WebSVN – Games.Chess Giants – Diff – /engine-stockfish/syzygy/tbprobe.cpp

 /*
+  Stockfish, a UCI chess playing engine derived from Glaurung 2.1
   Copyright (c) 2013 Ronald de Man
-  This file may be redistributed and/or modified without restrictions.
+  Copyright (C) 2016-2018 Marco Costalba, Lucas Braesch
+  Stockfish is free software: you can redistribute it and/or modify
-  tbprobe.cpp contains the Stockfish-specific routines of the
+  it under the terms of the GNU General Public License as published by
-  tablebase probing code. It should be relatively easy to adapt
+  the Free Software Foundation, either version 3 of the License, or
-  this code to other chess engines.
+  (at your option) any later version.
+  Stockfish is distributed in the hope that it will be useful,
+  but WITHOUT ANY WARRANTY; without even the implied warranty of
+  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+  GNU General Public License for more details.
+  You should have received a copy of the GNU General Public License
+  along with this program.  If not, see <http://www.gnu.org/licenses/>.
 */
-#define NOMINMAX
 #include <algorithm>
+#include <atomic>
+#include <cstdint>
+#include <cstring>   // For std::memset
+#include <deque>
+#include <fstream>
+#include <iostream>
+#include <list>
+#include <sstream>
+#include <type_traits>
+#include "../bitboard.h"
+#include "../movegen.h"
 #include "../position.h"
-#include "../movegen.h"
-#include "../bitboard.h"
 #include "../search.h"
+#include "../thread_win32.h"
+#include "../types.h"
 #include "tbprobe.h"
-#include "tbcore.h"
+#ifndef _WIN32
-#include "tbcore.cpp"
+#include <fcntl.h>
+#include <unistd.h>
+#include <sys/mman.h>
+#include <sys/stat.h>
+#else
+#define WIN32_LEAN_AND_MEAN
+#define NOMINMAX
+#include <windows.h>
+#endif
-namespace Zobrist {
+using namespace Tablebases;
-  extern Key psq[PIECE_NB][SQUARE_NB];
-}
-int Tablebases::MaxCardinality = 0;
+int Tablebases::MaxCardinality;
-// Given a position with 6 or fewer pieces, produce a text string
-// of the form KQPvKRP, where "KQP" represents the white pieces if
-// mirror == 0 and the black pieces if mirror == 1.
-static void prt_str(Position& pos, char *str, int mirror)
-{
-  Color color;
-  PieceType pt;
+namespace {
-  int i;
-  color = !mirror ? WHITE : BLACK;
-  for (pt = KING; pt >= PAWN; --pt)
-    for (i = popcount(pos.pieces(color, pt)); i > 0; i--)
+// Each table has a set of flags: all of them refer to DTZ tables, the last one to WDL tables
-      *str++ = pchr[6 - pt];
-  *str++ = 'v';
-  color = ~color;
-  for (pt = KING; pt >= PAWN; --pt)
-    for (i = popcount(pos.pieces(color, pt)); i > 0; i--)
+enum TBFlag { STM = 1, Mapped = 2, WinPlies = 4, LossPlies = 8, SingleValue = 128 };
-      *str++ = pchr[6 - pt];
-  *str++ = 0;
-}
-// Given a position, produce a 64-bit material signature key.
+inline WDLScore operator-(WDLScore d) { return WDLScore(-int(d)); }
-// If the engine supports such a key, it should equal the engine's key.
+inline Square operator^=(Square& s, int i) { return s = Square(int(s) ^ i); }
-static uint64 calc_key(Position& pos, int mirror)
+inline Square operator^(Square s, int i) { return Square(int(s) ^ i); }
-{
-  Color color;
-  PieceType pt;
-  int i;
-  uint64 key = 0;
-  color = !mirror ? WHITE : BLACK;
+// DTZ tables don't store valid scores for moves that reset the rule50 counter
-  for (pt = PAWN; pt <= KING; ++pt)
+// like captures and pawn moves but we can easily recover the correct dtz of the
-    for (i = popcount(pos.pieces(color, pt)); i > 0; i--)
+// previous move if we know the position's WDL score.
-      key ^= Zobrist::psq[make_piece(WHITE, pt)][i - 1];
+int dtz_before_zeroing(WDLScore wdl) {
-  color = ~color;
+    return wdl == WDLWin         ?  1   :
-  for (pt = PAWN; pt <= KING; ++pt)
+           wdl == WDLCursedWin   ?  101 :
-    for (i = popcount(pos.pieces(color, pt)); i > 0; i--)
+           wdl == WDLBlessedLoss ? -101 :
-      key ^= Zobrist::psq[make_piece(BLACK, pt)][i - 1];
+           wdl == WDLLoss        ? -1   : 0;
+}
+// Return the sign of a number (-1, 0, 1)
+template <typename T> int sign_of(T val) {
-  return key;
+    return (T(0) < val) - (val < T(0));
+}
-// Produce a 64-bit material key corresponding to the material combination
+// Numbers in little endian used by sparseIndex[] to point into blockLength[]
-// defined by pcs[16], where pcs[1], ..., pcs[6] is the number of white
+struct SparseEntry {
-// pawns, ..., kings and pcs[9], ..., pcs[14] is the number of black
+    char block[4];   // Number of block
-// pawns, ..., kings.
-static uint64 calc_key_from_pcs(int *pcs, int mirror)
+    char offset[2];  // Offset within the block
-{
+};
-  int color;
-  PieceType pt;
-  int i;
-  uint64 key = 0;
-  color = !mirror ? 0 : 8;
-  for (pt = PAWN; pt <= KING; ++pt)
-    for (i = 0; i < pcs[color + pt]; i++)
-      key ^= Zobrist::psq[make_piece(WHITE, pt)][i];
-  color ^= 8;
-  for (pt = PAWN; pt <= KING; ++pt)
-    for (i = 0; i < pcs[color + pt]; i++)
-      key ^= Zobrist::psq[make_piece(BLACK, pt)][i];
+static_assert(sizeof(SparseEntry) == 6, "SparseEntry must be 6 bytes");
-  return key;
+typedef uint16_t Sym; // Huffman symbol
-}
+struct LR {
+    enum Side { Left, Right, Value };
+    uint8_t lr[3]; // The first 12 bits is the left-hand symbol, the second 12
+                   // bits is the right-hand symbol. If symbol has length 1,
+                   // then the first byte is the stored value.
-bool is_little_endian() {
+    template<Side S>
+    Sym get() {
+        return S == Left  ? ((lr[1] & 0xF) << 8) | lr[0] :
+               S == Right ?  (lr[2] << 4) | (lr[1] >> 4) :
+               S == Value ?   lr[0] : (assert(false), Sym(-1));
-  union {
+    }
+};
+static_assert(sizeof(LR) == 3, "LR tree entry must be 3 bytes");
+const int TBPIECES = 6;
+struct PairsData {
-    int i;
+    int flags;
+    size_t sizeofBlock;            // Block size in bytes
+    size_t span;                   // About every span values there is a SparseIndex[] entry
+    int blocksNum;                 // Number of blocks in the TB file
+    int maxSymLen;                 // Maximum length in bits of the Huffman symbols
+    int minSymLen;                 // Minimum length in bits of the Huffman symbols
+    Sym* lowestSym;                // lowestSym[l] is the symbol of length l with the lowest value
+    LR* btree;                     // btree[sym] stores the left and right symbols that expand sym
+    uint16_t* blockLength;         // Number of stored positions (minus one) for each block: 1..65536
+    int blockLengthSize;           // Size of blockLength[] table: padded so it's bigger than blocksNum
+    SparseEntry* sparseIndex;      // Partial indices into blockLength[]
+    size_t sparseIndexSize;        // Size of SparseIndex[] table
+    uint8_t* data;                 // Start of Huffman compressed data
+    std::vector<uint64_t> base64;  // base64[l - min_sym_len] is the 64bit-padded lowest symbol of length l
+    std::vector<uint8_t> symlen;   // Number of values (-1) represented by a given Huffman symbol: 1..256
+    Piece pieces[TBPIECES];        // Position pieces: the order of pieces defines the groups
+    uint64_t groupIdx[TBPIECES+1]; // Start index used for the encoding of the group's pieces
+    int groupLen[TBPIECES+1];      // Number of pieces in a given group: KRKN -> (3, 1)
+};
+// Helper struct to avoid manually defining entry copy constructor as we
+// should because the default one is not compatible with std::atomic_bool.
