Use acquire() and release() for spinlocks
authorMarco Costalba <mcostalba@gmail.com>
Sun, 15 Mar 2015 06:39:10 +0000 (07:39 +0100)
committerMarco Costalba <mcostalba@gmail.com>
Mon, 16 Mar 2015 07:14:08 +0000 (08:14 +0100)
It is more idiomatick than lock() and unlock()

No functional change.

src/movepick.cpp
src/search.cpp
src/thread.cpp
src/thread.h

index d6a40d7..b82e68b 100644 (file)
@@ -162,9 +162,9 @@ void MovePicker::score<CAPTURES>() {
 
 template<>
 void MovePicker::score<QUIETS>() {
-  Square prevMoveSq = to_sq((ss-1)->currentMove);
-  Piece prevMovePiece = pos.piece_on(prevMoveSq);
-  const HistoryStats &cmh = counterMovesHistory[prevMovePiece][prevMoveSq];
+
+  Square prevSq = to_sq((ss-1)->currentMove);
+  const HistoryStats& cmh = counterMovesHistory[pos.piece_on(prevSq)][prevSq];
 
   for (auto& m : *this)
       m.value =  history[pos.moved_piece(m)][to_sq(m)]
index 374f6e6..948f526 100644 (file)
@@ -826,7 +826,7 @@ moves_loop: // When in check and at SpNode search starts from here
               continue;
 
           moveCount = ++splitPoint->moveCount;
-          splitPoint->mutex.unlock();
+          splitPoint->spinlock.release();
       }
       else
           ++moveCount;
@@ -895,7 +895,7 @@ 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;
           }
@@ -914,7 +914,7 @@ 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;
                   }
@@ -926,7 +926,7 @@ 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;
           }
@@ -1026,7 +1026,7 @@ 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;
       }
@@ -1630,7 +1630,7 @@ 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);
 
@@ -1659,7 +1659,7 @@ void Thread::idle_loop() {
           // 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.
@@ -1699,12 +1699,12 @@ void Thread::idle_loop() {
               sp = bestSp;
 
               // Recheck the conditions under lock protection
-              sp->mutex.lock();
+              sp->spinlock.acquire();
 
               if (   sp->allSlavesSearching
                   && sp->slavesMask.count() < MAX_SLAVES_PER_SPLITPOINT)
               {
-                  allocMutex.lock();
+                  spinlock.acquire();
 
                   if (can_join(sp))
                   {
@@ -1713,10 +1713,10 @@ void Thread::idle_loop() {
                       searching = true;
                   }
 
-                  allocMutex.unlock();
+                  spinlock.release();
               }
 
-              sp->mutex.unlock();
+              sp->spinlock.release();
           }
       }
   }
@@ -1767,7 +1767,7 @@ void check_time() {
           {
               SplitPoint& sp = th->splitPoints[i];
 
-              sp.mutex.lock();
+              sp.spinlock.acquire();
 
               nodes += sp.nodes;
 
@@ -1775,7 +1775,7 @@ void check_time() {
                   if (sp.slavesMask.test(idx) && Threads[idx]->activePosition)
                       nodes += Threads[idx]->activePosition->nodes_searched();
 
-              sp.mutex.unlock();
+              sp.spinlock.release();
           }
 
       if (nodes >= Limits.nodes)
index 71bf08d..4b5d845 100644 (file)
@@ -144,7 +144,7 @@ void Thread::split(Position& pos, Stack* ss, Value alpha, Value beta, Value* bes
   // Pick and init the next available split point
   SplitPoint& sp = splitPoints[splitPointsSize];
 
-  sp.mutex.lock(); // No contention here until we don't increment splitPointsSize
+  sp.spinlock.acquire(); // No contention here until we don't increment splitPointsSize
 
   sp.master = this;
   sp.parentSplitPoint = activeSplitPoint;
@@ -174,7 +174,7 @@ void Thread::split(Position& pos, Stack* ss, Value alpha, Value beta, Value* bes
   while (    sp.slavesMask.count() < MAX_SLAVES_PER_SPLITPOINT
          && (slave = Threads.available_slave(&sp)) != nullptr)
   {
-     slave->allocMutex.lock();
+     slave->spinlock.acquire();
 
       if (slave->can_join(activeSplitPoint))
       {
@@ -183,14 +183,14 @@ void Thread::split(Position& pos, Stack* ss, Value alpha, Value beta, Value* bes
           slave->searching = true;
       }
 
-      slave->allocMutex.unlock();
+      slave->spinlock.release();
   }
 
   // Everything is set up. The master thread enters the idle loop, from which
   // it will instantly launch a search, because its 'searching' flag is set.
   // The thread will return from the idle loop when all slaves have finished
   // their work at this split point.
-  sp.mutex.unlock();
+  sp.spinlock.release();
 
   Thread::idle_loop(); // Force a call to base class idle_loop()
 
@@ -205,7 +205,7 @@ void Thread::split(Position& pos, Stack* ss, Value alpha, Value beta, Value* bes
   // We have returned from the idle loop, which means that all threads are
   // finished. Note that decreasing splitPointsSize must be done under lock
   // protection to avoid a race with Thread::can_join().
-  sp.mutex.lock();
+  sp.spinlock.acquire();
 
   --splitPointsSize;
   activeSplitPoint = sp.parentSplitPoint;
@@ -214,7 +214,7 @@ void Thread::split(Position& pos, Stack* ss, Value alpha, Value beta, Value* bes
   *bestMove = sp.bestMove;
   *bestValue = sp.bestValue;
 
-  sp.mutex.unlock();
+  sp.spinlock.release();
 }
 
 
index 8f93014..7932ad4 100644 (file)
@@ -41,16 +41,17 @@ const size_t MAX_SPLITPOINTS_PER_THREAD = 8;
 const size_t MAX_SLAVES_PER_SPLITPOINT = 4;
 
 class Spinlock {
-  std::atomic_int _lock;
+
+  std::atomic_int lock;
 
 public:
-  Spinlock() { _lock = 1; } // Init here to workaround a bug with MSVC 2013
-  void lock() {
-      while (_lock.fetch_sub(1, std::memory_order_acquire) != 1)
-          for (int cnt = 0; _lock.load(std::memory_order_relaxed) <= 0; ++cnt)
+  Spinlock() { lock = 1; } // Init here to workaround a bug with MSVC 2013
+  void acquire() {
+      while (lock.fetch_sub(1, std::memory_order_acquire) != 1)
+          for (int cnt = 0; lock.load(std::memory_order_relaxed) <= 0; ++cnt)
               if (cnt >= 10000) std::this_thread::yield(); // Be nice to hyperthreading
   }
-  void unlock() { _lock.store(1, std::memory_order_release); }
+  void release() { lock.store(1, std::memory_order_release); }
 };
 
 
@@ -73,7 +74,7 @@ struct SplitPoint {
   SplitPoint* parentSplitPoint;
 
   // Shared variable data
-  Spinlock mutex;
+  Spinlock spinlock;
   std::bitset<MAX_THREADS> slavesMask;
   volatile bool allSlavesSearching;
   volatile uint64_t nodes;
@@ -97,6 +98,7 @@ struct ThreadBase {
 
   std::thread nativeThread;
   Mutex mutex;
+  Spinlock spinlock;
   ConditionVariable sleepCondition;
   volatile bool exit = false;
 };
@@ -127,7 +129,6 @@ struct Thread : public ThreadBase {
   SplitPoint* volatile activeSplitPoint;
   volatile size_t splitPointsSize;
   volatile bool searching;
-  Spinlock allocMutex;
 };