Integrating VLIW Processors with a Network on Chip

Abstract

Networks are a becoming a necessity to easily integrate multiple processors on a single chip. A crucial question here is whether it is good enough to reason about statistical performance as opposed to hard real-time performance constraints. Today’s processors often do not allow software design for hard real-time systems, caused by the design of the bus- and/or memory interfaces, thereby necessitating elaborate performance analysis through simulation.

In this presentation I will indicate what options a processor designer has, using Silicon Hive processor design tools, in specifying the interfaces and local memory sub-system in a processor. It allows a multitude of communication options to build either type of system: statistically bound or hard real-time bound performance.

Additionaly I will describe the multi-processor simulation and prototyping environment and touching on the processor design methodology.