Re: roundTiesToEven in precision 1

[Paul Zimmermann <Paul.Zimmermann@xxxxxxxx>]

       Hi John,

> And yet, I [...] never imagined a
> context in which anyone would really care about precision-1 rounding.

in MPFR, most of the roundoff error analyses are done in precision-1,
see for example http://www.mpfr.org/algorithms.pdf: we compute a relative
bound of the form 2^(-k) for the error corresponding to a given approximation,
and together with the target precision this enables us to decide whether we
can guarantee correct rounding, or if we need to compute another approximation
with a larger precision. (Of course we use rounding towards plus infinity and
not rounding to nearest in that context.)

Paul