Re: [gnugo-devel] tuning patch

[Arend Bayer]

On Gunnar wrote:

> Dan wrote:
> > Here's the results after Gunnar's patch. A net gain of 3 tests.

> I'm surprised this is so much worse than my preliminary results.

It's rather a gain of 5 tests (owl1:s.th. was a new test, nngs:240 see
below).

> > After making an obvious change to get it to compile, I added
> > arend_3_3.6 to the CVS and running the regressions on it.
> >
> > There are NO unexpected results.

> Can you recheck this? When I try nngs:240 with and without arend_3_3.6
> I get different results.

Here is a new version: I get a FAIL both with and without my patch
(without having applied Gunnar's patch) -- i.e., including 3.3-pre1.
Platform dependency again?

If there is a change due to my patch then that would be a bug, and I
should look into it.

Arend