Re: libgfortran array descriptor interface

[Daniel Kraft <d at domob dot eu>]

Paul Richard Thomas wrote:
> > > I'm actually writing a compiler for a subset of Fortran, to C++; so
> > > there's no escape :)
> > I'm curious, did you think about using gfortran's frontend and then
> > translating the intermediate gfc_code / gfc_namespace / gfc_symbol
> > representation into C++ instead of gcc's tree structure as we do?
>
> Tobi, I have often though that this would make a magnificent project
> for an SoC intern, although it would be better to translate to C,
> rather than C++.  That said, f2c direct to f2k2c???

Hi Paul,

I once asked the Ada guys if it would be possible to do a GNAT-internal 
tree translation to C as to gain a minimal C-to-Ada compiler.  My intent 
with that was the idea that we could provide C sources for GNAT that way 
and would no longer need a working Ada compiler to bootstrap Ada.

They meant that would be hard a job, but from what I know about 
gfortran's internals, it should be doable.  Another idea could be to do 
a new gcc-backend producing C-code instead of assembler (either by using 
the backend machinery and hijacking it (if possible), or by using the 
final RTL/Gimple structure to generate code from it).  That would have 
the benefit of being language independent, and could allow C++ to C or 
Ada to C as well :)

In any case, that's also quite a nice idea... Although it is probably 
not that useful compared to other things one could do ;)  But maybe as a 
GCC-general SoC project :D  I may have been interested in that if it 
would have been proposed last year.

Cheers,
Daniel

--
Done:  Arc-Bar-Cav-Rog-Sam-Val-Wiz
To go: Hea-Kni-Mon-Pri-Ran-Tou