+struct Atomic {
+    Atomic() = default;
+    Atomic(const Atomic& e) { ready = e.ready.load(); } // MSVC 2013 wants assignment within body
+    std::atomic_bool ready;
+};
+// We define types for the different parts of the WDLEntry and DTZEntry with
+// corresponding specializations for pieces or pawns.
+struct WDLEntryPiece {
+    PairsData* precomp;
+};
+struct WDLEntryPawn {
+    uint8_t pawnCount[2];     // [Lead color / other color]
+    WDLEntryPiece file[2][4]; // [wtm / btm][FILE_A..FILE_D]
+};
+struct DTZEntryPiece {
+    PairsData* precomp;
+    uint16_t map_idx[4]; // WDLWin, WDLLoss, WDLCursedWin, WDLBlessedLoss
+    uint8_t* map;
+};
+struct DTZEntryPawn {
+    uint8_t pawnCount[2];
-    char c[sizeof(int)];
+    DTZEntryPiece file[4];
+    uint8_t* map;
+};
+struct TBEntry : public Atomic {
+    void* baseAddress;
+    uint64_t mapping;
+    Key key;
+    Key key2;
+    int pieceCount;
+    bool hasPawns;
+    bool hasUniquePieces;
+};
+// Now the main types: WDLEntry and DTZEntry
+struct WDLEntry : public TBEntry {
+    WDLEntry(const std::string& code);
+   ~WDLEntry();
+    union {
+        WDLEntryPiece pieceTable[2]; // [wtm / btm]
+        WDLEntryPawn  pawnTable;
-  } x;
+    };
+};
+struct DTZEntry : public TBEntry {
+    DTZEntry(const WDLEntry& wdl);
+   ~DTZEntry();
+    union {
+        DTZEntryPiece pieceTable;
+        DTZEntryPawn  pawnTable;
-  x.i = 1;
+    };
+};
+typedef decltype(WDLEntry::pieceTable) WDLPieceTable;
+typedef decltype(DTZEntry::pieceTable) DTZPieceTable;
+typedef decltype(WDLEntry::pawnTable ) WDLPawnTable;
+typedef decltype(DTZEntry::pawnTable ) DTZPawnTable;
+auto item(WDLPieceTable& e, int stm, int  ) -> decltype(e[stm])& { return e[stm]; }
+auto item(DTZPieceTable& e, int    , int  ) -> decltype(e)& { return e; }
+auto item(WDLPawnTable&  e, int stm, int f) -> decltype(e.file[stm][f])& { return e.file[stm][f]; }
+auto item(DTZPawnTable&  e, int    , int f) -> decltype(e.file[f])& { return e.file[f]; }
+template<typename E> struct Ret { typedef int type; };
+template<> struct Ret<WDLEntry> { typedef WDLScore type; };
+int MapPawns[SQUARE_NB];
+int MapB1H1H7[SQUARE_NB];
+int MapA1D1D4[SQUARE_NB];
+int MapKK[10][SQUARE_NB]; // [MapA1D1D4][SQUARE_NB]
+// Comparison function to sort leading pawns in ascending MapPawns[] order
+bool pawns_comp(Square i, Square j) { return MapPawns[i] < MapPawns[j]; }
+int off_A1H8(Square sq) { return int(rank_of(sq)) - file_of(sq); }
+const Value WDL_to_value[] = {
+   -VALUE_MATE + MAX_PLY + 1,
+    VALUE_DRAW - 2,
+    VALUE_DRAW,
-  return x.c[0] == 1;
+    VALUE_DRAW + 2,
+    VALUE_MATE - MAX_PLY - 1
-}
+};
+const std::string PieceToChar = " PNBRQK  pnbrqk";
+int Binomial[6][SQUARE_NB];    // [k][n] k elements from a set of n elements
+int LeadPawnIdx[5][SQUARE_NB]; // [leadPawnsCnt][SQUARE_NB]
+int LeadPawnsSize[5][4];       // [leadPawnsCnt][FILE_A..FILE_D]
+enum { BigEndian, LittleEndian };
-static ubyte decompress_pairs(struct PairsData *d, uint64 idx)
+template<typename T, int Half = sizeof(T) / 2, int End = sizeof(T) - 1>
+inline void swap_byte(T& x)
+{
-  static const bool isLittleEndian = is_little_endian();
+    char tmp, *c = (char*)&x;
-  return isLittleEndian ? decompress_pairs<true >(d, idx)
+    for (int i = 0; i < Half; ++i)
-                        : decompress_pairs<false>(d, idx);
+        tmp = c[i], c[i] = c[End - i], c[End - i] = tmp;
+}
+template<> inline void swap_byte<uint8_t, 0, 0>(uint8_t&) {}
-// probe_wdl_table and probe_dtz_table require similar adaptations.
+template<typename T, int LE> T number(void* addr)
-static int probe_wdl_table(Position& pos, int *success)
+{
-  struct TBEntry *ptr;
+    const union { uint32_t i; char c[4]; } Le = { 0x01020304 };
-  struct TBHashEntry *ptr2;
+    const bool IsLittleEndian = (Le.c[0] == 4);
-  uint64 idx;
-  uint64 key;
-  int i;
-  ubyte res;
-  int p[TBPIECES];
-  // Obtain the position's material signature key.
-  key = pos.material_key();
+    T v;
+    if ((uintptr_t)addr & (alignof(T) - 1)) // Unaligned pointer (very rare)
+        std::memcpy(&v, addr, sizeof(T));
+    else
+        v = *((T*)addr);
+    if (LE != IsLittleEndian)
+        swap_byte(v);
-  // Test for KvK.
+    return v;
+}
+class HashTable {
+    typedef std::pair<WDLEntry*, DTZEntry*> EntryPair;
+    typedef std::pair<Key, EntryPair> Entry;
+    static const int TBHASHBITS = 10;
+    static const int HSHMAX     = 5;
+    Entry hashTable[1 << TBHASHBITS][HSHMAX];
+    std::deque<WDLEntry> wdlTable;
+    std::deque<DTZEntry> dtzTable;
+    void insert(Key key, WDLEntry* wdl, DTZEntry* dtz) {
+        Entry* entry = hashTable[key >> (64 - TBHASHBITS)];
+        for (int i = 0; i < HSHMAX; ++i, ++entry)
+            if (!entry->second.first || entry->first == key) {
-  if (key == (Zobrist::psq[W_KING][0] ^ Zobrist::psq[B_KING][0]))
+                *entry = std::make_pair(key, std::make_pair(wdl, dtz));
+                return;
-    return 0;
+            }
+        std::cerr << "HSHMAX too low!" << std::endl;
+        exit(1);
+    }
+public:
+    template<typename E, int I = std::is_same<E, WDLEntry>::value ? 0 : 1>
+    E* get(Key key) {
+      Entry* entry = hashTable[key >> (64 - TBHASHBITS)];
+      for (int i = 0; i < HSHMAX; ++i, ++entry)
+          if (entry->first == key)
+              return std::get<I>(entry->second);
-  ptr2 = TB_hash[key >> (64 - TBHASHBITS)];
-  for (i = 0; i < HSHMAX; i++)
-    if (ptr2[i].key == key) break;
-  if (i == HSHMAX) {
-    *success = 0;
-    return 0;
+      return nullptr;
+  }
-  ptr = ptr2[i].ptr;
+  void clear() {
-  if (!ptr->ready) {
+      std::memset(hashTable, 0, sizeof(hashTable));
-    LOCK(TB_mutex);
+      wdlTable.clear();
-    if (!ptr->ready) {
+      dtzTable.clear();
-      char str[16];
+  }
-      prt_str(pos, str, ptr->key != key);
+  size_t size() const { return wdlTable.size(); }
-      if (!init_table_wdl(ptr, str)) {
+  void insert(const std::vector<PieceType>& pieces);
+};
-        ptr2[i].key = 0ULL;
+HashTable EntryTable;
-        *success = 0;
+class TBFile : public std::ifstream {
-        UNLOCK(TB_mutex);
+    std::string fname;
-        return 0;
+public:
+    // Look for and open the file among the Paths directories where the .rtbw
+    // and .rtbz files can be found. Multiple directories are separated by ";"
+    // on Windows and by ":" on Unix-based operating systems.
