Toga II 4.0 release
Posted: Fri Dec 29, 2017 6:37 am
Dear all,
Happy New Year! It is my pleasure, after a break of several years from computer chess, to release Toga II 4.0.
Strength
Self-play results:
+80 elo at 4' + 0.04' on 1 thread after 4000 games
+85 elo at 10' + 0.1' on 4 threads after 2000 games
So while I doubt the rating lists will show +80 elo it should be significantly stronger than Toga II 3.0 or any other Fruit derivative and also one of the strongest array-based engines in the world. Needless to say, Stockfish and the other top engines still smash it though ...
Compiles
- The IntelPGO compiles should be the fastest. However, if there are any problems, GCC and Intel non-PGO compiles are also provided.
- The x64 compile runs about 10% faster than the w32 compile on my machines.
- At this stage I do not have Linux, Mac or Android compiles. Hopefully someone will provide these at some stage.
Major changes
For a full list of changes since Toga II 3.0, see readme.txt in the package. The major changes are:
- code tidy-up, removed a lot of unused / defunct code and fixed a few bugs.
- Toga II can now use up to 16 GB hash and up to 64 threads, and the lazy SMP design has been tweaked.
- fixed aspiration windows, and skip fail-high research for single-thread competitive games (same patch as SF tested a day ago).
- removed egbb support. Instead I plan to add support for Ronald De Man's excellent syzygy bases for future releases.
- added eval pruning (also called static null move pruning), and move count based pruning;
- improved null move: reduction (3 + depth/4) from Stockfish used, and double null move avoided;
- better usage of hash in quiescent search;
- bad captures reduced less;
- history aging.
Have fun!
Jerry
Happy New Year! It is my pleasure, after a break of several years from computer chess, to release Toga II 4.0.
Strength
Self-play results:
+80 elo at 4' + 0.04' on 1 thread after 4000 games
+85 elo at 10' + 0.1' on 4 threads after 2000 games
So while I doubt the rating lists will show +80 elo it should be significantly stronger than Toga II 3.0 or any other Fruit derivative and also one of the strongest array-based engines in the world. Needless to say, Stockfish and the other top engines still smash it though ...
Compiles
- The IntelPGO compiles should be the fastest. However, if there are any problems, GCC and Intel non-PGO compiles are also provided.
- The x64 compile runs about 10% faster than the w32 compile on my machines.
- At this stage I do not have Linux, Mac or Android compiles. Hopefully someone will provide these at some stage.
Major changes
For a full list of changes since Toga II 3.0, see readme.txt in the package. The major changes are:
- code tidy-up, removed a lot of unused / defunct code and fixed a few bugs.
- Toga II can now use up to 16 GB hash and up to 64 threads, and the lazy SMP design has been tweaked.
- fixed aspiration windows, and skip fail-high research for single-thread competitive games (same patch as SF tested a day ago).
- removed egbb support. Instead I plan to add support for Ronald De Man's excellent syzygy bases for future releases.
- added eval pruning (also called static null move pruning), and move count based pruning;
- improved null move: reduction (3 + depth/4) from Stockfish used, and double null move avoided;
- better usage of hash in quiescent search;
- bad captures reduced less;
- history aging.
Have fun!
Jerry