摘要
针对使用CPU统计加速器能谱图的过程中,消耗时间过长的问题,给出了一种完全由GPU实现粒子能量统计、最大值bin的查找、绘制能谱图的方法,可以降低CPU的负担,发挥GPU的效率,解决了CPU和GPU之间数据传输的瓶颈问题。实验结果表明,与只使用CPU和只使用GPU进行数据统计而不进行结果显示的2种方案相比,在处理大量粒子数据时,可以获得80倍以上的加速比。
Aiming to solve the problem encountered in the process of CPU- based spectroscopy generation,which usually consumes too much time,in this paper,a method fully based on GPU is proposed,in which GPU can alone implement particles statistics,seek maximum bin,and display spectroscopy on screen. By doing so,burden of CPU is decreased and efficiency of GPU is increased,the bottleneck of data delivering between CPU and GPU is overcomed. The experimental results indicate that,compared with the scheme which only uses CPU and other scheme performing statistics by GPU,this kind of scheme can achieve 80 times speedup when processing large amount of particle data.
出处
《核电子学与探测技术》
CAS
北大核心
2016年第1期52-55,共4页
Nuclear Electronics & Detection Technology
基金
国家自然科学基金(U1232123)
甘肃省高等学校科研项目(2013B-100)
关键词
能谱图
互操作
计算设备统一架构
并行计算
开源图像库
spectroscopy
interoperability
compute unified device architecture（CUDA）
parallel algorithm
OpenGL
