This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Fix thinko in cleanup_tree_cfg
On Fri, 2005-08-19 at 21:23 -0400, Daniel Berlin wrote:
> On Fri, 2005-08-19 at 14:50 -0400, Diego Novillo wrote:
> > On Fri, Aug 19, 2005 at 11:51:24AM -0400, Diego Novillo wrote:
> >
> > > I'm running timings now to see what effect, if any this has on
> > > compile time.
> > >
> > Nothing changed much. Now I remember that when I started
> > changing CFG cleanup I had gotten most of the speedups by
> > allowing the CFG cleanup to iterate a certain number of times
> > (the fix to PR 22550).
> >
> > Afterwards, I removed the iteration limit and let it iterate
> > until no more changes were necessary (fix to PR 23164). That was
> > the patch that broke cleanup_tree_cfg's behaviour. However, at
> > the time I didn't get any significant speedups. So, fixing it
> > hasn't produced any significant slowdowns. All the numbers I've
> > got for the usual code bases are in the -2% to +2% range, which
> > is the usual noise.
>
It also appears to have caused an mgrid and fma3d regression on ppc64
SPEC, at least on the SuSE tester :(
(Again, the only real patch i can see in the diffs was yours)