Re: [Bug-gnubg] How many threads can gnubg (reliably) handle?

[Michael Petch]

I committed a fix to cvs.

On 13/09/09 2:41 AM, "Massimiliano Maini" <address@hidden> wrote:

> I got some issues compilling without threads and without GUI:
> 
> Compiling analysis.c
> analysis.c: In function `CommandAnalyseGame':
> analysis.c:1259: error: too many arguments to function `AnalyzeGame'
> analysis.c: In function `CommandAnalyseMatch':
> analysis.c:1293: error: too many arguments to function `AnalyzeGame'
> analysis.c:1304: warning: implicit declaration of function `multi_debug'
> analysis.c:1305: warning: implicit declaration of function `MT_WaitForTasks'
> analysis.c:1305: error: `UpdateProgressBar' undeclared (first use in
> this function)
> analysis.c:1305: error: (Each undeclared identifier is reported only once
> 
> 
> MaX.
> 
> 2009/9/12 Jonathan Kinsey <address@hidden>:
>> This change will have no affect on rollouts. The speed increase will depend
>> on
>> the number of games and how big they are (lots of small games showing the
>> biggest increase). In general the "analyse match" feature should scale
>> better
>> and show better cpu utilisation.
>> 
>> Jon
>> 
>> Michael Petch wrote:
>>> 
>>> I did some 2 ply tests and I¹ll compile some of the stats after but it
>>> appears on my dual quadcore system with threads=cores, this minor change
>>> improved 2ply performance by 20%. I¹m not sure about the rollouts, will
>>> try later.
>>> 
>>> On 11/09/09 9:23 AM, "Jonathan Kinsey" wrote:
>>> 
>>> I've checked in a small change that should overcome this scaling
>>> issue, very
>>> limited testing so could easily contain problems...
>>> 
>>> Jon
>>> 
>> 
>> 
>> 
>> 
>> ________________________________
>> View your other email accounts from your Hotmail inbox. Add them now.
>> _______________________________________________
>> Bug-gnubg mailing list
>> address@hidden
>> http://lists.gnu.org/mailman/listinfo/bug-gnubg
>> 
>> 
>