Tucano

Tucano Chess Engine

This is my chess engine Tucano (Toucan in english). It is the name of a colorful bird with an oversized bill, and is popular in my country Brazil. I'm an IT professional that likes chess and programming, so I combined both with the development of the development of a chess engine. I used the information available on the internet, specially, other engines source code, such as Fruit, Crafty, Stockfish, Rodent, Olithink, Sungorus, Tscp, Ethereal, Demolito and Xiphos. Thanks to all developers that made this knowledge available. The main source of information for development of chess engines can is the Chess Programming Wiki Pages (https://www.chessprogramming.org/Main_Page). In case of any question please send me an email ([email protected]) or stop by talkchess.com.

Tucano can be downloaded from github: https://github.com/alcides-schulz/Tucano

Neural Network Evaluation

Starting with version 10 release, Tucano uses a neural network evaluation, which increases the engine strength. This network was trained on about 1.5 billion positions using Tucano's evaluation at depth 8 and Nodchip trainer's code (https://github.com/nodchip/Stockfish). The neural network access code is from Daniel Shawn nnue library (https://github.com/dshawul/nnue-probe). In order to use neural network you need to use an evaluation file, currently tucano_nn01.bin, that can be found in the Tucano’s github release section.

When running tucano, it will try to load the file from the same folder where tucano executable is. You can load a different file by using a new command line parameter, e.g.: tucano -eval_file tucano_nn01.bin. Also, you can send the file name through the new UCI option "name EvalFile type string".

It is important to make sure tucano can locate the eval file. To validate if the file is loaded you can start tucano on your system and see if it shows the message below:

    Tucano chess engine by Alcides Schulz - 10.00 (type 'help' for information)

    Eval file 'tucano_nn01.bin' loaded !
        hash table: 64 MB, threads: 1

It is recommend to run with the neural network evaluation, othwerwise the performance will not be much better than previous version 9.

Terms of use

Tucano is free, and distributed under the GNU General Public License (GPL). Essentially, this means that you are free to do almost exactly what you want with the program, including distributing it among your friends, making it available for download from your web site, selling it (either by itself or as part of some bigger software package), or using it as the starting point for a software project of your own.

The only real limitation is that whenever you distribute Tucano in some way, you must always include the full source code, or a pointer to where the source code can be found. If you make any changes to the source code, these changes must also be made available under the GPL.

For full details, read the copy of the GPL found in the file named copying.txt.

Notice: this is free software, there is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Have Fun.

Alcides Schulz.

Running Tucano

Version 8.0 and earlier supports XBoard protocol. Starting from version 9.00 it supports UCI protocol, which is the recommended protocol to use.

Syzygy endgame tablebases

Starting on version 8.00, tucano supports syzygy (option SyzygyPath). It leverages Fathom from Jon Dart.

- Syzygy: endgame tables bases by Ronald de Man

- Fathom: syzygy probing tool by Jon Dart.

UCI Protocol options

Tucano supports the following uci options:

option name Hash type spin default 64 min 8 max 65536
option name Threads type spin default 1 min 1 max 256
option name SyzygyPath type string default <empty>
option name Ponder type check default false
option name EvalFile type string default <empty>

Command Line options

tucano -hash N -threads N -syzygy_path F -eval_file E

-hash indicates the size of hash table, default = 64 MB, minimum: 8 MB, maximum: 65536 MB.
-threads indicates how many threads to use during search, minimum: 1, maximum: 256. Depends on how many cores your computer have.
-syzygy_path indicates the folder of syzygy endgame tablebase.
-eval_file indicates the location of neural network file used by evaluation.

Signature

If you compile tucano you can use the command "bench" to get a signature. Just start tucano and type "bench". Signature is a number generated after searching a couple of positions to indicate you have the correct compilation. If you don't get the correct signature it means that something is wrong with the compilation process and the program may not perform correctly.

10.00:  5734637 (with nn eval file loaded)
 9.00: 21898211
 8.00: 32406478

Executable

The release will come with executables for different plataforms, try to use the one that providers the most nodes per seconds. Start tucano and type the command "speed", you should see the following report:

    Tucano chess engine by Alcides Schulz - 10.00 (type 'help' for information)

    Eval file 'D:\ChessProg\tucano\tucano_nn01.bin' loaded !
       hash table: 64 MB, threads: 1

    speed
    running speed test 1 of 5...
    running speed test 2 of 5...
    running speed test 3 of 5...
    running speed test 4 of 5...
    running speed test 5 of 5...

    Speed test: nodes per second average = 1376k

Compilation

The main platform used for development is Windows. I try to use standard functions and make tucano portable, so it can be compiled on other platforms, but there's no warranty it will work on all of them. You can report issues with other platforms and I will try to address as possible.

Here are the commands used for compilation:

Windows (compiled using mingW version 7.2.0)

    AVX2
    gcc -o tucano_avx2.exe -DEGTB_SYZYGY -DTUCANNUE -O3 -Isrc -flto -m64 -mtune=generic -s -static -Wall -Wfatal-errors -DUSE_AVX2 -mavx2 -DUSE_SSE41 -msse4.1 -DUSE_SSSE3 -mssse3 -DUSE_SSE2 -msse2 -DUSE_SSE -msse src\*.c src\nnue\*.cpp src\fathom\tbprobe.c
    
    SSE4.1
    gcc -o tucano_sse4.exe -DEGTB_SYZYGY -DTUCANNUE -O3 -Isrc -flto -m64 -mtune=generic -s -static -Wall -Wfatal-errors -DUSE_SSE41 -msse4.1 -DUSE_SSSE3 -mssse3 -DUSE_SSE2 -msse2 -DUSE_SSE -msse src\*.c src\nnue\*.cpp src\fathom\tbprobe.c
    
    SSE3
    gcc -o tucano_sse4.exe -DEGTB_SYZYGY -DTUCANNUE -O3 -Isrc -flto -m64 -mtune=generic -s -static -Wall -Wfatal-errors -DUSE_SSSE3 -mssse3 -DUSE_SSE2 -msse2 -DUSE_SSE -msse src\*.c src\nnue\*.cpp src\fathom\tbprobe.c
    
    SSE2
    gcc -o tucano_sse4.exe -DEGTB_SYZYGY -DTUCANNUE -O3 -Isrc -flto -m64 -mtune=generic -s -static -Wall -Wfatal-errors -DUSE_SSE2 -msse2 -DUSE_SSE -msse src\*.c src\nnue\*.cpp src\fathom\tbprobe.c
    
    OLD
    gcc -o tucano_old.exe -DEGTB_SYZYGY -DTUCANNUE -O3 -Isrc -flto -m64 -mtune=generic -s -static -Wall -Wfatal-errors src\*.c src\nnue\*.cpp src\fathom\tbprobe.c

Linux and ARM V8 (using src/makefile):

cd src
make <architeture>
      <architecture>: avx2, sse4, sse3, sse2, old

Note: It is recommend to use AVX2 or the higher SSE in order to have a good performance with neural network evaluation.

//END