[computer-go] Re: Effective Go Library v0.101

[Łukasz Lew]

On 2/5/07, Hideki Kato <hideki_katoh at ybb.ne.jp> wrote:
> Łukasz Lew: <c55009e70702040911u8a41c2ewf1618dedc00e89b5 at mail.gmail.com>:
> >On 2/4/07, Hideki Kato <gg at nue.ci.i.u-tokyo.ac.jp> wrote:
>
> >> Łukasz Lew: <c55009e70702040513r107ae80fvdc1bc8ee870467d9 at mail.gmail.com>:
>
> >> >50 kpps on 1.4 GHz and
>
> >> >80 kpps on 2.2 GHz.
>
> >>
>
> >>
>
> >> I'd like to add a sample in hand (10^6 playouts):
>
> >> 115 kpps on 3.0 GHz Intel Core2Duo (E6700/Conroe).
>
> >
>
> >A pleasant number to look at ;)
>
>
> :)
>
>
> >> And I have a question. Why Komi is not float but int
> >For efficiency reasons ( :D )
>
> Sure, I've modified your source code to use float for komi and get 75
> (!) kpps on the same box above.

This is a way to much, please send me Your modification.

>
> >Draw means win
> >for white, so 7,5 is the same as 7.
>
> It's ok for almost all games but you can also use a classical technique
> of doubling komi that concludes having all internal scores be doubled.
>
> I've tried this and get 104 kpps but I'm not sure it's correct (very
> slower than expected).

There is no need to double all scores, because the convention that
white wins draws have the same effect.

>
> >You should use board->set_komi (float) it works OK.
> >I will add get_komi function and fix benchmark output..
>
>
> >Łukasz
> add get_komi function and fix benchmark output..
>
> >
>
> >Łukasz
>
> >
>
> >>
>
> >>
> - gg
>
> >> --
>
> >> Hideki
> Kato <mailto:hideki_katoh at ybb.ne.jp>
>
> >>
> _______________________________________________
>
> >>
> computer-go mailing list
>
> >>
> computer-go at computer-go.org
>
> >>
> http://www.computer-go.org/mailman/listinfo/computer-go/
>
> >>
>
>
> >---- inline file
> >_______________________________________________
> >computer-go mailing list
> >computer-go at computer-go.org
> >http://www.computer-go.org/mailman/listinfo/computer-go/
> --
> gg at nue.ci.i.u-tokyo.ac.jp (Kato)
> _______________________________________________
> computer-go mailing list
> computer-go at computer-go.org
> http://www.computer-go.org/mailman/listinfo/computer-go/
>