Survey on GPGPU and CUDA Unified Memory Research Status

In the context of big data, the rapid advancement of fields such as scientific computing and artificial inteliigence, there is an increasing demand for highcomputaional power across various domains. Theunique hardware architecture of the Graphics Processing Unit (GPU) makes it suitable for parallel computing. In recent years, the concurrent development of GPUs and iields such as ariiiicial inteliigence and scientiiic computing has enhanced GPU capabiiities, leading to the emergence of mature General-Purpose Graphics Processing Units (GPGPUs). Currently, GPGPUs are one of the most important co-processors for Central Processing Units (CPUs). However, the ixed hardware coniguration of the GPU after deiivery and its iimited memory capacity can signiiicantly hinder its performance, pariicularly when deaiing with large datasets. To address this issue, Compute Uniied Device Architecture (CUDA) 6. 0 introduces uniiied memory, allowing GPGPU and CPU to share a virtual memory space, thereby simpiifying heterogeneous programming and expanding the GPGPU-accessible memory space. Uniiied memory offers a solution for processing large datasets on GPGPUs and alleviates the constraints of iimited GPGPU memory capacity. However, the use of uniiied memory introduces performance issues. Effective datamanagement within uniiied memory is the key to enhancing performance. This article provides an overview of the development and appiication of CUDA uniiied memory. It covers topics such as the features and evolution of uniiied memory, its advantages and iimitafions, its appiications in ariicial inteliigence and big data processing systems, and its prospects. This aricle provides a valuable reference for future work on applying and optimizing CUDA uniied memory.