Research on Multi-GPUs Image Processing Acceleration Based CUDA

摘要：

Multiple-Graphics Processing Units (Multi-GPUs) means using two or more display adapters in the same PC to speed up the graphics animation for gamers. In this paper,we provide a way of using multi-GPUs in image processing acceleration based on NVIDIA CUDA technique that makes the processing progress more flexible and controllable. For demonstration we deal with canny edge detection using GTX295,which with the power of two GeForce GTX 200 GPUs on a single card,we show how to use multi-GPUs when accelerating image processing. Also,we did some optimization work. We found out that in image processing which contains mass data and complex algorithm,by using multi-GPUs the processing time almost can be reduced to half;however in the case have less data and simple algorithm,the advantage of multi-GPUs is not obvious even takes more time.

关键词： Multi-GPUs CUDA Image Processing Optimization

作者: Gao Song Gao Biao Xiao Qinkun Wang haiyun

作者单位: School of Electronic Information Engineering,Xian Technological University,Xian,710032,China

会议类型: 国际会议

会议名称: 2011 International Conference on Opto-Electronics Engineering and Information Science(2011光电电子工程与信息科学国际会议 ICOEIS 2011)

会议地点: 西安

会议语种:英文

页码: 434-437

在线出版日期: 2011-12-23（万方平台首次上网日期，不代表论文的发表时间）