Ecosystems for the development of multi-core and many-core SoC models

Multi-core and many-core Systems-on-Chip (SoC) are growing more complex than ever. Consequently, developing system models for such SoCs to guide and validate architectural and implementation decisions is becoming a daunting task. It consumes a huge amount of time and effort just to get the model up and running. Although these system models can be fairly abstracted, they still require the setup of a complicated platform to model a homogeneous or a heterogeneous mix of processing cores, a network-on-chip, cache memories, input-output interfaces as well as several other functional units. The abstraction levels for the individual constituent models can be represented at different abstraction levels forcing the need for special interfacing and simulation tools. Moreover, unified benchmarks are needed to be able to have a fair comparison between different architectures and different implementation options while at the same time target the specific application domain the SoC is being developed for. This paper explores the availability of such ecosystems and discusses the important features and capabilities that need to be provided in them. The level of abstraction and the software support in this model is also examined. © 2009 IEEE.