RE: [gnugo-devel] Persistent cache

[Portela Fernand]

Dan wrote:

> The string is marked if it is adjacent to an active string. We
> only have to mark it once, so as soon as such a neighbor is
> found, we mark it then exit the r loop.

Gotcha, I was just making a false assumption. Thanx for the explanation.
But, wouldn't it be faster to look for unmarked opponent strings around the
(currently defined) active area, rather than checking all opponent strings
of the board to see if they are in contact with the active area ?
Of course, this portion of code isn't time-critical, so why bother...

Btw, I tried to find out more about the case nngs3:230

- owl escape values are definitely involved.
- it would need a larger surface than the currently computed active area
  in order to detect a change (in the escape values map) which should
  trigger an invalidation of the cache entry.

/nando