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 AutoFDO/2]Treat ZERO as common profile probability/count


On Wed, Oct 31, 2018 at 5:11 PM Richard Biener
<richard.guenther@gmail.com> wrote:
>
> On Wed, Oct 31, 2018 at 7:30 AM bin.cheng <bin.cheng@linux.alibaba.com> wrote:
> >
> > Hi,
> > In new profile probability/count infra, we have different precision quality categories,
> > and probabilities/counts of different categories are not supposed to be compared or
> > calculated.  Though in general is an improvement, it introduces unexpected behavior.
> > Specifically, class profile_probablity and profile_count themselves are implemented
> > by comparing probabilities/counts against profile_count::zero().  while zero() is of
> > profile_precision category, it's always compared different to zero of other precision
> > categories including afdo.
> >
> > I can see two ways fixing this: 1) Treat zero as a common probability/count regardless
> > of its category; 2) Provide an "is_zero" method rather than relying on "==" comparison
> > against probability_count::zero().  2) requires lots of code changes so I went with 1)
> > in this patch set.  This patch doesn't handle "always" but it might be.
> >
> > This patch also corrects a minor issue where we try to invert an uninitialized value.
> >
> > Bootstrap and test on x86_64 in patch set.  Is it OK?
>
> I'll defer on the emit_store_flag_force change, likewise for the zero
> handling in
> compares - I don't think zeros of different qualities should compare equal.
> Would compares against ::always() not have the very same issue?
> Likewise ::even(),
> ::likely(), etc.?  Those always get guessed quality.
Yes, these values also affected if compared with precise category, but
zero is the major issue.  So 2) makes more sense when checking if a
profile count is_zero/is_likely/is_always etc. regardless of its categories.
Once with Honza's input, I can do some experiments.

Thanks,
bin
>
> The invert change looks OK to me.  The related change to the always() API would
> suggest to replace guessed_always() with always (guessed) and also do similar
> changes throughout the whole API...
>
> Honza?
>
> Thanks,
> Richard.
>
>
> > Thanks,
> > bin
> >
> > 2018-10-31  Bin Cheng  <bin.cheng@linux.alibaba.com>
> >
> >         * expmed.c (emit_store_flag_force): Use profile_probability::always.
> >         * profile-count.h (profile_probability::always): Add parameter.
> >         (profile_probability::operator==, profile_count::operator==): Treat
> >         ZERO as common probability/count regardless of its quality.
> >         (profile_probability::invert): Don't invert uninitialized probability.


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