Re: bignums

["Ehud Lamm" <mslamm@mscc.huji.ac.il>]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1157 bytes --]

Gisle S�lensminde <gisle@apal.ii.uib.no> wrote in message
news:slrn9lrdd8.t30.gisle@apal.ii.uib.no...
>
> I also wrote a binding for GMP, and have finished the integer part of it,
> including a thick binding, but not for rationals and floats. I had plans
> to develop the rest when I have made the thick binding for floats as
> well, but the project stop since I personally only needed integers, and
> the float stuff seems to be crap, so I stoped there without finishing the
> project. I if anybody are interested I can put the bits together and
> relese it.
>
>
> The problem is that each new GMP number is dynamically allocated,
> and for uing it with operator overloading you must make a controlled
> type and allocate a new new number for each operation. This leads to
> very slow code, compared to the thin binding. In my fibonacci test
> there were a 1:6 difference between them.
>


Still seems to me worth publishing it somewhere (adapower, I guess). I'd
post a fast version, not using controlled types - since I assume most users
of such things worry mostly about speed. But then, I guess they also run out
of memory sometimes...

Ehud Lamm