This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: How to check that two ssa registers have the same symbolic value
On Tue, May 29, 2007 at 01:03:24PM -0600, Jeffrey Law wrote:
> On Tue, 2007-05-29 at 21:39 +0300, Victor Kaplansky wrote:
> > "Daniel Berlin" <dberlin@dberlin.org> wrote on 29.05.2007 20:05:34:
> >
> > >
> > > tree level PRE knows it could hoist these things (and they are marked
> > > as such), but we simply don't do it because it's not a partial
> > > redundancy.
> > > IE they are all marked antic at the beginning of the block.
> > >
> > > I've never bothered to teach it there are other times it might be
> > > profitable to move calculations anyway.
> >
> > It may be worth to do when PRE is called in its FRE flavor.
> > Looks like it always should be helpful for decreasing code size without
> > increasing runtime.
> I believe the general consensus was that this kind of hoisting was
> beneficial from a runtime standpoint -- though IIRC nobody bothered
> to really investigate *WHY*, they merely noted that if the RTL
> hoisting code was enabled the resulting code ran faster.
>
Is this related to
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31460
H.J.