RE: [gnugo-devel] A speedup

[Portela Fernand]

Arend wrote :

> I wanted to try this, too!

:-)

> I think we should look for exact timing results, however.

Definitely. I would have done them myself if I could, but I have no idea
how to get these (with accuracy) under Win32.

> You are duplicating a lot of work of the move generation. (And
> note that many of the reading nodes you have saved will be extremely
> short ones, where the attacking code would return immediately.)
>
> A compromise might be to add shortcuts e.g. in defend4 as soon as we
> discover a move that gains 5 liberties.

Ok, I'm gonna try it, in defend2 and defend3 as well I think. And I'll check
with fastlib to see if it makes any difference.

/nando