Re: About gcc builtin atomics
On Sat, 1 Dec 2012 12:33:06 +0000 (UTC), Thorsten Glaser <firstname.lastname@example.org> wrote:
> Mikael Pettersson dixit:
> >and with two patches to reduce 'genattrtab' overheads; with all
> I think I asked already, but again, pretty please, care to share? ;-)
Sure, you can find three attrtab speedup patches I've backported from
gcc-4.8/trunk to gcc-4.6.3 in:
Two are generic, one is m68k-specific. The m68k-specific one makes a
huge difference, even when building a cross-compiler on a fast x86_64.
I haven't measured the other two but they probably help a bit too.
All three have been thoroughly regression-tested on multiple archs
(m68k, armv5te, x86_64, sparc64, powerpc64).
I also have corresponding patches for gcc-4.7.2, if you want them.