X-Git-Url: https://git.sesse.net/?p=stockfish;a=blobdiff_plain;f=src%2Fsearch.cpp;h=5ff23ad0fd894791c7b5e8ac625246c40343e1e5;hp=075aa9b4422d01d490e2a1a25196e8befe7d6002;hb=36c82b751ce227c05bfb;hpb=37e9802411fa122eb140ada2edfc9df3be55dbb8 diff --git a/src/search.cpp b/src/search.cpp index 075aa9b4..5ff23ad0 100644 --- a/src/search.cpp +++ b/src/search.cpp @@ -1,7 +1,7 @@ /* Stockfish, a UCI chess playing engine derived from Glaurung 2.1 Copyright (C) 2004-2008 Tord Romstad (Glaurung author) - Copyright (C) 2008-2012 Marco Costalba, Joona Kiiski, Tord Romstad + Copyright (C) 2008-2013 Marco Costalba, Joona Kiiski, Tord Romstad Stockfish is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by @@ -26,7 +26,6 @@ #include "book.h" #include "evaluate.h" -#include "history.h" #include "movegen.h" #include "movepick.h" #include "notation.h" @@ -87,7 +86,8 @@ namespace { TimeManager TimeMgr; int BestMoveChanges; Value DrawValue[COLOR_NB]; - History H; + History Hist; + Gains Gain; template Value search(Position& pos, Stack* ss, Value alpha, Value beta, Depth depth); @@ -98,9 +98,9 @@ namespace { void id_loop(Position& pos); Value value_to_tt(Value v, int ply); Value value_from_tt(Value v, int ply); - bool check_is_dangerous(Position& pos, Move move, Value futilityBase, Value beta); - bool yields_to_threat(const Position& pos, Move move, Move threat); - bool prevents_threat(const Position& pos, Move move, Move threat); + bool check_is_dangerous(const Position& pos, Move move, Value futilityBase, Value beta); + bool allows(const Position& pos, Move first, Move second); + bool refutes(const Position& pos, Move first, Move second); string uci_pv(const Position& pos, int depth, Value alpha, Value beta); struct Skill { @@ -182,19 +182,19 @@ void Search::think() { static PolyglotBook book; // Defined static to initialize the PRNG only once RootColor = RootPos.side_to_move(); - TimeMgr.init(Limits, RootPos.startpos_ply_counter(), RootColor); + TimeMgr.init(Limits, RootPos.game_ply(), RootColor); if (RootMoves.empty()) { RootMoves.push_back(MOVE_NONE); sync_cout << "info depth 0 score " - << score_to_uci(RootPos.in_check() ? -VALUE_MATE : VALUE_DRAW) + << score_to_uci(RootPos.checkers() ? -VALUE_MATE : VALUE_DRAW) << sync_endl; goto finalize; } - if (Options["OwnBook"] && !Limits.infinite) + if (Options["OwnBook"] && !Limits.infinite && !Limits.mate) { Move bookMove = book.probe(RootPos, Options["Book File"], Options["Best Book Move"]); @@ -208,7 +208,7 @@ void Search::think() { if (Options["Contempt Factor"] && !Options["UCI_AnalyseMode"]) { int cf = Options["Contempt Factor"] * PawnValueMg / 100; // From centipawns - cf = cf * MaterialTable::game_phase(RootPos) / PHASE_MIDGAME; // Scale down with phase + cf = cf * Material::game_phase(RootPos) / PHASE_MIDGAME; // Scale down with phase DrawValue[ RootColor] = VALUE_DRAW - Value(cf); DrawValue[~RootColor] = VALUE_DRAW + Value(cf); } @@ -218,7 +218,7 @@ void Search::think() { if (Options["Use Search Log"]) { Log log(Options["Search Log Filename"]); - log << "\nSearching: " << RootPos.to_fen() + log << "\nSearching: " << RootPos.fen() << "\ninfinite: " << Limits.infinite << " ponder: " << Limits.ponder << " time: " << Limits.time[RootColor] @@ -227,22 +227,25 @@ void Search::think() { << std::endl; } - Threads.wake_up(); + // Reset the threads, still sleeping: will be wake up at split time + for (size_t i = 0; i < Threads.size(); i++) + Threads[i]->maxPly = 0; + + Threads.sleepWhileIdle = Options["Use Sleeping Threads"]; // Set best timer interval to avoid lagging under time pressure. Timer is // used to check for remaining available thinking time. - if (Limits.use_time_management()) - Threads.set_timer(std::min(100, std::max(TimeMgr.available_time() / 16, - TimerResolution))); - else if (Limits.nodes) - Threads.set_timer(2 * TimerResolution); - else - Threads.set_timer(100); + Threads.timer->msec = + Limits.use_time_management() ? std::min(100, std::max(TimeMgr.available_time() / 16, TimerResolution)) : + Limits.nodes ? 