X-Git-Url: https://git.sesse.net/?p=stockfish;a=blobdiff_plain;f=src%2Fsearch.cpp;h=b0e2c950fe7354c784cd14689ab7c3cbb7bed6b4;hp=425c2c65d7e098d18b07e9d459c23511f6a2a701;hb=e51965aa57ddc50d04016e3622da49cf9f8e6238;hpb=1b947aafbf19fb35c71eb2201ad32ae072ab6658 diff --git a/src/search.cpp b/src/search.cpp index 425c2c65..b0e2c950 100644 --- a/src/search.cpp +++ b/src/search.cpp @@ -41,7 +41,7 @@ namespace Search { LimitsType Limits; RootMoveVector RootMoves; Position RootPos; - Time::point SearchTime; + TimePoint SearchTime; StateStackPtr SetupStates; } @@ -90,11 +90,52 @@ namespace { Move best = MOVE_NONE; }; + // EasyMoveManager struct is used to detect a so called 'easy move'; when PV is + // stable across multiple search iterations we can fast return the best move. + struct EasyMoveManager { + + void clear() { + stableCnt = 0; + expectedPosKey = 0; + pv[0] = pv[1] = pv[2] = MOVE_NONE; + } + + Move get(Key key) const { + return expectedPosKey == key ? pv[2] : MOVE_NONE; + } + + void update(Position& pos, const std::vector& newPv) { + + assert(newPv.size() >= 3); + + // Keep track of how many times in a row 3rd ply remains stable + stableCnt = (newPv[2] == pv[2]) ? stableCnt + 1 : 0; + + if (!std::equal(newPv.begin(), newPv.begin() + 3, pv)) + { + std::copy(newPv.begin(), newPv.begin() + 3, pv); + + StateInfo st[2]; + pos.do_move(newPv[0], st[0], pos.gives_check(newPv[0], CheckInfo(pos))); + pos.do_move(newPv[1], st[1], pos.gives_check(newPv[1], CheckInfo(pos))); + expectedPosKey = pos.key(); + pos.undo_move(newPv[1]); + pos.undo_move(newPv[0]); + } + } + + int stableCnt; + Key expectedPosKey; + Move pv[3]; + }; + size_t PVIdx; TimeManager TimeMgr; + EasyMoveManager EasyMove; double BestMoveChanges; Value DrawValue[COLOR_NB]; HistoryStats History; + CounterMovesHistoryStats CounterMovesHistory; GainsStats Gains; MovesStats Countermoves, Followupmoves; @@ -109,7 +150,6 @@ namespace { Value value_from_tt(Value v, int ply); void update_pv(Move* pv, Move move, Move* childPv); void update_stats(const Position& pos, Stack* ss, Move move, Depth depth, Move* quiets, int quietsCnt); - string uci_pv(const Position& pos, Depth depth, Value alpha, Value beta); } // namespace @@ -153,19 +193,19 @@ uint64_t Search::perft(Position& pos, Depth depth) { CheckInfo ci(pos); const bool leaf = (depth == 2 * ONE_PLY); - for (const ExtMove& ms : MoveList(pos)) + for (const auto& m : MoveList(pos)) { if (Root && depth <= ONE_PLY) cnt = 1, nodes++; else { - pos.do_move(ms.move, st, ci, pos.gives_check(ms.move, ci)); + pos.do_move(m, st, pos.gives_check(m, ci)); cnt = leaf ? MoveList(pos).size() : perft(pos, depth - ONE_PLY); nodes += cnt; - pos.undo_move(ms.move); + pos.undo_move(m); } if (Root) - sync_cout << UCI::move(ms.move, pos.is_chess960()) << ": " << cnt << sync_endl; + sync_cout << UCI::move(m, pos.is_chess960()) << ": " << cnt << sync_endl; } return nodes; } @@ -200,7 +240,7 @@ void Search::think() { if (RootMoves.empty()) { - RootMoves.push_back(MOVE_NONE); + RootMoves.push_back(RootMove(MOVE_NONE)); sync_cout << "info depth 0 score " << UCI::value(RootPos.checkers() ? -VALUE_MATE : VALUE_DRAW) << sync_endl; @@ -239,10 +279,13 @@ void Search::think() { } for (Thread* th : Threads) + { th->maxPly = 0; + th->notify_one(); // Wake up all the