[Coco] gcc-coco revisited

[James Dessart]

On Fri, 31 Oct 2003, David wrote:

> I haven't attempted a compile yet.  I just noted that the patch worked -
> although the "obsolete" addition placed the "| m6809-* |" addition in a
> "non-alphabetical" location.

Yeah, but that's the easiest sort of thing to fix. :)

> > There may be gcc 3.3 features that might not be supported...
> > heck, there may even be 3.1 not supported...
> 
> That is a definite possibility (with 3.1, even).  If there are - with
> either version - I'm sure they can be worked out if we continue to work
> with it.
> 
> Actually, it might be a bit safer to stick with 3.1.

I definitely advocate sticking with 3.1.  It's stable, it's tested, and I
think the main reason for moving to 3.3, for anyone else, is for newer C++
template support, and the new ABI that everyone's moving to, also for C++.
We don't really need different name mangling, because we wont be
interoperating with other C++ compilers.  In fact, I think the names only
appear in object files.

> I played around with this a bit, "m6809.md"? I don't think it would take
> a lot to get it "near-perfect".

Yeah, that's the machine description file.  That's where you'd need to add
support for the DP register, if you plan on making those modifications.

> >I'd also like to see
> > some soft-float support in there.  Anyone here know the IEEE standard?
> From glancing through the files, I thought this was already set up.  If
> not, we definitely need this.

I don't think it is.  It's be nice, but I don't think we can reasonably
support 32-bit IEEE floats, which is what most software expects.  Well, I
imagine we could, it would just be a big pain.

James