-      }
+    //
+    // Example:
-      // Memory barrier to ensure ptr->ready = 1 is not reordered.
+    // C:\tb\wdl345;C:\tb\wdl6;D:\tb\dtz345;D:\tb\dtz6
+    static std::string Paths;
+    TBFile(const std::string& f) {
-#ifdef _MSC_VER
+#ifndef _WIN32
-      _ReadWriteBarrier();
+        const char SepChar = ':';
 #else
-      __asm__ __volatile__ ("" ::: "memory");
+        const char SepChar = ';';
 #endif
+        std::stringstream ss(Paths);
-      ptr->ready = 1;
+        std::string path;
+        while (std::getline(ss, path, SepChar)) {
+            fname = path + "/" + f;
+            std::ifstream::open(fname);
+            if (is_open())
+                return;
+        }
+    }
-    UNLOCK(TB_mutex);
-  }
+    // Memory map the file and check it. File should be already open and will be
-  int bside, mirror, cmirror;
+    // closed after mapping.
+    uint8_t* map(void** baseAddress, uint64_t* mapping, const uint8_t* TB_MAGIC) {
-  if (!ptr->symmetric) {
+        assert(is_open());
+        close(); // Need to re-open to get native file descriptor
+#ifndef _WIN32
-    if (key != ptr->key) {
+        struct stat statbuf;
+        int fd = ::open(fname.c_str(), O_RDONLY);
-      cmirror = 8;
+        if (fd == -1)
+            return *baseAddress = nullptr, nullptr;
+        fstat(fd, &statbuf);
+        *mapping = statbuf.st_size;
+        *baseAddress = mmap(nullptr, statbuf.st_size, PROT_READ, MAP_SHARED, fd, 0);
-      mirror = 0x38;
+        ::close(fd);
-      bside = (pos.side_to_move() == WHITE);
+        if (*baseAddress == MAP_FAILED) {
+            std::cerr << "Could not mmap() " << fname << std::endl;
+            exit(1);
-    } else {
+        }
+#else
+        HANDLE fd = CreateFile(fname.c_str(), GENERIC_READ, FILE_SHARE_READ, nullptr,
+                               OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL, nullptr);
+        if (fd == INVALID_HANDLE_VALUE)
+            return *baseAddress = nullptr, nullptr;
-      cmirror = mirror = 0;
+        DWORD size_high;
+        DWORD size_low = GetFileSize(fd, &size_high);
+        HANDLE mmap = CreateFileMapping(fd, nullptr, PAGE_READONLY, size_high, size_low, nullptr);
+        CloseHandle(fd);
+        if (!mmap) {
+            std::cerr << "CreateFileMapping() failed" << std::endl;
+            exit(1);
+        }
+        *mapping = (uint64_t)mmap;
+        *baseAddress = MapViewOfFile(mmap, FILE_MAP_READ, 0, 0, 0);
+        if (!*baseAddress) {
+            std::cerr << "MapViewOfFile() failed, name = " << fname
+                      << ", error = " << GetLastError() << std::endl;
+            exit(1);
+        }
+#endif
+        uint8_t* data = (uint8_t*)*baseAddress;
+        if (   *data++ != *TB_MAGIC++
+            || *data++ != *TB_MAGIC++
+            || *data++ != *TB_MAGIC++
+            || *data++ != *TB_MAGIC) {
+            std::cerr << "Corrupted table in file " << fname << std::endl;
-      bside = !(pos.side_to_move() == WHITE);
+            unmap(*baseAddress, *mapping);
+            return *baseAddress = nullptr, nullptr;
+        }
+        return data;
+    }
-  } else {
-    cmirror = pos.side_to_move() == WHITE ? 0 : 8;
-    mirror = pos.side_to_move() == WHITE ? 0 : 0x38;
-    bside = 0;
-  }
-  // p[i] is to contain the square 0-63 (A1-H8) for a piece of type
-  // pc[i] ^ cmirror, where 1 = white pawn, ..., 14 = black king.
+    static void unmap(void* baseAddress, uint64_t mapping) {
-  // Pieces of the same type are guaranteed to be consecutive.
-  if (!ptr->has_pawns) {
+#ifndef _WIN32
-    struct TBEntry_piece *entry = (struct TBEntry_piece *)ptr;
-    ubyte *pc = entry->pieces[bside];
+        munmap(baseAddress, mapping);
-    for (i = 0; i < entry->num;) {
-      Bitboard bb = pos.pieces((Color)((pc[i] ^ cmirror) >> 3),
-                                      (PieceType)(pc[i] & 0x07));
-      do {
+#else
-        p[i++] = pop_lsb(&bb);
+        UnmapViewOfFile(baseAddress);
-      } while (bb);
+        CloseHandle((HANDLE)mapping);
+#endif
+    }
-    idx = encode_piece(entry, entry->norm[bside], p, entry->factor[bside]);
-    res = decompress_pairs(entry->precomp[bside], idx);
-  } else {
+};
-    struct TBEntry_pawn *entry = (struct TBEntry_pawn *)ptr;
+std::string TBFile::Paths;
-    int k = entry->file[0].pieces[0][0] ^ cmirror;
+WDLEntry::WDLEntry(const std::string& code) {
-    Bitboard bb = pos.pieces((Color)(k >> 3), (PieceType)(k & 0x07));
-    i = 0;
+    StateInfo st;
-    do {
+    Position pos;
-      p[i++] = pop_lsb(&bb) ^ mirror;
+    memset(this, 0, sizeof(WDLEntry));
-    } while (bb);
+    ready = false;
+    key = pos.set(code, WHITE, &st).material_key();
-    int f = pawn_file(entry, p);
+    pieceCount = popcount(pos.pieces());
-    ubyte *pc = entry->file[f].pieces[bside];
+    hasPawns = pos.pieces(PAWN);
-    for (; i < entry->num;) {
+    for (Color c = WHITE; c <= BLACK; ++c)
+        for (PieceType pt = PAWN; pt < KING; ++pt)
-      bb = pos.pieces((Color)((pc[i] ^ cmirror) >> 3),
+            if (popcount(pos.pieces(c, pt)) == 1)
-                                    (PieceType)(pc[i] & 0x07));
+                hasUniquePieces = true;
-      do {
+    if (hasPawns) {
+        // Set the leading color. In case both sides have pawns the leading color
+        // is the side with less pawns because this leads to better compression.
-        p[i++] = pop_lsb(&bb) ^ mirror;
+        bool c =   !pos.count<PAWN>(BLACK)
-      } while (bb);
+                || (   pos.count<PAWN>(WHITE)
+                    && pos.count<PAWN>(BLACK) >= pos.count<PAWN>(WHITE));
+        pawnTable.pawnCount[0] = pos.count<PAWN>(c ? WHITE : BLACK);
+        pawnTable.pawnCount[1] = pos.count<PAWN>(c ? BLACK : WHITE);
+    }
-    idx = encode_pawn(entry, entry->file[f].norm[bside], p, entry->file[f].factor[bside]);
-    res = decompress_pairs(entry->file[f].precomp[bside], idx);
-  }
-  return ((int)res) - 2;
+    key2 = pos.set(code, BLACK, &st).material_key();
+}
-static int probe_dtz_table(Position& pos, int wdl, int *success)
-{
-  struct TBEntry *ptr;
+WDLEntry::~WDLEntry() {
-  uint64 idx;
-  int i, res;
-  int p[TBPIECES];
-  // Obtain the position's material signature key.
