------- Additional Comments From law at redhat dot com 2004-12-22 19:28 -------
Subject: Re: [4.0 Regression] jump threading
on trees is slow with switch statements with large # of cases
On Wed, 2004-12-22 at 19:18 +0000, steven at gcc dot gnu dot org wrote:
> ------- Additional Comments From steven at gcc dot gnu dot org 2004-12-22
> 19:18 -------
> That may all be true, but the fact is that you're now talking about
> missed-optimizations, and not a real compile time hog. Yes we can
> win. No, this is not a hog anymore, it's just not as good as it
> could be. You're also talking about RTL, not tree jump threading.
I'm talking strictly about compile-time hogs. It's insanely dumb that
we're burning 5% of compilation time just trying to hoist the
fake GLOBAL_VAR variable out of empty loops. It's insanely dumb that
we're burning 5% of compilation time threading jumps at the RTL
level when they all should have been threaded at the tree level.
>
> In fact I think it's plain wrong to keep this bug open. It's just
> keeping the number of real regressions artificially high. We have
> compile time slowdowns, and that's a regression. But we don't need
> to keep bugs open when the issue of that bug was been fixed.
I stongly disagree. Please do not close this bug report.
jeff
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15524
