Re: [Bug-gnubg] What does Analyse Evaluation Speed Do?

[Gary Wong]

On Mon, Jan 20, 2003 at 05:19:00PM +0100, Nardy Pillards wrote:
> set priority highest
> 5000
> 
> set priority timecritical
> 6100

Oops, that's a bit annoying.  The code was supposed to measure elapsed
CPU time, not real time, so that the result would not be influenced by
CPU load, process priority, etc.  From clock(3):

} DESCRIPTION
}        The clock() function returns an approximation of processor
                                                          ^^^^^^^^^
}        time used by the program.
         ^^^^
} 
} RETURN VALUE
}        The value returned is the  CPU  time  used  so  far  as  a
                                    ^^^^^^^^^
}        clock_t;  to  get  the  number  of seconds used, divide by
}        CLOCKS_PER_SEC.

Apparently the C library under MS Windows is measuring something
slightly different (real time, presumably).  Under GNU/Linux, the
result seems to be unaffected by external processes.  For instance, on
a 1300MHz Celeron, gcc 3.2.1 -O3, two simultaneously running gnubg
processes (one "nice"d, the other not):

    Calibrating: 13566 static evaluations/second        
    Calibrating: 13621 static evaluations/second        

which is essentially the same as I get running on an otherwise unloaded
system.

To a first approximation from the results posted so far, it seems that
recent x86 processors handle about 10 evals/sec/MHz, or one evaluation
every 100,000 clock cycles.

Cheers,
Gary.
-- 
   Gary Wong           address@hidden           http://www.cs.arizona.edu/~gary/