> > In the FFT range, will it still hold that the time for MP_n will be > > almost the same as for a full n × n multiply, or what's the expected > > ratio? > > Yes, and in fact I expect the ratio is much closer to 1 than in the > karatsuba case. I confirm. What David has done (MP for Karatsuba and Toom-Cook) is really the hard work. Paul