Re: [Bug-gnubg] Can increasing the cache detract from evaluation/analysis quality?

[Michael Petch]

As long as there are no bugs we are aware of the cache should render the
same results. However I highly suggest that you use builds after September
5th, 2009. Versions prior to that had a number  of cache related bugs (Plies
> 3 and for some match scores greater than 32 away 32 away). For the past
month I have been running analyses with 0 cache and 168mb cache  on almost
every match to see if I can discover any problems. I haven't seen any. I
personally think as of September 5th the cache can be considered reliable
for rollouts and analysis.


On 26/09/09 4:21 AM, "Adi Kadmon" <address@hidden> wrote:

> Hi all,
>  
> Lately I increased the cache (from "Options" menu) from 0 to 5mb and
> afterwards also to the maximum 168mb. The result was in many positions a
> dramatic shortening of evaluation and analysis time, also of rollout time.
> Rollouts complete even 10 to 100 times faster!
>  
> So my question is, leaving aside the increased demand on my pc's memory and
> performances: Can the increased cache detract in any way from GNU's quality
> (or alter the results) in evaluation, analysis, rollouts? In other words, does
> the much faster performance has any "prices" from the angle of GNU's
> performance quality alone?
>  
> Thanks,
>  
> -- Adi
> 
> 
> _______________________________________________
> Bug-gnubg mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/bug-gnubg