2 * TimerResolution + : 100; + + Threads.timer->notify_one(); // Wake up the recurring timer id_loop(RootPos); // Let's start searching ! - Threads.set_timer(0); // Stop timer - Threads.sleep(); + Threads.timer->msec = 0; // Stop the timer + Threads.sleepWhileIdle = true; // Send idle threads to sleep if (Options["Use Search Log"]) { @@ -262,10 +265,15 @@ void Search::think() { finalize: // When we reach max depth we arrive here even without Signals.stop is raised, - // but if we are pondering or in infinite search, we shouldn't print the best - // move before we are told to do so. + // but if we are pondering or in infinite search, according to UCI protocol, + // we shouldn't print the best move before the GUI sends a "stop" or "ponderhit" + // command. We simply wait here until GUI sends one of those commands (that + // raise Signals.stop). if (!Signals.stop && (Limits.ponder || Limits.infinite)) - RootPos.this_thread()->wait_for_stop_or_ponderhit(); + { + Signals.stopOnPonderhit = true; + RootPos.this_thread()->wait_for(Signals.stop); + } // Best move could be MOVE_NONE when searching on a stalemate position sync_cout << "bestmove " << move_to_uci(RootMoves[0].pv[0], RootPos.is_chess960()) @@ -285,14 +293,14 @@ namespace { Stack ss[MAX_PLY_PLUS_2]; int depth, prevBestMoveChanges; Value bestValue, alpha, beta, delta; - bool bestMoveNeverChanged = true; memset(ss, 0, 4 * sizeof(Stack)); depth = BestMoveChanges = 0; bestValue = delta = -VALUE_INFINITE; ss->currentMove = MOVE_NULL; // Hack to skip update gains TT.new_search(); - H.clear(); + Hist.clear(); + Gain.clear(); PVSize = Options["MultiPV"]; Skill skill(Options["Skill Level"]); @@ -345,7 +353,7 @@ namespace { // we want to keep the same order for all the moves but the new // PV that goes to the front. Note that in case of MultiPV search // the already searched PV lines are preserved. - sort(RootMoves.begin() + PVIdx, RootMoves.end()); + std::stable_sort(RootMoves.begin() + PVIdx, RootMoves.end()); // Write PV back to transposition table in case the relevant // entries have been overwritten during the search. @@ -390,8 +398,10 @@ namespace { } // Sort the PV lines searched so far and update the GUI - sort(RootMoves.begin(), RootMoves.begin() + PVIdx + 1); - sync_cout << uci_pv(pos, depth, alpha, beta) << sync_endl; + std::stable_sort(RootMoves.begin(), RootMoves.begin() + PVIdx + 1); + + if (PVIdx + 1 == PVSize || Time::now() - SearchTime > 3000) + sync_cout << uci_pv(pos, depth, alpha, beta) << sync_endl; } // Do we need to pick now the sub-optimal best move ? @@ -405,9 +415,11 @@ namespace { << std::endl; } - // Filter out startup noise when monitoring best move stability - if (depth > 2 && BestMoveChanges) - bestMoveNeverChanged = false; + // Do we have found a "mate in x"? + if ( Limits.mate + && bestValue >= VALUE_MATE_IN_MAX_PLY + && VALUE_MATE - bestValue <= 2 * Limits.mate) + Signals.stop = true; // Do we have time for the next iteration? Can we stop searching now? if (Limits.use_time_management() && !Signals.stopOnPonderhit) @@ -428,8 +440,8 @@ namespace { if ( depth >= 12 && !stop && PVSize == 1 - && ( (bestMoveNeverChanged && pos.captured_piece_type()) - || Time::now() - SearchTime > (TimeMgr.available_time() * 40) / 100)) + && ( RootMoves.size() == 1 + || Time::now() - SearchTime > (TimeMgr.available_time() * 20) / 100)) { Value rBeta = bestValue - 2 * PawnValueMg; (ss+1)->excludedMove = RootMoves[0].pv[0]; @@ -477,7 +489,7 @@ namespace { Move movesSearched[64]; StateInfo st; const TTEntry *tte; - SplitPoint* sp; + SplitPoint* splitPoint; Key posKey; Move ttMove, move, excludedMove, bestMove, threatMove; Depth ext, newDepth; @@ -490,19 +502,19 @@ namespace { // Step 1. Initialize node Thread* thisThread = pos.this_thread(); moveCount = playedMoveCount = 0; - inCheck = pos.in_check(); + inCheck = pos.checkers(); if (SpNode) { - sp = ss->sp; - bestMove = sp->bestMove; - threatMove = sp->threatMove; - bestValue = sp->bestValue; + splitPoint = ss->splitPoint; + bestMove = splitPoint->bestMove; + threatMove = splitPoint->threatMove; + bestValue = splitPoint->bestValue; tte = NULL; ttMove = excludedMove = MOVE_NONE; ttValue = VALUE_NONE; - assert(sp->bestValue > -VALUE_INFINITE && sp->moveCount > 0); + assert(splitPoint->bestValue > -VALUE_INFINITE && splitPoint->moveCount > 0); goto split_point_start; } @@ -520,7 +532,7 @@ namespace { if (!RootNode) { // Step 2. Check for aborted search and immediate draw - if (Signals.stop || pos.is_draw() || ss->ply > MAX_PLY) + if (Signals.stop || pos.is_draw() || ss->ply > MAX_PLY) return DrawValue[pos.side_to_move()]; // Step 3. Mate distance pruning. Even if we mate at the next