An Introduction to the MISD Technology

Abstract

The growth of data volume, velocity, and variety will be the global IT challenges in the next decade. To overcome performance limits, the most effective innovations such as cognitive computing, GPU, FPGA acceleration, and heterogeneous computing have to be integrated with the traditional microprocessor technology. As the fundamental part of most computational challenges, the discrete mathematics should be supported both by the computer hardware and software. But for now, the optimization methods on graphs and big data sets are generally based on software technology, while hardware support is promising to give a better result. \ \ In this paper, the new computing technology with direct hardware support of discrete mathematic functions is presented. The new non-Von Neumann microprocessor named Structure Processing Unit (SPU) to perform operations over large data sets, data structures, and graphs was developed and verified in Bauman Moscow State Technical University. The basic principles of SPU implementation in the computer system with multiple instruction and single data stream (MISD) are presented. We then introduce the programming techniques for such a system with CPU and SPU included. The experimental results and performance tests for the universal MISD computer are shown.