Dimitrios S. Nikolopoulos (Queen's University of Belfast), Christos D. Antonopoulos (University of Thessaly)
Contact: cda [at] inf [dot] uth [dot] gr (subject: ParCo%202015%20Mini%20Symposium) (cda at inf.uth.gr )
Computing systems now expose power and reliability indicators to software. These indicators are in turn used to develop new energy-aware optimisations and new approaches to resilience that span all layers of the software stack, from firmware and operating systems, to compilers, runtime systems, languages and application software. Awareness of power and resilience are essential in large-scale parallel computing systems and datacentres, where mean time to failure rates are deteriorating and software executes under power caps that often manifest as non-deterministic performance variation. Power and resilience are perennially important for many-core embedded systems, where emerging applications push for higher performance and data processing ability on a limited energy budget, while component density and feature sizes compromise reliability.
The proposed minisymposium provides a forum to broadly explore the implications of energy and resilience on the principles and practice of parallel programming. The minisymposium aims to present original research that uses energy and resilience as first-class resources in parallel programming languages, libraries and models and demonstrates how parallel programming can improve energy-efficiency and resilience of large-scale computing systems or many-core embedded systems. The minisymposium also intends to explore how energy and resilience management in hardware or system software may affect the performance of existing parallel programming environments.
The topics explored by this minisymposium include but are not limited to:
- Parallel programming abstractions for energy and resilience
- Compiler and runtime support for energy- and resilience-aware execution
- Parallel libraries and auto-tuning for energy and resilience
- Performance of parallel programming languages, libraries and tools in energy-constrained environments
- Measurement and characterisation of resilience in parallel programming models, languages and libraries
- Measurement and characterisation of energy in parallel programming models, languages and libraries
- New computing paradigms for improving energy and resilience in parallel programming, such as approximate computing or near-threshold computing
Submission deadline: June 30, 2015
Notification of acceptance: July 17, 2015
Mini Symposium: September 3-4, 2015