I have written an in place, recursive FFT/SS algorithm for multiplying polynomials over ZZ, since it is more cache friendly than the iterative algorithm in NTL.
On my machine, NTL takes 17.7 seconds to multiply 1000 pairs of polynomials of degree 255 with coefficients of 1000 bits. The new algorithm takes 14 seconds even. Now I can start to implement some of the tricks that I spoke of in an earlier thread to bring this down by a couple of factors of 2. Funnily enough, some of the optimizations I used can be applied to Victor's original algorithm. Bill. --~--~---------~--~----~------------~-------~--~----~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://sage.scipy.org/sage/ and http://modular.math.washington.edu/sage/ -~----------~----~----~----~------~----~------~--~---
