摘要
提出一种基于图形处理器(GPU)的快速傅里叶变换算法和并行处理算法,给出了傅里叶变换分析法的GPU并行实现流程,通过对各像素点进行并行处理,使得处理速度大大加快。GPU和计算机的处理精度和运算速度计算结果表明:GPU能在不降低处理精度的前提下,保证其效率得到很大提升。
The conventional time--consuming Fourier transform analysis algorithm was described as well as a parallel data processing algorithm which used GPU in the real--time surface measure- ment. This parallel algorithm adopted GPU to deal with a series of image pixels simultaneously, so the procedure of processing data gets got to speed up. By analyzing the results calculated by GPU and cen- tral processing unit,we find that the precision of the results computed by GPU do not decrease while the processing efficiency is highly increased.
出处
《中国机械工程》
EI
CAS
CSCD
北大核心
2013年第6期791-794,共4页
China Mechanical Engineering
基金
国家重点基础研究发展计划(973计划)资助项目(2011CB706505)
机械设计及理论浙江省重中之重学科和浙江理工大学重点实验室开放基金资助项目(ZSTUMD2011A004)
关键词
表面形貌测量
白光谱线扫描干涉法
GPU
CUDA
surface measurement
wavelength scanning interferometry
graphics processing unit (GPU)
compute unified device architecture(CUDA)