Clereco

Cross-Layer Early Reliability Evaluation for the Computing cOntinuum

January 20, 2016, Prague

https://wapco.inf.uth.gr

Scope

Research in the last few years has focused on approximate computing as a means to overcome the energy scaling barrier of computer systems. Such savings can be achieved by utilizing the inherent error resilience of algorithms in many application domains such as signal processing, multimedia, data analytics and computational engineering, among others. Indeed, fully accurate arithmetic in specific phases of a computation in those applications may have only a marginal effect on output quality, especially if combined with error correction frameworks such as iterative refinement. Thus, accurate execution may be traded off with lower energy consumption by providing the ability to scale supply voltage below nominal values or to use lower precision arithmetic (i.e. 8 or 16 bit), thus, trading off low energy with quality of output results.

Rather than focusing on a single layer, designing such systems in a general-purpose computing environment requires a holistic view of all layers from algorithms, programming models, system software, and hardware down to the transistor level. This half-day workshop is an inter-disciplinary effort to bring together researchers from the areas of mathematics, computer science, computer and electrical engineering to discuss challenges, risks and opportunities of approximate computing in all design layers. Papers will not be published in proceedings, so submitting to WAPCO will not preclude future publication opportunities. We are soliciting original papers on topics that include but are not limited to the following:

• Formal and mathematical methods for approximate computing
• Programming languages and models for approximate computing
• Compiler and system software support for approximate computing
• Hardware support for approximate computing
• Hardware-software interaction for approximate computing
• Applications that can benefit from approximate computing
• Simulation and modeling techniques for approximate computing
• Position papers on the potential and limitations of approximate computing