论文部分内容阅读
通过模拟数据和实际资料处理的结果来看,时间S变换作为一种近似计算具有良好的时频特性,它不仅能提供付里叶变换所不具备的时间分辨率,而且还能克服频率S变换在滤波时的边缘效应.为了提高工作效率,缩短计算周期,本文研究实现了基于GPU的时间S变换,并对炮集记录使用时间S变换来压制面波.实验表明,时间S变换的GPU/CPU协同计算模式具有很好的加速效果,正变换使用全局内存加速比峰值可达一百三十多倍,使用零拷贝内存时达一百八十多倍,反变换使用全局内存达十四倍,而零拷贝内存达一百三十五倍.时间S变换优良的时频特性借助GPU的高性能加速,使得时间S变换在地震资料的时频分析中将有广阔的应用前景.
Through the simulation data and actual data processing results, the time S transform has good time-frequency characteristics as an approximate calculation, it can not only provide the time resolution not available Fourier transform, but also overcome the frequency S transform In order to improve the work efficiency and shorten the calculation period, this paper studies the realization of the GPU-based time S transform, and uses the time transform S to compress the surface wave.The experiment shows that the time S transform GPU / CPU collaborative computing mode has a good acceleration effect, is the use of global memory acceleration peak up to more than one hundred and thirty times, using zero copy memory up to more than one hundred and eighty times, the inverse transform using global memory up to four times , And zero copy memory up to 135 times the time S transform excellent time-frequency characteristics with the GPU’s high-performance acceleration, making the time S transform seismic data in time-frequency analysis will have broad application prospects.