The Procedure for Implementing the Operation of Multiplying Two Matrices Using the Residual Number System

Abstract

The report considers solution to the problem of improving the speed implementation of the operation of multiplying two square matrices of the same dimension. To carry out calculations and comparative analysis of the speed of the multiplication operation, we consider a computer system (CS) in the positional binary number system (PNS) and in the non-positional number system in the residual classes (the residual number system - RNS). A comparative analysis of the performance of the CS was carried out with the same characteristics of the computing system: equal lengths of bit grids, the same command systems, the same methods of addressing operands and instructions, the same clock speed of the processor, the equal number of program commands, etc. When calculating the speed of the matrix multiplication operation, the fastest data processing method in RNS was used, based on the tabular principle.