Toga II 3.1.2SE

[Shaun]

Toga II 3.1.2SE including source

Following the release of the source the SE versions are now included in the 40/4 rating list

Single-CPU engines

About Toga II 3.1.2SE

Modifications to Toga by Chris Formula

Readme

LEGAL DETAILS
-------------------------

This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or (at
your option) any later version.

This program is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
USA

See the file "copying.txt" for details.


GENERAL
----------------

The Special Edition or "SE" of Toga II based on Fruit 2.1 is being developed to
further increase the playing strength of the engine created by Thomas Gaksch.
Though Toga II is among the topnotch chess engines (and FREE) today, there
are still rooms for improvement on its search and evaluation. When Thomas
showed signs to quit computer chess, the "SE" project came to light to
continue what has been started by him. As of the moment, the "SE" project is
being done privately by Chris Formula.


EPD TESTING
----------------------

EET testsuite of Walter Eigenmann was used to evaluate improvement in the endgame
knowledge. Such testsuite was composed of 100 important test positions. Moreover, the
positional prowess of the engine was evaluated using the SSDF Positional testsuite of
Tony Hedlund. Such testsuite was composed of 16 important test positions. The total
points and the number of best move found (in parenthesis) were recorded. Below are
settings used and the results obtained during EPD testing. Amazing results!

Hash: 128M
Time Limit: 60 sec/move for EET; 600 sec/move for SSDF Positional
Software: Arena by Martin Blume
Bitbases: 3-4-man Scorpiobases by Daniel Shawul
Success Criteria:
(1) at least 50 points in EET testsuite
(2) at least 128 points in SSDF Positional testsuite

Version EET Testsuite SSDF Positional
1.3.1 48/100 116/160 (8/16)
1.3.4 45/100 105/160 (7/16)
2.1SE 51/100 117/160 (8/16)
2.2SE 53/100 88/160 (4/16)
3.1.2SE* 50/100* 128/160 (8/16)*


VERSION HISTORY
------------------------------

Toga II 3.1.2SE based on Fruit 2.1
- code was based on Toga II 1.3.1 based on Fruit 2.1
- improved search efficiency
- play style becomes more aggressive

Toga II 2.2SE based on Fruit 2.1
- code was based on Toga II 1.3.4 based on Fruit 2.1
- improved material evaluation

Toga II 2.1SE based on Fruit 2.1
- code was based on Toga II 1.3.1 based on Fruit 2.1
- improved material evaluation

Toga II 1.3.4 based on Fruit 2.1
Following changes and extensions made:
- improved history pruning
- improved futility pruning
- improved lazy evaluation
- new evaluation features (outpost knight etc.)
- MultiPV Mode
- This version performs similar level as version 1.3.1. However, its endgame is more accurate and uses EGBBs provided by Daniel Shawul, author of Scorpio.
- This version was previously known as 1.3x4b with some setting tweaks as follows.
a) Toga Extended History Pruning has been turned off
b) History Threshold has been adjusted to 70
- For Blitz play Toga Extended History Pruning should be turned on and the History Threshold adjusted to 80. These setting have shown to improve performance at blitz.

Toga II 1.3x4b based on Fruit 2.1
- bugfix of the endgame bitbases with help from Daniel Shawul

Toga II 1.3x4 based on Fruit 2.1
- improved evaluation
- improved search
- added support to endgame bitbases by Daniel Shawul

Toga II 1.3.1 based on Fruit 2.1
Following changes and extensions made:
- improved history pruning
- improved futility pruning
- improved lazy evaluation
- new evaluation features (outpost knight etc.)
- MultiPV Mode
- This is the best performing beta provided by Thomas and release with his permission.
- This version was previously known as 1.3beta1.
- For Blitz play Toga Extended History Pruning should be turned on and the History Threshold adjusted to 75. These settings have shown to improve performance at blitz.

Toga II 1.2.1a based on Fruit 2.1
- problems with internal opening book fixed
- playing strength is 100% identically to 1.2.1

Toga II 1.2.1 based on Fruit 2.1
- new settings
- some minor bugfixes

Toga II 1.2 based on Fruit 2.1
- more pruning
- more extensions
- new settings by Dieter Eberle


EGBBs Installation
----------------------------
Toga II uses the Scorpio endgame bitbases up to 5 pieces.

1. Download the 5-man bitbases from Leo Dijksman's WBEC site http://www.wbec-ridderkerk.nl . The egbbs are 340mb in size.
2. Store them anywhere in your computer. The default path is c:\egbb\ but you can change this in the uci options.
IMPORTANT: The egbbdll.dll must be in the same folder as the bitbase files.


