Winboard Forum

Hi,

I ported my latest engine version to Palm.
It plays exactly the same as the PC version. The only difference is that Palm version doesn't have some endgame bitbases which are integrated into the PC version.

This is a 'light' version made to support older Palm models with only 256KB available for the program execution (all global data + stack + heap have to fit in it). This version for instance uses only 32KB for the hash table.

I am working on the 'full' version, which will use much more memory, so it will probably be much stronger on the newer Palms.

Alex

Hi Alex,

Many thanks for the continued development of Naum for Palm (my favourite Naum!).

Naum on my Zire was a regular poolside companion during my recent holiday to France. A personal highlight being that I managed to beat it on 10s/move with permanent brain on! (hmm, better not tell my wife! ).

Best wishes

Chris

Hi Chris,

I guarantee you a much tougher job to beat this version. It's tactically much stronger.

Actually I am able to beat 1.2 version regulary on my Clie running at 33MHz (reaching only ply 4-5). I think I am going to continue playing against that one

It's a pitty I don't have an ARM PDA, so that I can develop an ARM version for your Zire.
Out of curiosity, what kind of depths are you getting at 10sec on Zire?
On Zire Naum runs in 68K emulation mode, which is light years slower then the native ARM mode.

Regards,
Alex

Naum wrote:Hi,

I ported my latest engine version to Palm.

Hi Alex,

This is great news! Thanks a lot for the Palm OS version.

There will be a Palm version of Glaurung as soon as there is a decent set of Palm OS development tools available for Mac OS X. At the moment, the Palm OS Simulator is Windows only, which makes it very tedious to develop from OS X. It's a great pity; Glaurung was designed with handheld devices in mind since the beginning. I have taken great care to keep the code and data size to a minimum.

Tord

Tord Romstad wrote: It's a great pity; Glaurung was designed with handheld devices in mind since the beginning. I have taken great care to keep the code and data size to a minimum.

Tord,
I wonder, if one really has to care about small code and data sizes anymore, nowadays. In the context of a chess engine, I mean. I cannot imagine multi megabyte code/data for the typical chess engine. Don't cellular phones already have many MBs of memory? Of course, there are many things in chess engines, where lots of memory is handy (TT tables, tables needed for TB support, ...). But for the "pure" chess engine, it is hard to imagine, that this really is important anymore. Sure, there will be things like cache size issues. But, would it really make a lot of difference?

When I started chess programming (purely private at that time), it was mostly on Atari ST, 1 MB of RAM. I never had any problems with RAM at that time (besides the RAM needed for the developement tools - the computer had no HD, and to develope and debug, the only practical method was, to also have a rather large RAM disk).

IIRC, stack space was limited to 64 k on the Atari. I believe to remember, that it was tricky at times, to not overflow that (or even to setup things, to be able to use more than some default 4k stack or so). I'd assume, that this is no problem anymore. But I have practically no experience with these new environment (cellular phones, Palm, Pocket PC, ...)

Regards,
Dieter

Dieter B?r?ner wrote:

Tord Romstad wrote: It's a great pity; Glaurung was designed with handheld devices in mind since the beginning. I have taken great care to keep the code and data size to a minimum.

Tord,
I wonder, if one really has to care about small code and data sizes anymore, nowadays. In the context of a chess engine, I mean. I cannot imagine multi megabyte code/data for the typical chess engine. Don't cellular phones already have many MBs of memory? Of course, there are many things in chess engines, where lots of memory is handy (TT tables, tables needed for TB support, ...). But for the "pure" chess engine, it is hard to imagine, that this really is important anymore. Sure, there will be things like cache size issues. But, would it really make a lot of difference?

At least for Palm OS devices, I think it does. Remember that there is no hard disk drive; all programs and data on the device are stored in RAM. Many of the older Palms have only 16 or 32 MB RAM. It is therefore really important to keep the programs as small as possible.

Tord

Tord Romstad wrote:

Dieter B?r?ner wrote:

Tord Romstad wrote: It's a great pity; Glaurung was designed with handheld devices in mind since the beginning. I have taken great care to keep the code and data size to a minimum.

