changing floating point number input.. (was realroots...)



On 8/27/06, Richard Fateman <fateman at cs.berkeley.edu> wrote:

>  We could adopt MPFR wholesale for all floats, and have
> special routines that take these numbers and map them down
> to double-floats when speed counts, e.g. crunching numbers
> for linear algebra. This is usually very reasonable, since the
> conversion is "linear time" and most of the other stuff is much slower.

That sounds pretty interesting. I'm not 100% sure we
should do it, but if you are willing to go into the details,
I'm interested to hear about it.

Robert