+    if (baseAddress)
-  uint64 key = pos.material_key();
+        TBFile::unmap(baseAddress, mapping);
-  if (DTZ_table[0].key1 != key && DTZ_table[0].key2 != key) {
-    for (i = 1; i < DTZ_ENTRIES; i++)
+    for (int i = 0; i < 2; ++i)
-      if (DTZ_table[i].key1 == key) break;
-    if (i < DTZ_ENTRIES) {
-      struct DTZTableEntry table_entry = DTZ_table[i];
-      for (; i > 0; i--)
+        if (hasPawns)
-        DTZ_table[i] = DTZ_table[i - 1];
+            for (File f = FILE_A; f <= FILE_D; ++f)
-      DTZ_table[0] = table_entry;
+                delete pawnTable.file[i][f].precomp;
-    } else {
+        else
-      struct TBHashEntry *ptr2 = TB_hash[key >> (64 - TBHASHBITS)];
+            delete pieceTable[i].precomp;
+}
-      for (i = 0; i < HSHMAX; i++)
+DTZEntry::DTZEntry(const WDLEntry& wdl) {
-        if (ptr2[i].key == key) break;
+    memset(this, 0, sizeof(DTZEntry));
-      if (i == HSHMAX) {
+    ready = false;
-        *success = 0;
+    key = wdl.key;
-        return 0;
+    key2 = wdl.key2;
-      }
-      ptr = ptr2[i].ptr;
+    pieceCount = wdl.pieceCount;
-      char str[16];
+    hasPawns = wdl.hasPawns;
-      int mirror = (ptr->key != key);
+    hasUniquePieces = wdl.hasUniquePieces;
-      prt_str(pos, str, mirror);
+    if (hasPawns) {
-      if (DTZ_table[DTZ_ENTRIES - 1].entry)
-        free_dtz_entry(DTZ_table[DTZ_ENTRIES-1].entry);
+        pawnTable.pawnCount[0] = wdl.pawnTable.pawnCount[0];
-      for (i = DTZ_ENTRIES - 1; i > 0; i--)
-        DTZ_table[i] = DTZ_table[i - 1];
+        pawnTable.pawnCount[1] = wdl.pawnTable.pawnCount[1];
-      load_dtz_table(str, calc_key(pos, mirror), calc_key(pos, !mirror));
+    }
-  }
+}
-  ptr = DTZ_table[0].entry;
+DTZEntry::~DTZEntry() {
-  if (!ptr) {
-    *success = 0;
-    return 0;
-  }
+    if (baseAddress)
+        TBFile::unmap(baseAddress, mapping);
+    if (hasPawns)
+        for (File f = FILE_A; f <= FILE_D; ++f)
+            delete pawnTable.file[f].precomp;
+    else
-  int bside, mirror, cmirror;
+        delete pieceTable.precomp;
+}
+void HashTable::insert(const std::vector<PieceType>& pieces) {
-  if (!ptr->symmetric) {
+    std::string code;
-    if (key != ptr->key) {
+    for (PieceType pt : pieces)
+        code += PieceToChar[pt];
+    TBFile file(code.insert(code.find('K', 1), "v") + ".rtbw"); // KRK -> KRvK
+    if (!file.is_open()) // Only WDL file is checked
-      cmirror = 8;
+        return;
+    file.close();
+    MaxCardinality = std::max((int)pieces.size(), MaxCardinality);
+    wdlTable.emplace_back(code);
+    dtzTable.emplace_back(wdlTable.back());
+    insert(wdlTable.back().key , &wdlTable.back(), &dtzTable.back());
+    insert(wdlTable.back().key2, &wdlTable.back(), &dtzTable.back());
+}
+// TB tables are compressed with canonical Huffman code. The compressed data is divided into
+// blocks of size d->sizeofBlock, and each block stores a variable number of symbols.
+// Each symbol represents either a WDL or a (remapped) DTZ value, or a pair of other symbols
+// (recursively). If you keep expanding the symbols in a block, you end up with up to 65536
+// WDL or DTZ values. Each symbol represents up to 256 values and will correspond after
+// Huffman coding to at least 1 bit. So a block of 32 bytes corresponds to at most
+// 32 x 8 x 256 = 65536 values. This maximum is only reached for tables that consist mostly
+// of draws or mostly of wins, but such tables are actually quite common. In principle, the
+// blocks in WDL tables are 64 bytes long (and will be aligned on cache lines). But for
+// mostly-draw or mostly-win tables this can leave many 64-byte blocks only half-filled, so
+// in such cases blocks are 32 bytes long. The blocks of DTZ tables are up to 1024 bytes long.
+// The generator picks the size that leads to the smallest table. The "book" of symbols and
+// Huffman codes is the same for all blocks in the table. A non-symmetric pawnless TB file
+// will have one table for wtm and one for btm, a TB file with pawns will have tables per
+// file a,b,c,d also in this case one set for wtm and one for btm.
+int decompress_pairs(PairsData* d, uint64_t idx) {
+    // Special case where all table positions store the same value
+    if (d->flags & TBFlag::SingleValue)
+        return d->minSymLen;
+    // First we need to locate the right block that stores the value at index "idx".
+    // Because each block n stores blockLength[n] + 1 values, the index i of the block
+    // that contains the value at position idx is:
+    //
+    //                    for (i = -1, sum = 0; sum <= idx; i++)
+    //                        sum += blockLength[i + 1] + 1;
+    //
+    // This can be slow, so we use SparseIndex[] populated with a set of SparseEntry that
+    // point to known indices into blockLength[]. Namely SparseIndex[k] is a SparseEntry
+    // that stores the blockLength[] index and the offset within that block of the value
+    // with index I(k), where:
+    //
+    //       I(k) = k * d->span + d->span / 2      (1)
+    // First step is to get the 'k' of the I(k) nearest to our idx, using definition (1)
+    uint32_t k = (uint32_t) (idx / d->span); // Pierre-Marie Baty -- added type cast
+    // Then we read the corresponding SparseIndex[] entry
+    uint32_t block = number<uint32_t, LittleEndian>(&d->sparseIndex[k].block);
+    int offset     = number<uint16_t, LittleEndian>(&d->sparseIndex[k].offset);
+    // Now compute the difference idx - I(k). From definition of k we know that
+    //
+    //       idx = k * d->span + idx % d->span    (2)
+    //
+    // So from (1) and (2) we can compute idx - I(K):
+    int diff = idx % d->span - d->span / 2;
+    // Sum the above to offset to find the offset corresponding to our idx
-      mirror = 0x38;
+    offset += diff;
+    // Move to previous/next block, until we reach the correct block that contains idx,
+    // that is when 0 <= offset <= d->blockLength[block]
+    while (offset < 0)
+        offset += d->blockLength[--block] + 1;
+    while (offset > d->blockLength[block])
-      bside = (pos.side_to_move() == WHITE);
+        offset -= d->blockLength[block++] + 1;
+    // Finally, we find the start address of our block of canonical Huffman symbols
+    uint32_t* ptr = (uint32_t*)(d->data + block * d->sizeofBlock);
+    // Read the first 64 bits in our block, this is a (truncated) sequence of
+    // unknown number of symbols of unknown length but we know the first one
+    // is at the beginning of this 64 bits sequence.
+    uint64_t buf64 = number<uint64_t, BigEndian>(ptr); ptr += 2;
+    int buf64Size = 64;
-    } else {
+    Sym sym;
+    while (true) {
+        int len = 0; // This is the symbol length - d->min_sym_len
+        // Now get the symbol length. For any symbol s64 of length l right-padded
+        // to 64 bits we know that d->base64[l-1] >= s64 >= d->base64[l] so we
+        // can find the symbol length iterating through base64[].
+        while (buf64 < d->base64[len])
+            ++len;
+        // All the symbols of a given length are consecutive integers (numerical
+        // sequence property), so we can compute the offset of our symbol of
+        // length len, stored at the beginning of buf64.
+        sym = (Sym) ((buf64 - d->base64[len]) >> (64 - len - d->minSymLen)); // Pierre-Marie Baty -- added type cast
+        // Now add the value of the lowest symbol of length len to get our symbol
+        sym += number<Sym, LittleEndian>(&d->lowestSym[len]);
+        // If our offset is within the number of values represented by symbol sym
-      cmirror = mirror = 0;
+        // we are done...
-      bside = !(pos.side_to_move() == WHITE);
+        if (offset < d->symlen[sym] + 1)
+            break;
+        // ...otherwise update the offset and continue to iterate
+        offset -= d->symlen[sym] + 1;
+        len += d->minSymLen; // Get the real length
+        buf64 <<= len;       // Consume the just processed symbol
+        buf64Size -= len;
+        if (buf64Size <= 32) { // Refill the buffer
+            buf64Size += 32;
+            buf64 |= (uint64_t)number<uint32_t, BigEndian>(ptr++) << (64 - buf64Size);
+        }
+    }
-  } else {
-    cmirror = pos.side_to_move() == WHITE ? 0 : 8;
-    mirror = pos.side_to_move() == WHITE ? 0 : 0x38;
-    bside = 0;
-  }
+    // Ok, now we have our symbol that expands into d->symlen[sym] + 1 symbols.