threads + } Threads.timer->run = true; - Threads.timer->notify_one(); // Wake up the recurring timer + Threads.timer->notify_one(); // Start the recurring timer id_loop(RootPos); // Let's start searching ! @@ -281,6 +324,9 @@ namespace { Depth depth; Value bestValue, alpha, beta, delta; + Move easyMove = EasyMove.get(pos.key()); + EasyMove.clear(); + std::memset(ss-2, 0, 5 * sizeof(Stack)); depth = DEPTH_ZERO; @@ -290,6 +336,7 @@ namespace { TT.new_search(); History.clear(); + CounterMovesHistory.clear(); Gains.clear(); Countermoves.clear(); Followupmoves.clear(); @@ -356,8 +403,8 @@ namespace { // the UI) before a re-search. if ( multiPV == 1 && (bestValue <= alpha || bestValue >= beta) - && Time::now() - SearchTime > 3000) - sync_cout << uci_pv(pos, depth, alpha, beta) << sync_endl; + && now() - SearchTime > 3000) + sync_cout << UCI::pv(pos, depth, alpha, beta) << sync_endl; // In case of failing low/high increase aspiration window and // re-search, otherwise exit the loop. @@ -387,10 +434,10 @@ namespace { if (Signals.stop) sync_cout << "info nodes " << RootPos.nodes_searched() - << " time " << Time::now() - SearchTime << sync_endl; + << " time " << now() - SearchTime << sync_endl; - else if (PVIdx + 1 == multiPV || Time::now() - SearchTime > 3000) - sync_cout << uci_pv(pos, depth, alpha, beta) << sync_endl; + else if (PVIdx + 1 == multiPV || now() - SearchTime > 3000) + sync_cout << UCI::pv(pos, depth, alpha, beta) << sync_endl; } // If skill level is enabled and time is up, pick a sub-optimal best move @@ -404,27 +451,44 @@ namespace { Signals.stop = true; // Do we have time for the next iteration? Can we stop searching now? - if (Limits.use_time_management() && !Signals.stop && !Signals.stopOnPonderhit) + if (Limits.use_time_management()) { - // Take some extra time if the best move has changed - if (depth > 4 * ONE_PLY && multiPV == 1) - TimeMgr.pv_instability(BestMoveChanges); - - // Stop the search if only one legal move is available or all - // of the available time has been used. - if ( RootMoves.size() == 1 - || Time::now() - SearchTime > TimeMgr.available_time()) + if (!Signals.stop && !Signals.stopOnPonderhit) { - // If we are allowed to ponder do not stop the search now but - // keep pondering until the GUI sends "ponderhit" or "stop". - if (Limits.ponder) - Signals.stopOnPonderhit = true; - else - Signals.stop = true; + // Take some extra time if the best move has changed + if (depth > 4 * ONE_PLY && multiPV == 1) + TimeMgr.pv_instability(BestMoveChanges); + + // Stop the search if only one legal move is available or all + // of the available time has been used or we matched an easyMove + // from the previous search and just did a fast verification. + if ( RootMoves.size() == 1 + || now() - SearchTime > TimeMgr.available_time() + || ( RootMoves[0].pv[0] == easyMove + && BestMoveChanges < 0.03 + && now() - SearchTime > TimeMgr.available_time() / 10)) + { + // If we are allowed to ponder do not stop the search now but + // keep pondering until the GUI sends "ponderhit" or "stop". + if (Limits.ponder) + Signals.stopOnPonderhit = true; + else + Signals.stop = true; + } } + + if (RootMoves[0].pv.size() >= 3) + EasyMove.update(pos, RootMoves[0].pv); + else + EasyMove.clear(); } } + // Clear any candidate easy move