Re: [Bug-gnubg] Problems using Stop rollout when one move appears to be best option

[Michael Petch]

On 23/08/09 2:54 PM, "Neil Robins" <address@hidden> wrote:

> 
> The progress bar jumps backwards when a move starts to reroll when it comes
> back within the JSD limit. In versions of GNU up to about a year ago a move
> that came back in would roll exclusively till it either caught up with the
> other rolls or went back outside the JSD limit.
> Now it just rolls permanently behind which can lead to the rollout finishing
> with the progress bar not at the end and that roll not either rolled to the
> full number of trials nor outside the JSD limit. In addition, just like not
> all rolls stop when they go outside the JSD limit, some rolls that have
> stopped do not restart at all if they come back within the limit.
> 

My progress bar has never ended early Even with 16 threads). The progress
bar will jump back then catch up and ultimately by the last trial it ends at
the right edge (100%). I've never had the progress bar stop before  reaching
the end.

What is curious is how given your position when analyzing the just the same
6 moves that one of your rolls (Number 4 in your list) went to 2592 and mine
never does (I have tried on a few different systems).

Can you do me a favor. Can you send me a copy of your gnubgautorc file (it
is inside your home directory in a subdirectory called .gnubg . I'm
wondering if there is something about your configuration that is different
than mine. As well if you are using ROL files to save your rollout settings
can you send a copy of the one you used for the position you gave in your
email? If you do send the info requested, how many total cores do you have
and what type of CPU's are you using? (I am just curious).

And one last request. Can you try rolling out the same position but set
Gnubg to use one thread only? (If you are using more than 1 thread
currently). I am curious if the output differ much from what you posted to
this list.

I just saw Christian's email about a possible race condition. That is what I
am thinking as well.