This is the mail archive of the gcc@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: Incorrect application of loop exit heuristic?


Jan Hubicka <hubicka@ucw.cz> wrote on 08/19/2006 07:51:42 PM:

> >
> > Hi,
> > thepatch limiting minimal probability to 2% seems to make sense to me,
> > so please submit it for review.  It would be nice to have the code to
> > compute maximal number of exits from loop too, but if it is really 9,
we
> > would end up with 1% (it might however make difference elsewhere).
> > If you would feed motivated to do so, please just do ;) Otheriwse I
will
> > add it to my ever growing TODO list.
>
> Hi,
> this patch at least hides the ugly details within some abstraction so we
> can eventally go for propagating reliability information across CFG if
> we conclude it is important.  How does it work for you?
>

I like your approach, thanks.  I've confirmed that the generated code no
longer contains the mis-predicted branch hints.  Will follow up with
results from SPEC run once I get them (hosed things up for myself, so
restarting from scratch) :-(

-Pat


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