[SM-Discuss] RFC: combining all gcc frontends into a single spell
Andraž 'ruskie' Levstik
ruskie at codemages.net
Mon Jul 20 14:32:25 EDT 2009
> I agree that combining the spells again is probably a good idea. The
> original separation happened mostly because we didn't have SUB_DEPENDS
> at the time, so this was the only way to make sure spells could depend
> on the compiler for the language they needed.
> If this makes it possible to make gmp and mpfr internal to gcc, that
> would be a welcome advantage, circular depends are nasty when you run
> into problems.
>
I have a merged gcc spell now ready(for all languages). Still need to test
cast all the options but gcc+g++ work. mpfr and gmp can be put into the
source tree and thus gcc will pick them up. I'll commit it into a
devel-merged-gcc branch first thing in the morning.
Some stats(both same flags, profiledbootstrap-lean):
gcc+g++
gaze time gcc g++
Casting gcc took 3h 3m
Casting g++ took 3h 31m
Total time: 6h 34m
unified gcc with g++ enabled
Casting gcc took 3h 32m
So no matter how one twists this it is faster in a single one.
--
Andraž ruskie Levstik
Source Mage GNU/Linux Games/Xorg grimoire guru
Re-Alpine Coordinator http://sourceforge.net/projects/re-alpine/
Geek/Hacker/Tinker
Be sure brain is in gear before engaging mouth.
More information about the SM-Discuss
mailing list