Colin Percival wrote:
At 21:46 +0100, 29 Dec 2003, Matthias Waldhauer wrote:

During my studies of FFT+NTT I got following idea:

In case of FFT we have to provide enough bits per word to safely hold
the squares of the values in frequency space without getting rounding
errors during the inverse transform. But actually we don't need that
many bits during the forward transform.


  Sorry.  Those bits are necessary; without them, the values being
squared aren't computed with sufficient precision.

Colin Percival

Thanks. I came to the same conclusion while visualizing the schoolboy method of multiplication. One shouldn't think about math problems during holidays ;)


But what about the NTT variant of this idea? Would a mapping of one field or modular ring to another one be possible while using a DWT?

Matthias Waldhauer

_________________________________________________________________________
Unsubscribe & list info -- http://www.ndatech.com/mersenne/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

Reply via email to