This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [patch] Enable fmove-loop-invariants by default
Steven Bosscher wrote:
> But almost entirely due to mcf, which is memory bound (hence, cache
> bound on ia64). So that benchmark probably doesn't have much to do
> with loop.c anyway. It certainly doesn't look like something to be
> worried about to me...
OK, I think we've done our due diligence.
Thus far, nobody has stepped forward to demonstrate any benefit of
keeping -floop-optimize. All of the data we have seems to prove that
it's approximately a wash performance wise.
A patch to remove -floop-optimize is therefore pre-approved. Please
make sure to remove flag_loop_optimize, all uses thereof (look in
backends/frontends just in case), and any references in the documentation.
--
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713