From: Marco Costalba Date: Sun, 2 Nov 2008 14:58:10 +0000 (+0100) Subject: Revert movepick optimizations before to release X-Git-Url: https://git.sesse.net/?p=stockfish;a=commitdiff_plain;h=1a158c0cf056930de3c08389ab2db0f75a3f6723 Revert movepick optimizations before to release More testing is needed and better do not risk just before release. Reverted: Disable LSN filtering as defualt for release Use MVV/LVA in score_evasions() Signed-off-by: Marco Costalba --- diff --git a/src/movepick.cpp b/src/movepick.cpp index 59855dba..187a3c51 100644 --- a/src/movepick.cpp +++ b/src/movepick.cpp @@ -206,22 +206,27 @@ void MovePicker::score_captures() { // Suprisingly, this appears to perform slightly better than SEE based // move ordering. The reason is probably that in a position with a winning // capture, capturing a more valuable (but sufficiently defended) piece - // first usually doesn't hurt. The opponent will have to recapture, and + // first usually doesn't hurt. The opponent will have to recapture, and // the hanging piece will still be hanging (except in the unusual cases // where it is possible to recapture with the hanging piece). Exchanging // big pieces before capturing a hanging piece probably helps to reduce - // the subtree size. Instead of calculating SEE here to filter out - // loosing captures, we delay the filtering in pick_move_from_list() + // the subtree size. Move m; + int seeValue; for (int i = 0; i < numOfMoves; i++) { m = moves[i].move; - if (move_promotion(m)) - moves[i].score = QueenValueMidgame; - else - moves[i].score = int(pos.midgame_value_of_piece_on(move_to(m))) - -int(pos.type_of_piece_on(move_from(m))); + seeValue = pos.see(m); + if (seeValue >= 0) + { + if (move_promotion(m)) + moves[i].score = QueenValueMidgame; + else + moves[i].score = int(pos.midgame_value_of_piece_on(move_to(m))) + -int(pos.type_of_piece_on(move_from(m))); + } else + moves[i].score = seeValue; } } @@ -251,26 +256,21 @@ void MovePicker::score_noncaptures() { } } -void MovePicker::score_evasions() { - - Move m; - int hs; - - // Use MVV/LVA ordering - for (int i = 0; i < numOfMoves; i++) - { - m = moves[i].move; - - if (m == ttMove) - hs = 2*HistoryMax; - else if (!pos.square_is_empty(move_to(m))) - hs = int(pos.midgame_value_of_piece_on(move_to(m))) - -int(pos.type_of_piece_on(move_from(m))) + HistoryMax; - else - hs = H.move_ordering_score(pos.piece_on(move_from(m)), m); - - moves[i].score = hs; - } +void MovePicker::score_evasions() { + + for (int i = 0; i < numOfMoves; i++) + { + Move m = moves[i].move; + if (m == ttMove) + moves[i].score = 2*HistoryMax; + else if (!pos.square_is_empty(move_to(m))) + { + int seeScore = pos.see(m); + moves[i].score = (seeScore >= 0)? seeScore + HistoryMax : seeScore; + } else + moves[i].score = H.move_ordering_score(pos.piece_on(move_from(m)), m); + } + // FIXME try psqt also here } void MovePicker::score_qcaptures() { @@ -325,21 +325,26 @@ Move MovePicker::pick_move_from_list() { while (movesPicked < numOfMoves) { - bestIndex = find_best_index(); - - if (bestIndex != -1) // Found a possibly good capture + int bestScore = -10000000; + bestIndex = -1; + for (int i = movesPicked; i < numOfMoves; i++) { - move = moves[bestIndex].move; - int seeValue = pos.see(move); - if (seeValue < 0) + if (moves[i].score < 0) { // Losing capture, move it to the badCaptures[] array assert(numOfBadCaptures < 63); - moves[bestIndex].score = seeValue; - badCaptures[numOfBadCaptures++] = moves[bestIndex]; - moves[bestIndex] = moves[--numOfMoves]; - continue; + badCaptures[numOfBadCaptures++] = moves[i]; + moves[i--] = moves[--numOfMoves]; } + else if (moves[i].score > bestScore) + { + bestIndex = i; + bestScore = moves[i].score; + } + } + if (bestIndex != -1) // Found a good capture + { + move = moves[bestIndex].move; moves[bestIndex] = moves[movesPicked++]; if ( move != ttMove && move != mateKiller