Tord,
I wonder, if one really has to care about small code and data sizes anymore, nowadays. In the context of a chess engine, I mean. I cannot imagine multi megabyte code/data for the typical chess engine. Don't cellular phones already have many MBs of memory? Of course, there are many things in chess engines, where lots of memory is handy (TT tables, tables needed for TB support, ...). But for the "pure" chess engine, it is hard to imagine, that this really is important anymore. Sure, there will be things like cache size issues. But, would it really make a lot of difference?

At least for Palm OS devices, I think it does. Remember that there is no hard disk drive; all programs and data on the device are stored in RAM. Many of the older Palms have only 16 or 32 MB RAM. It is therefore really important to keep the programs as small as possible.

Tord

Hi Tord

Well, i'm working on a PocketPC version of Homer. May there are comparable environment between PocketPC (Windows mobile) and Palm OS (less RAM and so on) and, for example, i must changed my hashcode for better results. Allso my code for input/output is completly different und works different.
After all i must say it was more work as i expected and i'm not ready right now.

Tord Romstad wrote:

Naum wrote:Hi,

I ported my latest engine version to Palm.

Hi Alex,

This is great news! Thanks a lot for the Palm OS version.

There will be a Palm version of Glaurung as soon as there is a decent set of Palm OS development tools available for Mac OS X. At the moment, the Palm OS Simulator is Windows only, which makes it very tedious to develop from OS X. It's a great pity; Glaurung was designed with handheld devices in mind since the beginning. I have taken great care to keep the code and data size to a minimum.

Tord

Hi Tord,

Size of the code is not the big problem since the code is stored in the permanent memory. However, there are some restrictions regarding the code. You will have to arrange your code in segments not larger then 64K since the Palm cannot do longer jumps.
Main problem is with the data that is kept in the dynamic heap area during the program execution. This data includes all global data + stack + anything you allocate (hash tables etc.). I think that global data also includes the GUI resources (bitmaps, forms, etc.).
For instance, if dynamic heap size is 256K (that was maximum on older Palms), I have only 32K left for all the hash tables (TT and pawn eval hash).

The good news is that all newer Palms have at least 1MB of dynamic heap. I don't think there are many of us left with old Palm models (mine is 5 years old). I am not going to support those old Palms any more, and this will make my life much easier.

One more thing. The endian order of bytes on Palm is opposite to the PC. This is actually causing me most of the work, since I have some speed tricks that depend on the endian order. And to make things more interesting, the new ARM processor uses PC endians, so you have to do the swaping when executing in ARM native mode.

I could send you the Naum's Palm GUI and book source code if you think it would help. You may be able to plug-in your engine code into it. Engine's source can work with almost no changes. You would probably need some work integrating with my time management (which is very primitive on my Palm version). Then I can try to compile it on my machine.

Regards,
Alex

Naum wrote:Size of the code is not the big problem since the code is stored in the permanent memory. However, there are some restrictions regarding the code. You will have to arrange your code in segments not larger then 64K since the Palm cannot do longer jumps.

This will not be a big problem for Glaurung, I think.

Main problem is with the data that is kept in the dynamic heap area during the program execution. This data includes all global data + stack + anything you allocate (hash tables etc.). I think that global data also includes the GUI resources (bitmaps, forms, etc.).
For instance, if dynamic heap size is 256K (that was maximum on older Palms), I have only 32K left for all the hash tables (TT and pawn eval hash).

This is a more difficult challenge, but I think it should still be possible to solve. Except for the TT and the pawn hash table, I don't have any big global data structures.

The good news is that all newer Palms have at least 1MB of dynamic heap. I don't think there are many of us left with old Palm models (mine is 5 years old). I am not going to support those old Palms any more, and this will make my life much easier.

I also don't intend to support old Palms. If it is not too difficult, I will probably make an ARM executable.

One more thing. The endian order of bytes on Palm is opposite to the PC. This is actually causing me most of the work, since I have some speed tricks that depend on the endian order. And to make things more interesting, the new ARM processor uses PC endians, so you have to do the swaping when executing in ARM native mode.

