Re: Need Fortran Cross-Compiler Solaris->m68020

[Joel Sherrill <joel dot sherrill at OARcorp dot com>]

Kai Ruottu wrote:
> 
> Joel Sherrill wrote:
> 
> > Kai Ruottu wrote:
> > >
> > > Rod <rod@tolosa.com> wrote:
> > > > I've tried building the C/C++ compilers, then cross-compilers, with
> > > > trouble, and never made it to the Fortran cross-compiler.
> > >
> > >  Funny, I had no problems at all with producing a 'sparc-solaris2.7'
> > > hosted and embedded 'm68k-elf' targeted cross-G77 (among the normal
> > > C and C++ stuff) on my RedHat 7.1 build host (trying this just for a
> > > fun...)
> >
> > Did it actually build all the support libraries?  I haven't tried
> > Fortran for embedded targets since 2.95 but some languages/targets
> > don't build the language support libraries.
> 
>  Yes, all those 'm68k-elf/*/libf2c/libg2c.a's were built for
> 'm68000/m68020/m68040/m68060/m5200' additionally with the
> default 'mcpu32' (set in my sources).
> 
>  Your "embedded targets since 2.95" maybe is the key, I used the
> gcc-2.95.3 sources because I didn't see the new C++ things in gcc-3.x
> giving any benefits in Fortran77. 

For RTEMS, we built all languages for nearly all targets for gcc 2.95.
But 3.0 was such a major change that many embedded targets broke.  I
did not even consider 3.0 useable for RTEMS' purposes.  3.1 was better
and I am pretty pleased with 3.2 for C and C++.  We provide Ada for
a few CPUs and I would like to provide java and fortran but we
haven't yet. 

> And I remembered the 'g77' being somehow 'frozen' since gcc-2.95.  

I don't watch fortran that much but I don't think there was a formal
freeze.  That doesn't mean activity slowed down or that focus was 
on different non-obvious things.

> And finally, building the 3.x
> sources is not fun if not having a very quick machine, the resulted
> GCC is much slower than the gcc-2.95.3 and that must be used to
> compile the target libs.  Generally I don't know what C/F77 benefits
> the gcc-3.x could bring for m68k... Besides those new 'features'
> available only with patches, like the '-mstrict-align/-mno-strict-
> align' and the '-mpcrel' support.

For sure, the m68k has taken a while to settle out but I think many
of the 2.95 m68k patches have now been merged.  If you have some
patches that you are sure are not in the 3.x mainline, let me know
and we can try to get them addressed.  

>  Maybe there are serious problems now with the gcc-3.x and the
> 'frozen libf2c' ? sources...

I don't think so in general terms.  I think it gets tested 
automatically and regularly.

>  The last time I did something serious with Fortran77 was in 1986,
> but I remember compiling something using GCC with 'f2c' and with
> 'g77' later and those compiled & linked things worked, on PC :
> DOS/Windows/Linux.

I did the same think around 1993 targeting RTEMS/m68020.  Worked
nicely for the code we used (GEOTRANS as I recall) but I wouldn't
have sworn about 100% correctness either. :)

>  But how well those compiled and linked F77-apps then run on an
> embedded m68020-card is unclear for me, but at least with the
> gcc-2.95.3 sources the 'm68k-elf-g77' + 'f771' + 'libg2c.a's
> should be produced...

Yep.  If someone can produce host g77 results for something like
m68k-netbsd, I would feel better personally.  But I would give it
a shot anyway.

> Cheers, Kai

-- 
Joel Sherrill, Ph.D.             Director of Research & Development
joel@OARcorp.com                 On-Line Applications Research
Ask me about RTEMS: a free RTOS  Huntsville AL 35805
Support Available                (256) 722-9985

------
Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sources.redhat.com