Mail Archives: djgpp-workers/2002/03/03/13:28:33
> From: Martin Str|mberg <ams AT ludd DOT luth DOT se>
> Date: Sun, 3 Mar 2002 18:23:03 +0100 (MET)
>
> According to Eli Zaretskii:
> > against standard.results. Does the new compile only add new
> > inaccuracies, or does it add some and remove others?
>
> It adds many and removes some.
That might mean that code generation was changed in significant ways.
Someone ought to look into this, but even if not, inaccuracies in the
last 2-3 bits of the mantissa are not very grave.
> > > log2 has two in bit 12 and one in bit 0!?
> >
> > What results are those? Are they finite or Inf/NaN?
>
> Bit 12: (-Inf NaN)
> Bit 0: (NaN NaN)
These are known problems: we deviate (deliberately) from what the
Cygwin suite expects.
> > > log2 also has errno wrong-
> >
> > What value of errno? If log2 is documented to produce that value,
> > it's okay.
>
> It says:
> In log2() using log2_vec.c:146, errno wrong: 0 1
> In log2() using log2_vec.c:147, errno wrong: 0 1
Ah, this one is in standard.results, and is another known deviation.
> > > sqrt has two in bit 0!?
> >
> > That's a sign bit? A NaN, perhaps? (Our NaN has its sign bit set.)
>
> Yes, (NaN NaN).
Known; disregard it.
- Raw text -