+    // We binary-search for our value recursively expanding into the left and
+    // right child symbols until we reach a leaf node where symlen[sym] + 1 == 1
+    // that will store the value we need.
-  if (!ptr->has_pawns) {
+    while (d->symlen[sym]) {
-    struct DTZEntry_piece *entry = (struct DTZEntry_piece *)ptr;
+        Sym left = d->btree[sym].get<LR::Left>();
-    if ((entry->flags & 1) != bside && !entry->symmetric) {
+        // If a symbol contains 36 sub-symbols (d->symlen[sym] + 1 = 36) and
+        // expands in a pair (d->symlen[left] = 23, d->symlen[right] = 11), then
+        // we know that, for instance the ten-th value (offset = 10) will be on
+        // the left side because in Recursive Pairing child symbols are adjacent.
+        if (offset < d->symlen[left] + 1)
-      *success = -1;
+            sym = left;
-      return 0;
+        else {
+            offset -= d->symlen[left] + 1;
+            sym = d->btree[sym].get<LR::Right>();
+        }
+    }
-    ubyte *pc = entry->pieces;
-    for (i = 0; i < entry->num;) {
-      Bitboard bb = pos.pieces((Color)((pc[i] ^ cmirror) >> 3),
-                                    (PieceType)(pc[i] & 0x07));
-      do {
-        p[i++] = pop_lsb(&bb);
-      } while (bb);
-    }
-    idx = encode_piece((struct TBEntry_piece *)entry, entry->norm, p, entry->factor);
-    res = decompress_pairs(entry->precomp, idx);
-    if (entry->flags & 2)
-      res = entry->map[entry->map_idx[wdl_to_map[wdl + 2]] + res];
+    return d->btree[sym].get<LR::Value>();
+}
+bool check_dtz_stm(WDLEntry*, int, File) { return true; }
+bool check_dtz_stm(DTZEntry* entry, int stm, File f) {
+    int flags = entry->hasPawns ? entry->pawnTable.file[f].precomp->flags
+                                : entry->pieceTable.precomp->flags;
+    return   (flags & TBFlag::STM) == stm
+          || ((entry->key == entry->key2) && !entry->hasPawns);
+}
+// DTZ scores are sorted by frequency of occurrence and then assigned the
+// values 0, 1, 2, ... in order of decreasing frequency. This is done for each
+// of the four WDLScore values. The mapping information necessary to reconstruct
+// the original values is stored in the TB file and read during map[] init.
+WDLScore map_score(WDLEntry*, File, int value, WDLScore) { return WDLScore(value - 2); }
+int map_score(DTZEntry* entry, File f, int value, WDLScore wdl) {
+    const int WDLMap[] = { 1, 3, 0, 2, 0 };
+    int flags = entry->hasPawns ? entry->pawnTable.file[f].precomp->flags
+                                : entry->pieceTable.precomp->flags;
+    uint8_t* map = entry->hasPawns ? entry->pawnTable.map
+                                   : entry->pieceTable.map;
+    uint16_t* idx = entry->hasPawns ? entry->pawnTable.file[f].map_idx
+                                    : entry->pieceTable.map_idx;
+    if (flags & TBFlag::Mapped)
-    if (!(entry->flags & pa_flags[wdl + 2]) || (wdl & 1))
+        value = map[idx[WDLMap[wdl + 2]] + value];
+    // DTZ tables store distance to zero in number of moves or plies. We
+    // want to return plies, so we have convert to plies when needed.
+    if (   (wdl == WDLWin  && !(flags & TBFlag::WinPlies))
+        || (wdl == WDLLoss && !(flags & TBFlag::LossPlies))
+        ||  wdl == WDLCursedWin
+        ||  wdl == WDLBlessedLoss)
-      res *= 2;
+        value *= 2;
+    return value + 1;
+}
+// Compute a unique index out of a position and use it to probe the TB file. To
+// encode k pieces of same type and color, first sort the pieces by square in
+// ascending order s1 <= s2 <= ... <= sk then compute the unique index as:
+//
+//      idx = Binomial[1][s1] + Binomial[2][s2] + ... + Binomial[k][sk]
+//
+template<typename Entry, typename T = typename Ret<Entry>::type>
+T do_probe_table(const Position& pos, Entry* entry, WDLScore wdl, ProbeState* result) {
+    const bool IsWDL = std::is_same<Entry, WDLEntry>::value;
+    Square squares[TBPIECES];
+    Piece pieces[TBPIECES];
+    uint64_t idx;
+    int next = 0, size = 0, leadPawnsCnt = 0;
-  } else {
+    PairsData* d;
+    Bitboard b, leadPawns = 0;
+    File tbFile = FILE_A;
+    // A given TB entry like KRK has associated two material keys: KRvk and Kvkr.
+    // If both sides have the same pieces keys are equal. In this case TB tables
+    // only store the 'white to move' case, so if the position to lookup has black
+    // to move, we need to switch the color and flip the squares before to lookup.
+    bool symmetricBlackToMove = (entry->key == entry->key2 && pos.side_to_move());
+    // TB files are calculated for white as stronger side. For instance we have
+    // KRvK, not KvKR. A position where stronger side is white will have its
+    // material key == entry->key, otherwise we have to switch the color and
+    // flip the squares before to lookup.
-    struct DTZEntry_pawn *entry = (struct DTZEntry_pawn *)ptr;
+    bool blackStronger = (pos.material_key() != entry->key);
+    int flipColor   = (symmetricBlackToMove || blackStronger) * 8;
+    int flipSquares = (symmetricBlackToMove || blackStronger) * 070;
+    int stm         = (symmetricBlackToMove || blackStronger) ^ pos.side_to_move();
+    // For pawns, TB files store 4 separate tables according if leading pawn is on
+    // file a, b, c or d after reordering. The leading pawn is the one with maximum
+    // MapPawns[] value, that is the one most toward the edges and with lowest rank.
+    if (entry->hasPawns) {
+        // In all the 4 tables, pawns are at the beginning of the piece sequence and
+        // their color is the reference one. So we just pick the first one.
-    int k = entry->file[0].pieces[0] ^ cmirror;
+        Piece pc = Piece(item(entry->pawnTable, 0, 0).precomp->pieces[0] ^ flipColor);
+        assert(type_of(pc) == PAWN);
-    Bitboard bb = pos.pieces((Color)(k >> 3), (PieceType)(k & 0x07));
+        leadPawns = b = pos.pieces(color_of(pc), PAWN);
-    i = 0;
+        do
+            squares[size++] = pop_lsb(&b) ^ flipSquares;
+        while (b);
+        leadPawnsCnt = size;
+        std::swap(squares[0], *std::max_element(squares, squares + leadPawnsCnt, pawns_comp));
+        tbFile = file_of(squares[0]);
+        if (tbFile > FILE_D)
+            tbFile = file_of(squares[0] ^ 7); // Horizontal flip: SQ_H1 -> SQ_A1
+        d = item(entry->pawnTable , stm, tbFile).precomp;
+    } else
+        d = item(entry->pieceTable, stm, tbFile).precomp;
+    // DTZ tables are one-sided, i.e. they store positions only for white to
+    // move or only for black to move, so check for side to move to be stm,
+    // early exit otherwise.
+    if (!IsWDL && !check_dtz_stm(entry, stm, tbFile))
+        return *result = CHANGE_STM, T();
+    // Now we are ready to get all the position pieces (but the lead pawns) and
+    // directly map them to the correct color and square.
+    b = pos.pieces() ^ leadPawns;
     do {
-      p[i++] = pop_lsb(&bb) ^ mirror;
+        Square s = pop_lsb(&b);
+        squares[size] = s ^ flipSquares;
+        pieces[size++] = Piece(pos.piece_on(s) ^ flipColor);
-    } while (bb);
+    } while (b);
+    assert(size >= 2);
+    // Then we reorder the pieces to have the same sequence as the one stored
+    // in precomp->pieces[i]: the sequence that ensures the best compression.
