Optimize an expression in endgame.cpp

[protonspring]

This is non-functional. I believe using foward_file_bb here is fewer instructions.

GOOD) fewer instructions and probably more clear (debatable).

BAD) Possible that a lookup is slower than a few local operations, but the forward_file_bb table is probably used often enough that it is always cached.

STC
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 21004 W: 4263 L: 4141 D: 12600
http://tests.stockfishchess.org/tests/view/5b1cad830ebc5902ab9c6239

[protonspring]

One thought. This code is only used for a specialized ending (KRKP). I guess it's possible that too many simplifications here could pass if the games don't use these endings much (it would be no different from master in this case). Is there any way to specifically test these kinds of endings to make sure there really is no elo loss?

[Rocky640]

Looks good to me, However, I would prefer
if (forward_file_bb(WHITE, wksq) & psq)

[ceebo]

The patch doesn't look correct to me. psq is relative to the strong side so you can't use it to test the squares of a Bitboard.

Many of the endgame functions are written in this style. I guess the reason is that most of the code had its origins in Glaurung 1.0 which was a mailbox engine (?)

[Rocky640]

@ceebo

I might be missing something, but my understanding is that
a) The following lines insures that everything is "normalized" for the strongSide changed to WHITE

  Square wksq = relative_square(strongSide, pos.square<KING>(strongSide));
  Square bksq = relative_square(strongSide, pos.square<KING>(weakSide));
  Square rsq  = relative_square(strongSide, pos.square<ROOK>(strongSide));
  Square psq  = relative_square(strongSide, pos.square<PAWN>(weakSide));

b) After above normalisation, the next line shows that master is expecting the weakside pawn to move towards the first rank

 Square queeningSq = make_square(file_of(psq), RANK_1);

c) Finally the single line change

if (forward_file_bb(WHITE, wksq) & psq) 

seems logically equivalent to this

  if (wksq < psq && file_of(wksq) == file_of(psq))

[ceebo]

@Rocky640 Sorry, you are right. I missed that wksq is also a relative square so everything does work out. I just saw a Bitboard being anded with a relative square and immediately assumed it as wrong.

[snicolet]

Merged via fc3af7c, thanks.