Make ONE_PLY value independent again

[stockfish] / src / tt.cpp

diff --git a/src/tt.cpp b/src/tt.cpp
index 9f0bc4ba63091919a123fad9a2185f5f8e1e4fa6..33768ca4117d099e8fc0a46284a14ea803c60195 100644 (file)
--- a/src/tt.cpp
+++ b/src/tt.cpp
@@ -2,7 +2,7 @@
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
   Copyright (C) 2004-2008 Tord Romstad (Glaurung author)
   Copyright (C) 2008-2015 Marco Costalba, Joona Kiiski, Tord Romstad
-  Copyright (C) 2015-2018 Marco Costalba, Joona Kiiski, Gary Linscott, Tord Romstad
+  Copyright (C) 2015-2019 Marco Costalba, Joona Kiiski, Gary Linscott, Tord Romstad
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -24,11 +24,36 @@
 
 #include "bitboard.h"
 #include "misc.h"
+#include "thread.h"
 #include "tt.h"
 #include "uci.h"
 
 TranspositionTable TT; // Our global transposition table
 
+/// TTEntry::save populates the TTEntry with a new node's data, possibly
+/// overwriting an old position. Update is not atomic and can be racy.
+
+void TTEntry::save(Key k, Value v, bool pv, Bound b, Depth d, Move m, Value ev) {
+
+  assert(d / ONE_PLY * ONE_PLY == d);
+
+  // Preserve any existing move for the same position
+  if (m || (k >> 48) != key16)
+      move16 = (uint16_t)m;
+
+  // Overwrite less valuable entries
+  if (  (k >> 48) != key16
+      || d / ONE_PLY > depth8 - 4
+      || b == BOUND_EXACT)
+  {
+      key16     = (uint16_t)(k >> 48);
+      value16   = (int16_t)v;
+      eval16    = (int16_t)ev;
+      genBound8 = (uint8_t)(TT.generation8 | uint8_t(pv) << 2 | b);
+      depth8    = (int8_t)(d / ONE_PLY);
+  }
+}
+
 
 /// TranspositionTable::resize() sets the size of the transposition table,
 /// measured in megabytes. Transposition table consists of a power of 2 number
@@ -36,6 +61,8 @@ TranspositionTable TT; // Our global transposition table
 
 void TranspositionTable::resize(size_t mbSize) {
 
+  Threads.main()->wait_for_search_finished();
+
   clusterCount = mbSize * 1024 * 1024 / sizeof(Cluster);
 
   free(mem);
@@ -53,26 +80,29 @@ void TranspositionTable::resize(size_t mbSize) {
 }
 
 
-/// TranspositionTable::clear() overwrites the entire transposition table
-/// with zeros. It is called whenever the table is resized, or when the
-/// user asks the program to clear the table (from the UCI interface).
-/// It starts as many threads as allowed by the Threads option.
+/// TranspositionTable::clear() initializes the entire transposition table to zero,
+//  in a multi-threaded way.
 
 void TranspositionTable::clear() {
 
-  const size_t stride = clusterCount / Options["Threads"];
   std::vector<std::thread> threads;
-  for (size_t idx = 0; idx < Options["Threads"]; idx++)
+
+  for (size_t idx = 0; idx < Options["Threads"]; ++idx)
   {
-      const size_t start =  stride * idx,
-                   len =    idx != Options["Threads"] - 1 ?
-                            stride :
-                            clusterCount - start;
-      threads.push_back(std::thread([this, idx, start, len]() {
-          if (Options["Threads"] >= 8)
+      threads.emplace_back([this, idx]() {
+
+          // Thread binding gives faster search on systems with a first-touch policy
+          if (Options["Threads"] > 8)
               WinProcGroup::bindThisThread(idx);
+
+          // Each thread will zero its part of the hash table
+          const size_t stride = clusterCount / Options["Threads"],
+                       start  = stride * idx,
+                       len    = idx != Options["Threads"] - 1 ?
+                                stride : clusterCount - start;
+
           std::memset(&table[start], 0, len * sizeof(Cluster));
-      }));
+      });
   }
 
   for (std::thread& th: threads)
@@ -94,8 +124,7 @@ TTEntry* TranspositionTable::probe(const Key key, bool& found) const {
   for (int i = 0; i < ClusterSize; ++i)
       if (!tte[i].key16 || tte[i].key16 == key16)
       {
-          if ((tte[i].genBound8 & 0xFC) != generation8 && tte[i].key16)
-              tte[i].genBound8 = uint8_t(generation8 | tte[i].bound()); // Refresh
+          tte[i].genBound8 = uint8_t(generation8 | (tte[i].genBound8 & 0x7)); // Refresh
 
           return found = (bool)tte[i].key16, &tte[i];
       }
@@ -104,11 +133,11 @@ TTEntry* TranspositionTable::probe(const Key key, bool& found) const {
   TTEntry* replace = tte;
   for (int i = 1; i < ClusterSize; ++i)
       // Due to our packed storage format for generation and its cyclic
-      // nature we add 259 (256 is the modulus plus 3 to keep the lowest
-      // two bound bits from affecting the result) to calculate the entry
+      // nature we add 263 (256 is the modulus plus 7 to keep the unrelated
+      // lowest three bits from affecting the result) to calculate the entry
       // age correctly even after generation8 overflows into the next cycle.
-      if (  replace->depth8 - ((259 + generation8 - replace->genBound8) & 0xFC) * 2
-          >   tte[i].depth8 - ((259 + generation8 -   tte[i].genBound8) & 0xFC) * 2)
+      if (  replace->depth8 - ((263 + generation8 - replace->genBound8) & 0xF8)
+          >   tte[i].depth8 - ((263 + generation8 -   tte[i].genBound8) & 0xF8))
           replace = &tte[i];
 
   return found = false, replace;
@@ -121,12 +150,9 @@ TTEntry* TranspositionTable::probe(const Key key, bool& found) const {
 int TranspositionTable::hashfull() const {
 
   int cnt = 0;
-  for (int i = 0; i < 1000 / ClusterSize; i++)
-  {
-      const TTEntry* tte = &table[i].entry[0];
-      for (int j = 0; j < ClusterSize; j++)
-          if ((tte[j].genBound8 & 0xFC) == generation8)
-              cnt++;
-  }
-  return cnt;
+  for (int i = 0; i < 1000 / ClusterSize; ++i)
+      for (int j = 0; j < ClusterSize; ++j)
+          cnt += (table[i].entry[j].genBound8 & 0xF8) == generation8;
+
+  return cnt * 1000 / (ClusterSize * (1000 / ClusterSize));
 }