Author: Christophe Theron
Date: 22:04:28 12/15/02
Go up one level in this thread
On December 15, 2002 at 10:27:23, Tord Romstad wrote:
>On December 13, 2002 at 11:08:47, Christophe Theron wrote:
>
>>OK, so I need to explain again.
>>
>>Current Palm programs are going to be faster on the Tungsten only if:
>>
>>1) they make a lot of calls to the system.
>>2) they are recompiled in native ARM mode.
>>
>>Chess Tiger (current version) and ChessGenius (current version) are in none of
>>the above category.
>
>I was aware of this, of course.
>
>>That means that at this time your ARM processor is just executing 68000 code.
>>So it's extremely slow.
>
>The advertising I have seen want us to believe that the ARM Palms are usually
>faster than the 68000 Palms, even when executing 68000 code in emulation. I
>was probably a bit too optimistic when I believed in this.
It is the case when the application calls the system frequently (for example to
display graphics).
When the application does not call the system very often, everything is
processed in emulation mode. It's already a miracle that emulation is so fast!
>>In the future, when Chess Tiger is recompiled in native ARM mode, you are going
>>to see an incredible speed performance. My guess is 10x faster.
>>
>>For ChessGenius it's a different story. It is written in 68000 assembly, so it
>>cannot be recompiled in ARM native mode. So it will probably continue to run in
>>emulation mode on the PalmOS 5 handhelds. Unless Richard Lang rewrites it
>>completely.
>
>I think this will not be as difficult as you expect -- remember that Richard
>already has a version of Chess Genius running on Pocket PCs, which (IIRC) are
>also ARM-based.
Actually switching to faster processors will in my opinion make the gap between
ChessGenius and Chess Tiger even bigger. I mean when they both are running in
native ARM code.
From the SSDF list:
70 Chess Tiger 11.8 Pentium 90 MHz 2383 43 -43 261 50% 2384
77 Genius 5.0 DOS Pentium 90 MHz 2330 18 -18 1558 47% 2349
That was with an old version of Chess Tiger using a small opening book. I
believe that Chess Tiger has improved a lot more than the latest versions of
Chess Genius.
I am firmly prepared to defend Chess Tiger's position as the strongest chess
engine for the PalmOS platform. Both DragonBall and ARM cores.
>>>I also have a question regarding the (non-ARM-optimized) CT15. Will the new
>>>version support the 320x320 resolution? CT14.9 looks somewhat ugly compared
>>>to CG1.6 on my hi-res screen.
>>
>>CT15 for Palm supports only the 160x160 resolution.
>>
>>You can see screenshots at http://www.chesstiger.com. I don't find it ugly...
>
>Sorry, "ugly" was probably too harsh a word. But have you seen the new CG1.6
>running in 320x320? It looks fantastic, and I would very much like to see
>CT in the same high resolution.
>
>I have now played with Chess Tiger for a couple of days, and I am very
>impressed by the number of features you have been able to pack into such a
>small program,
Easy: the same source code is used for both Chess Tiger for PC and for Chess
Tiger for Palm.
All the work done on the PC version is instantly available on PalmOS.
There are even features that are already available in the PalmOS engine, but
that are not used by the PalmOS GUI.
For example, your Chess Tiger for Palm includes code for generating M-best moves
(not only the best move in a given position, but also alternative slightly
inferior moves). The code is in the engine, but the GUI does not take advantage
of this feature at this time!!!
> and that the program is able to play a decent game of chess
>despite searching at such a low speed (around 300--400 nps on the Tungsten T,
>it typically reaches 3 or 4 plies in the middle game at 5 minutes/game). My
>estimate is that the playing strength is around 1700-1800 elo on my Palm (while
>Genius feels more like 1400-1500).
Chess Tiger for Palm has been evaluated at 2100 SSDF elo. That should be close
to 2100 FIDE elo.
Even if it is twice as slow on your Palm, it should still be above 2000 elo.
>The main weakness of the program is that it (like Genius) appears to have a
>rather limited understanding of king safety. The simple strategy of castling
>opposite sides and throwing pawns against Tiger's king is very effective.
>
>One concrete type of mistake which occurs very often in Tiger's games is
>this: It often plays moves like ...h6 in the opening, and a couple of moves
>later castles kingside while the opponent has already castled queenside (or
>has not yet castled), inviting a quick g2-g4-g5.
On slow processors weaknesses like this one become very annoying, indeed.
>I have also noticed one minor annoyance in the user inteface: Holding down
>the center button of the navigator pad at the lower end of my Palm for a
>couple of seconds is supposed to switch to the Application Launcher. This
>works in all other programs I am running on my Palm, but _not_ in Chess Tiger.
I have no idea why it does not work!
>Hence I have to uncover the Grafitti area of my Palm every time I want to
>switch to another app, which is rather annoying.
>
>Finally, I would like to point out that I have not had so much fun playing a
>computer program for many years -- Chess Tiger is a very enjoyable opponent.
>Thank you for giving us this program, Christophe!
You are welcome!
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.