ABSTRACTS OF ARTICLES OF THE JOURNAL "INFORMATION TECHNOLOGIES".
No. 11. Vol. 25. 2019

DOI: 10.17587/it.25.682-690

A. Yu. Popov, Ph. D., Associate Professor of the Department "Computer Systems and Networks", e-mail: alexpopov@bmstu.ru, Bauman Moscow State Technical University, Moscow

Application of a Heterogeneous Computing System with a Discrete Mathematics Instruction Set to Solve Large-Scale Graphs Problems

The interest to the complex data models analysis has grown significantly in the last decade. Graphs known as the most adequate form of the real data representing in many areas from social networks, computer programs structure, topology of integrated circuits, to the bioinformatics and many others. As the size of datasets increases the need to find more efficient methods and tools for analyzing large graphs, including on the basis of more advanced hardware technical solutions, becomes obvious. Modern GPUs have great parallelism and performance, but cannot solve the fundamental problems of the graph processing: the data dependencies problem, the distribution of irregular graphs computing workload on multiple processor devices, as well as conflicts with memory access for many cores. We introduce a specialized microprocessor Leonhard x64 and the Discrete mathematics Instruction Set Computer (DISC) for large dimension graph processing, which was developed in the Bauman University. This article provides information about the discrete mathematics operators and the corresponding Leonhard x64 instructions, and justifies the architecture principles of a heterogeneous computing system based on it. With examples of Dijkstra and Belman-Ford algorithms we show the difference between the implementation of graph processing and the efficiency of universal computing systems, heterogeneous computing systems based on GPGPU, as well as Leonhard x64 microprocessor-based system.

Keywords: discrete mathematics instruction set computer, Leonhard microprocessor, graph, data structure, GPGPU

P. 682–690

To the contents