This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: loop optimizations, mark 12[a]
- To: law at cygnus dot com, Richard Henderson <rth at cygnus dot com>
- Subject: Re: loop optimizations, mark 12[a]
- From: Richard Henderson <rth at cygnus dot com>
- Date: Thu, 2 Jul 1998 00:02:34 -0700
- Cc: egcs-patches at cygnus dot com
- References: <19980627234611.A17278@dot.cygnus.com> <10832.899072440@hurl.cygnus.com>
- Reply-To: Richard Henderson <rth at cygnus dot com>
On Sun, Jun 28, 1998 at 04:20:40PM -0600, Jeffrey A Law wrote:
> Runtime Specmark
> Without: 34.5 266.7
> With mark12a 88.4 104.1
>
> Something bad is happening.
Something is fishy, cause I don't see a regression, but rather a minor
improvement. Admittedly, the spec92 test harness is mysteriously bombing
out on me, but running it by hand I get
Without:
% cumulative self self total
time seconds seconds calls ms/call ms/call name
50.11 23.32 23.32 1 23320.00 46420.00 twldrv_
43.85 43.73 20.41 271614 0.08 0.08 fpppp_
4.83 45.98 2.25 139560 0.02 0.02 efill_
0.95 46.42 0.44 158803 0.00 0.00 fmtgen_
0.26 46.54 0.12 pow_dd
0.00 46.54 0.00 7 0.00 0.00 intowp_
0.00 46.54 0.00 4 0.00 0.00 ilsw_
0.00 46.54 0.00 2 0.00 0.00 fmtset_
0.00 46.54 0.00 2 0.00 0.00 lclear_
0.00 46.54 0.00 1 0.00 46420.00 MAIN__
0.00 46.54 0.00 1 0.00 0.00 aclear_
0.00 46.54 0.00 1 0.00 46420.00 d2esp_
0.00 46.54 0.00 1 0.00 1.12 gamgen_
With 12a:
% cumulative self self total
time seconds seconds calls ms/call ms/call name
53.37 22.89 22.89 271614 0.08 0.08 fpppp_
39.92 40.01 17.12 1 17120.00 42770.00 twldrv_
5.55 42.39 2.38 139560 0.02 0.02 efill_
0.89 42.77 0.38 158803 0.00 0.00 fmtgen_
0.28 42.89 0.12 pow_dd
0.00 42.89 0.00 7 0.00 0.00 intowp_
0.00 42.89 0.00 4 0.00 0.00 ilsw_
0.00 42.89 0.00 2 0.00 0.00 fmtset_
0.00 42.89 0.00 2 0.00 0.00 lclear_
0.00 42.89 0.00 1 0.00 42770.00 MAIN__
0.00 42.89 0.00 1 0.00 0.00 aclear_
0.00 42.89 0.00 1 0.00 42770.00 d2esp_
0.00 42.89 0.00 1 0.00 0.97 gamgen_
So. Can you run the test again? Or see what I've done wrong
configuring spec92 in /nfs/happy/smile/rth/spec92/?
Grr.. didn't have these sorts of problems with spec95...
r~