基于OpenCL的显卡加速射线能谱数据处理  

ACCELERATED SPECTROSCOPY DATA PROCESSING WITH GPU BASE ON OPENCL

作　　者：刘端[1] 葛良全[1] 张庆贤[1] 谷懿[1] 

机构地区：[1]成都理工大学,成都610059

出　　处：《物探化探计算技术》2013年第2期242-245,122,共4页Computing Techniques For Geophysical and Geochemical Exploration

基　　金：国家自然科学基金项目(41074093)

摘　　要：极大似然算法在核探测谱线中能够进行全谱解析,可以有效分解能量差为2/3FWHM重叠峰,但是由于该算法计算量大、耗时多而制约实际应用。这里采用OpenCL(Open Compu-ting Language,开放运算语言)平台,在异构系统中利用GPU的并行运算能力,实现RL算法的并行计算。通过测试,在特定平台下该算法可以有效地提高运行效率,运行时间为原有的1/30,提高了该算法的实际可行性。The maximum likelihood(ML) solution is used to resolve the full spectrum of nuclear detection,which can be used to decompose the overlapping peaks with the energy difference of 2/3 FWHM.However,the practical application of the algorithm is,due to high and time-consuming,restricted.In this paper,the authors used the OpenCL platform in a heterogeneous system,to achieve the parallel computation of the RL algorithm by the parallel computing capabilities of GPU.The results show that the method can effectively improve operational efficiency.The running time is thirtieth of the original one.

关 键 词：极大似然估计法 OPENCL 并行计算 射线能谱 

分 类 号：O242.2[理学—计算数学]