Merge pull request official-stockfish#1022 from IIvec/master

IIvec · web-flow · commit c2b4638e5c85 · 2019-03-12T21:00:35.000+01:00
nm
diff --git a/src/bitboard.cpp b/src/bitboard.cpp
@@ -19,6 +19,7 @@
 */
 
 #include <algorithm>
+#include <bitset>
 
 #include "bitboard.h"
 #include "misc.h"
@@ -49,14 +50,6 @@ namespace {
   Bitboard BishopTable[0x1480]; // To store bishop attacks
 
   void init_magics(Bitboard table[], Magic magics[], Direction directions[]);
-
-  // popcount16() counts the non-zero bits using SWAR-Popcount algorithm
-  unsigned popcount16(unsigned u) {
-    u -= (u >> 1) & 0x5555U;
-    u = ((u >> 2) & 0x3333U) + (u & 0x3333U);
-    u = ((u >> 4) + u) & 0x0F0FU;
-    return (u * 0x0101U) >> 8;
-  }
 }
 
 
@@ -85,7 +78,7 @@ const std::string Bitboards::pretty(Bitboard b) {
 void Bitboards::init() {
 
   for (unsigned i = 0; i < (1 << 16); ++i)
-      PopCnt16[i] = (uint8_t)popcount16(i);
+      PopCnt16[i] = std::bitset<16>(i).count();
 
   for (Square s = SQ_A1; s <= SQ_H8; ++s)
       SquareBB[s] = (1ULL << s);
diff --git a/src/search.cpp b/src/search.cpp
@@ -299,7 +299,6 @@ void Thread::search() {
   MainThread* mainThread = (this == Threads.main() ? Threads.main() : nullptr);
   double timeReduction = 1.0;
   Color us = rootPos.side_to_move();
-  bool failedLow;
 
   std::memset(ss-7, 0, 10 * sizeof(Stack));
   for (int i = 7; i > 0; i--)
@@ -310,7 +309,7 @@ void Thread::search() {
   beta = VALUE_INFINITE;
 
   if (mainThread)
-      mainThread->bestMoveChanges = 0, failedLow = false;
+      mainThread->bestMoveChanges = 0;
 
   size_t multiPV = Options["MultiPV"];
   Skill skill(Options["Skill Level"]);
@@ -351,7 +350,7 @@ void Thread::search() {
 
       // Age out PV variability metric
       if (mainThread)
-          mainThread->bestMoveChanges *= 0.517, failedLow = false;
+          mainThread->bestMoveChanges *= 0.517;
 
       // Save the last iteration's scores before first PV line is searched and
       // all the move scores except the (new) PV are set to -VALUE_INFINITE.
@@ -431,7 +430,6 @@ void Thread::search() {
                   if (mainThread)
                   {
                       failedHighCnt = 0;
-                      failedLow = true;
                       mainThread->stopOnPonderhit = false;
                   }
               }
@@ -483,19 +481,19 @@ void Thread::search() {
           && !Threads.stop
           && !mainThread->stopOnPonderhit)
       {
-          double fallingEval = (306 + 119 * failedLow + 6 * (mainThread->previousScore - bestValue)) / 581.0;
+          double fallingEval = (306 + 9 * (mainThread->previousScore - bestValue)) / 581.0;
           fallingEval        = std::max(0.5, std::min(1.5, fallingEval));
 
           // If the bestMove is stable over several iterations, reduce time accordingly
           timeReduction = lastBestMoveDepth + 10 * ONE_PLY < completedDepth ? 1.95 : 1.0;
+          double reduction = std::pow(mainThread->previousTimeReduction, 0.528) / timeReduction;
 
           // Use part of the gained time from a previous stable move for the current move
           double bestMoveInstability = 1.0 + mainThread->bestMoveChanges;
-          bestMoveInstability *= std::pow(mainThread->previousTimeReduction, 0.528) / timeReduction;
 