that wasn't stable for the last search + // iterations; the second condition prevents consecutive fast moves. + if (EasyMove.stableCnt < 6 || now() - SearchTime < TimeMgr.available_time()) + EasyMove.clear(); + // If skill level is enabled, swap best PV line with the sub-optimal one if (skill.enabled()) std::swap(RootMoves[0], *std::find(RootMoves.begin(), @@ -521,7 +585,7 @@ namespace { ss->ttMove = ttMove = RootNode ? RootMoves[PVIdx].pv[0] : ttHit ? tte->move() : MOVE_NONE; ttValue = ttHit ? value_from_tt(tte->value(), ss->ply) : VALUE_NONE; - // At non-PV nodes we check for a fail high/low. We don't probe at PV nodes + // At non-PV nodes we check for a fail high/low. We don't prune at PV nodes if ( !PvNode && ttHit && tte->depth() >= depth @@ -688,14 +752,14 @@ namespace { assert((ss-1)->currentMove != MOVE_NONE); assert((ss-1)->currentMove != MOVE_NULL); - MovePicker mp(pos, ttMove, History, pos.captured_piece_type()); + MovePicker mp(pos, ttMove, History, CounterMovesHistory, pos.captured_piece_type()); CheckInfo ci(pos); while ((move = mp.next_move()) != MOVE_NONE) if (pos.legal(move, ci.pinned)) { ss->currentMove = move; - pos.do_move(move, st, ci, pos.gives_check(move, ci)); + pos.do_move(move, st, pos.gives_check(move, ci)); value = -search(pos, ss+1, -rbeta, -rbeta+1, rdepth, !cutNode); pos.undo_move(move); if (value >= rbeta) @@ -727,7 +791,7 @@ moves_loop: // When in check and at SpNode search starts from here Move followupmoves[] = { Followupmoves[pos.piece_on(prevOwnMoveSq)][prevOwnMoveSq].first, Followupmoves[pos.piece_on(prevOwnMoveSq)][prevOwnMoveSq].second }; - MovePicker mp(pos, ttMove, depth, History, countermoves, followupmoves, ss); + MovePicker mp(pos, ttMove, depth, History, CounterMovesHistory, countermoves, followupmoves, ss); CheckInfo ci(pos); value = bestValue; // Workaround a bogus 'uninitialized' warning under gcc improving = ss->staticEval >= (ss-2)->staticEval @@ -766,7 +830,7 @@ moves_loop: // When in check and at SpNode search starts from here continue; moveCount = ++splitPoint->moveCount; - splitPoint->mutex.unlock(); + splitPoint->spinlock.release(); } else ++moveCount; @@ -775,7 +839,7 @@ moves_loop: // When in check and at SpNode search starts from here { Signals.firstRootMove = (moveCount == 1); - if (thisThread == Threads.main() && Time::now() - SearchTime > 3000) + if (thisThread == Threads.main() && now() - SearchTime > 3000) sync_cout << "info depth " << depth / ONE_PLY << " currmove " << UCI::move(move, pos.is_chess960()) << " currmovenumber " << moveCount + PVIdx << sync_endl; @@ -835,7 +899,7 @@ moves_loop: // When in check and at SpNode search starts from here && moveCount >= FutilityMoveCounts[improving][depth]) { if (SpNode) - splitPoint->mutex.lock(); + splitPoint->spinlock.acquire(); continue; } @@ -854,7 +918,7 @@ moves_loop: // When in check and at SpNode search starts from here if (SpNode) { - splitPoint->mutex.lock(); + splitPoint->spinlock.acquire(); if (bestValue > splitPoint->bestValue) splitPoint->bestValue = bestValue; } @@ -866,14 +930,14 @@ moves_loop: // When in check and at SpNode search starts from here if (predictedDepth < 4 * ONE_PLY && pos.see_sign(move) < VALUE_ZERO) { if (SpNode) - splitPoint->mutex.lock(); + splitPoint->spinlock.acquire(); continue; } } // Speculative