RE: [computer-go] Chains and liberties - performance

["Mark Boon" <tesuji@xxxxxxxxx>]

> -----Original Message-----
> From: computer-go-bounces@xxxxxxxxxxxxxxxxx
> [mailto:computer-go-bounces@xxxxxxxxxxxxxxxxx]On Behalf Of Arend Bayer
> Sent: Sunday, November 14, 2004 22:12
> To: computer-go
> Subject: [computer-go] Chains and liberties - performance
>
>
>
>
> On Sun, 14 Nov 2004, Mark Boon wrote:
>
>
> Just as a data point:
> For a life and death test set (owl.tst) of GNU Go, I get 296000 moves
> per second. According to a profile run, gnugo spends 14.2 % of that time
> updating the board data, and 3.3% in the undo function. So that would be
> 1.7 Million moves per second for the board library alone.
> This is on a Pentium M 1.4GHz (appr. as fast as a Pentium 2.4GHz on
> usual benchmarks, but I dont know about gnugo).

Apparently the desktop version of my CPU should be equivalent to 2.4Ghz
Pentium as well. But I have a laptop...

It seems I don't have the appropriate hardware to do speed-tests. Between
runs I'm experiencing large variations in performance, and I think I've
finally figured out why. When I looked at the 'My Computer' info a while ago
it said 900Mhz. I checked again,and now it says 1.8Ghz. So I think it does
speed-stepping to conserve power. I have no idea if and how I can tell it
not to do that (when the adapter is connected for example) but until I do,
trying to do optimization based on running speed-tests isn't going to give
me usuable data. So please ignore any of the numbers I published before.



_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/