Shifted Bitboards

From Chessprogramming wiki
Revision as of 14:34, 23 January 2021 by GerdIsenberg (talk | contribs)
Jump to: navigation, search

Home * Board Representation * Bitboards * Sliding Piece Attacks * Shifted Bitboards

Shifted Bitboards,
an method to determine sliding piece attacks introduced by Neels Groenewald as implemented in his engine NagaSkaki [1]. The idea is original and does not need huge memory tables. However, with the proposed 56 64-bit operations for either rook and bishop attacks its space-time tradeoff seems not that advantageous with respect to time, which looks more in the range of set-wise fill algorithms for multiple sliders, like dumb7fill or its parallel prefix Kogge-Stone pendant.

Description

Shifted Bitboards work ray-wise and uses pre-calculated ray-attacks on the otherwise empty board for each of the eight ray-directions for all origin squares, and intersects one of them with the occupancy to determine the blockers on the attacking ray in question, quite similar to the Classical Approach. While the Classical Approach performs a bitscan, either forward or reverse to determine the first blocker (if any) for the covered ray-attack exclusion by a ray-square lookup, Shifted Bitboards performs a fill-like union of all six direction shifts of the blocker(s) from one to six (the maximum amount of covered squares behind a blocker), which were then excluded from the initial empty board ray-wise attack set.

U64 rayAttacks[8][64]; // requires initialization

U64 getRightRayAttacks(U64 occupied, enumSquare square) {
   U64 attacks = rayAttacks[right][square];
   U64 blocker = attacks & occupied;
   if ( blocker ) {
      blocker = (blocker << 1) | (blocker << 2)
              | (blocker << 3) | (blocker << 4)
              | (blocker << 5) | (blocker << 6);
      attacks ^= (blocker & attacks);
   }
   return attacks;
}

See also

External Links

References