A new chess engine : m8 (comming not so soon)

[mar]

I'm pretty sure you meant

7. Lazy smp

It scales above 16 cores and works even at ultra-fast TC (1" per game).
YBWC already contradicts point 1 (lazy smp can be implemented and debugged in two weeks or perhaps even less).

[Evert]

Perhaps you may try chess variant, say capablanca, 8 rows x 10 columns. Two pieces each are added in each side, one piece is a combination of knight and rook the other piece is a combination of knight and bishop. It is fun and interesting to explore the possibilities of these two new pieces over the expanded board.

I can only add my support for this suggestion. If you want to stick with 8x8, consider Seirawan Chess: it adds the same two pieces as Capablanca chess, and regular chess is just a subset Seirawan Chess anyway.

EDIT: think of it this way: another chess engine is just that, they're a dime a dozen. Another chess variant engine is something special. There may be fewer people interested in chess variant engines, but because there are so few of them out there, people who are interested will be interested in yours.

[Modern Times]

6. Fisher Random Chess

I’ll implement FRC from the start. It’s not too hard to implement and since there is not many FRC engine it could make m8 more appealing.

Excellent idea

[mathmoi]

Perhaps you may try chess variant, say capablanca, 8 rows x 10 columns. Two pieces each are added in each side, one piece is a combination of knight and rook the other piece is a combination of knight and bishop. It is fun and interesting to explore the possibilities of these two new pieces over the expanded board.

I'd like to implements other variants beside FRC but m8 will be a bitboard engine and I don't think there is an easy way to implement a 8 x 10 variant as a bitboard engine. Is there?

[mathmoi]

I'm pretty sure you meant

7. Lazy smp

It scales above 16 cores and works even at ultra-fast TC (1" per game).
YBWC already contradicts point 1 (lazy smp can be implemented and debugged in two weeks or perhaps even less).

Isn't lazy SMP the technique that use the TT as a communication device between threads that search the same tree? My understanding was that it gave little speedup passed 2 CPU. Am I missing something?

You are right, YBWC is not really KISS compliant. I was thinking about it because it seems to be not too complicated (compared to DTS) and still giving interesting results. This being said, before I implement SMP I will need to do some research to make sure I choose a good approach.

[mathmoi]

Perhaps you may try chess variant, say capablanca, 8 rows x 10 columns. Two pieces each are added in each side, one piece is a combination of knight and rook the other piece is a combination of knight and bishop. It is fun and interesting to explore the possibilities of these two new pieces over the expanded board.

I can only add my support for this suggestion. If you want to stick with 8x8, consider Seirawan Chess: it adds the same two pieces as Capablanca chess, and regular chess is just a subset Seirawan Chess anyway.

EDIT: think of it this way: another chess engine is just that, they're a dime a dozen. Another chess variant engine is something special. There may be fewer people interested in chess variant engines, but because there are so few of them out there, people who are interested will be interested in yours.

Hi Evert,

I'll look into this. As I said, I'm interested in implementing search variant, but I am probably limited to 8 x 8 variants since I want to use bitboards.

[hgm]

There are zillions of variants on 8x8 boards. E.g. Seirawan Chess, Makruk, Spartan Chess, Knightmate, Mighty-Lion Chess, all flavors of Chess with Different Armies, King of the Hill, Team Mate Chess...

Things like the opening book should not be in binary format if possible (I say if possible because I’m not sure it would be efficient to have a plain text opening book). An opening book in text format can be read and modified by a human. If text format is not an option an open binary format should be used (Think SQLite).

Actually Polyglot books (an open binary format) are much easier to read and modify by a human than plain text. Because you can use a Chess GUI like WinBoard to edit them, so that you can enter the new moves by playing them, rather than typing them, they are automatically checked for legality, you can walk through the book by just clicking in the list of moves from the current position, etc.

[mar]

Isn't lazy SMP the technique that use the TT as a communication device between threads that search the same tree?

Basically yes, but it has been improved since the old days (also we're using larger TTs today):
resync on each ID iteration, early termination when one of the helpers finishes and run each other thread on depth+1 (as proposed by Dan Homan).

My understanding was that it gave little speedup passed 2 CPU. Am I missing something?

Where did you read that? There's evidence that it works and in fact it works very well.
If claim sun in blue will you believe it? If I keep repeating it over and over again? Or will you trust what you see?