move our score @@ -576,15 +588,9 @@ namespace { else if (tte) { - // Following asserts are valid only in single thread condition because - // TT access is always racy and its contents cannot be trusted. - assert(tte->static_value() != VALUE_NONE || Threads.size() > 1); - assert(ttValue != VALUE_NONE || tte->type() == BOUND_NONE || Threads.size() > 1); - - ss->staticEval = eval = tte->static_value(); - ss->evalMargin = tte->static_value_margin(); - - if (eval == VALUE_NONE || ss->evalMargin == VALUE_NONE) // Due to a race + // Never assume anything on values stored in TT + if ( (ss->staticEval = eval = tte->eval_value()) == VALUE_NONE + ||(ss->evalMargin = tte->eval_margin()) == VALUE_NONE) eval = ss->staticEval = evaluate(pos, ss->evalMargin); // Can ttValue be used as a better position evaluation? @@ -609,7 +615,7 @@ namespace { && type_of(move) == NORMAL) { Square to = to_sq(move); - H.update_gain(pos.piece_on(to), to, -(ss-1)->staticEval - ss->staticEval); + Gain.update(pos.piece_on(to), to, -(ss-1)->staticEval - ss->staticEval); } // Step 6. Razoring (is omitted in PV nodes) @@ -659,12 +665,12 @@ namespace { if (eval - PawnValueMg > beta) R += ONE_PLY; - pos.do_null_move(st); + pos.do_null_move(st); (ss+1)->skipNullMove = true; nullValue = depth-R < ONE_PLY ? -qsearch(pos, ss+1, -beta, -alpha, DEPTH_ZERO) : - search(pos, ss+1, -beta, -alpha, depth-R); (ss+1)->skipNullMove = false; - pos.do_null_move(st); + pos.undo_null_move(); if (nullValue >= beta) { @@ -696,7 +702,7 @@ namespace { if ( depth < 5 * ONE_PLY && (ss-1)->reduction && threatMove != MOVE_NONE - && yields_to_threat(pos, (ss-1)->currentMove, threatMove)) + && allows(pos, (ss-1)->currentMove, threatMove)) return beta - 1; } } @@ -719,7 +725,7 @@ namespace { assert((ss-1)->currentMove != MOVE_NONE); assert((ss-1)->currentMove != MOVE_NULL); - MovePicker mp(pos, ttMove, H, pos.captured_piece_type()); + MovePicker mp(pos, ttMove, Hist, pos.captured_piece_type()); CheckInfo ci(pos); while ((move = mp.next_move()) != MOVE_NONE) @@ -739,7 +745,7 @@ namespace { && ttMove == MOVE_NONE && (PvNode || (!inCheck && ss->staticEval + Value(256) >= beta))) { - Depth d = (PvNode ? depth - 2 * ONE_PLY : depth / 2); + Depth d = depth - 2 * ONE_PLY - (PvNode ? DEPTH_ZERO : depth / 4); ss->skipNullMove = true; search(pos, ss, alpha, beta, d); @@ -751,7 +757,7 @@ namespace { split_point_start: // At split points actual search starts from here - MovePicker mp(pos, ttMove, depth, H, ss, PvNode ? -VALUE_INFINITE : beta); + MovePicker mp(pos, ttMove, depth, Hist, ss, PvNode ? -VALUE_INFINITE : beta); CheckInfo ci(pos); value = bestValue; // Workaround a bogus 'uninitialized' warning under gcc singularExtensionNode = !RootNode @@ -783,8 +789,8 @@ split_point_start: // At split points actual search starts from here if (!pos.pl_move_is_legal(move, ci.pinned)) continue; - moveCount = ++sp->moveCount; - sp->mutex.unlock(); + moveCount = ++splitPoint->moveCount; + splitPoint->mutex.unlock(); } else moveCount++; @@ -793,7 +799,7 @@ split_point_start: // At split points actual search starts from here { Signals.firstRootMove = (moveCount == 1); - if (thisThread == Threads.main_thread() && Time::now() - SearchTime > 2000) + if (thisThread == Threads.main_thread() && Time::now() - SearchTime > 3000) sync_cout << "info depth " << depth / ONE_PLY << " currmove " << move_to_uci(move, pos.is_chess960()) << " currmovenumber " << moveCount + PVIdx << sync_endl; @@ -839,7 +845,7 @@ split_point_start: // At split points actual search starts from here ss->excludedMove = MOVE_NONE; if (value < rBeta) - ext = rBeta >= beta ? ONE_PLY + ONE_PLY / 2 : ONE_PLY; + ext = ONE_PLY; } // Update current move (this must be done after singular extension search) @@ -851,16 +857,15 @@ split_point_start: // At split points actual search starts from here && !inCheck && !dangerous && move != ttMove - && (bestValue > VALUE_MATED_IN_MAX_PLY || ( bestValue == -VALUE_INFINITE - && alpha > VALUE_MATED_IN_MAX_PLY))) + && bestValue > VALUE_MATED_IN_MAX_PLY) { // Move count based pruning if ( depth < 16 * ONE_PLY && moveCount >= FutilityMoveCounts[depth] - && (!threatMove || !prevents_threat(pos, move, threatMove))) + && (!threatMove || !refutes(pos, move, threatMove))) { if (SpNode) - sp->mutex.lock(); + splitPoint->mutex.lock(); continue; } @@ -870,35 +875,40 @@ split_point_start: // At split points actual search starts from here // but fixing this made program slightly weaker. Depth predictedDepth = newDepth - reduction(depth, moveCount); futilityValue = ss->staticEval + ss->evalMargin + futility_margin(predictedDepth, moveCount) - + H.gain(pos.piece_moved(move), to_sq(move)); + + 2 * Gain[pos.piece_moved(move)][to_sq(move)]; if (futilityValue < beta) { - if (SpNode) - sp->mutex.lock(); + bestValue = std::max(bestValue, futilityValue); + if (SpNode) + { + splitPoint->mutex.lock(); + if (bestValue > splitPoint->bestValue) + splitPoint->bestValue = bestValue; + } continue; } // Prune moves with negative SEE at low depths - if ( predictedDepth < 2 * ONE_PLY + if ( predictedDepth < 4 * ONE_PLY && pos.see_sign(move) < 0) { if (SpNode) - sp->mutex.lock(); + splitPoint->mutex.lock(); continue; } } // Check for legality only before to do the move - if (!pos.pl_move_is_legal(move, ci.pinned)) + if (!RootNode && !SpNode && !pos.pl_move_is_legal(move, ci.pinned)) { moveCount--; continue; } - pvMove = PvNode ? moveCount == 1 : false; + pvMove = PvNode && moveCount == 1; ss->currentMove = move; if (!SpNode && !captureOrPromotion && playedMoveCount < 64) movesSearched[playedMoveCount++] = move; @@ -912,12 +922,14 @@ split_point_start: // At split points actual search starts from here && !pvMove && !captureOrPromotion && !dangerous - && ss->killers[0] != move - && ss->killers[1] != move) + && move != ttMove + && move != ss->killers[0] + && move != ss->killers[1]) { ss->reduction = reduction(depth, moveCount); Depth d = std::max(newDepth - ss->reduction, ONE_PLY); - alpha = SpNode ? sp->alpha : alpha; + if (SpNode) + alpha = splitPoint->alpha; value = -search(pos, ss+1, -(alpha+1), -alpha, d); @@ -930,7 +942,9 @@ split_point_start: // At split points actual search starts from here // Step 16. Full depth search, when LMR is skipped or fails high if (doFullDepthSearch) { - alpha = SpNode ? sp->alpha : alpha; + if (SpNode) + alpha = splitPoint->alpha; + value = newDepth < ONE_PLY ? givesCheck ? -qsearch(pos, ss+1, -(alpha+1), -alpha, DEPTH_ZERO) : -qsearch(pos, ss+1, -(alpha+1), -alpha, DEPTH_ZERO) @@ -953,9 +967,9 @@ split_point_start: // At split points actual search starts from here // Step 18. Check for new best move if (SpNode) { - sp->mutex.lock(); - bestValue = sp->bestValue; - alpha = sp->alpha; + splitPoint->mutex.lock(); + bestValue = splitPoint->bestValue; + alpha = splitPoint->alpha; } // Finished searching the move. If Signals.stop is true, the search @@ -990,22 +1004,21 @@ split_point_start: // At split points actual search starts from here if (value > bestValue) { - bestValue = value; - if (SpNode) sp->bestValue = value; + bestValue = SpNode ? splitPoint->bestValue = value : value; if (value > alpha) { - bestMove = move; - if (SpNode) sp->bestMove = move; + bestMove = SpNode ? splitPoint->bestMove = move : move; - if (PvNode && value < beta) - { - alpha = value; // Update alpha here! Always alpha < beta - if (SpNode) sp->alpha = value; - } - else // Fail high + if (PvNode && value < beta) // Update alpha! Always alpha < beta + alpha = SpNode ? splitPoint->alpha = value : value; + else { - if (SpNode) sp->cutoff = true; + assert(value >= beta); // Fail high + + if (SpNode) + splitPoint->cutoff = true; + break; } } @@ -1013,12 +1026,14 @@ split_point_start: // At split points actual search starts from here // Step 19. Check for splitting the search if ( !SpNode - && depth >= Threads.min_split_depth() - && bestValue < beta - && Threads.available_slave_exists(thisThread)) + && depth >= Threads.minimumSplitDepth + && Threads.available_slave(thisThread) + && thisThread->splitPointsSize < MAX_SPLITPOINTS_PER_THREAD) { - bestValue = Threads.split(pos, ss, alpha, beta, bestValue, &bestMove, - depth, threatMove, moveCount, mp, NT); + assert(bestValue < beta); + + thisThread->split(pos, ss, alpha, beta, &bestValue, &bestMove, + depth, threatMove, moveCount, &mp, NT); if (bestValue >= beta) break; } @@ -1061,13 +1076,13 @@ split_point_start: // At split points actual search starts from here // Increase history value of the cut-off move Value bonus = Value(int(depth) * int(depth)); - H.add(pos.piece_moved(bestMove), to_sq(bestMove), bonus); + Hist.update(pos.piece_moved(bestMove), to_sq(bestMove), bonus); // Decrease history of all the other