prefetch as early as possible - prefetch((char*)TT.first_entry(pos.key_after(move))); + prefetch(TT.first_entry(pos.key_after(move))); // Check for legality just before making the move if (!RootNode && !SpNode && !pos.legal(move, ci.pinned)) @@ -887,7 +951,7 @@ moves_loop: // When in check and at SpNode search starts from here quietsSearched[quietCount++] = move; // Step 14. Make the move - pos.do_move(move, st, ci, givesCheck); + pos.do_move(move, st, givesCheck); // Step 15. Reduced depth search (LMR). If the move fails high it will be // re-searched at full depth. @@ -966,7 +1030,7 @@ moves_loop: // When in check and at SpNode search starts from here // Step 18. Check for new best move if (SpNode) { - splitPoint->mutex.lock(); + splitPoint->spinlock.acquire(); bestValue = splitPoint->bestValue; alpha = splitPoint->alpha; } @@ -1011,6 +1075,12 @@ moves_loop: // When in check and at SpNode search starts from here if (value > alpha) { + // If there is an easy move for this position, clear it if unstable + if ( PvNode + && EasyMove.get(pos.key()) + && (move != EasyMove.get(pos.key()) || moveCount > 1)) + EasyMove.clear(); + bestMove = SpNode ? splitPoint->bestMove = move : move; if (PvNode && !RootNode) // Update pv even in fail-high case @@ -1035,7 +1105,9 @@ moves_loop: // When in check and at SpNode search starts from here && Threads.size() >= 2 && depth >= Threads.minimumSplitDepth && ( !thisThread->activeSplitPoint - || !thisThread->activeSplitPoint->allSlavesSearching) + || !thisThread->activeSplitPoint->allSlavesSearching + || ( Threads.size() > MAX_SLAVES_PER_SPLITPOINT + && thisThread->activeSplitPoint->slavesMask.count() == MAX_SLAVES_PER_SPLITPOINT)) && thisThread->splitPointsSize < MAX_SPLITPOINTS_PER_THREAD) { assert(bestValue > -VALUE_INFINITE && bestValue < beta); @@ -1191,7 +1263,7 @@ moves_loop: // When in check and at SpNode 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, History, to_sq((ss-1)->currentMove)); + MovePicker mp(pos, ttMove, depth, History, CounterMovesHistory, to_sq((ss-1)->currentMove)); CheckInfo ci(pos); // Loop through the moves until no moves remain or a beta cutoff occurs @@ -1239,7 +1311,7 @@ moves_loop: // When in check and at SpNode search starts from here continue; // Speculative prefetch as early as possible - prefetch((char*)TT.first_entry(pos.key_after(move))); + prefetch(TT.first_entry(pos.key_after(move))); // Check for legality just before making the move if (!pos.legal(move, ci.pinned)) @@ -1248,7 +1320,7 @@ moves_loop: // When in check and at SpNode search starts from here ss->currentMove = move; // Make and search the move - pos.do_move(move, st, ci, givesCheck); + pos.do_move(move, st, givesCheck); value = givesCheck ? -qsearch(pos, ss+1, -beta, -alpha, depth - ONE_PLY) : -qsearch(pos, ss+1, -beta, -alpha, depth - ONE_PLY); pos.undo_move(move); @@ -1330,8 +1402,8 @@ moves_loop: // When in check and at SpNode search starts from here *pv = MOVE_NONE; } - // update_stats() updates killers, history, countermoves and followupmoves stats after a fail-high - // of a quiet move. + // update_stats() updates killers, history, countermoves and followupmoves + // stats after a fail-high of a quiet move. void update_stats(const Position& pos, Stack* ss, Move move, Depth