[seqfan] Re: Base conversion question

[victor miller]

I remember the Chudnovsky brothers telling me many years ago that when they
were calculating pi to billions of decimal places they ended up using
multiprecision in a base of 10^k (for some convenient k) since they had
found that it took much longer to convert from the result in binary than it
did to actually compute it!
Victor

On Tue, Jul 21, 2009 at 9:21 PM, Joerg Arndt <arndt at jjj.de> wrote:

> Binary splitting is your friend, cf. section
> 32.1.8 "Fast radix conversion" of
>  http://www.jjj.de/fxt/#fxtbook
>
> For the large multiplication near the end of the
> algorithm some fast scheme is to be used.
>
> The "direct FFT" idea (later mail) does not seem to work,
> at least no one managed so far.
>
>
> * David Wilson <dwilson at gambitcomm.com> [Jul 22. 2009 11:00]:
> > I suppose this question is better addressed to math-fun, however, I do
> > not think my work self belongs to math-fun.
> >
> > At any rate, does anyone know anything about the efficiency of base
> > conversion between non-homomorphic bases, e.g., 2 and 10?
> >
> >
> > _______________________________________________
> >
> > Seqfan Mailing list - http://list.seqfan.eu/
>
>
> _______________________________________________
>
> Seqfan Mailing list - http://list.seqfan.eu/
>