Doch 1.3.1

[Don]

I want to make all of this work on any UCI interface. I am basically a unix person so I'm not very clear on certain windows glitches and/or practices. I had forgotten that in windows, the extension to a file is often assumed, this is much less so on unix.

The personality file can have any name and any extension and it is not assumed or put there for you. I use ".per" just as a convention to organize what I'm doing and so that I know at a glance it is a personality file.

Also, the program does not require there to be a personality file and it knows nothing about sample.per. After I put in the personality support in Doch I later made the decision to put in a sample file in the distribution and sample.per was an arbitrary name I made up.

I asked Jim Ablett to make yet another version that has better error checking and sends the info string command to the user interface to reassure the user that the file loaded and also to inform the user of how many parameters were actually loaded (and the total expected.) I don't want to release that version until I know it's right. So if anyone with windows wants to help check it out, assuming Jim is willing to continue to go to the trouble and produces yet another version, let us know.

Alternately, I can give you 32 and 64 bit versions that are compiled with mingw but these compiles are not as fast as the Jim Ablett versions. I'm not sure solving problems with mingw always translates to the JA version.

Don Dailey

The 32 bit version 1.3.2 does not work in Vista-32, in Arena 201 or Fritz 10. Apparently, the engine doesn't start it's calculations (no Doch process visible in the Process Explorer). My cpu is Intel D945 (~dual P4).

Version 1.3 works bascially, with the exception that it crashed in Fritz after I removed the filename of a test.per from the engine properties. Can that input field be empty, or does it have to contain "<empty>" if I don't want to specify a custom .per file?

Also, the console mode seems not to work with Doch 1.3/1.3.2 in Vista 32 bit. The console window appears, but does not accept or display any keyboard input. I don't know if this is a missing feature or another bug.

Anyway, thanks to the programmers and compilers! I understand that there is big effort behind it, to make engines run so smoothly as they usually do. From my viewpoint, no need to hurry whatsoever. The price-performance ratio is unrivaled.

The 32 bit version 1.3.2 works fine for me in WinXP and Arena 1.99. The personality filename must be specified as well as the extension, for example "sample.per" or "test.per".

Sample game, from the start position, default settings -

[Event "Zappa - Doch"]
[Site "?"]
[Date "2010.01.05"]
[Round "6"]
[White "Doch 1.3.2"]
[Black "Zappa Mexico II"]
[Result "1-0"]
[ECO "A09"]
[Opening "Reti Opening"]
[Time "23:26:26"]
[Variation "Advance Variation, Anglo-Polish Attack"]
[TimeControl "40/300:40/300:40/300"]
[Termination "normal"]

{C:\Arena\PgnBook\Basic.pgn 3/31 (?-?, ?, ????, *)} 1.Nf3 {+0.18/13} d5 {-0.14/12} 2.c4 {+0.42/13} d4 {+0.08/11} 3.b4 {+0.40/14} c5 {+0.10/11} 4.bxc5 {+0.27/13} Nc6 {+0.19/13} 5.d3 {+0.29/14} e5 {+0.38/12} 6.Qa4 {+0.30/13} Bd7 {+0.33/12} 7.Nbd2 {+0.32/12} Nf6 {+0.40/12} 8.g4 {+0.18/12} Nxg4 {+0.61/11} 9.Ne4 {+0.22/12} f5 {+1.00/12} 10.Nd6+ {-0.13/9} Bxd6 {+1.00/11} 11.cxd6 {-0.16/10} e4 {+1.00/11} 12.h3 {-0.90/12} Nxf2 {+1.14/12} 13.Bg5 {-1.00/13} Ne5 {+1.38/12} 14.Nxe5 {-0.76/15} Bxa4 {+0.97/12} 15.Bxd8 {-0.74/11} Nxh1 {+0.97/12} 16.Bh4 {-0.71/14} Rg8 {+0.96/12} 17.Bg5 {-0.48/14} Ng3 {+1.04/12} 18.Bg2 {-0.40/14} h6 {+1.07/12} 19.Be7 {-0.45/15} Rc8 {+1.02/12} 20.Kf2 {-0.61/14} Nh5 {+1.16/12} 21.dxe4 {-0.91/14} Nf4 {+1.17/12} 22.exf5 {-0.86/14} Nxg2 {+1.69/11} 23.Kxg2 {-0.93/12} Rc5 {+1.69/11} 24.Nf3 {-0.95/15} Rxc4 {+1.65/12} 25.Kg3 {-0.84/14} Bd7 {+1.50/11} 26.Kf4 {-0.62/14} Ra4 {+1.42/11} 27.Ne5 {-0.45/13} Bc6 {+1.03/11} 28.Nxc6 {-0.52/13} bxc6 {+1.03/10} 29.Rb1 {-0.47/14} Kd7 {+0.87/11} 30.Ke5 {-0.48/15} Rxa2 {+1.02/11} 31.Rb7+ {+0.77/16} Kc8 {+1.22/6} 32.Rc7+ {+1.10/16} Kb8 {+1.24/7} 33.Ke6 {+1.11/17} Rxe2+ {-1.02/10} 34.Kf7 {+1.35/17} Rh8 {-1.02/10} 35.Rxc6 {+1.34/16} Re5 {-0.90/11} 36.Kxg7 {+0.93/15} Rc8 {-1.27/12} 37.Rxc8+ {+0.95/15} Kxc8 {-1.27/12} 38.f6 {+0.63/18} Rf5 {-1.27/12} 39.f7 {+1.28/20} Rxf7+ {-1.27/11} 40.Kxf7 {+1.61/18} Kd7 {-1.27/11} 41.Bf6 {+3.12/23} d3 {-4.18/15} 42.Bc3 {+3.47/25} Kxd6 {-4.90/17} 43.h4 {+3.65/26} Kc6 {-4.90/17} 44.h5 {+3.68/26} Kb5 {-4.90/17} 45.Ke6 {+3.87/26} a5 {-5.42/18} 46.Kd5 {+6.30/28} a4 {-5.57/18} 47.Kd4 {+6.33/29} Kc6 {-6.70/18} 48.Kxd3 {+6.33/23} Kd7 {-12.45/18} 49.Bg7 {+5.03/24} Ke7 {-12.45/18} 50.Bxh6 {+6.36/28} Kf6 {-6.70/16} 51.Kc4 {+6.48/28} Kf5 {-7.10/17} 52.Bd2 {+6.48/27} a3 {-10.18/17} 53.Kb3 {+6.35/27} a2 {-15.00/18} 54.Kb2 {+6.31/26} Ke4 {-14.99/17} 55.h6 {+6.50/17} Kd3 {-M13/16} 56.Ba5 {+M16/17} a1=Q+ {-M11/16} {Black resigns} 1-0