depth, Move* quiets, int quietsCnt) { @@ -1341,27 +1413,40 @@ moves_loop: // When in check and at SpNode search starts from here ss->killers[0] = move; } - // Increase history value of the cut-off move and decrease all the other - // played quiet moves. Value bonus = Value((depth / ONE_PLY) * (depth / ONE_PLY)); + + Square prevSq = to_sq((ss-1)->currentMove); + HistoryStats& cmh = CounterMovesHistory[pos.piece_on(prevSq)][prevSq]; + History.update(pos.moved_piece(move), to_sq(move), bonus); - for (int i = 0; i < quietsCnt; ++i) + if (is_ok((ss-1)->currentMove)) { - Move m = quiets[i]; - History.update(pos.moved_piece(m), to_sq(m), -bonus); + Countermoves.update(pos.piece_on(prevSq), prevSq, move); + cmh.update(pos.moved_piece(move), to_sq(move), bonus); } - if (is_ok((ss-1)->currentMove)) + // Decrease all the other played quiet moves + for (int i = 0; i < quietsCnt; ++i) { - Square prevMoveSq = to_sq((ss-1)->currentMove); - Countermoves.update(pos.piece_on(prevMoveSq), prevMoveSq, move); + History.update(pos.moved_piece(quiets[i]), to_sq(quiets[i]), -bonus); + + if (is_ok((ss-1)->currentMove)) + cmh.update(pos.moved_piece(quiets[i]), to_sq(quiets[i]), -bonus); } if (is_ok((ss-2)->currentMove) && (ss-1)->currentMove == (ss-1)->ttMove) { - Square prevOwnMoveSq = to_sq((ss-2)->currentMove); - Followupmoves.update(pos.piece_on(prevOwnMoveSq), prevOwnMoveSq, move); + Value bonus2 = Value(((depth+1) / ONE_PLY) * ((depth+1) / ONE_PLY)); + + Square prevPrevSq = to_sq((ss-2)->currentMove); + Followupmoves.update(pos.piece_on(prevPrevSq), prevPrevSq, move); + + Square prevMoveSq = to_sq((ss-1)->currentMove); + Piece prevMovePiece = pos.piece_on(prevMoveSq); + + HistoryStats& cmh2 = CounterMovesHistory[pos.piece_on(prevPrevSq)][prevPrevSq]; + cmh2.update(prevMovePiece, prevMoveSq, -bonus2); } } @@ -1372,7 +1457,7 @@ moves_loop: // When in check and at SpNode search starts from here Move Skill::pick_best(size_t multiPV) { // PRNG sequence should be non-deterministic, so we seed it with the time at init - static PRNG rng(Time::now()); + static PRNG rng(now()); // RootMoves are already sorted by score in descending order int variance = std::min(RootMoves[0].score - RootMoves[multiPV - 1].score, PawnValueMg); @@ -1384,75 +1469,77 @@ moves_loop: // When in check and at SpNode search starts from here // then we choose the move with the resulting highest score. for (size_t i = 0; i < multiPV; ++i) { - int score = RootMoves[i].score; - // This is our magic formula - score += ( weakness * int(RootMoves[0].score - score) - + variance * (rng.rand() % weakness)) / 128; + int push = ( weakness * int(RootMoves[0].score - RootMoves[i].score) + + variance * (rng.rand() % weakness)) / 128; - if (score > maxScore) + if (RootMoves[i].score + push > maxScore) { - maxScore = score; + maxScore = RootMoves[i].score + push; best = RootMoves[i].pv[0]; } } return best; } +} // namespace + - // uci_pv() formats PV information according to the UCI protocol. UCI - // requires that all (if any) unsearched PV lines are sent using a previous - // search score. +/// UCI::pv() formats PV information according to the UCI protocol. UCI requires +/// that all (if any) unsearched PV lines are sent using a previous search score. - string uci_pv(const Position& pos, Depth