Impact of Data-Oriented and Object-Oriented Design on Performance and Cache Utilization with Artificial Intelligence Algorithms in Multi-Threaded CPUs
By: Gabriel M. Arantes, Richard F. Pinto, Bruno L. Dalmazo, Eduardo N. Borges, Giancarlo Lucca, Viviane L. D. de Mattos, Fabian C. Cardoso, Rafael A. Berri
Published: 2025-12-09
View on arXiv →#cs.AI
Abstract
This study provides a comprehensive performance analysis of Data Oriented Design (DOD) versus traditional Object-Oriented Design (OOD), focusing on cache utilization and efficiency in multi-threaded environments for AI algorithms. It highlights DOD's architectural superiority for maximizing hardware efficiency in complex, large-scale AI and parallel computing tasks.