Session 1 : Heterogeneous Multi/Manycore Processing for Energy-Efficient Execution
Organizers : Abdoulaye Gamatie - Abdoulaye.Gamatie [at] lirmm.fr , Fernando Magno Quintao Pereira - fernando.quintao-pereira [at] lirmm.fr
Heterogeneous Multi/Manycore Processing (HMP) is nowadays an important trend for addressing the energy- efficiency challenge faced in both embedded and high-performance computing domains. In the last decade, CPUs and accelerators such as GPUs and FPGAs have been massively mixed in modern architectures. Mopre recently, another heterogeneous architecture paradigm has emerged under the name of ARM big.LITTLE technology, in which high-performance (big) cores suitable for heavy workloads are combined with low- power (LITTLE) cores dedicated to lightweight tasks. Dynamic runtime systems are in charge of allocat- ing/scheduling workloads on the different cores. For this purpuse, they typically apply frequency/voltage scaling to minimize the dissipated power whenever possible.
HMP brings to both industry and academia a number of design challenges along different directions such as programmability, runtime environments, operating systems and computer architectures. Given this context, the motivation of this special session is to bring together researchers working on heterogeneous multicore/manycore architectures and workload management for energy-efficient systems, providing them with a unique opportunity to discuss recent progress on these topics.
Key topics include, but are not limited to:
- programming models
- compilation techniques for heterogeneous embedded and high-performance systems
- adaptive runtime management
- machine learning techniques for intelligent system management - heterogeneous multicore architectures
- memory and communication technologies for HMP
- power consumption modeling and management
- energy-efficient design
- applications and case studies on HMP
Session 2 : Adaptive Processor Architectures
Organizers: Michael Huebner - Michael.Huebner [at] ruhr-uni-bochum.de , Stephan Wong - J.S.S.M.Wong [at] tudelft.nl
Adaptive processors are different from their static counterparts, which are optimized for either performance, flexibility or power/energy consumption. This means a tradeoff must be made between the mentioned three axes. Moreover, security and fault-tolerance are topics that are gaining importance in recent years and cannot be covered by static processors if their design had not taken these into account. Adaptive processors are moving away from design-time adaptivity towards run-time adaptivity to adapt towards dynamic environments that computing platforms are increasingly more placed in. In this special session, we address issues encountered in designing such processors from technology to architectures to compilers. Furthermore, we invite contributions that describe application domains that require the use of (dynamic) adaptive processors.
Key topics include, but are not limited to Adaptive Processor Architectures for:
- image processing
- industry automation
- real-time processing
- computing in space
- industrial imaging