ACKNOWLEDGMENT
--------------------------------

Huge thanks to Fabien Letouzey (France) for the great source code of the program Fruit 2.1.
Huge thanks to Thomas Gaksch (Germany) for the great source code of the program Toga II based on Fruit 2.1.

Special thanks to:
Dieter Eberle, Karl-Heinz Söntges, Shaun Brewer for testing Beta Versions.
A big thanks to Dieter Eberle for his setting.
Orlando Mouchel for bugfixes and new ideas.
Wilhelm Hudetz for the Logo.
Daniel Shawul for sharing his great endgame bitbases and his dll.
Walter Eigenmann for the EET testsuite.
Tony Hedlund for the SSDF Positional testsuite.
Martin Blume for the Arena software.
Denis Mendoza for the compile support for different platforms.

Without their help, the improved playing strength of Toga II based on Fruit 2.1 would not be possible.
Thank you! Maraming salamat!

Regards,
Thomas Gaksch (Germany)
Chris Formula (Philippines)

[Denis P. Mendoza]

Thanks Shaun. This code has no added algorithm but an improved revision of material evaluation and search efficiency (similar to 1.3.4 without egbb). Though I only saw small changes on the code from version 1.3.1, this one sure brought a big leap in strength on the original Toga. It worked, so I congratulate Chris Formula for this.

Just to add something, Intel pc users should use the original compile by Chris based on his feedbacks that his compile performed better than mine. AMD users could benefit on my ICC compiles. I maybe wrong, but there's plenty of exe to choose from .

[Denis P. Mendoza]

From our discussion Shaun, here are my new compiles for Toga 3.1.2SE for 64-bit. I only posted these as they showed about 8-10% speed increase on my preliminary tests. Other testers could try these builds and submit their findings. Here they are : http://mihd.net/lfn96r
---------------------------------
Quoted from my last email:
---------------------------------
Here are the attached files.

1.Toga64_SSE2: for P4 and above, compatible with Intel/non-Intel processors (correction from last email)
2.Toga64_SSE3: for P4 and above, compatible with Intel/non-Intel processors (correction from last email)
3.Toga64_Intel_C2_SSE3: for Intel Core 2 Duo and Other Intel processors based on Intel® CoreT
microarchitecture or Intel NetBurst® microarchitecture
4.Toga64_Intel_C2Q_SSE3: for Intel Core 2 Quad and Other Intel processors based on Intel® CoreTmicroarchitecture or Intel NetBurst® microarchitecture

I have no idea if the "Core 2" compiles are ok since it doesn't run on my AMD pc. This is quite
a good project for me as my previous 64-bit compiles of the Toga and Fruit codes were similar to
Bryan Hoffman's findings - slow speeds on 64-bit. I'll also try some builds on the last updates for v1.3.1 and 1.3.4 and see what happens.
----------------------------------------------------------------------------------------------------------------------------------------------------------------------
Thanks.

[Denis P. Mendoza]

I just reviewed my Toga archives right this minute to verify something after compiling and comparing my 64-bit Toga 1.3.1 with my submitted 32-bit (DM1/2 versions). With all these Toga compiled versions I entered in Arena GUI, I noticed that I compared the newly compiled Toga 3.1.2SE 64-bit with Chris's 32-bit compile, which should be my own 32-bit compile using the same command switches in the Intel compiler. On my pc. Chris's compile was much slower than my 32 bit build. My 32 and 64 bit SSE2 compiles were almost the same on my speed tests. My test procedure was wrong. My mistake pal as I didn't verify which is which at first. My excitement covered the truth at that period. Just informing!

[Eelco de Groot]

Hello CCRL,

Thanks for all the interesting news about Toga, Shaun thanks for your testing of the Toga II 3.2.1 SE and other Toga versions, Charles for the new version and Denis for your compiles of it!

Here is my own latest Beta version with Toga Checkov settings and some minor changes in the sourcecode, I hope it is alright to upload it here, because attachments are possible in this forum and I can't do that in CCC.

It is just a general compile and just a Beta, not optimized with Profile Guided Optimization or something like that, it should work on any AMD or Intel machine with regular X86 processor but don't expect it to be as fast as Denis' compiles.


Eelco

[Shaun]

Hi Eelco,

I was thinking of giving this a try but I get the following error running the exe stand alone...

I am running Windows XP 64-bit on this machine an Intel Core 2 Duo E4300.

