Chessprogramming: how to get a single move out of a bitboard attack-mask most efficently

Question

How can I get efficiently a single move out of an attack mask, that looks like this:

....1...
1...1...
.1..1..1
..1.1.1.
...111..
11111111
..1.11..
.1..1.1.

for a queen.

What I've done in the past, is to get the square-indices of every single possible move from the queen by counting the trailing zeros (bitScanForward) and after I generated the new move i removed this square from the attack mask and continued with the next attack-square. Is there any technic to get the single attack bits directly?

Answer 1

I think what you are describing is already the most efficient way. Looping over the bitboard until it is zero and pick one move at a time.

To sketch the idea with some code, it could look like this:

using Bitboard = uint64_t; // 64 bit unsigned integer

pMoves createAllMoves(Bitboard mask, int from_sq, Move* pMoves) {
  while(moves != 0) {
    int to_sq = findAndClearSetBit(mask);
   *pMoves++ = createMove(from_sq, to_sq);
  }
  return pMoves;
}

The findAndClearSetBit function can choose any set bit, but commonly on today's hardware, finding the least significant bit is most efficient. If you are using GCC or Clang, you can use __builtin_ctzll which should be optimized to the specific hardware:

int findAndClearSetBit(Bitboard& mask) {
   int sq = __builtin_ctzll(mask); // find least significant bit
   mask &= mask - 1; // clear least significant bit
   return sq;
}

If I am not mistaken, your existing function bitScanForward is already an implementation to find the least significant bit. So, you can use it to get a portable version.