| Title |
Analysis of CUDA Memory Usage Patterns using Open Source in a Single GPU Environment |
| DOI |
https://doi.org/10.5573/ieie.2025.62.2.31 |
| Keywords |
GPU; CUDA; Memory allocation; Open GPU kernel modules |
| Abstract |
With the growth of the data industry, computing technologies related to it are gaining significant attention. Among these, CUDA stands out as a parallel computing platform designed to maximize GPU performance, where memory optimization is essential for efficient memory usage and performance enhancement. This study aims to analyze CUDA memory usage patterns in a single GPU environment by leveraging NVIDIA's open GPU kernel modules, in an effort to find ways to improve memory efficiency. Through this, we aim to identify GPU memory allocation and deallocation patterns, as well as continuous page access patterns, to enhance the performance of CUDA-based applications. |