Glaurung makes no assumptions about endianness. I use exactly the same source code on Intel and PowerPC processors, without any #ifdefs depending on the platform. Of course, it is still possible that the swapping when switching between ARM and 68k modes could cause problems.

I could send you the Naum's Palm GUI and book source code if you think it would help. You may be able to plug-in your engine code into it. Engine's source can work with almost no changes. You would probably need some work integrating with my time management (which is very primitive on my Palm version). Then I can try to compile it on my machine.

Thanks a lot for your generous offer! Yes, I would very much like to have a look at your GUI and book code, and I will see if it is feasible to plug in my engine code. I am not sure I can use this for a public release, though. Glaurung is a GPL program, which means that I am probably not allowed to release a Glaurung for Palm OS with a non-GPL GUI.

Tord

Thanks a lot for your generous offer! Yes, I would very much like to have a look at your GUI and book code, and I will see if it is feasible to plug in my engine code. I am not sure I can use this for a public release, though. Glaurung is a GPL program, which means that I am probably not allowed to release a Glaurung for Palm OS with a non-GPL GUI.

Tord

Hi Tord,

Send me your email to naum_chess@yahoo.com and I will reply you with the code.
GPL and other licences are giving me a headache, so I don't think about them. You can do with the code whatever you like.

Alex

Hi Alex,

Unfortunately, it doesn't take much to give me a good beating and existing 1.2 is a good challenge for me at blitz.

I have Zire 21 as it has good battery life due to b&w display but not as fast as latest Tungsten models. On my Palm, I generally get 4 or perhaps 5 ply at this time control (10s). As an example, playing 1 h4 (I don't normally do this!) - Naum analysis is 5, 19, 9,5079.

Best wishes

Chris

Hi Chris,

Very interesting.
On your 120MHz Zire you are getting the same NPS as I am getting on my 33MHz Clie.

This would mean that if I manage to create an ARM version, it would probably run 3-4 times faster on your Zire. This is well worthed my time.

Thanks,
Alex

Hi Alex,

Go for it....

Best wishes

Chris

For those of you who are interested in this light version (made primarily for older Palms), I fixed some memory leaks that were causing crashes.

I also added more time control options and adjustable hash size.

New version can be found at http://www.geocities.com/naum_chess/palm

Alex

Hello ,

here are the first two games of Naum1.8 vs Chesstiger 15.1....

Naum was very close to a draw in th e second game...
enjoys :

[Event "Defi Naum 1.8."]
[Site "Palm Zire 21"]
[Date "2005.08.24"]
[Round "1"]
[White "Naum 1.8"]
[Black "Chess Tiger 15.1"]
[Result "0-1"]

1. e4 e5 2. Nf3 Nf6 3. Nxe5 d6 4. Nf3 Nxe4 5. Qe2
{
Naum choose the drawish variation, and Chess Tiger will be set out of his book.
}
5... Qe7 6. d3 Nf6 7. Bg5 Bg4 8. Bxf6 gxf6 9. Nbd2 Nc6 10. O-O-O O-O-O
{
Naum have a sligth advantage with his pawns but Chess Tiger have his two Bishop available
}
11. Kb1 Qd7 12. Ne4 Re8 13. Nxf6
{
the sharpest way to get a small advantage.
}
13... Rxe2 14. Nxd7 Rxf2 15. Nxf8 Rxf8 16. Rd2 Rxd2 17. Nxd2 Re8
{
chess tiger have weack pawn king side , but his pieces are active....
}
18. b3 Re1+ 19. Kb2 f5 20. Rg1
{
i prefer 20.g3 Bh3 21.Bg2
}
20... Bd1 21. c3
{
this move avoids either 21.Nb4 or 21.Nd4
}
21... Ne5 22. d4 Nd3+ 23. Ka3 Kd7 24. Ka4 Be2
{
pins the bishop in f1 and allows Rd1
}
25. c4 Nc1
{
25.Ka3 Rd1
}
26. Kb4 Nxa2+ 27. Ka3 Nc3 28. g4
{
to gain activity
}
28... fxg4 29. Rg3 Bxf1 30. Rxc3 Bh3 31. Rg3 Re2 32. Nf3 gxf3 33. Rxh3 f2
34. Rf3 c6 35. Kb4 h6 36. h4 Rc2
{
Naum have no more good move left.
}
37. Rf6 b6 38. Rf5 c5+ 39. dxc5 dxc5+ 40. Ka3 Ke6 41. Rf8 Ke5 42. Ka4 Kd4
43. Kb5 Rb2 44. Ka6 Ke3 45. Kxa7 Rxb3 46. Re8+ Kd4 47. Rf8 Rb2
0-1