Shaun

[Eelco de Groot]

Hi Shaun,

Hmmm that is a dificult question I think for me to answer. I think it is caused because I compiled it with MSVC 2005 and I can find some other posts that this sometimes happens. The compiled program I think can't find certain DLLs on your PC that are actually present in most cases, but you would have to look them up. It would probably be better if maybe Denis could make a compile especially for a Core 2 Duo for you, it would be a bit faster and he has a better compiler I think.

I could try to make a new .exe with my free MVSC 2005 Express that incorporates more .DLLs but I'm very much a newbie at this, I don't know if I have enough compiler experience to do that right and if the 2005 Express version supports Windows XP 64-bit well enough for that, and if I could find the right instructions on the web.

Regards, Eelco

[Denis P. Mendoza]

Eelco/Shaun,
I just downlloaded your code. I'll start working on it and post it here. FYI.

[Denis P. Mendoza]

Here it is! Toga Checkov was built with Intel C/C++ v10.1 compiler. I made 4 versions from Pentium 3 to Core 2 Duo. All are compatible with Intel pc (of course ). Only P3 and P4 have command options to run on non-Intel processors like AMD. Just a reminder. Happy testing guys, feel free to notify for any comments.

I'll build an extra compile for older pc as well later (below Pentium III). I just need to prepare my old pc for this (for verification). I'm making it a standard in the future as there are still many computerchess afficionados using softwares on old hardware.

[Denis P. Mendoza]

Hi Shaun,

Hmmm that is a dificult question I think for me to answer. I think it is caused because I compiled it with MSVC 2005 and I can find some other posts that this sometimes happens. The compiled program I think can't find certain DLLs on your PC that are actually present in most cases, but you would have to look them up. It would probably be better if maybe Denis could make a compile especially for a Core 2 Duo for you, it would be a bit faster and he has a better compiler I think.

I could try to make a new .exe with my free MVSC 2005 Express that incorporates more .DLLs but I'm very much a newbie at this, I don't know if I have enough compiler experience to do that right and if the 2005 Express version supports Windows XP 64-bit well enough for that, and if I could find the right instructions on the web.

Regards, Eelco

With your stated problem, you just need a very useful utility to find which dll is called to run each application, etc. Try googling "Dependency Walker". It is free and dependable.

[Eelco de Groot]

Here it is! Toga Checkov was built with Intel C/C++ v10.1 compiler. I made 4 versions from Pentium 3 to Core 2 Duo. All are compatible with Intel pc (of course ). Only P3 and P4 have command options to run on non-Intel processors like AMD. Just a reminder. Happy testing guys, feel free to notify for any comments.

I'll build an extra compile for older pc as well later (below Pentium III). I just need to prepare my old pc for this (for verification). I'm making it a standard in the future as there are still many computerchess afficionados using softwares on old hardware.

Thanks very much Denis! I hope the Core 2 Duo compile can solve some of the problems that Core 2 Duo owners Shaun and Eduard reported with the original. I think these compiles should be a bit faster than mine, I did not use any of the options in the make file or any other optimizations!

I would like to stress that it is amazing how good Fabien Letouzy's original code is holding up, as I wrote to you I did not yet add any new instruction in the code, only some of the numbers were changed. Thomas Gaksch has not made much changes to the evaluation either as far as I know, at least not in Toga II 1.3.1 that was the basis for the 3.1.2 SE. More search changes that can make Toga's searches deeper. So the evaluation and all the basic program structures are still pretty much the same as in Fruit 2.1. And it is still going strong in 2008! No need for any rewrites by Fabien

Regards, Eelco

P.S. thanks for the tip about Dependency Walker, I believe I saw it mentioned before in some thread about this problem, but I am just new to compiling, just glad I actually got MSVC 2005 to work. It is fascinating stuff though!

[Shaun]

Here it is! Toga Checkov was built with Intel C/C++ v10.1 compiler. I made 4 versions from Pentium 3 to Core 2 Duo. All are compatible with Intel pc (of course ). Only P3 and P4 have command options to run on non-Intel processors like AMD. Just a reminder. Happy testing guys, feel free to notify for any comments.

I'll build an extra compile for older pc as well later (below Pentium III). I just need to prepare my old pc for this (for verification). I'm making it a standard in the future as there are still many computerchess afficionados using softwares on old hardware.

Thanks Denis!!!

Shaun

[Vempele]

I think it is caused because I compiled it with MSVC 2005 and I can find some other posts that this sometimes happens. The compiled program I think can't find certain DLLs on your PC that are actually present in most cases, but you would have to look them up.

It's because MSVC's default is to link the CRT dynamically. Add /MT to the command line or a link to the Visual C++ 2005 redistributable package (a ~2.5MB download) in the readme.

