On 7/9/07, Jonathan Kinsey <[EMAIL PROTECTED]> wrote:
Christian Anthon wrote:

Hi,

I'm not too familiar with this code - where is the bit you're talking about
(cube rollout std dev calculation)?


Hi Jon,

I looked at it some more over the weekend. The problem is that the
cube rollouts have shared rollout contexts between double/take and
no-double. But the threads need not complete in a specific order so we
cannot use the shared prc ->nGamesDone for counting. The code ensures
that the game count is correct for alt=1, but not for alt=0 where it
may be off by working-threads minus 1. I have fixed this myself and
just need a little testing.

Christian.


_______________________________________________
Bug-gnubg mailing list
Bug-gnubg@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-gnubg

Reply via email to