Optimize the most common update accumalator cases w/o tiling

[stockfish] / src / misc.h

diff --git a/src/misc.h b/src/misc.h
index 9761da8addd93b5ec4a41223af7742de26582fe2..52595fb906bdd448472332de91e6e63f515c0482 100644 (file)
--- a/src/misc.h
+++ b/src/misc.h
@@ -21,12 +21,13 @@
 
 #include <cassert>
 #include <chrono>
-#include <ostream>
-#include <string>
-#include <vector>
+#include <cstddef>
 #include <cstdint>
+#include <iosfwd>
+#include <string>
 
-#include "types.h"
+#define stringify2(x) #x
+#define stringify(x) stringify2(x)
 
 namespace Stockfish {
 
@@ -45,21 +46,13 @@ void dbg_stdev_of(int64_t value, int slot = 0);
 void dbg_correl_of(int64_t value1, int64_t value2, int slot = 0);
 void dbg_print();
 
-typedef std::chrono::milliseconds::rep TimePoint; // A value in milliseconds
+using TimePoint = std::chrono::milliseconds::rep; // A value in milliseconds
 static_assert(sizeof(TimePoint) == sizeof(int64_t), "TimePoint should be 64 bits");
 inline TimePoint now() {
   return std::chrono::duration_cast<std::chrono::milliseconds>
         (std::chrono::steady_clock::now().time_since_epoch()).count();
 }
 
-template<class Entry, int Size>
-struct HashTable {
-  Entry* operator[](Key key) { return &table[(uint32_t)key & (Size - 1)]; }
-
-private:
-  std::vector<Entry> table = std::vector<Entry>(Size); // Allocate on the heap
-};
-
 
 enum SyncCout { IO_LOCK, IO_UNLOCK };
 std::ostream& operator<<(std::ostream&, SyncCout);
@@ -86,32 +79,6 @@ static inline const union { uint32_t i; char c[4]; } Le = { 0x01020304 };
 static inline const bool IsLittleEndian = (Le.c[0] == 4);
 
 
-// RunningAverage : a class to calculate a running average of a series of values.
-// For efficiency, all computations are done with integers.
-class RunningAverage {
-  public:
-
-      // Reset the running average to rational value p / q
-      void set(int64_t p, int64_t q)
-        { average = p * PERIOD * RESOLUTION / q; }
-
-      // Update average with value v
-      void update(int64_t v)
-        { average = RESOLUTION * v + (PERIOD - 1) * average / PERIOD; }
-
-      // Test if average is strictly greater than rational a / b
-      bool is_greater(int64_t a, int64_t b) const
-        { return b * average > a * (PERIOD * RESOLUTION); }
-
-      int64_t value() const
-        { return average / (PERIOD * RESOLUTION); }
-
-  private :
-      static constexpr int64_t PERIOD     = 4096;
-      static constexpr int64_t RESOLUTION = 1024;
-      int64_t average;
-};
-
 template <typename T, std::size_t MaxSize>
 class ValueList {
 
@@ -120,6 +87,7 @@ public:
   void push_back(const T& value) { values_[size_++] = value; }
   const T* begin() const { return values_; }
   const T* end() const { return values_ + size_; }
+  const T& operator[](int index) const { return values_[index]; }
 
 private:
   T values_[MaxSize];
@@ -165,14 +133,14 @@ public:
 
 inline uint64_t mul_hi64(uint64_t a, uint64_t b) {
 #if defined(__GNUC__) && defined(IS_64BIT)
-    __extension__ typedef unsigned __int128 uint128;
-    return ((uint128)a * (uint128)b) >> 64;
+    __extension__ using uint128 = unsigned __int128;
+    return (uint128(a) * uint128(b)) >> 64;
 #else
-    uint64_t aL = (uint32_t)a, aH = a >> 32;
-    uint64_t bL = (uint32_t)b, bH = b >> 32;
+    uint64_t aL = uint32_t(a), aH = a >> 32;
+    uint64_t bL = uint32_t(b), bH = b >> 32;
     uint64_t c1 = (aL * bL) >> 32;
     uint64_t c2 = aH * bL + c1;
-    uint64_t c3 = aL * bH + (uint32_t)c2;
+    uint64_t c3 = aL * bH + uint32_t(c2);
     return aH * bH + (c2 >> 32) + (c3 >> 32);
 #endif
 }