[gnugo-devel] nando_3.9.4c ?

[Portela Fernand]

Hi,

A point of information about my work on this patch and a question.

I've made some performance tests on the patch, and it appears that it
increases the owl nodes consumption on the regression tests by an amount of
about 6%. I think this is pretty heavy, specially if we don't get more
concrete results than the very few passes observed with nando_3_9.4b

Mistakenly, I informed some participants of this list about further
developments of the patch, but you can find the details in Dan's mail (which
I can't find in the list archive, weird). Briefly, it's about some
improvements I got by tuning and extending a bit this patch.

1. this "better" tuning doesn't cost additional owl nodes (actually it saves
   a few of them)
2. the regression breakage shows 9 fails (I have already corrected 2 of
   them) and 14 passes, including the "interesting" ones in safety.tst

The question : if I can confirm that the engine's behavior is favorably
modified (regarding groups safety) by this patch, do you think it is worth
the 6% additional owl nodes or not ?

/nando