diff --git a/src/search.cpp b/src/search.cpp index ffe36126..29820a1f 100644 --- a/src/search.cpp +++ b/src/search.cpp @@ -1042,7 +1042,9 @@ 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 + || ( int(Threads.size()) > MAX_SLAVES_PER_SPLITPOINT + && thisThread->activeSplitPoint->slavesCount == MAX_SLAVES_PER_SPLITPOINT)) && thisThread->splitPointsSize < MAX_SPLITPOINTS_PER_THREAD) { assert(bestValue > -VALUE_INFINITE && bestValue < beta); @@ -1587,6 +1589,11 @@ void Thread::idle_loop() { // Try to late join to another split point if none of its slaves has // already finished. if (Threads.size() > 2) + { + SplitPoint *bestSp = NULL; + int bestThread = 0; + int bestScore = INT_MAX; + for (size_t i = 0; i < Threads.size(); ++i) { const int size = Threads[i]->splitPointsSize; // Local copy @@ -1594,26 +1601,42 @@ void Thread::idle_loop() { if ( sp && sp->allSlavesSearching + && sp->slavesCount < MAX_SLAVES_PER_SPLITPOINT && available_to(Threads[i])) { - // Recheck the conditions under lock protection - Threads.mutex.lock(); - sp->mutex.lock(); + int score = sp->spLevel * 256 * 256 + sp->slavesCount * 256 - sp->depth * 1; - if ( sp->allSlavesSearching - && available_to(Threads[i])) + if (score < bestScore) { - sp->slavesMask.set(idx); - activeSplitPoint = sp; - searching = true; + bestSp = sp; + bestThread = i; + bestScore = score; } - - sp->mutex.unlock(); - Threads.mutex.unlock(); - - break; // Just a single attempt } } + + if (bestSp) + { + sp = bestSp; + + // Recheck the conditions under lock protection + Threads.mutex.lock(); + sp->mutex.lock(); + + if ( sp->allSlavesSearching + && sp->slavesCount < MAX_SLAVES_PER_SPLITPOINT + && available_to(Threads[bestThread])) + { + sp->slavesMask.set(idx); + sp->slavesCount++; + activeSplitPoint = sp; + searching = true; + } + + sp->mutex.unlock(); + Threads.mutex.unlock(); + } + } } // Grab the lock to avoid races with Thread::notify_one() diff --git a/src/thread.cpp b/src/thread.cpp index b8571dcd..02cb4562 100644 --- a/src/thread.cpp +++ b/src/thread.cpp @@ -154,7 +154,9 @@ void Thread::split(Position& pos, Stack* ss, Value alpha, Value beta, Value* bes sp.masterThread = this; sp.parentSplitPoint = activeSplitPoint; + sp.spLevel = activeSplitPoint ? activeSplitPoint->spLevel + 1 : 0; sp.slavesMask = 0, sp.slavesMask.set(idx); + sp.slavesCount = 1; sp.depth = depth; sp.bestValue = *bestValue; sp.bestMove = *bestMove; @@ -182,9 +184,11 @@ void Thread::split(Position& pos, Stack* ss, Value alpha, Value beta, Value* bes Thread* slave; - while ((slave = Threads.available_slave(this)) != NULL) + while ( sp.slavesCount < MAX_SLAVES_PER_SPLITPOINT + && (slave = Threads.available_slave(this)) != NULL) { sp.slavesMask.set(slave->idx); + sp.slavesCount++; slave->activeSplitPoint = &sp; slave->searching = true; // Slave leaves idle_loop() slave->notify_one(); // Could be sleeping diff --git a/src/thread.h b/src/thread.h index 9750ed7b..949f87ad 100644 --- a/src/thread.h +++ b/src/thread.h @@ -33,6 +33,7 @@ struct Thread; const int MAX_THREADS = 128; const int MAX_SPLITPOINTS_PER_THREAD = 8; +const int MAX_SLAVES_PER_SPLITPOINT = 4; /// Mutex and ConditionVariable struct are wrappers of the low level locking /// machinery and are modeled after the corresponding C++11 classes. @@ -72,6 +73,7 @@ struct SplitPoint { const Position* pos; Search::Stack* ss; Thread* masterThread; + int spLevel; Depth depth; Value beta; int nodeType; @@ -84,6 +86,7 @@ struct SplitPoint { // Shared variable data Mutex mutex; std::bitset slavesMask; + int slavesCount; volatile bool allSlavesSearching; volatile uint64_t nodes; volatile Value alpha;