Micronet: An operational model for System Level Intergration
Micronet
Micronet[1][2] is a network of entities which operate concurrently and communicate asynchronously. The entities can be one of the following types: computational elements (usually digital), sensors (Analogue/Mixed Signal) or actuators (MEMS).
Micronets present a uniform model at the different levels of abstraction, i.e. a fractal model of system design: network of sub-systems, each sub-system, in turn, is a network of functional units, down to network of transistors
Control is layered: threads, compounds, instructions, microoperations, datapath control, handshaking protocol, signal transitions.
Micronets distribute control locally - behaviour can be decomposed to run on architectural clusters tailored with the optimal mix of computational elements.
Micronets make a clean separation between computation and communication, and, between behaviour and timing, which leads to a compositional design style.
The entities can be either clocked or self-timed; implementations using the latter exhibit lower power consumption[11] and better EMI characteristics.
Micronet-based Asynchronous Processor (MAP) have been designed for scalar[1], supers calar[5][6], VLIW[4] and multithreaded architectures[7] together with their parallelising compilers[12].
Behaviour-Achitecture Co-design: It is our belief that future Integration Platforms will be overwhelmingly programmable, consisting of soft-programmable forms - as instruction set architectures, and hard-programmable forms - realised as field programmable logic. The Integration Platform is composed of networks (micronets) of heterogeneous computational entities that operate in a multi-threaded fashion. The Application is composed of behavioural blocks - some pre-defined such as communication protocols; others, are more specific to the application. The first step in the co-design is to recognise concurrent operations and optimise communications at different levels and map them to the platform. The second step is to explore the trade-off between programmability (both soft and hard), and performance (in terms of speed and energy consumption) of the applications executing on the Integration Platforms.
Power point slides of a recent lecture (Feb. 2002)
Selected Publications
DK Arvind.