[gnugo-devel] Arend's owl speedup

[Daniel Bump]

I ran the regressions on two GNU Go processes, with
and without Arend's owl speedup patch. There were
no changed results.

About timing, there was a real improvement. As with
his other patch, ran the two processes simultaneously
on a dual Celeron machine, and Arend has pointed out
that the results of this were ambiguous for the
previous patch. This time they are unambiguous,
showing this patch is an impressive speedup. In
fact, when I got up this morning, the patched version
was done, and the unpatched version was still
cranking away.

Patched:

real  579m2.566s
user  421m31.890s
sys   0m18.460s

time to complete: 34743 seconds

Unpatched:

real    624m40.755s
user    466m19.620s
sys     0m18.190s

time to finish: 37480 seconds

Dan