This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Your SPECfp2000 run - date/time stamp.
- From: Andreas Jaeger <aj at suse dot de>
- To: Toon Moene <moene at knmi dot nl>
- Cc: gcc at gcc dot gnu dot org
- Date: Fri, 08 Feb 2002 13:15:05 +0100
- Subject: Re: Your SPECfp2000 run - date/time stamp.
- References: <3C63AC29.BA3DCABD@knmi.nl>
Toon Moene <moene@knmi.nl> writes:
> Hi Andreas,
>
> Between the last and the previous run of your SPECfp2000 suite, some
> update was made to GCC that helped wupwise, swim, and apsi (a little
> bit). I assume that it's Richard Henderson's fix to alias.c and loop.c
> at 7/2 00:57 PST.
The changes are noticable for wupwise and swim, but there's also a
tiny regression for some of the others, e.g. art
>
> However, the date/time stamp on the graphics page says 7/2 10:20 UTC for
> the earlier run and 7/2 23:00 UTC for the last run, so that would not be
> the explanation - unless these date/times do not indicate the moment you
> CVS'd the compiler used for that run.
>
> What is the exact meaning of the date/time stamp ?
It should be the date/time the scripts were started. But you can have
a look at the exact diff, e.g. at:
http://www.suse.de/~aj/SPEC/CFP/d-permanent/200202071020.fp/gcc-200202071020.diff.bz2
to check the changes (go there via spec result files). The ChangeLog contains:
+ 2002-02-07 Richard Henderson <rth@redhat.com>
+
+ * alias.c (find_base_value): Recall base values for fixed hard regs.
+ * loop.c (loop_regs_update): Don't use single_set on non-insns.
I don't know why there's a slight time difference between Richard's
fix and the start of my run.
Note that I rsync now all stuff and use that internal rsync tree for
the branches. This doesn't matter for the fp stuff but it's relevant
for others.
Andreas
--
Andreas Jaeger
SuSE Labs aj@suse.de
private aj@arthur.inka.de
http://www.suse.de/~aj