Mattia Petroni (2011) Gene regulatory networks as dataflow processing platforms. EngD thesis.
The research and study on pure dataflow (DF) computer architectures has almost reached a dead point since early '90, when several technological issues became insurmountable. Although, the pioneering work of J.B. Dennis at MIT in early seventies, showed to engineers all the potential of the fine grained parallelism, implicit in dataflow computing, applied to several prototype machines, still many issues inside architecture model remain partially unsolved. Because of increasing complexity which grew in solving these issues, the dataflow architecture became quickly economically disadvantageous in comparison to von Neumann architecture, so it remained alive only as laboratory project. With the recently works of many computer engineers and biotechnologists, which use biological systems, precisely gene regulatory networks (GRNs), to build logic circuits, a new point of view arises concerning this technology and dataflow computation. Several properties of GRNs allow to view the memory system and the information processing of dynamic dataflow architecture in entirely different way, so that some constraints may fall. Although these properties may improve solutions to the main dataflow architecture issues, several technological (biochemical) problems remain to solve. The work presents a general review of the most studied models of dataflow architectures and their analogies with GRNs. At the end, a few possible implementations of fundamental constructs of dataflow computing using GRNs are presented.
Actions (login required)