17 Jul 2000 18:24:54 GMT, Marcin 'Qrczak' Kowalczyk <[EMAIL PROTECTED]> pisze:
> > * If the -fstrictfp Option is given, *all* calculations (at
> > compile time and runtime) are done with the precision/range
> > of the type in question.
> Wouldn't it be simpler
[...]
I did not realize that it affects runtime calculations through gcc
-ffloat-store. Forget the previous comment: maybe it's not simpler,
but stability of floating point computations is usually not guaranteed
anyway. It's not easy to be stable by default.
As far as I know C, the default means that it's possible that
x /= x
even for not-NaN values of x::Double. It might fire a computation
and pass one x through registers and the other through memory :-)
IMHO GHC's documentation should clearly warn that programmers should
not depend on even basic stability and exactness of floating point
computations, and only stability is provided by -fstrictfp.
--
__("< Marcin Kowalczyk * [EMAIL PROTECTED] http://qrczak.ids.net.pl/
\__/ GCS/M d- s+:-- a23 C+++$ UL++>++++$ P+++ L++>++++$ E-
^^ W++ N+++ o? K? w(---) O? M- V? PS-- PE++ Y? PGP+ t
QRCZAK 5? X- R tv-- b+>++ DI D- G+ e>++++ h! r--%>++ y-