depth, Value alpha, Value beta) { +string UCI::pv(const Position& pos, Depth depth, Value alpha, Value beta) { - std::stringstream ss; - Time::point elapsed = Time::now() - SearchTime + 1; - size_t uciPVSize = std::min((size_t)Options["MultiPV"], RootMoves.size()); - int selDepth = 0; + std::stringstream ss; + TimePoint elapsed = now() - SearchTime + 1; + size_t multiPV = std::min((size_t)Options["MultiPV"], RootMoves.size()); + int selDepth = 0; - for (Thread* th : Threads) - if (th->maxPly > selDepth) - selDepth = th->maxPly; + for (Thread* th : Threads) + if (th->maxPly > selDepth) + selDepth = th->maxPly; - for (size_t i = 0; i < uciPVSize; ++i) - { - bool updated = (i <= PVIdx); + for (size_t i = 0; i < multiPV; ++i) + { + bool updated = (i <= PVIdx); - if (depth == ONE_PLY && !updated) - continue; + if (depth == ONE_PLY && !updated) + continue; - Depth d = updated ? depth : depth - ONE_PLY; - Value v = updated ? RootMoves[i].score : RootMoves[i].previousScore; + Depth d = updated ? depth : depth - ONE_PLY; + Value v = updated ? RootMoves[i].score : RootMoves[i].previousScore; - bool tb = TB::RootInTB && abs(v) < VALUE_MATE - MAX_PLY; - v = tb ? TB::Score : v; + bool tb = TB::RootInTB && abs(v) < VALUE_MATE - MAX_PLY; + v = tb ? TB::Score : v; - if (ss.rdbuf()->in_avail()) // Not at first line - ss << "\n"; + if (ss.rdbuf()->in_avail()) // Not at first line + ss << "\n"; - ss << "info depth " << d / ONE_PLY - << " seldepth " << selDepth - << " multipv " << i + 1 - << " score " << UCI::value(v); + ss << "info" + << " depth " << d / ONE_PLY + << " seldepth " << selDepth + << " multipv " << i + 1 + << " score " << UCI::value(v); - if (!tb && i == PVIdx) - ss << (v >= beta ? " lowerbound" : v <= alpha ? " upperbound" : ""); + if (!tb && i == PVIdx) + ss << (v >= beta ? " lowerbound" : v <= alpha ? " upperbound" : ""); - ss << " nodes " << pos.nodes_searched() - << " nps " << pos.nodes_searched() * 1000 / elapsed - << " tbhits " << TB::Hits - << " time " << elapsed - << " pv"; + ss << " nodes " << pos.nodes_searched() + << " nps " << pos.nodes_searched() * 1000 / elapsed; - for (size_t j = 0; j < RootMoves[i].pv.size(); ++j) - ss << " " << UCI::move(RootMoves[i].pv[j], pos.is_chess960()); - } + if (elapsed > 1000) // Earlier makes little sense + ss << " hashfull " << TT.hashfull(); - return ss.str(); + ss << " tbhits " << TB::Hits + << " time " << elapsed + << " pv"; + + for (Move m : RootMoves[i].pv) + ss << " " << UCI::move(m, pos.is_chess960()); } -} // namespace + return ss.str(); +} /// RootMove::insert_pv_in_tt() is called at the end of a search iteration, and @@ -1462,22 +1549,22 @@ moves_loop: // When in check and at SpNode search starts from here void RootMove::insert_pv_in_tt(Position& pos) { StateInfo state[MAX_PLY], *st = state; - size_t idx = 0; + bool ttHit; - for ( ; idx < pv.size(); ++idx) + for (Move m : pv) { - bool ttHit; - TTEntry* tte = TT.probe(pos.key(), ttHit); + assert(MoveList(pos).contains(m)); - if (!ttHit || tte->move() != pv[idx]) // Don't overwrite correct entries - tte->save(pos.key(), VALUE_NONE, BOUND_NONE, DEPTH_NONE, pv[idx], VALUE_NONE, TT.generation()); + TTEntry* tte = TT.probe(pos.key(), ttHit); - assert(MoveList(pos).contains(pv[idx])); + if (!ttHit || tte->move() != m) // Don't