Analysis of Multicore Neural Processing Structures Based on Hardware and Software Co-Design

The paper suggests analysis of multicore neuroprocessing structures based on co-design of hardware and software. The development of the task is achieved by using of set-theoretic analyses of all internal algorithms with a goal to open up in a step of the compilation independently of equivalence classes in the type of clusters. As a result, it solves a problem of opening up the set of all possible cluster structures, also it defines the analysis of explicit parallelism and used for choosing the MCNPS for designing a strategy for CPS realization. The suggested co-design is based on the formalization of one of the type models of knowledge, which is a production form for expert system design on one or more technical parameters. As an example of the developed methodology, an analysis of algorithms for monitoring natural and anthropogenic emergencies or forest-fire monitoring is proposed and the expert system user interface and intellectual telecommunications structures on the base module MC 127.04 in base SBIS 1879BM8Y neuroprocessor is described.