一种基于Kepler架构GPU的通信仿真加速方法

发布时间：2019-03-26 17:07

【摘要】：提出了一种在Kepler架构GPU(graphics processing unit,图形处理器)上利用CUDA(compute unified device architecture,统一计算设备架构)技术加速通信仿真中DFT(discrete Fourier transform,离散傅里叶变换)处理过程的方法。该方法的核心思想是利用线程级并行技术实现单条收发链路内部DFT运算的并行加速,并利用动态并行和Hyper-Q技术实现不同收发用户对之间链路处理过程的并行加速,从而最终达到加速仿真中DFT处理过程的目的。实验结果表明,相对单核单线程CPU程序和上一代Fermi架构GPU程序,该方法分别能够将DFT处理速度提升300倍和3倍,具有较好的加速效果。
[Abstract]:This paper presents a method to accelerate the process of DFT (discrete Fourier transform, discrete Fourier transform (DFT) in communication simulation using CUDA (compute unified device architecture, Unified Computing equipment Architecture (CUDA (compute unified device architecture, Unified Computing equipment Architecture) based on Kepler architecture GPU (graphics processing unit, graphics processor (GPU). The core idea of this method is to use thread-level parallel technology to realize the parallel acceleration of DFT operation in a single transceiver link, and to use dynamic parallel and Hyper-Q technology to realize the parallel acceleration of the process of link processing between different transceiver pairs. Finally, the purpose of accelerating the process of DFT processing in simulation is achieved. The experimental results show that compared with the single-core and single-thread CPU program and the previous generation Fermi architecture GPU program, this method can increase the processing speed of DFT 300 times and 3 times, respectively, and has a good acceleration effect.
【作者单位】：中国信息通信研究院;北京邮电大学;
【基金】：国家科技重大专项基金资助项目(No.2014ZX03003011-003)~~
【分类号】：TP332

【相似文献】