played non-capture moves for (int i = 0; i < playedMoveCount - 1; i++) { Move m = movesSearched[i]; - H.add(pos.piece_moved(m), to_sq(m), -bonus); + Hist.update(pos.piece_moved(m), to_sq(m), -bonus); } } } @@ -1092,7 +1107,7 @@ split_point_start: // At split points actual search starts from here const bool PvNode = (NT == PV); assert(NT == PV || NT == NonPV); - assert(InCheck == pos.in_check()); + assert(InCheck == !!pos.checkers()); assert(alpha >= -VALUE_INFINITE && alpha < beta && beta <= VALUE_INFINITE); assert(PvNode || (alpha == beta - 1)); assert(depth <= DEPTH_ZERO); @@ -1101,17 +1116,27 @@ split_point_start: // At split points actual search starts from here const TTEntry* tte; Key posKey; Move ttMove, move, bestMove; - Value bestValue, value, ttValue, futilityValue, futilityBase; + Value bestValue, value, ttValue, futilityValue, futilityBase, oldAlpha; bool givesCheck, enoughMaterial, evasionPrunable; Depth ttDepth; + // To flag BOUND_EXACT a node with eval above alpha and no available moves + if (PvNode) + oldAlpha = alpha; + ss->currentMove = bestMove = MOVE_NONE; ss->ply = (ss-1)->ply + 1; // Check for an instant draw or maximum ply reached - if (pos.is_draw() || ss->ply > MAX_PLY) + if (pos.is_draw() || ss->ply > MAX_PLY) return DrawValue[pos.side_to_move()]; + // Decide whether or not to include checks, this fixes also the type of + // TT entry depth that we are going to use. Note that in qsearch we use + // only two types of depth in TT: DEPTH_QS_CHECKS or DEPTH_QS_NO_CHECKS. + ttDepth = InCheck || depth >= DEPTH_QS_CHECKS ? DEPTH_QS_CHECKS + : DEPTH_QS_NO_CHECKS; + // Transposition table lookup. At PV nodes, we don't use the TT for // pruning, but only for move ordering. posKey = pos.key(); @@ -1119,11 +1144,6 @@ split_point_start: // At split points actual search starts from here ttMove = tte ? tte->move() : MOVE_NONE; ttValue = tte ? value_from_tt(tte->value(),ss->ply) : VALUE_NONE; - // Decide whether or not to include checks, this fixes also the type of - // TT entry depth that we are going to use. Note that in qsearch we use - // only two types of depth in TT: DEPTH_QS_CHECKS or DEPTH_QS_NO_CHECKS. - ttDepth = InCheck || depth >= DEPTH_QS_CHECKS ? DEPTH_QS_CHECKS - : DEPTH_QS_NO_CHECKS; if ( tte && tte->depth() >= ttDepth && ttValue != VALUE_NONE // Only in case of TT access race @@ -1146,19 +1166,11 @@ split_point_start: // At split points actual search starts from here { if (tte) { - assert(tte->static_value() != VALUE_NONE || Threads.size() > 1); - - ss->staticEval = bestValue = tte->static_value(); - ss->evalMargin = tte->static_value_margin(); - - if (ss->staticEval == VALUE_NONE || ss->evalMargin == VALUE_NONE) // Due to a race + // Never assume anything on values stored in TT + if ( (ss->staticEval = bestValue = tte->eval_value()) == VALUE_NONE + ||(ss->evalMargin = tte->eval_margin()) == VALUE_NONE) ss->staticEval = bestValue = evaluate(pos, ss->evalMargin); } - else if ((ss-1)->currentMove == MOVE_NULL) - { - ss->staticEval = bestValue = -(ss-1)->staticEval; - ss->evalMargin = VALUE_ZERO; // Hack, we really don't know the value - } else ss->staticEval = bestValue = evaluate(pos, ss->evalMargin); @@ -1183,7 +1195,7 @@ split_point_start: // At split points actual search starts from here // to search the moves. Because the depth is <= 0 here, only captures, // queen promotions and checks (only if depth >= DEPTH_QS_CHECKS) will // be generated. - MovePicker mp(pos, ttMove, depth, H, to_sq((ss-1)->currentMove)); + MovePicker mp(pos, ttMove, depth, Hist, to_sq((ss-1)->currentMove)); CheckInfo ci(pos); // Loop through the moves until no moves remain or a beta cutoff occurs @@ -1208,9 +1220,7 @@ split_point_start: // At split points actual search starts from here if (futilityValue < beta) { - if (futilityValue > bestValue) - bestValue = futilityValue; - + bestValue = std::max(bestValue, futilityValue); continue; } @@ -1218,7 +1228,10 @@ split_point_start: // At split points actual search starts from here if ( futilityBase < beta && depth < DEPTH_ZERO && pos.see(move) <= 0) + { + bestValue = std::max(bestValue, futilityBase); continue; + } } // Detect non-capture evasions that are candidate to be pruned @@ -1289,7 +1302,7 @@ split_point_start: // At split points actual search starts from here return mated_in(ss->ply); // Plies to mate from the root