overwrite correct entries + tte->save(pos.key(), VALUE_NONE, BOUND_NONE, DEPTH_NONE, m, VALUE_NONE, TT.generation()); - pos.do_move(pv[idx], *st++); + pos.do_move(m, *st++, pos.gives_check(m, CheckInfo(pos))); } - while (idx) pos.undo_move(pv[--idx]); + for (size_t i = pv.size(); i > 0; ) + pos.undo_move(pv[--i]); } @@ -1486,22 +1573,25 @@ void RootMove::insert_pv_in_tt(Position& pos) { /// root. We try hard to have a ponder move to return to the GUI, otherwise in case of /// 'ponder on' we have nothing to think on. -Move RootMove::extract_ponder_from_tt(Position& pos) +bool RootMove::extract_ponder_from_tt(Position& pos) { StateInfo st; - bool found; + bool ttHit; assert(pv.size() == 1); - pos.do_move(pv[0], st); - TTEntry* tte = TT.probe(pos.key(), found); - Move m = found ? tte->move() : MOVE_NONE; - if (!MoveList(pos).contains(m)) - m = MOVE_NONE; - + pos.do_move(pv[0], st, pos.gives_check(pv[0], CheckInfo(pos))); + TTEntry* tte = TT.probe(pos.key(), ttHit); pos.undo_move(pv[0]); - pv.push_back(m); - return m; + + if (ttHit) + { + Move m = tte->move(); // Local copy to be SMP safe + if (MoveList(pos).contains(m)) + return pv.push_back(m), true; + } + + return false; } @@ -1511,21 +1601,21 @@ void Thread::idle_loop() { // Pointer 'this_sp' is not null only if we are called from split(), and not // at the thread creation. This means we are the split point's master. - SplitPoint* this_sp = splitPointsSize ? activeSplitPoint : nullptr; + SplitPoint* this_sp = activeSplitPoint; - assert(!this_sp || (this_sp->masterThread == this && searching)); + assert(!this_sp || (this_sp->master == this && searching)); - while (!exit) + while (!exit && !(this_sp && this_sp->slavesMask.none())) { // If this thread has been assigned work, launch a search while (searching) { - Threads.mutex.lock(); + spinlock.acquire(); assert(activeSplitPoint); SplitPoint* sp = activeSplitPoint; - Threads.mutex.unlock(); + spinlock.release(); Stack stack[MAX_PLY+4], *ss = stack+2; // To allow referencing (ss-2) and (ss+2) Position pos(*sp->pos, this); @@ -1533,7 +1623,7 @@ void Thread::idle_loop() { std::memcpy(ss-2, sp->ss-2, 5 * sizeof(Stack)); ss->splitPoint = sp; - sp->mutex.lock(); + sp->spinlock.acquire(); assert(activePosition == nullptr); @@ -1559,66 +1649,81 @@ void Thread::idle_loop() { sp->allSlavesSearching = false; sp->nodes += pos.nodes_searched(); - // Wake up the master thread so to allow it to return from the idle - // loop in case we are the last slave of the split point. - if ( this != sp->masterThread - && sp->slavesMask.none()) - { - assert(!sp->masterThread->searching); - sp->masterThread->notify_one(); - } - // After releasing the lock we can't access any SplitPoint related data // in a safe way because it could have been released under our feet by // the sp master. - sp->mutex.unlock(); + sp->spinlock.release(); // Try to late join to another split point if none of its slaves has // already finished. - if (Threads.size() > 2) - for (size_t i = 0; i < Threads.size(); ++i) + SplitPoint* bestSp = NULL; + int minLevel = INT_MAX; + + for (Thread* th : Threads) + { + const size_t size = th->splitPointsSize; // Local copy + sp = size ? &th->splitPoints[size - 1] : nullptr; + + if ( sp + && sp->allSlavesSearching + && sp->slavesMask.count() < MAX_SLAVES_PER_SPLITPOINT + && can_join(sp)) { - const int size = Threads[i]->splitPointsSize; // Local copy - sp = size ? &Threads[i]->splitPoints[size - 1] : nullptr; + assert(this != th); + assert(!