The application of energy-saving technologies in parallel computing systems

Abstract

This scientific paper provides a detailed analysis of the architecture of parallel computing systems, their operational principles, and methods for organizing parallel processing. In particular, it investigates the dynamic performance and energy consumption of the OpenBLAS DGEMM function using software tools and applications designed to evaluate energy usage on multi-core processors. Throughout the study, effective methods for ensuring energy efficiency are developed, and mathematical models for predicting energy consumption are applied to assess and optimize the energy costs associated with computational workloads. Additionally, the paper analyzes the performance metrics of high-efficiency parallel computing blocks and their impact on energy efficiency. Based on the research findings, practical recommendations are proposed to reduce energy consumption in parallel systems without compromising computational performance.