-    int f = pawn_file((struct TBEntry_pawn *)entry, p);
+    for (int i = leadPawnsCnt; i < size; ++i)
+        for (int j = i; j < size; ++j)
-    if ((entry->flags[f] & 1) != bside) {
+            if (d->pieces[i] == pieces[j])
+            {
+                std::swap(pieces[i], pieces[j]);
+                std::swap(squares[i], squares[j]);
-      *success = -1;
+                break;
-      return 0;
+            }
+    // Now we map again the squares so that the square of the lead piece is in
+    // the triangle A1-D1-D4.
+    if (file_of(squares[0]) > FILE_D)
+        for (int i = 0; i < size; ++i)
+            squares[i] ^= 7; // Horizontal flip: SQ_H1 -> SQ_A1
+    // Encode leading pawns starting with the one with minimum MapPawns[] and
+    // proceeding in ascending order.
+    if (entry->hasPawns) {
+        idx = LeadPawnIdx[leadPawnsCnt][squares[0]];
+        std::sort(squares + 1, squares + leadPawnsCnt, pawns_comp);
+        for (int i = 1; i < leadPawnsCnt; ++i)
+            idx += Binomial[i][MapPawns[squares[i]]];
+        goto encode_remaining; // With pawns we have finished special treatments
+    }
+    // In positions withouth pawns, we further flip the squares to ensure leading
-    ubyte *pc = entry->file[f].pieces;
+    // piece is below RANK_5.
+    if (rank_of(squares[0]) > RANK_4)
-    for (; i < entry->num;) {
+        for (int i = 0; i < size; ++i)
-      bb = pos.pieces((Color)((pc[i] ^ cmirror) >> 3),
+            squares[i] ^= 070; // Vertical flip: SQ_A8 -> SQ_A1
+    // Look for the first piece of the leading group not on the A1-D4 diagonal
+    // and ensure it is mapped below the diagonal.
-                            (PieceType)(pc[i] & 0x07));
+    for (int i = 0; i < d->groupLen[0]; ++i) {
+        if (!off_A1H8(squares[i]))
-      do {
+            continue;
+        if (off_A1H8(squares[i]) > 0) // A1-H8 diagonal flip: SQ_A3 -> SQ_C3
-        p[i++] = pop_lsb(&bb) ^ mirror;
+            for (int j = i; j < size; ++j)
+                squares[j] = Square(((squares[j] >> 3) | (squares[j] << 3)) & 63);
-      } while (bb);
+        break;
+    }
-    idx = encode_pawn((struct TBEntry_pawn *)entry, entry->file[f].norm, p, entry->file[f].factor);
-    res = decompress_pairs(entry->file[f].precomp, idx);
+    // Encode the leading group.
+    //
+    // Suppose we have KRvK. Let's say the pieces are on square numbers wK, wR
+    // and bK (each 0...63). The simplest way to map this position to an index
+    // is like this:
+    //
+    //   index = wK * 64 * 64 + wR * 64 + bK;
+    //
+    // But this way the TB is going to have 64*64*64 = 262144 positions, with
+    // lots of positions being equivalent (because they are mirrors of each
+    // other) and lots of positions being invalid (two pieces on one square,
-    if (entry->flags[f] & 2)
+    // adjacent kings, etc.).
+    // Usually the first step is to take the wK and bK together. There are just
+    // 462 ways legal and not-mirrored ways to place the wK and bK on the board.
+    // Once we have placed the wK and bK, there are 62 squares left for the wR
+    // Mapping its square from 0..63 to available squares 0..61 can be done like:
+    //
+    //   wR -= (wR > wK) + (wR > bK);
+    //
-      res = entry->map[entry->map_idx[f][wdl_to_map[wdl + 2]] + res];
+    // In words: if wR "comes later" than wK, we deduct 1, and the same if wR
+    // "comes later" than bK. In case of two same pieces like KRRvK we want to
+    // place the two Rs "together". If we have 62 squares left, we can place two
+    // Rs "together" in 62 * 61 / 2 ways (we divide by 2 because rooks can be
+    // swapped and still get the same position.)
+    //
+    // In case we have at least 3 unique pieces (inlcuded kings) we encode them
+    // together.
+    if (entry->hasUniquePieces) {
-    if (!(entry->flags[f] & pa_flags[wdl + 2]) || (wdl & 1))
+        int adjust1 =  squares[1] > squares[0];
-      res *= 2;
+        int adjust2 = (squares[2] > squares[0]) + (squares[2] > squares[1]);
-  }
+        // First piece is below a1-h8 diagonal. MapA1D1D4[] maps the b1-d1-d3
+        // triangle to 0...5. There are 63 squares for second piece and and 62
+        // (mapped to 0...61) for the third.
+        if (off_A1H8(squares[0]))
+            idx = (   MapA1D1D4[squares[0]]  * 63
+                   + (squares[1] - adjust1)) * 62
+                   +  squares[2] - adjust2;
+        // First piece is on a1-h8 diagonal, second below: map this occurence to
+        // 6 to differentiate from the above case, rank_of() maps a1-d4 diagonal
+        // to 0...3 and finally MapB1H1H7[] maps the b1-h1-h7 triangle to 0..27.
+        else if (off_A1H8(squares[1]))
+            idx = (  6 * 63 + rank_of(squares[0]) * 28
+                   + MapB1H1H7[squares[1]])       * 62
+                   + squares[2] - adjust2;
+        // First two pieces are on a1-h8 diagonal, third below
+        else if (off_A1H8(squares[2]))
+            idx =  6 * 63 * 62 + 4 * 28 * 62
+                 +  rank_of(squares[0])        * 7 * 28
+                 + (rank_of(squares[1]) - adjust1) * 28
+                 +  MapB1H1H7[squares[2]];
+        // All 3 pieces on the diagonal a1-h8
-  return res;
+        else
+            idx = 6 * 63 * 62 + 4 * 28 * 62 + 4 * 7 * 28
+                 +  rank_of(squares[0])         * 7 * 6
+                 + (rank_of(squares[1]) - adjust1)  * 6
+                 + (rank_of(squares[2]) - adjust2);
+    } else
+        // We don't have at least 3 unique pieces, like in KRRvKBB, just map
+        // the kings.
+        idx = MapKK[MapA1D1D4[squares[0]]][squares[1]];
+encode_remaining:
+    idx *= d->groupIdx[0];
+    Square* groupSq = squares + d->groupLen[0];
+    // Encode remainig pawns then pieces according to square, in ascending order
+    bool remainingPawns = entry->hasPawns && entry->pawnTable.pawnCount[1];
+    while (d->groupLen[++next])
+    {
+        std::sort(groupSq, groupSq + d->groupLen[next]);
+        uint64_t n = 0;
+        // Map down a square if "comes later" than a square in the previous
+        // groups (similar to what done earlier for leading group pieces).
+        for (int i = 0; i < d->groupLen[next]; ++i)
+        {
+            auto f = [&](Square s) { return groupSq[i] > s; };
+            auto adjust = std::count_if(squares, groupSq, f);
+            n += Binomial[i + 1][groupSq[i] - adjust - 8 * remainingPawns];
+        }
+        remainingPawns = false;
+        idx += n * d->groupIdx[next];
+        groupSq += d->groupLen[next];
+    }
+    // Now that we have the index, decompress the pair and get the score
+    return map_score(entry, tbFile, decompress_pairs(d, idx), wdl);
+}
-// Add underpromotion captures to list of captures.
+// Group together pieces that will be encoded together. The general rule is that
+// a group contains pieces of same type and color. The exception is the leading
-static ExtMove *add_underprom_caps(Position& pos, ExtMove *stack, ExtMove *end)
+// group that, in case of positions withouth pawns, can be formed by 3 different
+// pieces (default) or by the king pair when there is not a unique piece apart
+// from the kings. When there are pawns, pawns are always first in pieces[].
+//
+// As example KRKN -> KRK + N, KNNK -> KK + NN, KPPKP -> P + PP + K + K
-{
+//
+// The actual grouping depends on the TB generator and can be inferred from the
-  ExtMove *moves, *extra = end;
+// sequence of pieces in piece[] array.