           // Stop the search if we have only one legal move, or if available time elapsed
           if (   rootMoves.size() == 1
-              || Time.elapsed() > Time.optimum() * bestMoveInstability * fallingEval)
+              || Time.elapsed() > Time.optimum() * fallingEval * reduction * bestMoveInstability)
           {
               // If we are allowed to ponder do not stop the search now but
               // keep pondering until the GUI sends "ponderhit" or "stop".
@@ -924,12 +922,13 @@ namespace {
           &&  move == ttMove
           && !rootNode
           && !excludedMove // Avoid recursive singular search
-          &&  ttValue != VALUE_NONE
+      /*  &&  ttValue != VALUE_NONE Already implicit in the next condition */
+          &&  abs(ttValue) < VALUE_KNOWN_WIN
           && (tte->bound() & BOUND_LOWER)
           &&  tte->depth() >= depth - 3 * ONE_PLY
           &&  pos.legal(move))
       {
-          Value singularBeta = std::max(ttValue - 2 * depth / ONE_PLY, -VALUE_MATE);
+          Value singularBeta = ttValue - 2 * depth / ONE_PLY;
           ss->excludedMove = move;
           value = search<NonPV>(pos, ss, singularBeta - 1, singularBeta, depth / 2, cutNode);
           ss->excludedMove = MOVE_NONE;
diff --git a/src/syzygy/tbprobe.cpp b/src/syzygy/tbprobe.cpp
@@ -32,7 +32,7 @@
 #include "../movegen.h"
 #include "../position.h"
 #include "../search.h"
-#include "../thread_win32.h"
+#include "../thread_win32_osx.h"
 #include "../types.h"
 #include "../uci.h"
 
diff --git a/src/thread.h b/src/thread.h
@@ -32,7 +32,7 @@
 #include "pawns.h"
 #include "position.h"
 #include "search.h"
-#include "thread_win32.h"
+#include "thread_win32_osx.h"
 
 
 /// Thread class keeps together all the thread-related stuff. We use
@@ -46,7 +46,7 @@ class Thread {
   ConditionVariable cv;
   size_t idx;
   bool exit = false, searching = true; // Set before starting std::thread
-  std::thread stdThread;
+  NativeThread stdThread;
 
 public:
   explicit Thread(size_t);
diff --git a/src/thread_win32_osx.h b/src/thread_win32_osx.h
@@ -18,8 +18,8 @@
   along with this program.  If not, see <http://www.gnu.org/licenses/>.
 */
 
-#ifndef THREAD_WIN32_H_INCLUDED
-#define THREAD_WIN32_H_INCLUDED
+#ifndef THREAD_WIN32_OSX_H_INCLUDED
+#define THREAD_WIN32_OSX_H_INCLUDED
 
 /// STL thread library used by mingw and gcc when cross compiling for Windows
 /// relies on libwinpthread. Currently libwinpthread implements mutexes directly
@@ -33,6 +33,7 @@
 
 #include <condition_variable>
 #include <mutex>
+#include <thread>
 
 #if defined(_WIN32) && !defined(_MSC_VER)
 
@@ -67,4 +68,45 @@ typedef std::condition_variable ConditionVariable;
 
 #endif
 
-#endif // #ifndef THREAD_WIN32_H_INCLUDED
+/// On OSX threads other than the main thread are created with a reduced stack
+/// size of 512KB by default, this is dangerously low for deep searches, so
+/// adjust it to TH_STACK_SIZE. The implementation calls pthread_create() with
+/// proper stack size parameter.
+
+#if defined(__APPLE__)
+
+#include <pthread.h>
+
+static const size_t TH_STACK_SIZE = 2 * 1024 * 1024;
+
+template <class T, class P = std::pair<T*, void(T::*)()>>
+void* start_routine(void* ptr)
+{
+   P* p = reinterpret_cast<P*>(ptr);
+   (p->first->*(p->second))(); // Call member function pointer
+   delete p;
+   return NULL;
+}
+
+class NativeThread {
+
+   pthread_t thread;
+
+public:
+  template<class T, class P = std::pair<T*, void(T::*)()>>
+  explicit NativeThread(void(T::*fun)(), T* obj) {
+    pthread_attr_t attr_storage, *attr = &attr_storage;
+    pthread_attr_init(attr);
+    pthread_attr_setstacksize(attr, TH_STACK_SIZE);
+    pthread_create(&thread, attr, start_routine<T>, new P(obj, fun));
+  }
+  void join() { pthread_join(thread, NULL); }
+};
+
+#else // Default case: use STL classes
+
+typedef std::thread NativeThread;
+
+#endif
+
+#endif // #ifndef THREAD_WIN32_OSX_H_INCLUDED