(this_sp && this_sp->slavesMask.none())); + assert(Threads.size() > 2); + + // Prefer to join to SP with few parents to reduce the probability + // that a cut-off occurs above us, and hence we waste our work. + int level = 0; + for (SplitPoint* p = th->activeSplitPoint; p; p = p->parentSplitPoint) + level++; - if ( sp - && sp->allSlavesSearching - && available_to(Threads[i])) + if (level < minLevel) { - // Recheck the conditions under lock protection - Threads.mutex.lock(); - sp->mutex.lock(); - - if ( sp->allSlavesSearching - && available_to(Threads[i])) - { - sp->slavesMask.set(idx); - activeSplitPoint = sp; - searching = true; - } - - sp->mutex.unlock(); - Threads.mutex.unlock(); - - break; // Just a single attempt + bestSp = sp; + minLevel = level; } } - } + } - // Grab the lock to avoid races with Thread::notify_one() - std::unique_lock lk(mutex); + if (bestSp) + { + sp = bestSp; - // If we are master and all slaves have finished then exit idle_loop - if (this_sp && this_sp->slavesMask.none()) - { - assert(!searching); - break; + // Recheck the conditions under lock protection + sp->spinlock.acquire(); + + if ( sp->allSlavesSearching + && sp->slavesMask.count() < MAX_SLAVES_PER_SPLITPOINT) + { + spinlock.acquire(); + + if (can_join(sp)) + { + sp->slavesMask.set(idx); + activeSplitPoint = sp; + searching = true; + } + + spinlock.release(); + } + + sp->spinlock.release(); + } } - // If we are not searching, wait for a condition to be signaled instead of - // wasting CPU time polling for work. - if (!searching && !exit) - sleepCondition.wait(lk); + // If search is finished then sleep, otherwise just yield + if (!Threads.main()->thinking) + { + assert(!this_sp); + + std::unique_lock lk(mutex); + while (!exit && !Threads.main()->thinking) + sleepCondition.wait(lk); + } + else + std::this_thread::yield(); // Wait for a new job or for our slaves to finish } } @@ -1629,12 +1734,12 @@ void Thread::idle_loop() { void check_time() { - static Time::point lastInfoTime = Time::now(); - Time::point elapsed = Time::now() - SearchTime; + static TimePoint lastInfoTime = now(); + TimePoint elapsed = now() - SearchTime; - if (Time::now() - lastInfoTime >= 1000) + if (now() - lastInfoTime >= 1000) { - lastInfoTime = Time::now(); + lastInfoTime = now(); dbg_print(); } @@ -1657,18 +1762,17 @@ void check_time() { else if (Limits.nodes) { - Threads.mutex.lock(); - int64_t nodes = RootPos.nodes_searched(); // Loop across all split points and sum accumulated SplitPoint nodes plus // all the currently active positions nodes. + // FIXME: Racy... for (Thread* th : Threads) - for (int i = 0; i < th->splitPointsSize; ++i) + for (size_t i = 0; i < th->splitPointsSize; ++i) { SplitPoint& sp = th->splitPoints[i]; - sp.mutex.lock(); + sp.spinlock.acquire(); nodes += sp.nodes; @@ -1676,11 +1780,9 @@ void check_time() { if (sp.slavesMask.test(idx) && Threads[idx]->activePosition) nodes += Threads[idx]->activePosition->nodes_searched(); - sp.mutex.unlock(); + sp.spinlock.release(); } - Threads.mutex.unlock(); - if (nodes >= Limits.nodes) Signals.stop = true; }