[Denis P. Mendoza]

Yes, Vempelle is right. In most cases, you just need the updated Microsoft Visual C++ 2005 Redistributable Package (x86) and for x64 as well if you use it.

[Denis P. Mendoza]

As promised, I built a couple of fast compiles for older pc (Pentium1 and PentiumII). I hope older AMD computers are compatible with these files too. Just pick what best standard fits your pc.

[Denis P. Mendoza]

I'm still unsure if my packer works fine, so I'll just have to upload the uncompressed files just to be sure.

Shaun, better test the PI version as I'm amaze of its speed!

[Eelco de Groot]

Hello Denis and others,

Here are the sources for the Toga 1.4 Checkov Beta 3 version. It is still a Beta version and not well tested, but I thought it would be nice to share it. My philosophy is that in the end a strong open source program will not hurt sales of the best commercial programs because any worthwhile ideas should be implementable in other programs as well, and the commercial programmers can probably get more out of it than the amateurs, for instance by using bitboard versions. The commercial programs will maybe only have an incremental advantage but I think customers are willing to pay for that advantage.

-Eelco

These Beta 3 sources are rather messy as far as I made any changes, but there are not that many changes that it should be confusing I think. Work in progress! I do hope the MP functionality was not affected by changes made, I can not test that on my computer. I guess I will hear it soon if it does not work...

[Denis P. Mendoza]

Hello Denis and others,

Here are the sources for the Toga 1.4 Checkov Beta 3 version. It is still a Beta version and not well tested, but I thought it would be nice to share it. My philosophy is that in the end a strong open source program will not hurt sales of the best commercial programs because any worthwhile ideas should be implementable in other programs as well, and the commercial programmers can probably get more out of it than the amateurs, for instance by using bitboard versions. The commercial programs will maybe only have an incremental advantage but I think customers are willing to pay for that advantage.

-Eelco

These Beta 3 sources are rather messy as far as I made any changes, but there are not that many changes that it should be confusing I think. Work in progress! I do hope the MP functionality was not affected by changes made, I can not test that on my computer. I guess I will hear it soon if it does not work...

Here is your Toga1.4Checkov_beta3 for further testing. Multithreading works well for me. I'll just attach it. FYI.

[Denis P. Mendoza]

I've been testing some new compiling switches which made some additional boosts on my single-cpu builds. On my Athlon 64 pc, It worked fine, so I made an improved version of Toga 3.1.2SE. It's faster than my previous sets. Thanks.

[jd1]

Hi,

I have made a new Toga version that has done well in my tests.
Would anyone be interested in this?
If so, I can attached it here.

Thanks,
J Donald

[Shaun]

Hi J,

there are currently a lot of new Toga versions - I would suggest contacting Thomas or Chis and try and work collaboratively...

All the best

Shaun

[Gabor Szots]

Hi J,

there are currently a lot of new Toga versions - I would suggest contacting Thomas or Chis and try and work collaboratively...

All the best

Shaun

Hi all,

I am going to release my Toga III 6.5.9 FE (4th edition) engine before long. I suggest you guys do the same else you would be the only ones who will not have their own Toga.

[jd1]

Hi all,

For those interested, here is my Toga II 1.4.2JD source.
I would appreciate if someone could compile it for me as I can't without disabling the egbb support!
I hope the Mp-support still works.

All the best,
J Donald

[Eelco de Groot]

Hi all,

For those interested, here is my Toga II 1.4.2JD source.
I would appreciate if someone could compile it for me as I can't without disabling the egbb support!
I hope the Mp-support still works.

All the best,
J Donald

Hello Jerry,

Nice experiment with Toga! I think the bitbases access should not be a problem with your code changes, I made a standard compile with MSVC 2005 Express and it loads the Scorpio bibases okay:

Code: Select all

 
Toga II 1.4.2JD UCI based on Fruit 2.1 by Thomas Gaksch and Fabien Letouzey. Settings by Dieter Eberle
Experimental engine by J Donald. Code was based on Toga II 1.4.1 SE by Thomas Gaksch and Chris Formula
EgbbProbe 3.0 by Daniel Shawul
Cache Size = 16 Mb (2070 entries)
Egbbs loaded !