[Event "Defi Naum 1.8"]
[Site "Bures Sur Yvette"]
[Date "2005.08.25"]
[Round "2"]
[White "Chess Tiger 15.1"]
[Black "Naum 1.8"]
[Result "1-0"]

1. d4 d5 2. c4 e6 3. Nf3 Nf6 4. Bg5 Be7 5. Nc3 O-O 6. e3 h6 7. Bh4 b6 8.
Bd3 Bb7 9. Bxf6 Bxf6 10. cxd5 exd5 11. O-O c5 12. Rc1
{
a very long book variation for both engines.... now Naum begins to think .
}
12... c4
{
now naum can try to get a pawn majority Queenside.
}
13. Bf5 g6 14. Bb1
{
the bishop stay on the main b1-h7 diag
}
14... Nc6 15. e4 dxe4 16. Nxe4 Bxd4
{
simplifying the game.... i prefer 16...b5
}
17. Nxd4 Qxd4 18. Nd6 Ba6 19. Nxc4 Bxc4 20. Qxd4 Nxd4 21. Rxc4 Ne2+ 22. Kh1
Rfe8
{
and both engines have his chance.
}
23. Rc7 Nf4 24. f3 Nd5 25. Rd7 Red8 26. Rb7 Ne3
{
good ! Naum activate his pieces.
}
27. Re1 Rd2
{
once again a good choice for naum.. the second row is invaded
}
28. g3 Rxb2 29. Bxg6 fxg6 30. Rxe3 Rd8
{
and naum gain a checkmate threat
}
31. Re1 Rxa2
{
naum gain a pawn but the position are aqual...
It is very difficult now for both engine to get a better position..
The white king and rook are blocked at the first row , and the black rook and king are blocked at the 8th row...
Naum may have some chance with his two passed pawns queenside , but there are impossible to support without deleting the checkmate threat.
This is why i think thiis game should be a draw .
}
32. f4 Rc8 33. Kg1
{
chess tiger rook in b7 cannot move , because then the b-pawn can advance freely...
The king is coming close to the black rook to free the rook in e1
}
33... Kh8 34. h4 Kg8 35. Kf1 Kf8 36. Rd1 h5 37. Rh7 Kg8 38. Rb7 Kh8
{
the king is safer in g8....
}
39. Re1 Rd8
{
i think that Kg8 is better
}
40. Rc1 Kg8 41. Ke1 Re8+ 42. Kf1 Kf8
{
42...Rd8 and this is a draw.
}
43. Rcc7
{
as the checkmate threat is over Chesstiger can invade the 7th row.
}
43... Ra8
{
a bad move....43...Rd8 is probably always a draw for naum.
43...Rd8 44.Ke1 Re8+ 45.Kd1 Rd8+ 46.Kc1 Re8
}
44. Rh7 Ra1+ 45. Kg2 Ra2+ 46. Kh3 Kg8 47. Rbg7+ Kf8 48. Rxg6 b5 49. Rg5 Re8
50. Rxb5
{
and chess tiger is better now.
}
50... Kg8 51. Rhxh5 Re3 52. Rhg5+ Kh8 53. Rb8+ Kh7 54. Rb7+ Kh6 55. Kg4 Ra4
56. Rh5+ Kg6 57. Rhh7 Rxg3+
{
to avoid a checkmate.
}
58. Kxg3 a5 59. h5+ Kf5 60. Rhf7+ Ke4 61. h6 Ra3+ 62. Kg4 Ra1 63. Rfe7+ Kd4
64. h7 Rg1+ 65. Kf5 Rh1 66. Re8 a4 67. h8=Q+ Rxh8 68. Rxh8
1-0



regards

Nicolas

Hi Nicolas,

Thanks for the test. It's a pitty, we can't run auto-tests on Palm with many games. I wonder how far Naum is from the CT.

