[PATCH] Remove GMP in favor of MPFR

[Tim Prince]

At 08:43 PM 7/4/2004, =?gb2312?q?Feng=20Wang?= wrote:

>--- Richard Henderson <rth@redhat.com> µÄÕýÎÄ£º> On Sun, Jul 04, 2004 at
>10:57:26AM -0700, Tim Prince wrote:
> > > >      /* Recall that IEEE numbers are interpreted as 1.F x 2**exp,
> > > >         whereas the intermediate representation is 0.F x 2**exp.
> > >
> > > That should be 0.5F x 2**exp
> >
> > No, the "1" is implicit in the IEEE format.  There is no implicit
> > bit in the real.c representation.  Thus "0.F" is correct.
>It's the same with fortran.
>
>In fact, IEEE just describes the storing format of floating point. 8 bits hold
>the binary exponent with the range [-126, 127](normalized). In normalized 
>form,
>there is an implicit '1' in the fraction. Fortran uses "0.F x 2**e" REAL
>concept model. The exponent range should be [-125,128].
>

Which is also the C exponent range [FLT_MIN_EXP, FLT_MAX_EXP] for IEEE-754 
float type, 1 off from the IEEE definition.


Tim Prince