+template<typename T>
+void set_groups(T& e, PairsData* d, int order[], File f) {
+    int n = 0, firstLen = e.hasPawns ? 0 : e.hasUniquePieces ? 3 : 2;
+    d->groupLen[n] = 1;
+    // Number of pieces per group is stored in groupLen[], for instance in KRKN
+    // the encoder will default on '111', so groupLen[] will be (3, 1).
-  for (moves = stack; moves < end; moves++) {
+    for (int i = 1; i < e.pieceCount; ++i)
+        if (--firstLen > 0 || d->pieces[i] == d->pieces[i - 1])
+            d->groupLen[n]++;
+        else
+            d->groupLen[++n] = 1;
+    d->groupLen[++n] = 0; // Zero-terminated
+    // The sequence in pieces[] defines the groups, but not the order in which
+    // they are encoded. If the pieces in a group g can be combined on the board
+    // in N(g) different ways, then the position encoding will be of the form:
+    //
+    //           g1 * N(g2) * N(g3) + g2 * N(g3) + g3
+    //
+    // This ensures unique encoding for the whole position. The order of the
+    // groups is a per-table parameter and could not follow the canonical leading
+    // pawns/pieces -> remainig pawns -> remaining pieces. In particular the
+    // first group is at order[0] position and the remaining pawns, when present,
+    // are at order[1] position.
+    bool pp = e.hasPawns && e.pawnTable.pawnCount[1]; // Pawns on both sides
-    Move move = moves->move;
+    int next = pp ? 2 : 1;
+    int freeSquares = 64 - d->groupLen[0] - (pp ? d->groupLen[1] : 0);
+    uint64_t idx = 1;
+    for (int k = 0; next < n || k == order[0] || k == order[1]; ++k)
+        if (k == order[0]) // Leading pawns or pieces
+        {
+            d->groupIdx[0] = idx;
+            idx *=         e.hasPawns ? LeadPawnsSize[d->groupLen[0]][f]
+                  : e.hasUniquePieces ? 31332 : 462;
+        }
+        else if (k == order[1]) // Remaining pawns
+        {
+            d->groupIdx[1] = idx;
+            idx *= Binomial[d->groupLen[1]][48 - d->groupLen[0]];
+        }
+        else // Remainig pieces
+        {
+            d->groupIdx[next] = idx;
+            idx *= Binomial[d->groupLen[next]][freeSquares];
+            freeSquares -= d->groupLen[next++];
+        }
+    d->groupIdx[n] = idx;
+}
+// In Recursive Pairing each symbol represents a pair of childern symbols. So
+// read d->btree[] symbols data and expand each one in his left and right child
+// symbol until reaching the leafs that represent the symbol value.
+uint8_t set_symlen(PairsData* d, Sym s, std::vector<bool>& visited) {
-    if (type_of(move) == PROMOTION && !pos.empty(to_sq(move))) {
+    visited[s] = true; // We can set it now because tree is acyclic
+    Sym sr = d->btree[s].get<LR::Right>();
+    if (sr == 0xFFF)
+        return 0;
+    Sym sl = d->btree[s].get<LR::Left>();
+    if (!visited[sl])
-      (*extra++).move = (Move)(move - (1 << 12));
+        d->symlen[sl] = set_symlen(d, sl, visited);
+    if (!visited[sr])
-      (*extra++).move = (Move)(move - (2 << 12));
+        d->symlen[sr] = set_symlen(d, sr, visited);
+    return d->symlen[sl] + d->symlen[sr] + 1;
+}
+uint8_t* set_sizes(PairsData* d, uint8_t* data) {
+    d->flags = *data++;
+    if (d->flags & TBFlag::SingleValue) {
-      (*extra++).move = (Move)(move - (3 << 12));
+        d->blocksNum = d->blockLengthSize = 0;
+        d->span = d->sparseIndexSize = 0; // Broken MSVC zero-init
+        d->minSymLen = *data++; // Here we store the single value
+        return data;
+    }
-  }
+    // groupLen[] is a zero-terminated list of group lengths, the last groupIdx[]
+    // element stores the biggest index that is the tb size.
+    uint64_t tbSize = d->groupIdx[std::find(d->groupLen, d->groupLen + 7, 0) - d->groupLen];
+    d->sizeofBlock = 1ULL << *data++;
+    d->span = 1ULL << *data++;
+    d->sparseIndexSize = (size_t) ((tbSize + d->span - 1) / d->span); // Round up // Pierre-Marie Baty -- added type cast
+    int padding = number<uint8_t, LittleEndian>(data++);
+    d->blocksNum = number<uint32_t, LittleEndian>(data); data += sizeof(uint32_t);
+    d->blockLengthSize = d->blocksNum + padding; // Padded to ensure SparseIndex[]
+                                                 // does not point out of range.
+    d->maxSymLen = *data++;
+    d->minSymLen = *data++;
+    d->lowestSym = (Sym*)data;
+    d->base64.resize(d->maxSymLen - d->minSymLen + 1);
+    // The canonical code is ordered such that longer symbols (in terms of
+    // the number of bits of their Huffman code) have lower numeric value,
+    // so that d->lowestSym[i] >= d->lowestSym[i+1] (when read as LittleEndian).
+    // Starting from this we compute a base64[] table indexed by symbol length
+    // and containing 64 bit values so that d->base64[i] >= d->base64[i+1].
+    // See http://www.eecs.harvard.edu/~michaelm/E210/huffman.pdf
+    for (int i = d->base64.size() - 2; i >= 0; --i) {
+        d->base64[i] = (d->base64[i + 1] + number<Sym, LittleEndian>(&d->lowestSym[i])
+                                         - number<Sym, LittleEndian>(&d->lowestSym[i + 1])) / 2;
+        assert(d->base64[i] * 2 >= d->base64[i+1]);
+    }
+    // Now left-shift by an amount so that d->base64[i] gets shifted 1 bit more
+    // than d->base64[i+1] and given the above assert condition, we ensure that
+    // d->base64[i] >= d->base64[i+1]. Moreover for any symbol s64 of length i
+    // and right-padded to 64 bits holds d->base64[i-1] >= s64 >= d->base64[i].
+    for (size_t i = 0; i < d->base64.size(); ++i)
+        d->base64[i] <<= 64 - i - d->minSymLen; // Right-padding to 64 bits
+    data += d->base64.size() * sizeof(Sym);
+    d->symlen.resize(number<uint16_t, LittleEndian>(data)); data += sizeof(uint16_t);
-  return extra;
+    d->btree = (LR*)data;
+    // The comrpession scheme used is "Recursive Pairing", that replaces the most
+    // frequent adjacent pair of symbols in the source message by a new symbol,
+    // reevaluating the frequencies of all of the symbol pairs with respect to
+    // the extended alphabet, and then repeating the process.
+    // See http://www.larsson.dogma.net/dcc99.pdf
+    std::vector<bool> visited(d->symlen.size());
+    for (Sym sym = 0; sym < d->symlen.size(); ++sym)
+        if (!visited[sym])
+            d->symlen[sym] = set_symlen(d, sym, visited);
+    return data + d->symlen.size() * sizeof(LR) + (d->symlen.size() & 1);
+}
+template<typename T>
-static int probe_ab(Position& pos, int alpha, int beta, int *success)
+uint8_t* set_dtz_map(WDLEntry&, T&, uint8_t*, File) { return nullptr; }
-{
-  int v;
-  ExtMove stack[64];
+template<typename T>
-  ExtMove *moves, *end;
+uint8_t* set_dtz_map(DTZEntry&, T& p, uint8_t* data, File maxFile) {
-  StateInfo st;
-  // Generate (at least) all legal non-ep captures including (under)promotions.
-  // It is OK to generate more, as long as they are filtered out below.
-  if (!pos.checkers()) {
+    p.map = data;
-    end = generate<CAPTURES>(pos, stack);
-    // Since underpromotion captures are not included, we need to add them.