CT is commerical, and probably optimized for low think depths on Palm, so it should be stronger then Naum. Naum is a simple port of the PC version. No adjustments done.
Time management is also not well adjusted to Palm, so Naum usually runs out of time and starts playing fast in the endgame.

Alex

Hello,

here are two more games of the Naum Chess Tiger Gauntlet.

in the first game , Naum was slowly outclassed by Chess Tiger , and his available space become smaller and smaller....
Then it seems to find a way to the draw ( or a better position ) but i think he miss something with 47.Rf1.

In the second game , Naum was succesfull in the middle game and achieve a drawish position in the endgame ( One rook and Three pawns on the same side for both engines) .

In four games Naum score only 1/2 point . This seems very severe , as Naum seems outclassed "only" in the endgame.... But i will continue to check Naum Strength!

enjoys the game :


[Event "Defi Naum 1.8"]
[Site "Zire 21"]
[Date "2005.08.26"]
[Round "3"]
[White "Naum 1.8"]
[Black "Chess Tiger 15.1"]
[Result "0-1"]

1. e4 Nc6 2. d4 d5 3. e5 f6 4. Nf3 fxe5 5. dxe5 e6 6. Bd3 Bc5 7. O-O Nge7
8. Nbd2 O-O 9. Nb3 Bb6 10. Bg5 h6
{
Naum's last move cause a weackening of black castle.
but i continue to prefer 10.Ng5 h6 11.Bh7+ Kh8 12.Qh5
}
11. Bd2 Nf5 12. c3
{
freing a square for the white Bishop and controling d4
}
12... a6 13. Qe2 Qe8 14. a4
{
to chase away the bishop naum weacken his pawn structure ...
i guess that to reach an equal position , Naum can try Bxf5 followed by Nbd4
}
14... Qf7 15. Bxf5 Qxf5 16. a5 Ba7 17. Nbd4 Qg4 18. Nxc6 bxc6 19. h3
{
Naum was tempted by this move.... but then the queen gain one tempi to become more active.
i suppose that Naum wants to avoids the following line ....
19.Be3 c5 and he cannot play 20.b4 because of the queen in g4
}
19... Qg6 20. b3 c5 21. Ra4
{
an original and good way to deploy the rook.
}
21... Rb8 22. Rg4 Qh7 23. Qd1 Bd7 24. Qc1
{
once again Naum threats Bxh6
}
24... h5 25. Ng5 Qg6
{
white mobility : 30
black mobility : 39
}
26. Rh4
{
here i prefer 26.Rg3 because Naum keep a threats vs the queen and a line such as :
Nd4-f6+ is then still possible
}
26... Rxb3 27. Rf4
{
naum will change his most actives pieces...
}
27... Bb5 28. Rxf8+ Kxf8 29. Rd1 Qf5 30. Bf4 Bc4
{
White Mobility : 26
Black mobility : 32
}
31. Kh1 Rb5 32. Qa1
{
the queen can move and leave the Bishop unprotected because of the line
32..... Qxf4?? 33.Nxe6+
}
32... c6 33. h4 Bb8 34. Re1 Bc7
{
chess tiger seems to take slowly th ebetter vs naum..
every advanced pawn or pieces are under attack.
}
35. Qc1 Bxa5 36. Rd1
{
simply avoid 36...d4
}
36... Rb3 37. Be3 Bxc3 38. Bxc5+ Kg8 39. f4 Be2 40. Rg1
{
Chess tiger invades Naum territory and reduces greatly his mobility
White mobility : 34
Black mobility : 41
}
40... Bb2
{
probably an error ! the white queen is now able to be activated
}
41. Qe1 Bd3 42. Qa5 Rb8
{
nearly a forced move.
}
43. Bd6 Qxf4 44. Nh3 Qxh4 45. Bxb8 Be4
{
Chess tiger have Four Pawn for the Rook, and his pieces are very active.... Not so bad...
}
46. Kh2 Bd4 47. Rf1
{
i didn't check but Qe1 may be better
}
47... Bxg2 48. Kxg2 Qg4+ 49. Kh2 Qe2+ 50. Rf2 Bxf2 51. Qd8+ Kh7 52. Ng5+
{
perhaps Qg5
}
52... Kg6 53. Qe8+ Kxg5 54. Qe7+ Kf4 55. Qf7+ Ke4 56. Qg6+ Kd4 57. Qxe6
Be3+ 58. Kh3 Qf1+ 59. Kh4 Qf4+ 60. Kh3 Qf3+ 61. Kh2 Bf4+ 62. Kg1 Qg4+
{
since several move Chess Tiger plan to change the queen.. the positin is now clearly a win for black.
}
63. Qxg4 hxg4 64. e6 Bg5 65. Bd6 c5 66. e7 Bxe7 67. Bxe7 a5 68. Kf2 a4 69.
Kg3 a3 70. Bf8 Kc4 71. Bxg7 d4
0-1

