> > perhaps, this is of more interest, when I also mention that I'd like
> > to fix the Sparc native code generator. I will not be able to do it
> > very quickly, because I have to read (and understand) a lot of stuff
> > beforehand, so if anybody else is quicker, no problem. Otherwise, I'd
> > be happy if someone could point me to more docs worth reading besides
> > the commentary (and the code, of course) and perhaps give some hints
> > in what sense it is 'bitrotted' (like it says in ticket #606).
I will add that I have spent the last two months building shiny new
tools that are intended in part to make it easier to generate native
code. I don't really understand the internals of the NCG but if you
want to take a swing at using the new stuff, let me know. A good
place to start would be as follows:
1. Define two algebraic data types to represent the native Sparc
instructions you need: one for computational instructions and one
for control transfers.
2. Define a translation from ZipCfgCmmRep down to native
instructions.
If you want to define a tiny fragment of part 1, I can try to get you
launched with part 2----I'll write a type class that will help.
Norman
_______________________________________________
Cvs-ghc mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/cvs-ghc
