Rybka source code - this time for real!

[Vempele]

Well, a tiny part of it anyway.

All right, fine - here's white knight captures :

Code: Select all

for (bb_t knights = Board.pieces [WN]; knights; knights &= knights-1)
{
    int knight_sq = bit_scan (knights);
    for (bb_t captures = knight_moves [knight_sq] & opponent_pieces; captures; captures &= captures - 1)
    {
        int capture_sq = bit_scan (captures);
        *moves ++ = move (knight_sq, capture_sq);
        *values ++ = Board.sq [capture_sq] * 256 + 192;
    }
}

http://rybkaforum.net/cgi-bin/rybkaforu ... ;#pid20132

[Onno Garms]

Not a big surprise but the straightforward way to create knight captures in a bitboard engine. My code amounts to almost exactly the same, though the style is a bit different. Nice to see that Rybka also has an array Board.sq, in addition to the bitboards. Seems to be indexed by 0..63, so without dummy squares (otherwise you could not use the bitscan result directly to access it).

The only thing in the Rybka fragment that not everyone does is to create the values (seems to be mvv-lva) along with the moves rather than delaying it. As I think over it, this could mean a small speedup for my engine too.

[Onno Garms]

Another thing is that there is "Board" rather then "board" This might be just a coincidence, but it might also indicate that there is one global board rather than a parameter. Of course this speeds up the engine, but it makes debugging and modifications more difficult. I'm not sure if I would like to have it.

[sje]

Symbolic merges white and black capture generation; here's the analogous knight code:

Code: Select all

    case CTPieceKnight:
      {
        CTBB theToBB = RefAttackFrSq(theFrSq) & thePasLocusBB;
        CTSq theToSq;

        CTScanBB(theToSq, theToBB)
        {
          theMove.PutToSq(theToSq);
          theMove.PutToMan(theBoard.GetSqMan(theToSq));
          theMLPtr->ATM(theMove);
        };
        theMove.PutToMan(CTManVacant);
      };
      break;