Use threads to clear the TT faster.

[stockfish] / src / tt.cpp

diff --git a/src/tt.cpp b/src/tt.cpp
index f9674601373319bdd32003429d000a8dc99ac448..5fa1290b9f721400cdb70d785277bc9fd5a8b736 100644 (file)
--- a/src/tt.cpp
+++ b/src/tt.cpp
@@ -2,6 +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
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -19,9 +20,12 @@
 
 #include <cstring>   // For std::memset
 #include <iostream>
+#include <thread>
 
 #include "bitboard.h"
+#include "misc.h"
 #include "tt.h"
+#include "uci.h"
 
 TranspositionTable TT; // Our global transposition table
 
@@ -32,7 +36,7 @@ TranspositionTable TT; // Our global transposition table
 
 void TranspositionTable::resize(size_t mbSize) {
 
-  size_t newClusterCount = size_t(1) << msb((mbSize * 1024 * 1024) / sizeof(Cluster));
+  size_t newClusterCount = mbSize * 1024 * 1024 / sizeof(Cluster);
 
   if (newClusterCount == clusterCount)
       return;
@@ -40,7 +44,7 @@ void TranspositionTable::resize(size_t mbSize) {
   clusterCount = newClusterCount;
 
   free(mem);
-  mem = calloc(clusterCount * sizeof(Cluster) + CacheLineSize - 1, 1);
+  mem = malloc(clusterCount * sizeof(Cluster) + CacheLineSize - 1);
 
   if (!mem)
   {
@@ -50,18 +54,35 @@ void TranspositionTable::resize(size_t mbSize) {
   }
 
   table = (Cluster*)((uintptr_t(mem) + CacheLineSize - 1) & ~(CacheLineSize - 1));
+  clear();
 }
 
 
 /// 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.
 
 void TranspositionTable::clear() {
 
-  std::memset(table, 0, clusterCount * sizeof(Cluster));
-}
+  const size_t stride = clusterCount / Options["Threads"];
+  std::vector<std::thread> threads;
+  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)
+              WinProcGroup::bindThisThread(idx);
+          std::memset(&table[start], 0, len * sizeof(Cluster));
+      }));
+  }
 
+  for (std::thread& th: threads)
+      th.join();
+}
 
 /// TranspositionTable::probe() looks up the current position in the transposition
 /// table. It returns true and a pointer to the TTEntry if the position is found.
@@ -78,7 +99,7 @@ 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].key16)
+          if ((tte[i].genBound8 & 0xFC) != generation8 && tte[i].key16)
               tte[i].genBound8 = uint8_t(generation8 | tte[i].bound()); // Refresh
 
           return found = (bool)tte[i].key16, &tte[i];
@@ -91,19 +112,19 @@ TTEntry* TranspositionTable::probe(const Key key, bool& found) const {
       // 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
       // age correctly even after generation8 overflows into the next cycle.
-      if (  replace->depth8 - ((259 + generation8 - replace->genBound8) & 0xFC) * 2 * ONE_PLY
-          >   tte[i].depth8 - ((259 + generation8 -   tte[i].genBound8) & 0xFC) * 2 * ONE_PLY)
+      if (  replace->depth8 - ((259 + generation8 - replace->genBound8) & 0xFC) * 2
+          >   tte[i].depth8 - ((259 + generation8 -   tte[i].genBound8) & 0xFC) * 2)
           replace = &tte[i];
 
   return found = false, replace;
 }
 
 
-/// Returns an approximation of the hashtable occupation during a search. The
-/// hash is x permill full, as per UCI protocol.
+/// TranspositionTable::hashfull() returns an approximation of the hashtable
+/// occupation during a search. The hash is x permill full, as per UCI protocol.
+
+int TranspositionTable::hashfull() const {
 
-int TranspositionTable::hashfull() const
-{
   int cnt = 0;
   for (int i = 0; i < 1000 / ClusterSize; i++)
   {