gnugo-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[gnugo-devel] remaining patches


From: bump
Subject: [gnugo-devel] remaining patches
Date: Sat, 21 Dec 2002 22:18:19 -0800

I'm thinking the remaining patches can be left out of 3.3.13,
though none of them are to be rejected, just postponed. Here
is my reasoning:

nando_3_13.3b Remaining part of nando_3_13.3. (Owl tuning.)
Nando proposed removing several patterns. Only one was
actually removed.  Three of the remaining patterns can be
deleted with no effect on the regressions, but also with
no measurable speedup. But I'm very reluctant to eliminate
A107. I've explained my reasons for this previously.

paul_3_13.8b. The patch paul_3_13.8a uncovered some bugs,
probably two different ones, and we should try to
understand and fix those. There is no rush to put this
patch in the CVS.

arend_3_13.2. SPEC stuff, not urgent.

evan_3_13.8. Owl tuning. The proposed pattern A424 is very
reminiscent of A1120 that Nando took out in 3.3.11. Perhaps 
Nando can comment on whether this patch is at odds with
the philosophy he explained in connection with
nando_3_13.3 and other patches.

evan_3_13.4. Attack1. I am running the regressions on evan_3_13.4
now, the atari1 ko patch. Maybe we should put an upper bound on
the number of trymoves this piece of code can generate. As 
written it looks a little dangerous.

My inclination is to put up 3.3.13 without any of these five
patches, and leave them all classified as pending.

Comments?

In other news, the archiving is working again.

Dan



reply via email to

[Prev in Thread] Current Thread [Next in Thread]