]> git.sesse.net Git - stockfish/blobdiff - src/search.cpp
Disable spinlocks
[stockfish] / src / search.cpp
index 7f733a950f73bb0904fe8abaa35fd458f8c9044a..681acb8141a0e1b5d7c8def9569ce28e8bb15ffc 100644 (file)
@@ -41,7 +41,7 @@ namespace Search {
   LimitsType Limits;
   RootMoveVector RootMoves;
   Position RootPos;
-  Time::point SearchTime;
+  TimePoint SearchTime;
   StateStackPtr SetupStates;
 }
 
@@ -158,7 +158,7 @@ uint64_t Search::perft(Position& pos, Depth depth) {
           cnt = 1, nodes++;
       else
       {
-          pos.do_move(m, st, ci, pos.gives_check(m, ci));
+          pos.do_move(m, st, pos.gives_check(m, ci));
           cnt = leaf ? MoveList<LEGAL>(pos).size() : perft<false>(pos, depth - ONE_PLY);
           nodes += cnt;
           pos.undo_move(m);
@@ -355,7 +355,7 @@ namespace {
                 // the UI) before a re-search.
                 if (   multiPV == 1
                     && (bestValue <= alpha || bestValue >= beta)
-                    && Time::now() - SearchTime > 3000)
+                    && now() - SearchTime > 3000)
                     sync_cout << UCI::pv(pos, depth, alpha, beta) << sync_endl;
 
                 // In case of failing low/high increase aspiration window and
@@ -386,9 +386,9 @@ 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)
+            else if (PVIdx + 1 == multiPV || now() - SearchTime > 3000)
                 sync_cout << UCI::pv(pos, depth, alpha, beta) << sync_endl;
         }
 
@@ -412,7 +412,7 @@ namespace {
             // 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())
+                || now() - SearchTime > TimeMgr.available_time())
             {
                 // If we are allowed to ponder do not stop the search now but
                 // keep pondering until the GUI sends "ponderhit" or "stop".
@@ -694,7 +694,7 @@ namespace {
             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<NonPV, false>(pos, ss+1, -rbeta, -rbeta+1, rdepth, !cutNode);
                 pos.undo_move(move);
                 if (value >= rbeta)
@@ -765,7 +765,7 @@ moves_loop: // When in check and at SpNode search starts from here
               continue;
 
           moveCount = ++splitPoint->moveCount;
-          splitPoint->mutex.unlock();
+          splitPoint->spinlock.release();
       }
       else
           ++moveCount;
@@ -774,7 +774,7 @@ 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;
@@ -834,7 +834,7 @@ 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;
           }
@@ -853,7 +853,7 @@ 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;
                   }
@@ -865,7 +865,7 @@ 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;
           }
@@ -886,7 +886,7 @@ 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.
@@ -965,7 +965,7 @@ 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;
       }
@@ -1034,7 +1034,9 @@ 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);
@@ -1247,7 +1249,7 @@ 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<NT,  true>(pos, ss+1, -beta, -alpha, depth - ONE_PLY)
                          : -qsearch<NT, false>(pos, ss+1, -beta, -alpha, depth - ONE_PLY);
       pos.undo_move(move);
@@ -1371,7 +1373,7 @@ 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);
@@ -1405,7 +1407,7 @@ moves_loop: // When in check and at SpNode search starts from here
 string UCI::pv(const Position& pos, Depth depth, Value alpha, Value beta) {
 
   std::stringstream ss;
-  Time::point elapsed = Time::now() - SearchTime + 1;
+  TimePoint elapsed = now() - SearchTime + 1;
   size_t multiPV = std::min((size_t)Options["MultiPV"], RootMoves.size());
   int selDepth = 0;
 
@@ -1474,7 +1476,7 @@ void RootMove::insert_pv_in_tt(Position& pos) {
       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(m, *st++);
+      pos.do_move(m, *st++, pos.gives_check(m, CheckInfo(pos)));
   }
 
   for (size_t i = pv.size(); i > 0; )
@@ -1494,7 +1496,7 @@ bool RootMove::extract_ponder_from_tt(Position& pos)
 
     assert(pv.size() == 1);
 
-    pos.do_move(pv[0], st);
+    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]);
 
@@ -1515,21 +1517,22 @@ 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)
   {
       // If this thread has been assigned work, launch a search
       while (searching)
       {
-          Threads.mutex.lock();
+          Threads.spinlock.acquire();
 
           assert(activeSplitPoint);
+
           SplitPoint* sp = activeSplitPoint;
 
-          Threads.mutex.unlock();
+          Threads.spinlock.release();
 
           Stack stack[MAX_PLY+4], *ss = stack+2; // To allow referencing (ss-2) and (ss+2)
           Position pos(*sp->pos, this);
@@ -1537,7 +1540,7 @@ 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);
 
@@ -1565,51 +1568,74 @@ void Thread::idle_loop() {
 
           // 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())
+          if (this != sp->master && sp->slavesMask.none())
           {
-              assert(!sp->masterThread->searching);
-              sp->masterThread->notify_one();
+              assert(!sp->master->searching);
+
+              sp->master->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;
                   }
               }
+          }
+
+          if (bestSp)
+          {
+              sp = bestSp;
+
+              // Recheck the conditions under lock protection
+              Threads.spinlock.acquire();
+              sp->spinlock.acquire();
+
+              if (   sp->allSlavesSearching
+                  && sp->slavesMask.count() < MAX_SLAVES_PER_SPLITPOINT
+                  && can_join(sp))
+              {
+                  sp->slavesMask.set(idx);
+                  activeSplitPoint = sp;
+                  searching = true;
+              }
+
+              sp->spinlock.release();
+              Threads.spinlock.release();
+          }
       }
 
-      // Grab the lock to avoid races with Thread::notify_one()
+      // Avoid races with notify_one() fired from last slave of the split point
       std::unique_lock<std::mutex> lk(mutex);
 
       // If we are master and all slaves have finished then exit idle_loop
@@ -1633,12 +1659,12 @@ 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();
   }
 
@@ -1661,18 +1687,18 @@ void check_time() {
 
   else if (Limits.nodes)
   {
-      Threads.mutex.lock();
+      Threads.spinlock.acquire();
 
       int64_t nodes = RootPos.nodes_searched();
 
       // Loop across all split points and sum accumulated SplitPoint nodes plus
       // all the currently active positions nodes.
       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;
 
@@ -1680,10 +1706,10 @@ 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();
+      Threads.spinlock.release();
 
       if (nodes >= Limits.nodes)
           Signals.stop = true;