基于超大点数FFT优化算法的研究与实现

针对应用系统对超大点数快速傅里叶变换(FFT)的性能需求不断提升，以及现有处理平台的资源对实现超大点数FFT的制约问题，该文提出一种超大点数FFT的实现方法。该方法通过优化铰链因子存储，采用行列号方式访问2维矩阵避免了3次显性转置，从而节省了内存资源；同时，通过分析处理器的分级存储结构特点，优化了矩阵行列划分规则，进而提高了行列访问效率。实验结果表明，该方法节约了近一半的内存资源，且有效提高了超大点数FFT的执行速度。

Research and Implementation of Ultra Large Sequences FFT Optimized Algorithm

For the increasing requirements of ultra long sequences FFTs efficient implementation in application systems, an implementation method is proposed to realize ultra long sequences FFT. The method can solve resources limitation of existed processing platforms, which can save mass memory resources by optimizing the twiddle factor store and adopting the way of rows and columns instead of three explicitly matrix transposition to access the two-dimensional matrix. The characteristics of processors hierarchical memory structure is analysed to optimize the rules of matrix partition, which improves the access efficiency. Experiment results show that the performance of execution efficiency of ultra long sequences FFT is improved nearly half of the memory resources.