Performance Evaluation of Parallel Programming Models for K-means Clustering on Parallel Computing Systems
Year: 2026
Type: conference-paper
Abstract: High-performance computing and parallel algorithms provide an opportunity to quickly solve timeconsuming problems in digital processing of large amounts of data. In this research work, the K-means algorithm was implemented in parallel in a single-node parallel computing environment using various parallel programming technologies. A hybrid model that can be used in combination with Open Multi-Processing Message Passing Interface for central processor-based systems, as well as Open Computing Language and Compute Unified Device Architecture technologies that enable parallel programming for graphics processor-based systems, and their efficiency in maximizing the use of computing resources were analyzed in depth. The results of the study showed that the central processor-based models showed efficiency in the range of ~78-84\%. The graphics processorbased Compute Unified Device Architecture model achieved ~94\% speedup, and the Open Computing Language model achieved $\boldsymbol{\sim} \mathbf{9 1} \boldsymbol{\%}$ speedup and gave significantly higher results than the central processor-based models. At the end of the study, the hybrid model on central processor-based systems and the Compute Unified Device Architecture parallel programming model on graphics processor-based systems provided more effective speedup under these experimental conditions.
Institution: Tashkent University of Information Technology
Language: English
Subfield: Hardware and Architecture
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7202251991
