2009/3/19 Michel Hack
<hack@xxxxxxxxxxxxxx>
P.S. Minor point: In MidRad representation, the radius may be
coarser, but must cover the same range as the midpoint.
So if the midpoint uses Binary64, the radius cannot use
Binary32, but it could use a truncated Binary64.
I do not fully get this.
MidRad is used in many cases for narrow intervals. If this is indeed how we end up defining them in the standard, why couldn't the radius use a Binary32 with its smaller range? Am I missing something else?
Thanks
--
Hossam A. H. Fahmy
Electronics and Communications Engineering Department
Cairo University
Egypt