ANALYSIS OF THE EFFECTIVENESS OF APPLYING STRUCTURAL-PARAMETRIC SYNTHESIS METHODS FOR THE DEVELOPMENT OF HETEROGENEOUS COMPUTING SYSTEMS

Abstract

The paper investigates the possibilities of applying various methodological approaches to the construction of heterogeneous computing systems (HCS) taking into account the required quality indicators. The possibility of applying structural-parametric synthesis (SPS) methods in the development of HCS is considered. The idea is to describe the HCS as a structure with corresponding parameters and constraints, followed by solving an optimization problem. To solve the synthesis problem, the effectiveness of applying classical and combinatorial methods, morphological analysis, evolutionary and metaheuristic algorithms is analyzed. Objective: to analyze methods and tools for structural-parametric synthesis of heterogeneous computing systems that ensure the selection of rational structures and parameters while accounting for multi-criteria requirements. Results: A classification of HCS by scale, node composition, computation organization, and application domain is proposed; a formalized description of the main types of computing nodes (CPU, GPU, FPGA, ASIC) is provided; a mathematical formulation of the SPS problem with a graph model of the HCS, a queuing system model, and formalization of quality criteria is developed. Practical relevance: lies in determining the possibility of effective application of SPS methods for the development of HCS.