[Event "Defi Naum 1.8"]
[Site "Zire 21"]
[Date "2005.08.26"]
[Round "4"]
[White "Chess Tiger 15.1"]
[Black "Naum 1.8"]
[Result "1/2-1/2"]

1. c4 Nf6 2. d4 e6 3. Nf3 d5 4. Bg5 Be7 5. Nc3 h6 6. Bxf6 Bxf6 7. e3 O-O 8.
Rc1 c6 9. Bd3 Bd7
{
Naum move his bishop but it is still blocked by the pawn.. i would prefer to prepare the e5 advance with either Re8 and Nd7
}
10. O-O Qb6 11. Na4 Qa5 12. b4 Qxb4 13. Rb1 Qa3 14. Rxb7 Qxa2 15. Qa1 Qxa1
16. Rxa1
{
chess Tiger gave a pawn , but Naum have now two weack pawns in a7 and c6
}
16... Bc8 17. Rb3 Ba6 18. Nb2 c5 19. Ra5 cxd4 20. cxd5 Rc8 21. Bxa6 Nxa6
22. Nd3
{
22.Rxa6?? Rc1+ and the game is over
}
22... Rcb8 23. Rxb8+ Nxb8 24. exd4 Bd8 25. Rb5 exd5 26. Rxd5
{
which pawn is the best ? Not so easy to say
}
26... Be7 27. Rb5 a5 28. Nde5 f6 29. Nc4 Bb4 30. Nb6 Ra7 31. Kf1 Rc7 32. d5
Na6 33. d6 Bxd6 34. Rxa5
{
now the game is a draw.
}
34... Rc1+ 35. Ke2 Rc2+ 36. Kf1 Nc5 37. Ra8+ Kh7 38. Rc8 Kg6 39. Nd7 Rc4
40. Rc6 Rc1+ 41. Ke2 Be7 42. Nd4 Rc3 43. h3 Kf7 44. Nxc5 Rxc5 45. Rb6 Re5+
46. Kd3 Bc5 47. Rb7+ Kg6 48. f4 Rd5 49. f5+ Kh5 50. Kc4 Rxd4+ 51. Kxc5 Rf4
52. Kd6 Kh4
{
now Rxf5 is possible
}
53. Ke6 Re4+ 54. Kf7 Re2 55. Kxg7 Rxg2+ 56. Kxf6 Rg3 57. Ke5 Re3+ 58. Kf4
Re1 59. f6 Rf1+ 60. Ke5 Kxh3 61. f7 h5 62. Ke6 h4
{
the game is a draw and chess tiger compute this result .. not Naum.
}
63. Rc7 Kg4 64. Rc4+ Kg5 65. Rc8 Rxf7 66. Kxf7
{
Now Naum see the draw
}
66... h3 67. Rh8 Kg4 68. Ke7 Kg3 69. Rg8+ Kh4 70. Ke6 h2 71. Rh8+ Kg3 72.
Kd5 Kg2 73. Rg8+ Kf3 74. Rf8+ Kg2 75. Rg8+ Kf3 76. Rf8+ Kg2 77. Rg8+
1/2-1/2



regards

Nicolas