[Mike S.]

The personality filename must be specified as well as the extension, for example "sample.per" or "test.per".

Thanks. I tried it once more but it doesn't work. I think Arena read the uci_about string correctly, which appears in the enigne output window. But as soon as the engine should start to calculate, it seems to crash somehow, also with a .per file correctly specified.

So it is either a Vista(32) problem, or some kind of individual problem. But version 1.3 works fine here.

Edit: Carlos' posting he just wrote, suggests that it could be a problem related to the Pentium D cpu we both have. Maybe it has to do with cpu related compiler options or the like, different to 1.3?

P.S. After 13 moves of your test game, both queens, three knights and the rook h1 were en prise at the same time! That seems unique:

[D]r2qk2r/pp1b2pp/3P4/4npB1/Q1Ppp3/3P1N1P/P3Pn2/R3KB1R w KQkq - 0 14

[Carlos777]

Doch 1.3.2 w32 by JA does not work here either on Arena 2.0.1.

Pentium D - Windows XP sp2 32 bits.

Using or not the personality file does not make a difference. It just does not work.

Carlos

[Paloma]

Hi Don,

have you noticed the 64-bit version to appear as >DOCH32 1.3.2 JA< in the Engine window?

[Carlos777]

Thanks. I tried it once more but it doesn't work. I think Arena read the uci_about string correctly, which appears in the enigne output window. But as soon as the engine should start to calculate, it seems to crash somehow, also with a .per file correctly specified.

So it is either a Vista(32) problem, or some kind of individual problem. But version 1.3 works fine here.

Edit: Carlos' posting he just wrote, suggests that it could be a problem related to the Pentium D cpu we both have. Maybe it has to do with cpu related compiler options or the like, different to 1.3?

Might be. OTOH, this new version works fine for me:

Here is a Windows 32 bit compile of Doch 1.3.2 for you to test.
It is compiled using the Linux src code & Gcc-cygwin compiler.

Speed in comparable to Intel 32 bit compile - maybe slightly slower.

Download:
http://www.mediafire.com/?gfvlyf1bgdw

Jim.

Regards,
Carlos

[Graham Banks]

Is anybody having a problem in any GUI other than Arena?
Fritz and ChessGUI seem to be fine? Not sure about Shredder or others.
Definitely stronger than Doch 1.2.

[Don]

Is anybody having a problem in any GUI other than Arena?
Fritz and ChessGUI seem to be fine? Not sure about Shredder or others.
Definitely stronger than Doch 1.2.

Just please don't load any personalities, even the default one.

How must stronger are you showing now? I'm sure the error margin is high, but I'm curious.

[Graham Banks]

Is anybody having a problem in any GUI other than Arena?
Fritz and ChessGUI seem to be fine? Not sure about Shredder or others.
Definitely stronger than Doch 1.2.

Just please don't load any personalities, even the default one.

How must stronger are you showing now? I'm sure the error margin is high, but I'm curious.

I haven't tried to load any personalities. I've only tested using what shows in the parameter screen at load up (except changing the hash). The personalities option shows up as empty and I just leave it at that.

So far in my replayed games of Super Tournament X, Doch 1.3.2 has 13.0/24, a much better performance. Small number of games, but I've watched many and it is definitely stronger.

In the gauntlets that I'm running, Doch has been similarly solid and difficult to crack (probably the wrong word to use considering what's been going on around here lately).

Should be around 100 games played by Friday's rating list update. Can't put a figure on numbers for sure, but I'm picking that it will be a 40+ ELO improvement.

Keep up the good work.

Cheers,
Graham.

[MirceaH]

Hi D!

Doch 1.3.2 32bit unloads in 1'+1" games after getting out of book moves (or PGN moves from a test suites). I use Arena 2.0.1. The 64bit version does not have this problem.

However normal time controls work well for the 32bit version, so I was "forced" to start the tests using a weird 40/1' repeated time control.

Anyway, I must say that I consider Doch (besides Stockfish and Spark) the most wonderful presents for us CC "freaks" since Fruit appeared. (Well, the appearance of Ruffian and of course Rybka were also very big steps ahead).

The first impressions are that Doch 1.3.2 x64 seem close to or above the Thinker league (the latter x64 2 cores).

[Jouni]

Still no current move display?

Jouni

(500 posts )