8/4k3/8/8/8/8/3BB3/4K3 w - -


8/4k3/8/8/8/8/3BB3/4K3 w - -

Engine: Toga II 1.4.2JD bb5men (256 MB)
by Thomas Gaksch and Fabien Letouzey

33/24 2:03 +41.99 1.Kf2 Kd6 2.Ke3 Kd5 3.Kf4 Kd6 4.Ke4 Ke6
5.Bb4 Kd7 6.Kd5 Kc7 7.Bc4 Kd7 8.Bb5+ Kc7
9.Bc5 Kd8 10.Kd6 Kc8 11.Bb6 Kb7 (93.314.854) 1515

34/26 2:48 +41.99 1.Kf2 Kd6 2.Ke3 Kd5 3.Kf4 Kd6 4.Ke4 Ke6
5.Bb4 Kd7 6.Kd5 Kc7 7.Bc4 Kc8 8.Kc6 Kd8
9.Bb3 Ke8 10.Bc5 Kd8 11.Bf7 Kc8 (127.643.401) 1513

34/26 3:30 +42.00 1.Bc4 Kf6 2.Ke2 Kf5 3.Bc3 Ke4 4.Kf2 Kf5
5.Kf3 Kg5 6.Bd3 Kh5 7.Kf4 Kh4 8.Bd2 Kh5
9.Kf3 Kh4 10.Bg6 Kh3 11.Bg5 Kh2 (160.905.270) 1526

35/26 4:15 +40.41 1.Bc4 Kf6 2.Ke2 Kf5 3.Bc3 Ke4 4.Kf2 Kf5
5.Kf3 Kg5 6.Bd3 Kh5 7.Kf4 Kh4 8.Be2 Kh3
9.Bd4 Kg2 10.Kg4 Kh2 11.Bf1 Kh1
12.Kf3 (195.011.076) 1516

36/28 6:30 +41.20 1.Bc4 Kf6 2.Ke2 Kf5 3.Kf3 Ke5 4.Be3 Kd6
5.Ke4 Ke7 6.Ke5 Kd7 7.Bb5+ Ke7
8.Bc5+ Kf7 9.Kf5 Kg7 10.Be8 Kh6
11.Kf6 Kh7 12.Be3 Kg8 (299.557.572) 1534

37/28 7:54 +41.19 1.Bc4 Kf6 2.Ke2 Kf5 3.Kf3 Ke5 4.Be3 Kd6
5.Ke4 Kd7 6.Ke5 Ke7 7.Bc5+ Kd7
8.Bb5+ Kc7 9.Ke6 Kc8 10.Kd6 Kb7
11.Kd7 Ka8 12.Kd8 Kb7 (365.989.576) 1544

38/27 10:52 +39.60 1.Bc4 Kd6 2.Ke2 Ke7 3.Ke3 Kd6 4.Kd4 Ke7
5.Ke5 Kd7 6.Bb5+ Ke7 7.Bb4+ Kf7
8.Kf5 Kg7 9.Bc4 Kh6 10.Bf7 Kg7
11.Ke6 Kh6 12.Kf6 Kh7 13.Bf8 (493.415.315) 1512

38/27 18:40 +39.61 1.Kf2 Kf6 2.Ke3 Kf5 3.Kd4 Ke6 4.Ke4 Kd6
5.Bb4+ Kc6 6.Ke5 Kc7 7.Kd5 Kd7
8.Bb5+ Kc7 9.Ba5+ Kb7 10.Kd6 Kc8
11.Ba6+ Kb8 12.Bb6 Ka8 13.Kc6 (776.245.931) 1385

best move: Ke1-f2 time: 20:10.109 min n/s: 1.383.725 CPU 100.0% n/s(1CPU): 1.383.725 nodes: 1.674.210.000

With (or without) bitbases Toga does not see mate with two Bishops very quickly, Kf2 is the best move, it is a mate in 16 here says Shredder Online Endgame Database but the bitbase access seems okay. Jerry if you use Microsoft Visual C++ in the Solution properties I think all you have to change is under Project Defaults -> Character Set: Not Set so the .exe does not expect only ANSI or Unicode characters as input. With other compilers maybe a similar difficulty could arise but this I suppose could have been the problem? I get still 56 Warnings from MVSC but for my own version it is typically 65 Warnings in Release mode so you are better already!

Maybe Dennis Mendoza can make a MP compile for you! I only could test your code on a single CPU Athlon.

Regards, Eelco

[jd1]

Hi Eelco,

Actually I am not using MVSC, and when I try to compile it crashes on "dlopen". I might give MVSC a try! Regarding a Mp compile, how could I get in touch with Denis? Thank you very much.

All the best,
Jerry

P.S. Just ran a gauntlet with Toga 1.4.2JD and Toga 1.4.1SE:
Toga 1.4.2JD 162.5/300
Toga 1.4.1SE 152/300