The HiPEAC Vision 2017 - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Ouvrage (Y Compris Édition Critique Et Traduction) Année : 2017

The HiPEAC Vision 2017

Résumé

Information technology is one of the cornerstones of modern society and it is evolving rapidly: while the main challenges identified in the HiPEAC Vision 2015 remain valid and have even increased in importance, new challenges are ahead of us. Computers are disappearing from view. They are taking on new forms, such as cars, smart meters, thermostats, and so on. They communicate with their users using voice, sound, pictures and video, closely resembling human interaction. We are entering the Artificial Intelligence era. This will not only change how we interact with machines, but it will also redefine how we instruct a machine what to do: less programming and more learning. The function of the computer is shifting from carrying out computational tasks to provide answers to numerical problems, to working together with humans (what we call the beginning of the Centaur Era), augmenting reality to assist us, or even creating virtual worlds for us to explore: the cyber-physical entanglement between the physical and virtual world. Computers will increasingly interact with the physical world, leading to a expansion from security into safety. Humans need to trust both the machines and the information that they keep about us, and therefore enforcement of security and privacy is of paramount importance. For compute-intensive tasks, we will continue to use the cloud and supercomputers (HPC); this means that connectivity is crucial, yet local processing is becoming increasingly important. The increasing computational requirements are making computer system architects look for accelerators for specialized tasks, diverting in many cases from the traditional Von Neumann architecture. Energy efficiency of computing systems remains a major challenge for the coming years. As the cost per transistor is no longer decreasing, we might see diversified paths for using silicon technology: many designs will not use the latest technology node, but the more mature (and cheaper) one. It is also the right time to revisit the basic assumptions in order to open new tracks and approaches and to eventually reinvent computing. With the flood of new systems and new system architectures, increasing attention must be paid to composability and interoperability between systems. The complexity of the new systems will be so high that human designers will only be able to master it with the help of computers using AI-based techniques. Innovative approaches will be required to ensure that the systems will do what they are supposed to do, both at the functional and at the non-functional level (e.g. timing requirement or reliability). We need to develop design techniques that go beyond predictability by design and allow the building of reliable systems from unreliable parts. Finally, holistic approaches, implying multi-disciplinary techniques, will be needed in order to meet all the requirements of trustability, efficiency and cost.
Fichier principal
Vignette du fichier
HiPEAC_Vision_2017.pdf (17.76 Mo) Télécharger le fichier
Origine : Accord explicite pour ce dépôt

Dates et versions

hal-01491758 , version 1 (17-03-2017)

Identifiants

  • HAL Id : hal-01491758 , version 1

Citer

Marc Duranton, Koen de Bosschere, Christian Gamrat, Jonas Maebe, Harm Munk, et al. (Dir.). The HiPEAC Vision 2017. Duranton, Marc; De Bosschere, Koen; Gamrat, Christian; Maebe, Jonas; Munk, Harm; Zendra, Olivier. HiPEAC network of excellence, pp.138, 2017, 978-90-9030182-2. ⟨hal-01491758⟩
750 Consultations
2154 Téléchargements

Partager

Gmail Facebook X LinkedIn More