------- Comment #21 from kargl at gcc dot gnu dot org 2007-06-15 22:22 -------
(In reply to comment #20)
> Created an attachment (id=13709)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=13709&action=view) [edit]
> Specific example where libm, libcrlibm, and mpfr differ
>
> Here is a specific example of three different math libraries providing three
> different answers to the same question. The number does not use very many
> decimal places and thus __could__ come up in common use.
>
> It is not really how likely the number would be used that is important but
> what
> the result of using the answer would be. If your life is important don't be it
> on this number.
>
> This is just one number. How many more could there be, how will you prove you
> are correct and deduce the actual correct answer in those instances. This is
> what you must answer. This is why we need a fast, simple, library that is
> accurate and comes with "proof".
>
What does this have to do with GCC developement?
GCC does not supply libm. That's an OS vendor library.
GCC can't include crlibm at this time. See Joseph's
comment. Additionally, crlibm does not support float
or long double.
GCC does not supply mpfr. It uses mpfr to do constant
folding. AFAIK, sufficient guard digits are used to achieve
the desired accuracy.
Among all the noise that you've generated, I've become lost
in the waht exactly you think is broken.
On x86_64-*-*freebsd, dpara.f from netlib with gfortran 4.3,
I get
No failures, defects nor flaws have been discovered.
Rounding appears to conform to the proposed IEEE standard P754
The arithmetic diagnosed appears to be Excellent!
End of Test.
with
gfc4x -o z dpara.f
gfc4x -o z -O dpara.f
gfc4x -o z -O2 dpara.f
I get similar results with spara.f from netlib.
What, in 25 or fewer words, am I missing?
--
kargl at gcc dot gnu dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |kargl at gcc dot gnu dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32180
