Experimenting with tree inliner parameter settings

[Richard Guenther]

On 5 May 2003, Steven Bosscher wrote:

> Op ma 05-05-2003, om 09:26 schreef Richard Guenther:
> > On Sun, 4 May 2003, John David Anglin wrote:
> >
> > > > So the question is: What do these numbers do for your code,
> > > > both in terms of compile time and in runtime performance?
> > > > Richard, Dave, Gerald, can you please test the settings
> > > > below on your favorite code and on the test case for
> > > > PR 10160 and PR 8361 (and maybe PR 10316)?
> > > >
> > > > max-inline-insns-single=230
> > > > max-inline-insns-auto=230
> > > > max-inline-insns=500
> > > > min-inline-insns=80
> > > > max-inline-slope=20
> > >
> >
> > Performance for those params is bad, as expected, but it was the
> > same before.
>
> How bad in terms of your "performance indicator", and compared to
> earlier GCC's?

g++-3.3 (GCC) 3.3 20030428 (prerelease) with the above parameters, the
performance indicator is

2.23542e-06 s/iteration

with the default params

1.93928e-06 s/iteration

with my personal favorite (min-inline-insns=250)

1.26497e-06 s/iteration

compared to g++-3.2 (GCC) 3.2.3

1.97868e-06 s/iteration

and g++-3.0 3.0.5

4.49811e-06 s/iteration

and g++-3.4 (GCC) 3.4 20030422 (experimental) with defaults

1.83345e-06 s/iteration

with my personal favorite (min-inline-insns=250)

1.29759e-06 s/iteration


That is, with the new proposed params we're ~1.8 times slower than
the best param I found for my <em>particular special</em> C++ application.
Of course this makes production runs of my application without tuned
inlining parameters unattractive (run times are usually a few days up to
a month).

Hope this helps in any way.

Richard.

--
Richard Guenther <richard dot guenther at uni-tuebingen dot de>
WWW: http://www.tat.physik.uni-tuebingen.de/~rguenth/