Author: Bas Hamstra
Date: 02:34:52 04/25/00
Go up one level in this thread
On April 25, 2000 at 03:23:58, Dann Corbit wrote: >On April 25, 2000 at 03:18:30, Tom Kerrigan wrote: >>On April 25, 2000 at 02:18:20, Dann Corbit wrote: >> >>>On April 25, 2000 at 01:32:48, Tom Kerrigan wrote: >>> >>>>On April 25, 2000 at 01:20:11, Bas Hamstra wrote: >>>> >>>>>(Currently I myself have started all over trying to make a really fast rotated >>>>>BB program. Looks promising so far: make/unmake=2M/sec and movegeneration=12M >>>>>moves/sec on a Celeron 466) >>>> >>>>My program does gen/make/unmake at 2M/sec, and just gen at ~10M/sec. >>>> >>>>I'm not using any sort of bitboards. >>>> >>>>Maybe something to consider... >>> >>>What general algorithm are you using? >> >>Pretty straightforward stuff: >> >>* 16x8 board so I can do the 0x88 trick >> >>* Non-linked piece lists (I've discovered that linked lists are _bad_!) > >I have seen another engine that does remarkably well and also uses 16x8 board >layout (Gullydeckel 2 -- 300KNPS on a lame PII 300 MHz actual search for one >sparse board in WAC.EPD). > >When 64 bit CPU's become the rage, I think this approach will be left behind >(unless it somehow magically scales). I am not so sure. It seems that at the moment the real node burners are all 0x88. I am targetting 500k nps for only material+ps on a celeron466. But according to what I hear for a 0x88 program 1M is possible doing the same on the same hardware. So there is something to catch up. Don't know how they do it, though. Possibly in assembler. And probably a 1 move at a time generator. My own 0x88 version was slower than BB (mat+ps). Regards, Bas Hamstra.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.