The focus is on doing the computation in-place and using as few multiplies as possible. (That would be the mid- to late-1980s.) Here's a typical textbook sample from the 1989 edition of Discrete-Time Signal Processing by Oppenheim and Schafer): But it makes me think back to all the FFT code that I saw back when I was first learning digital signal processing. Like Cleve, I really love this little bit of code. Then ffttx does a litte extra computation at the end to produce the final result. Each recursive call is on half of the original input values. Here's the key code fragment of Cleve's textbook function ffttx: % Recursive divide and conquerÄŻunction ffttx calls itself recursively, twice. Cleve described a very compact, elegant implementation of the divide-and-conquer idea that's central to fast Fourier transform algorithms. I have three little sort-of stories to tell, the first of which was inspired directly by Cleve's post. The Slowest Way (Ever?) to Fill a Vector with NaNs.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |