Session 5: Dynamic architectures & adaptive management for image & signal processing

Abstract

Nowadays applications require a high level of flexibility (adaptive scheduling and mapping in multi-core architectures, data driven execution, various operating modes, application scenarios, adaptation to environment, dynamic voltage and frequency scaling, workload balance…), with high constraints in terms of performances and low-power requirements. This poses the need for novel and efficient adaptive architectures and runtime management techniques. Partial and dynamical reconfiguration is supported in hardware for a decade, with different granularity, but still mainly relies on external intervention. Run-time management paradigms must now be defined, taking into account the dynamic evolution of applications. The special session contributions address the following design challenges: • Novel hardware architectures supporting dynamic management; • Software-/middleware-based dynamic management algorithms; • Methodologies for design and management of dynamically re-configurable systems.