This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Fix PR 1687 (C tree inliner related compile time 3.xregression)


Op za 03-05-2003, om 16:31 schreef Andreas Jaeger:
> Mark Mitchell <mark@codesourcery.com> writes:
> 
> > On Fri, 2003-05-02 at 02:49, Steven Bosscher wrote:
> >> Hi,
> >> 
> >> This fixes exponential behavior in tree-inline.c by walking tree nodes
> >> only once. 
> >
> > I couldn't resist this patch; it's too simple, and with too much
> > benefit, to ignore.  I checked it into 3.3.
> 
> What about mainline?  Shouldn't it go there also?

Please don't commit that yes.  I am trying something different for
mainline that unifies those three tree walks.

It turns out that for code with lots of very small inline functions,
creating/destroying the hash table three times gives you a slight
performance degradation (Less than 1% for Qt on my computer, but
still...).

For 3.3, so be it, it fixes a horrible regression.  But for mainline I
would like to avoid that little slowdown.

Greetz
Steven



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]