This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug rtl-optimization/31360] [4.3 Regression] rtl loop invariant is broken
- From: "rakdver at atrey dot karlin dot mff dot cuni dot cz" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: 26 Mar 2007 17:00:49 -0000
- Subject: [Bug rtl-optimization/31360] [4.3 Regression] rtl loop invariant is broken
- References: <bug-31360-6528@http.gcc.gnu.org/bugzilla/>
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
------- Comment #4 from rakdver at atrey dot karlin dot mff dot cuni dot cz 2007-03-26 18:00 -------
Subject: Re: [4.3 Regression] rtl loop invariant is broken
> > I guess the cost of loading zero is cheaper then?
>
> Cheaper than loading 0xDEADBEEF but not cheap enough not to pull out of the
> loop. All the "easy" (one instruction) constants are not pulled out.
>
> Really I think rtl loop invariant should pull everything it can pull out of the
> loop and then let the RA "fix" up the loop when it runs out of registers.
I would agree, if we had RA capable of that (which I am not quite sure
whether we do or not, although this seems simple enough), or better,
RA doing better job under high register pressure.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31360