This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: real.c on unicosmk
On Thu, May 23, 2002 at 12:32:27PM -0400, Paul Koning wrote:
> Excerpt of message (sent 22 May 2002) by Zack Weinberg:
> > On Wed, May 22, 2002 at 03:40:10PM -0700, Joe Buck wrote:
> > > > > In 3.2, REAL_ARITHMETIC is unconditionally on all the time. It's
> > > > > amazing how much crap vanishes when there is only one supported method
> > > > > of doing floating-point calculations.
> > > >
> > > > What do you intend to do about these eleven other GCC ports, then?
> > > > They had not been fixed to use REAL_ARITHMETIC, and for some (like
> > > > convex) the fix might be difficult.
> > > >
> > > > 1750a a29k clipper convex d30v elxsi fr30 i860 ns32k pdp11 romp
> > >
> > > About half of these need to go away: for example, there is not a single
> > > elxsi machine in the world.
> >
> > Of this list, only d30v fr30 ns32k pdp11 romp are still supported
> > targets. None of these claim to use anything other than IEEE floating
> > point. Where that is inaccurate, it needs to be fixed.
>
> It certainly is inaccurate for the PDP11, which predates IEEE float by
> many years. It uses VAX float formats. I recently changed gas to
> know that.
>
> This sounds like one of the lesser pdp11 target problems, but thanks
> for noting it...
The fix should be as simple as adding
#define TARGET_FLOAT_FORMAT VAX_FLOAT_FORMAT
to pdp11.h.
(The PDP-11 came before the VAX, so perhaps it would be appropriate to
rename VAX_FLOAT_FORMAT to PDP11_FLOAT_FORMAT.)
zw