Re: [Bug-gnubg] Cache question

[Michael Petch]

On 03/09/09 3:19 AM, "address@hidden"
<address@hidden> wrote:

> I ran some brief tests using rollouts with different cache settings and larger
> cache produced faster results (not linear).
> Are there any known issues (bugs) running gnubg with cache set to max for
> evals and rollouts for plies up to 2?
> If not are there any reasons to not set cache to max?

With the current release problems arise mainly when ply > 3 when you use a
cache > 0. In those cases the bigger the cache the more problems you have
(And it will apply to both rollouts and evals). There is also a bug I
discovered that will manifest itself on pontentially all plies where either
player exceeds a 31 away situation or more (So matches >= 32 will have
potential cache issues).

Those are the known ones, although I believe there is another one that I
can't easily reproduce (I got it one day this week, and haven't seen it
since)... Either it was human error and I was insane or there was an issue.
Probably a bit of both lol.

We have a fix that has been tested and works well for the > 3 ply and cache
> 0 and will make it into the next release.