez, INRIA, France
Gerhard Joubert, TU Clausthal, Germany
Alain Lichnewsky, GENCI, France
Frans Peters, Philips Research, Netherlands
Thierry Priol, INRIA, France
31 December 2009