• Jaidev Singh
• Nandini Agrawal
• Prajwal
• Dr. Jayanthi P N

Optimizing CPU and GPU performance is critical for improving computational efficiency and reducing energy consumption in modern computing systems. Traditional resource allocation and workload scheduling techniques primarily rely on static or heuristic-based approaches, which often fail to adapt to dynamically changing workloads. This paper presents a novel machine learning (ML)-based approach that integrates Reinforcement Learning (RL) and XGBoost to optimize CPU and GPU resource allocation. The proposed RL model learns an optimal scheduling strategy by interacting with the system and generating a dataset containing key performance metrics such as CPU/GPU utilization, execution time, and energy consumption. The collected dataset is then utilized by an XGBoost model to predict the most efficient workload distribution based on reward values, enabling real-time optimization of computational resources. To validate the effectiveness of our approach, we conduct extensive simulations using real-world workload traces from SPEC CPU 2017, PARSEC, and Google TPU workloads. Performance evaluations demonstrate that our ML-driven resource allocation method significantly outperforms conventional static scheduling algorithms in terms of execution time, energy efficiency, and overall system throughput. The proposed method adapts dynamically to workload variations, leading to optimized power consumption without compromising performance. This research provides a scalable and adaptable solution for intelligent resource management in heterogeneous computing environments, making it applicable to high-performance computing (HPC), cloud computing, and embedded systems.

CPU-GPU Optimization, Reinforcement Learning, XGBoost, Resource Allocation, Machine Learning, Performance Efficiency.

"ML - Based CPU and GPU utilization", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.12, Issue 2, page no.h320-h328, February-2025, Available :http://www.jetir.org/papers/JETIR2502741.pdf

"ML - Based CPU and GPU utilization", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.12, Issue 2, page no. pph320-h328, February-2025, Available at : http://www.jetir.org/papers/JETIR2502741.pdf