TT.store(posKey, value_to_tt(bestValue, ss->ply), - PvNode && bestMove != MOVE_NONE ? BOUND_EXACT : BOUND_UPPER, + PvNode && bestValue > oldAlpha ? BOUND_EXACT : BOUND_UPPER, ttDepth, bestMove, ss->staticEval, ss->evalMargin); assert(bestValue > -VALUE_INFINITE && bestValue < VALUE_INFINITE); @@ -1325,7 +1338,7 @@ split_point_start: // At split points actual search starts from here // check_is_dangerous() tests if a checking move can be pruned in qsearch() - bool check_is_dangerous(Position& pos, Move move, Value futilityBase, Value beta) + bool check_is_dangerous(const Position& pos, Move move, Value futilityBase, Value beta) { Piece pc = pos.piece_moved(move); Square from = from_sq(move); @@ -1359,39 +1372,40 @@ split_point_start: // At split points actual search starts from here } - // yields_to_threat() tests whether the move at previous ply yields to the so - // called threat move (the best move returned from a null search that fails - // low). Here 'yields to' means that the move somehow made the threat possible - // for instance if the moving piece is the same in both moves. + // allows() tests whether the 'first' move at previous ply somehow makes the + // 'second' move possible, for instance if the moving piece is the same in + // both moves. Normally the second move is the threat (the best move returned + // from a null search that fails low). - bool yields_to_threat(const Position& pos, Move move, Move threat) { + bool allows(const Position& pos, Move first, Move second) { - assert(is_ok(move)); - assert(is_ok(threat)); - assert(color_of(pos.piece_on(from_sq(threat))) == ~pos.side_to_move()); + assert(is_ok(first)); + assert(is_ok(second)); + assert(color_of(pos.piece_on(from_sq(second))) == ~pos.side_to_move()); + assert(color_of(pos.piece_on(to_sq(first))) == ~pos.side_to_move()); - Square mfrom = from_sq(move); - Square mto = to_sq(move); - Square tfrom = from_sq(threat); - Square tto = to_sq(threat); + Square m1from = from_sq(first); + Square m2from = from_sq(second); + Square m1to = to_sq(first); + Square m2to = to_sq(second); - // The piece is the same or threat's destination was vacated by the move - if (mto == tfrom || tto == mfrom) + // The piece is the same or second's destination was vacated by the first move + if (m1to == m2from || m2to == m1from) return true; - // Threat moves through the vacated square - if (between_bb(tfrom, tto) & mfrom) + // Second one moves through the square vacated by first one + if (between_bb(m2from, m2to) & m1from) return true; - // Threat's destination is defended by the move's piece - Bitboard matt = pos.attacks_from(pos.piece_on(mto), mto, pos.pieces() ^ tfrom); - if (matt & tto) + // Second's destination is defended by the first move's piece + Bitboard m1att = pos.attacks_from(pos.piece_on(m1to), m1to, pos.pieces() ^ m2from); + if (m1att & m2to) return true; - // Threat gives a discovered check through the move's checking piece - if (matt & pos.king_square(pos.side_to_move())) + // Second move gives a discovered check through the first's checking piece + if (m1att & pos.king_square(pos.side_to_move())) { - assert(between_bb(mto, pos.king_square(pos.side_to_move())) & tfrom); + assert(between_bb(m1to, pos.king_square(pos.side_to_move())) & m2from); return true; } @@ -1399,51 +1413,49 @@ split_point_start: // At split points actual search starts from here } - // prevents_threat() tests whether a move is able to defend against the so - // called threat move (the best move returned from a null search that fails - // low). In this case will not be pruned. + // refutes() tests whether a 'first' move is able to defend against a 'second' + // opponent's move. In this case will not be pruned. Normally the second move + // is the threat (the best move returned from a null search that fails low). - bool prevents_threat(const Position& pos, Move move, Move threat) { + bool refutes(const Position& pos, Move first, Move second) { - assert(is_ok(move)); - assert(is_ok(threat)); - assert(!pos.is_capture_or_promotion(move)); - assert(!pos.is_passed_pawn_push(move)); + assert(is_ok(first)); + assert(is_ok(second)); - Square mfrom = from_sq(move); - Square mto = to_sq(move); - Square tfrom = from_sq(threat); - Square tto = to_sq(threat); + Square m1from = from_sq(first); + Square m2from = from_sq(second); + Square m1to = to_sq(first); + Square m2to = to_sq(second); // Don't prune moves of the threatened piece - if (mfrom == tto) + if (m1from == m2to) return true; // If the threatened piece has value less than or equal to the value of the // threat piece, don't prune moves which defend it. - if ( pos.is_capture(threat) - && ( PieceValue[MG][pos.piece_on(tfrom)] >= PieceValue[MG][pos.piece_on(tto)] - || type_of(pos.piece_on(tfrom)) == KING)) + if ( pos.is_capture(second) + && ( PieceValue[MG][pos.piece_on(m2from)] >= PieceValue[MG][pos.piece_on(m2to)] + || type_of(pos.piece_on(m2from)) == KING)) { // Update occupancy as if the piece and the threat are moving - Bitboard occ = pos.pieces() ^ mfrom ^ mto ^ tfrom; - Piece piece = pos.piece_on(mfrom); + Bitboard occ = pos.pieces() ^ m1from ^ m1to ^ m2from; + Piece piece = pos.piece_on(m1from); // The moved piece attacks the square 'tto' ? - if (pos.attacks_from(piece, mto, occ) & tto) + if (pos.attacks_from(piece, m1to, occ) & m2to) return true; // Scan for possible X-ray attackers behind the moved piece - Bitboard xray = (attacks_bb< ROOK>(tto, occ) & pos.pieces(color_of(piece), QUEEN, ROOK)) - | (attacks_bb(tto, occ) & pos.pieces(color_of(piece), QUEEN, BISHOP)); + Bitboard xray = (attacks_bb< ROOK>(m2to, occ) & pos.pieces(color_of(piece), QUEEN, ROOK)) + | (attacks_bb(m2to, occ) & pos.pieces(color_of(piece), QUEEN, BISHOP)); // Verify attackers are triggered by our move and not already existing - if (xray && (xray ^ (xray & pos.attacks_from(tto)))) + if (xray && (xray ^ (xray & pos.attacks_from(m2to)))) return true; } // Don't prune safe moves which block the threat path - if ((between_bb(tfrom, tto) & mto) && pos.see_sign(move) >= 0) + if ((between_bb(m2from, m2to) & m1to) && pos.see_sign(first) >= 0) return true; return false; @@ -1504,8 +1516,8 @@ split_point_start: // At split points actual search starts from here int selDepth = 0; for (size_t i = 0; i < Threads.size(); i++) - if (Threads[i].maxPly > selDepth) - selDepth = Threads[i].maxPly; + if (Threads[i]->maxPly > selDepth) + selDepth = Threads[i]->maxPly; for (size_t i = 0; i < uciPVSize; i++) { @@ -1556,7 +1568,8 @@ void RootMove::extract_pv_from_tt(Position& pos) { do { pv.push_back(m); - assert(pos.move_is_legal(pv[ply])); + assert(MoveList(pos).contains(pv[ply])); + pos.do_move(pv[ply++], *st++); tte = TT.probe(pos.key()); @@ -1564,7 +1577,7 @@ void RootMove::extract_pv_from_tt(Position& pos) { && pos.is_pseudo_legal(m = tte->move()) // Local copy, TT could change && pos.pl_move_is_legal(m, pos.pinned_pieces()) && ply < MAX_PLY - && (!pos.is_draw() || ply < 2)); + && (!pos.is_draw() || ply < 2)); pv.push_back(MOVE_NONE); // Must be zero-terminating @@ -1581,22 +1594,15 @@ void RootMove::insert_pv_in_tt(Position& pos) { StateInfo state[MAX_PLY_PLUS_2], *st = state; TTEntry* tte; int ply = 0; - Value v, m; do { tte = TT.probe(pos.key()); if (!tte || tte->move() != pv[ply]) // Don't overwrite correct entries - { - if (pos.in_check()) - v = m = VALUE_NONE; - else - v = evaluate(pos, m); + TT.store(pos.key(), VALUE_NONE, BOUND_NONE, DEPTH_NONE, pv[ply], VALUE_NONE, VALUE_NONE); - TT.store(pos.key(), VALUE_NONE, BOUND_NONE, DEPTH_NONE, pv[ply], v, m); - } + assert(MoveList(pos).contains(pv[ply])); - assert(pos.move_is_legal(pv[ply])); pos.do_move(pv[ply++], *st++); } while (pv[ply] != MOVE_NONE); @@ -1609,33 +1615,29 @@ void RootMove::insert_pv_in_tt(Position& pos) { void Thread::idle_loop() { - // Pointer 'sp_master', if non-NULL, points to the active SplitPoint - // object for which the thread is the master. - const SplitPoint* sp_master = splitPointsCnt ? curSplitPoint : NULL; + // Pointer 'this_sp' is not null only if we are called from split(), and not + // at the thread creation. So it means we are the split point's master. + SplitPoint* this_sp = splitPointsSize ? activeSplitPoint : NULL; - assert(!sp_master || (sp_master->master == this && is_searching)); + assert(!this_sp || (this_sp->masterThread == this && searching)); - // If this thread is the master of a split point and all slaves have - // finished their work at this split point, return from the idle loop. - while (!sp_master || sp_master->slavesMask) + while (true) { - // If we are not searching, wait for a condition to be signaled - // instead of wasting CPU time polling for work. - while ( do_sleep - || do_exit - || (!is_searching && Threads.use_sleeping_threads())) + // If we are not searching, wait for a condition to be signaled instead of + // wasting CPU time polling for work. + while ((!searching && Threads.sleepWhileIdle) || exit) { - if (do_exit) + if (exit) { - assert(!sp_master); + assert(!this_sp); return; } - // Grab the lock to avoid races with Thread::wake_up() + // Grab the lock to avoid races with Thread::notify_one() mutex.lock(); - // If we are master and all slaves have finished don't go to sleep - if (sp_master && !sp_master->slavesMask) + // If we are master and all slaves have finished then exit idle_loop + if (this_sp && !this_sp->slavesMask) { mutex.unlock(); break; @@ -1643,23 +1645,23 @@ void Thread::idle_loop() { // Do sleep after retesting sleep conditions under lock protection, in // particular we need to avoid a deadlock in case a master thread has, - // in the meanwhile, allocated us and sent the wake_up() call before we - // had the chance to grab the lock. - if (do_sleep || !is_searching) + // in the meanwhile, allocated us and sent the notify_one() call before + // we had the chance to grab the lock. + if (!searching && !exit) sleepCondition.wait(mutex); mutex.unlock(); } // If this thread has been assigned work, launch a search - if (is_searching) + if (searching) { - assert(!do_sleep && !do_exit); + assert(!exit); Threads.mutex.lock(); - assert(is_searching); - SplitPoint* sp = curSplitPoint; + assert(searching); + SplitPoint* sp = activeSplitPoint; Threads.mutex.unlock(); @@ -1667,38 +1669,43 @@ void Thread::idle_loop() { Position pos(*sp->pos, this); memcpy(ss, sp->ss - 1, 4 * sizeof(Stack)); - (ss+1)->sp = sp; + (ss+1)->splitPoint = sp; sp->mutex.lock(); - assert(sp->activePositions[idx] == NULL); + assert(activePosition == NULL); - sp->activePositions[idx] = &pos; + activePosition = &pos; - if (sp->nodeType == Root) + switch (sp->nodeType) { + case Root: search(pos, ss+1, sp->alpha, sp->beta, sp->depth); - else if (sp->nodeType == PV) + break; + case PV: search(pos, ss+1, sp->alpha, sp->beta, sp->depth); - else if (sp->nodeType == NonPV) + break; + case NonPV: search(pos, ss+1, sp->alpha, sp->beta, sp->depth); - else + break; + default: assert(false); + } - assert(is_searching); + assert(searching); - is_searching = false; - sp->activePositions[idx] = NULL; + searching = false; + activePosition = NULL; sp->slavesMask &= ~(1ULL << idx); sp->nodes += pos.nodes_searched(); - // Wake up master thread so to allow it to return from the idle loop in - // case we are the last slave of the split point. - if ( Threads.use_sleeping_threads() - && this != sp->master + // Wake up master thread so to allow it to return from the idle loop + // in case we are the last slave of the split point. + if ( Threads.sleepWhileIdle + && this != sp->masterThread && !sp->slavesMask) { - assert(!sp->master->is_searching); - sp->master->wake_up(); + assert(!sp->masterThread->searching); + sp->masterThread->notify_one(); } // After releasing the lock we cannot access anymore any SplitPoint @@ -1707,6 +1714,17 @@ void Thread::idle_loop() { // unsafe because if we are exiting there is a chance are already freed. sp->mutex.unlock(); } + + // If this thread is the master of a split point and all slaves have finished + // their work at this split point, return from the idle loop. + if (this_sp && !this_sp->slavesMask) + { + this_sp->mutex.lock(); + bool finished = !this_sp->slavesMask; // Retest under lock protection + this_sp->mutex.unlock(); + if (finished) + return; + } } } @@ -1736,11 +1754,11 @@ void check_time() { nodes = RootPos.nodes_searched(); // Loop across all split points and sum accumulated SplitPoint nodes plus - // all the currently active slaves positions. + // all the currently active positions nodes. for (size_t i = 0; i < Threads.size(); i++) - for (int j = 0; j < Threads[i].splitPointsCnt; j++) + for (int j = 0; j < Threads[i]->splitPointsSize; j++) { - SplitPoint& sp = Threads[i].splitPoints[j]; + SplitPoint& sp = Threads[i]->splitPoints[j]; sp.mutex.lock(); @@ -1748,8 +1766,9 @@ void check_time() { Bitboard sm = sp.slavesMask; while (sm) { - Position* pos = sp.activePositions[pop_lsb(&sm)]; - nodes += pos ? pos->nodes_searched() : 0; + Position* pos = Threads[pop_lsb(&sm)]->activePosition; + if (pos) + nodes += pos->nodes_searched(); } sp.mutex.unlock();