Re: [bug-gawk] New multiplication algorighm

[Peter Brooks]

Thank you - that's most helpful. I'm pleased that I asked. I certainly
don't find myself working with numbers with over a thousand digits!

On Fri, 19 Apr 2019 at 22:58, Nelson H. F. Beebe <address@hidden>
wrote:

> The new O(n log n) multiplication algorithm has a large constant of
> proportionality, so it may not matter in practice.  In any event, I
> posted a note about it to the gmp developers lists a week ago: see the
> thread archived at
>
>         https://gmplib.org/list-archives/gmp-devel/2019-April/thread.html
>
> The mpfr library sits on top of gmp, and so is unlikely to be
> interested in the low-level multiplication algorithm.  Multiple
> precision libraries are often tuned to use different algorithms for
> multiply, divide, and square root, depending on the number of digits
> needed: the Fast Fourier Transform algorithm for multiplication is one
> such, but experiments in various multiple precision packages show that
> one may need to be working with a thousand or more digits before it is
> faster than the conventional schoolbook algorithm that takes O(n**2)
> time.
>
>
>
> -------------------------------------------------------------------------------
> - Nelson H. F. Beebe                    Tel: +1 801 581 5254
>     -
> - University of Utah                    FAX: +1 801 581 4148
>     -
> - Department of Mathematics, 110 LCB    Internet e-mail:
> address@hidden  -
> - 155 S 1400 E RM 233                       address@hidden
> address@hidden -
> - Salt Lake City, UT 84112-0090, USA    URL:
> http://www.math.utah.edu/~beebe/ -
>
> -------------------------------------------------------------------------------
>
>

-- 
Peter Brooks


Mobile: +27 82 717 6404
Skype:  Fustbariclation
Twitter: Fustbariclation
Author Page: amazon.com/author/peter_brooks