[petero2]

Isn't lazy SMP the technique that use the TT as a communication device between threads that search the same tree?

Basically yes, but it has been improved since the old days (also we're using larger TTs today):
resync on each ID iteration, early termination when one of the helpers finishes and run each other thread on depth+1 (as proposed by Dan Homan).

My understanding was that it gave little speedup passed 2 CPU. Am I missing something?

Where did you read that? :shock: There's evidence that it works and in fact it works very well.
If claim sun in blue will you believe it? If I keep repeating it over and over again? Or will you trust what you see?

I have tested this algorithm using Cheng 0.38 on a 16 core Dell PowerEdge T620 computer. Both hyperthreading and turbo boost are enabled. The computer runs Fedora 19. Here are the results, computed by bayeselo:

Code: Select all

Cheng 4c vs Cheng 1c, 1+0.08:
Rank Name             Elo    +    - games score oppo. draws 
   1 cheng4_038_4c     68    8    8  1457   70%   -68   35% 
   2 cheng4_038       -68    8    8  1457   30%    68   35% 

Cheng 4c vs Cheng 1c, 8+0.64:
Rank Name             Elo    +    - games score oppo. draws 
   1 cheng4_038_4c     66   10   10  1011   71%   -66   42% 
   2 cheng4_038       -66   10   10  1011   29%    66   42% 

Cheng 8c vs Cheng 4c, 1+0.08:
Rank Name             Elo    +    - games score oppo. draws 
   1 cheng4_038_8c     28    8    8  1475   59%   -28   44% 
   2 cheng4_038_4c    -28    8    8  1475   41%    28   44% 

Cheng 16c vs Cheng 8c, 1+0.08:
Rank Name             Elo    +    - games score oppo. draws 
   1 cheng4_038_16c     4    8    8  1400   51%    -4   46% 
   2 cheng4_038_8c     -4    8    8  1400   49%     4   46% 

Cheng 16c vs Cheng 8c, 2+0.16:
Rank Name             Elo    +    - games score oppo. draws 
   1 cheng4_038_16c    13    8    8  1433   54%   -13   50% 
   2 cheng4_038_8c    -13    8    8  1433   46%    13   50% 

Cheng 16c vs Cheng 8c, 4+0.32:
Rank Name             Elo    +    - games score oppo. draws 
   1 cheng4_038_16c     6    8    8  1151   52%    -6   53% 
   2 cheng4_038_8c     -6    8    8  1151   48%     6   53% 

Cheng 16c vs Cheng 8c, 180+1:
Rank Name             Elo    +    - games score oppo. draws 
   1 cheng4_038_16c    17   20   20   184   56%   -17   53% 
   2 cheng4_038_8c    -17   20   20   184   44%    17   53% 

Even at hyper bullet speed (1s+0.08s/move) it scales well up to 8 cores. +130 elo from 1 to 4 cores, and +56 elo from 4 to 8 cores. At 16 cores it seems hyper bullet speed is too fast for the algorithm to be effective, but it still seems to work well, +34 elo, at 3m+1s/move time control, even though the number of games (184) is too low to be really sure.

Feeding all 16 vs 8 cores games into bayeselo gives +16 elo and LOS=99.9%, so it is pretty clear that the algorithm works also for 16 cores even if it is unclear exactly how well it works.

Some other results for comparison, even though the low number of games and different 1 core ratings make it hard to draw any definite conclusions:

Code: Select all

Texel 16c vs Texel 8c, 180+1:
Rank Name          Elo    +    - games score oppo. draws 
   1 Texel16c       21   19   18   190   58%   -21   69% 
   2 Texel8c       -21   18   19   190   42%    21   69% 

Texel 16c vs Texel 16c_nonuma, 180+1:
Rank Name          Elo    +    - games score oppo. draws 
   1 Texel16c        7   19   19   175   53%    -7   71% 
   2 Texel16cnn     -7   19   19   175   47%     7   71% 

Komodo 8 16c vs Komodo 8 8c, 180+1:
Rank Name          Elo    +    - games score oppo. draws 
   1 komodo8_16c    12   18   18   196   55%   -12   78% 
   2 komodo8_8c    -12   18   18   196   45%    12   78% 

[mar]

Thanks for the data Peter
Not bad for something that's lousy beyond 2 cores I guess.