知识卡片
FFTW规划器试选计算分解
FFTW针对同一规模需要重复计算的DFT,先以试运行选择该机器上合适的FFT分解,再复用计划执行变换。计划还能利用缓存与小规模直线代码;这种选择与机器环境有关。
这张卡片看不懂?
这张卡片的反馈已满 10 条,暂时无法继续提交。
已有反馈 · 0 / 10 条
还没有人提交反馈。
在整书笔记中的位置
原书出处 · 1 处
- 第30章 多项式与快速傅里叶变换 · 查看原书依据
Frigo and Johnson [161] developed a fast and flexible implementation of the FFT, called FFTW (“fastest Fourier transform in the West”). FFTW is designed for situations requiring multiple DFT computations on the same problem size. Before actually computing the DFTs, FFTW executes a “planner,” which, by a series of trial runs, determines how best to decompose the FFT computation for the given problem size on the host machine. FFTW adapts to use the hardware cache efficiently, and once subproblems are small enough,
参考来源
- 位置:《算法导论(第4版)》第30章《多项式与快速傅里叶变换》(OEBPS/chapter030.xhtml,文本字符[40000,44125))。
- 结论依据:原文说明位逆序、FFT电路、精确变换或数值稳定性;相应图和公式已核对。
- 原始内容:Frigo and Johnson [ 161 ] developed a fast and flexible implementation of the FFT, called FFTW (“fastest Fourier transform in the West”). FFTW is designed for situations requiring multiple DFT computations on the same problem size. Before actually computing the DFTs, FFTW executes a “planner,” which, by a series of trial runs, determines how best to decompose the FFT computation for the given problem size on the host machine. FFTW adapts to use the hardware cache efficiently, and once subproblems are small enough,