基于图形处理器(GPU)的通用计算.pdfVIP

基于图形处理器(GPU)的通用计算 吴恩华1), 2) 柳有权1) 1)中国科学院软件研究所计算机科学重点实验室 北京 100080 2)澳门大学科学技术学院电脑与资讯科学系 澳门 ehwu@umac.mo, lyq@ 摘 要 伴随着PC 级微机的崛起和普及, 多年来计算机图形的大部分应用发生了从工作站向微机 的大转移, 而这种转移甚至于发生在象虚拟现实﹑计算机仿真这样的实时( 中﹑小规模)应用中。 这一切的发生从很大程度上源自于图形处理硬件的发展和革新。近年来伴随着图形处理器(GPU) 性能的大幅度提高以及可编程特性的发展，人们首先开始将图形流水线的某些处理阶段以及某些 图形算法从CPU 向 GPU 转移。而不可思议的是，除了计算机图形学本身的应用，涉及到其它 领域的计算，以至于通用计算近2-3 年来成为GPU 的应用之一，并成为研究热点。本文将从若 干图形硬件发展的历史开始，介绍和分析最新GPU 在通用计算方面的应用及其技术原理和发展 状况，并结合我们自己的实践讨论和探索其发展前景。 关键词 图形硬件, 图形处理器(GPU), 可编程性, 实时计算 General Purpose Computation on GPU Enhua Wu1), 2) Youquan Liu1) 1) Laboratory of Computer Science, Institute of Software Chinese Academy of Sciences, Beijing, 100080 2) Department of Computer and Information Science, Faculty of Science and Technology University of Macau, Macao Abstract Along with the rapid development of personal computers, various applications in computer graphics, even real-time applications in middle or small scale such as those in virtual reality and computer simulation, are migrating from workstations to PCs. To a large extent no doubt, the migration comes from the innovation of graphics hardware. For many years, with the advancement of Graphics Processing Unit(GPU) and the creation of its new feature of programmability, scientists begin to transfer some of the processing stages in the graphics output pipeline or some graphics algorithms from CPU to GPU. To our surprise, except from those applications in graphics itself, GPU finds its applications in general purpose computations in other fields, and it comes up as a hot topic for research in recent 2-3 ye