This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Pretty good gcc-3.1 on sparc32-linux
On Thu, Apr 18, 2002 at 01:02:33PM -0700, Richard Henderson wrote:
> On Thu, Apr 18, 2002 at 08:39:53PM +0200, Christian Jönsson wrote:
> > Well, that could of course very well be the case. Do you suggest that
> > I try something special to find something out for all of us or should
> > I just drop the issue?
>
> The delayed branch scheduling pass is a known problem child. It's
> been waiting to be rewritten for a few years now.
>
> I suppose you could compile that test case by hand with -Q to verify
> that that is where the bulk of the time is being spent, but...
I didn't do that but I timed the tests (memory usage was up to about 16%
of 320 MB in most cases but up to 36% for -Os early):
-O0:
99.80user 3.88system 1:43.58elapsed 100%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (2931major+20234minor)pagefaults 0swaps
-O1:
254.14user 3.45system 4:18.17elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (2745major+24005minor)pagefaults 0swaps
-O2:
345.29user 9.24system 5:54.47elapsed 100%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (2796major+69498minor)pagefaults 0swaps
-O3 -fomit-frame-pointer:
352.51user 8.62system 6:01.28elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (2797major+69499minor)pagefaults 0swaps
-O3 -g:
629.73user 10.15system 10:39.86elapsed 100%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (2830major+69525minor)pagefaults 0swaps
-Os:
423.80user 12.99system 7:16.87elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (2796major+93248minor)pagefaults 0swaps
Cheers,
/ChJ