Author: Christophe Theron
Date: 09:26:00 01/03/01
Go up one level in this thread
On January 03, 2001 at 07:48:12, Tord Romstad wrote: >On January 02, 2001 at 21:08:29, Christophe Theron wrote: > >>On January 02, 2001 at 17:29:46, Tord Romstad wrote: >> >>>On January 02, 2001 at 09:06:20, Christophe Theron wrote: >>> > >>>>I take Lang's programs as reference as they are known to be well optimized for >>>>680x0 processors. >>>> >>>>The DragonBall processor is close to a 68000 (at the same clock frequency). So >>>>the Mephisto Lyon @36MHz would be above, and the Mephisto Berlin is clearly >>>>below. >>>> >>>>My best guess is the rating of the Berlin on 68000 @12MHz + the elo difference >>>>you get by speeding up the processor to 33MHz (that makes >>>>70*log(33/12)/log(2)=102 elo points). >>>> >>>>So it would be rated 2117 elo (SSDF). This does not account for software >>>>improvements that have been done since the Berlin, maybe 50 to 100 additional >>>>points? I don't know... >>> >>>I am afraid the truth is precisely the opposite --- the chess engine in >>>Chess Genius for Palm is probably significantly weaker than the one in >>>Mephisto Berlin. The engine in Chess Genius, I have read, is the engine >>>of the very old Mephisto Roma (1987, I think), which was the last Lang >>>program which did not use transposition tables. The engine in Mepisto >>>Berlin is basically a port of Chess Genius 1 for the PC (released some time >>>during the years 1991--1993). >>> >>>The best way to estimate the strength of Genius for Palm is proably to >>>look up the rating of Mephisto Roma on the SSDF list and add some points >>>for the faster hardware. >>> >>>Please note that this is all based 100% on what I have read from various >>>sources, I have never seen Genius for Palm in action myself. >>> >>>Tord >> >> >> >>You are absolutely right in what you say, but I think you should keep in mind >>that I was simply trying to answer the following question: >> >>"In theory , what could a program be rated (SSDF) on a device like >> this , given the physical limitations ? (33MHz , and if the total >> program isn't over 4-5 mg) ?" > >I apologise for the misunderstanding, I was reading too fast. However, I >still think you are a little bit too optimistic. Mephisto Berlin did use >transposition tables. If I understand correctly, transposition tables are >difficult or impossible to implement in PalmOS. > > >Tord Hash tables slow the program down a little bit under PalmOS (about 5%), but there is no problem to implement them. Christophe
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.