Re: Wow! - ReadList.com

10 msg	flag_complex_method = 2 in C++?
2 msg	Help on emitting static constant arrays
2 msg	GCC 4.2.1 RC1
25 msg	About the is_gimple_min_invariant predicate
8 msg	Decimal float and the Cygwin build of GFortran.
1 msg	GCC 4.2.1 Status Report (2007-07-03)
1 msg	How should gcc handle '-shared -pie' and '-pie ...
3 msg	Question regarding getting .rodata into the .da...
2 msg	[patch] conditionally declare bswap functions d...
7 msg	bootstrap broken on powerpc?
5 msg	autoconf 2.61 on debian etch and GCC
2 msg	4.3 x86_64 Bootstrap breaks
2 msg	MPFR 2.3.0 Release Candidate
1 msg	FYI: today's regressions in libstdc++
1 msg	gcc-4.1-20070702 is now available
2 msg	Fwd: sub-optimal code for packed boolean arrays...
7 msg	[GSoC: DDG export][RFC] Current status

Wow!
\ Uros Bizjak (2 Jul 2007)
. \ Daniel Berlin (2 Jul 2007)
. . \ Uros Bizjak (2 Jul 2007)
. . . \ Daniel Berlin (3 Jul 2007)
. . . . \ Richard Guenther (3 Jul 2007)
. . . . . \ Jeffrey Law (3 Jul 2007)
. . . . . . \ Richard Guenther (3 Jul 2007)
. \ J.C. Pizarro (3 Jul 2007)

89 msg	no_new_pseudos
1 msg	Is it still true that message-lenght == 72 for ...

Subject:	Re: Wow!
Group:	Gcc
From:	Richard Guenther
Date:	3 Jul 2007

On 7/3/07, Jeffrey Law <law> wrote:
> On Tue, 2007-07-03 at 13:07 +0200, Richard Guenther wrote:
> > On 7/3/07, Daniel Berlin <dberlin> wrote:
> > > On 7/2/07, Uros Bizjak <ubizjak> wrote:
> > > > Daniel Berlin wrote:
> > > > > I'm curious if it was the pre/fre changes. can you try -fno-tree-pre
> > > > > and -fno-tree-fre and see if it slows down again?
> > > > Adding -fno-tree-pre slows aermod back to 33.942s. Nice optimization. ;)
> > > Thanks.
> > > (You really should thank richard guenther and steven bosscher, who
> > > continually beat me up till i finished it :P)
> > >
> >
> > Unfortunately, this "improvement" is a miscompare:
> >
> > > Value= 2190.9357900 Target= 2191.1145000 Tolerance=0.10000000000E-02
> > FAIL <<<<<<<<<<<<<<<<<<<< <<<<<<<<<<<<<<<<<<<< <<<<<<<<<<<<<<<<<<<< <<<<<<<<
> > > Value= 34310.421880 Target= 34310.421880 Tolerance=0.30000000000E-01
> > > Value= 4260.5888700 Target= 4260.9716800 Tolerance=0.50000000000E-01
> > FAIL <<<<<<<<<<<<<<<<<<<< <<<<<<<<<<<<<<<<<<<< <<<<<<<<<<<<<<<<<<<< <<<<<<<<
> > > Value= 37093.921880 Target= 37094.375000 Tolerance=0.20000000000E-01
> > FAIL <<<<<<<<<<<<<<<<<<<< <<<<<<<<<<<<<<<<<<<< <<<<<<<<<<<<<<<<<<<< <<<<<<<<
> > > Value= 7924.6811500 Target= 7924.8842800 Tolerance=0.30000000000E-01
> > FAIL <<<<<<<<<<<<<<<<<<<< <<<<<<<<<<<<<<<<<<<< <<<<<<<<<<<<<<<<<<<< <<<<<<<<
> > > Value= 37093.921880 Target= 37094.375000 Tolerance= 1.0000000000
> >
> > aermod FAILED 4 fails and 2 passes
> Having been down this path so many times -- any time a benchmark shows
> a big improvement, my first reaction is that something has been mis-
> compiled. If not, then I worry that it's just some ugly secondary
> cache effect from variables moving around on the stack or something
> similar. Only once I can convince myself that neither of those is the
> case do I get excited about the improvement.
>
> Perhaps I've become somewhat jaded through the years :-)

Of course in this case (the above is with -ffast-math) one could say
I got what I asked for. Fast but maybe (slightly) inaccurate. Of course
we miss the tolerances by more than an order of magnitude.

Richard.