@article{ART003348229},
author={Kim Min-Soo and Kim Jae Woong and Jang Mi-young},
title={Analysis of Portability Performance of Large-Scale Grid Processing Based on GPU Memory Architecture Differences},
journal={ Journal of Software Forensics},
issn={3092-541X},
year={2026},
volume={22},
number={2},
pages={143-154},
doi={10.29056/jsf.2026.06.13}
TY - JOUR
AU - Kim Min-Soo
AU - Kim Jae Woong
AU - Jang Mi-young
TI - Analysis of Portability Performance of Large-Scale Grid Processing Based on GPU Memory Architecture Differences
JO - Journal of Software Forensics
PY - 2026
VL - 22
IS - 2
PB - Korea Software Assessment and Valuation Society
SP - 143
EP - 154
SN - 3092-541X
AB - This paper analyses the performance characteristics observed when porting large-scale grid-resampling CUDA code from a PC discrete-GPU environment to a Jetson AGX Orin-based UMA GPU environment.
In a PC environment, CPU memory and dedicated GPU memory are physically separated, whereas the Jetson shares physical memory between the CPU and GPU. Therefore, the same memory method does not necessarily produce equivalent performance across the two platforms. Explicit-copy memory (Device Memory), Unified Memory, and Zero-copy were evaluated under linear, local-offset, and scatter access patterns. The results showed that Device Memory provided the most stable performance on the PC. On the Jetson, Unified Memory and Zero-copy showed potential advantages depending on data size and memory-management conditions. Scatter access degraded performance on both platforms, indicating that both memory architecture and data-access patterns should be considered when porting CUDA code.
KW - GPU Memory Architecture;CUDA Memory Management;Zero-copy;Large-scale Grid Processing;Portability Performance Analysis
DO - 10.29056/jsf.2026.06.13
ER -
Kim Min-Soo, Kim Jae Woong and Jang Mi-young. (2026). Analysis of Portability Performance of Large-Scale Grid Processing Based on GPU Memory Architecture Differences. Journal of Software Forensics, 22(2), 143-154.
Kim Min-Soo, Kim Jae Woong and Jang Mi-young. 2026, "Analysis of Portability Performance of Large-Scale Grid Processing Based on GPU Memory Architecture Differences", Journal of Software Forensics, vol.22, no.2 pp.143-154. Available from: doi:10.29056/jsf.2026.06.13
Kim Min-Soo, Kim Jae Woong, Jang Mi-young "Analysis of Portability Performance of Large-Scale Grid Processing Based on GPU Memory Architecture Differences" Journal of Software Forensics 22.2 pp.143-154 (2026) : 143.
Kim Min-Soo, Kim Jae Woong, Jang Mi-young. Analysis of Portability Performance of Large-Scale Grid Processing Based on GPU Memory Architecture Differences. 2026; 22(2), 143-154. Available from: doi:10.29056/jsf.2026.06.13
Kim Min-Soo, Kim Jae Woong and Jang Mi-young. "Analysis of Portability Performance of Large-Scale Grid Processing Based on GPU Memory Architecture Differences" Journal of Software Forensics 22, no.2 (2026) : 143-154.doi: 10.29056/jsf.2026.06.13
Kim Min-Soo; Kim Jae Woong; Jang Mi-young. Analysis of Portability Performance of Large-Scale Grid Processing Based on GPU Memory Architecture Differences. Journal of Software Forensics, 22(2), 143-154. doi: 10.29056/jsf.2026.06.13
Kim Min-Soo; Kim Jae Woong; Jang Mi-young. Analysis of Portability Performance of Large-Scale Grid Processing Based on GPU Memory Architecture Differences. Journal of Software Forensics. 2026; 22(2) 143-154. doi: 10.29056/jsf.2026.06.13
Kim Min-Soo, Kim Jae Woong, Jang Mi-young. Analysis of Portability Performance of Large-Scale Grid Processing Based on GPU Memory Architecture Differences. 2026; 22(2), 143-154. Available from: doi:10.29056/jsf.2026.06.13
Kim Min-Soo, Kim Jae Woong and Jang Mi-young. "Analysis of Portability Performance of Large-Scale Grid Processing Based on GPU Memory Architecture Differences" Journal of Software Forensics 22, no.2 (2026) : 143-154.doi: 10.29056/jsf.2026.06.13