-    end = add_underprom_caps(pos, stack, end);
-  } else
-    end = generate<EVASIONS>(pos, stack);
-  for (moves = stack; moves < end; moves++) {
+    for (File f = FILE_A; f <= maxFile; ++f) {
-    Move capture = moves->move;
-    if (!pos.capture(capture) || type_of(capture) == ENPASSANT
+        if (item(p, 0, f).precomp->flags & TBFlag::Mapped)
-                        || !pos.legal(capture))
+            for (int i = 0; i < 4; ++i) { // Sequence like 3,x,x,x,1,x,0,2,x,x
-      continue;
-    pos.do_move(capture, st, pos.gives_check(capture));
+                item(p, 0, f).map_idx[i] = (uint16_t)(data - p.map + 1);
-    v = -probe_ab(pos, -beta, -alpha, success);
-    pos.undo_move(capture);
-    if (*success == 0) return 0;
+                data += *data + 1;
-    if (v > alpha) {
-      if (v >= beta) {
-        *success = 2;
-        return v;
+            }
-      }
-      alpha = v;
+    }
-  }
-  v = probe_wdl_table(pos, success);
+    return data += (uintptr_t)data & 1; // Word alignment
-  if (*success == 0) return 0;
-  if (alpha >= v) {
-    *success = 1 + (alpha > 0);
-    return alpha;
-  } else {
-    *success = 1;
-    return v;
-  }
+}
-// Probe the WDL table for a particular position.
-// If *success != 0, the probe was successful.
-// The return value is from the point of view of the side to move:
-// -2 : loss
-// -1 : loss, but draw under 50-move rule
+template<typename Entry, typename T>
-//  0 : draw
-//  1 : win, but draw under 50-move rule
+void do_init(Entry& e, T& p, uint8_t* data) {
-//  2 : win
-int Tablebases::probe_wdl(Position& pos, int *success)
-{
-  int v;
-  *success = 1;
-  v = probe_ab(pos, -2, 2, success);
+    const bool IsWDL = std::is_same<Entry, WDLEntry>::value;
-  // If en passant is not possible, we are done.
-  if (pos.ep_square() == SQ_NONE)
-    return v;
+    PairsData* d;
-  if (!(*success)) return 0;
-  // Now handle en passant.
+    enum { Split = 1, HasPawns = 2 };
-  int v1 = -3;
-  // Generate (at least) all legal en passant captures.
-  ExtMove stack[192];
-  ExtMove *moves, *end;
-  StateInfo st;
-  if (!pos.checkers())
-    end = generate<CAPTURES>(pos, stack);
+    assert(e.hasPawns        == !!(*data & HasPawns));
-  else
-    end = generate<EVASIONS>(pos, stack);
+    assert((e.key != e.key2) == !!(*data & Split));
-  for (moves = stack; moves < end; moves++) {
-    Move capture = moves->move;
+    data++; // First byte stores flags
-    if (type_of(capture) != ENPASSANT
+    const int Sides = IsWDL && (e.key != e.key2) ? 2 : 1;
-          || !pos.legal(capture))
-      continue;
-    pos.do_move(capture, st, pos.gives_check(capture));
+    const File MaxFile = e.hasPawns ? FILE_D : FILE_A;
-    int v0 = -probe_ab(pos, -2, 2, success);
+    bool pp = e.hasPawns && e.pawnTable.pawnCount[1]; // Pawns on both sides
-    pos.undo_move(capture);
+    assert(!pp || e.pawnTable.pawnCount[0]);
-    if (*success == 0) return 0;
+    for (File f = FILE_A; f <= MaxFile; ++f) {
-    if (v0 > v1) v1 = v0;
-  }
-  if (v1 > -3) {
-    if (v1 >= v) v = v1;
+        for (int i = 0; i < Sides; i++)
-    else if (v == 0) {
-      // Check whether there is at least one legal non-ep move.
+            item(p, i, f).precomp = new PairsData();
-      for (moves = stack; moves < end; moves++) {
+        int order[][2] = { { *data & 0xF, pp ? *(data + 1) & 0xF : 0xF },
-        Move capture = moves->move;
-        if (type_of(capture) == ENPASSANT) continue;
+                           { *data >>  4, pp ? *(data + 1) >>  4 : 0xF } };
-        if (pos.legal(capture)) break;
+        data += 1 + pp;
-      }
-      if (moves == end && !pos.checkers()) {
-        end = generate<QUIETS>(pos, end);
+        for (int k = 0; k < e.pieceCount; ++k, ++data)
-        for (; moves < end; moves++) {
+            for (int i = 0; i < Sides; i++)
-          Move move = moves->move;
+                item(p, i, f).precomp->pieces[k] = Piece(i ? *data >>  4 : *data & 0xF);
-          if (pos.legal(move))
+        for (int i = 0; i < Sides; ++i)
-            break;
-        }
-      }
-      // If not, then we are forced to play the losing ep capture.
+            set_groups(e, item(p, i, f).precomp, order[i], f);
-      if (moves == end)
-        v = v1;
+    }
-  }
+    data += (uintptr_t)data & 1; // Word alignment
+    for (File f = FILE_A; f <= MaxFile; ++f)
+        for (int i = 0; i < Sides; i++)
+            data = set_sizes(item(p, i, f).precomp, data);
+    if (!IsWDL)
+        data = set_dtz_map(e, p, data, MaxFile);
+    for (File f = FILE_A; f <= MaxFile; ++f)
+        for (int i = 0; i < Sides; i++) {
+            (d = item(p, i, f).precomp)->sparseIndex = (SparseEntry*)data;
+            data += d->sparseIndexSize * sizeof(SparseEntry);
-  return v;
+        }
+    for (File f = FILE_A; f <= MaxFile; ++f)
+        for (int i = 0; i < Sides; i++) {
+            (d = item(p, i, f).precomp)->blockLength = (uint16_t*)data;
+            data += d->blockLengthSize * sizeof(uint16_t);
+        }
+    for (File f = FILE_A; f <= MaxFile; ++f)
+        for (int i = 0; i < Sides; i++) {
+            data = (uint8_t*)(((uintptr_t)data + 0x3F) & ~0x3F); // 64 byte alignment
+            (d = item(p, i, f).precomp)->data = data;
+            data += d->blocksNum * d->sizeofBlock;
+        }
+}
-// This routine treats a position with en passant captures as one without.
+template<typename Entry>
-static int probe_dtz_no_ep(Position& pos, int *success)
+void* init(Entry& e, const Position& pos) {
-{
-  int wdl, dtz;
-  wdl = probe_ab(pos, -2, 2, success);
+    const bool IsWDL = std::is_same<Entry, WDLEntry>::value;
-  if (*success == 0) return 0;
-  if (wdl == 0) return 0;
+    static Mutex mutex;
+    // Avoid a thread reads 'ready' == true while another is still in do_init(),
+    // this could happen due to compiler reordering.
-  if (*success == 2)
+    if (e.ready.load(std::memory_order_acquire))
-    return wdl == 2 ? 1 : 101;
+        return e.baseAddress;
-  ExtMove stack[192];
-  ExtMove *moves, *end = NULL;
+    std::unique_lock<Mutex> lk(mutex);
-  StateInfo st;
-  if (wdl > 0) {
-    // Generate at least all legal non-capturing pawn moves
+    if (e.ready.load(std::memory_order_relaxed)) // Recheck under lock
-    // including non-capturing promotions.
-    if (!pos.checkers())
+        return e.baseAddress;
-      end = generate<NON_EVASIONS>(pos, stack);
-    else
-      end = generate<EVASIONS>(pos, stack);
-    for (moves = stack; moves < end; moves++) {
+    // Pieces strings in decreasing order for each color, like ("KPP","KR")
-      Move move = moves->move;
+    std::string fname, w, b;
-      if (type_of(pos.moved_piece(move)) != PAWN || pos.capture(move)
+    for (PieceType pt = KING; pt >= PAWN; --pt) {
-                || !pos.legal(move))
-        continue;
-      pos.do_move(move, st, pos.gives_check(move));
+        w += std::string(popcount(pos.pieces(WHITE, pt)), PieceToChar[pt]);
-      int v = -Tablebases::probe_wdl(pos, success);
+        b += std::string(popcount(pos.pieces(BLACK, pt)), PieceToChar[pt]);
-      pos.undo_move(move);
-      if (*success == 0) return 0;
-      if (v == wdl)
-        return v == 2 ? 1 : 101;
+    }
-  }
-  dtz = 1 + probe_dtz_table(pos, wdl, success);
+    const uint8_t TB_MAGIC[][4] = { { 0xD7, 0x66, 0x0C, 0xA5 },
-  if (*success >= 0) {
-    if (wdl & 1) dtz += 100;
-    return wdl >= 0 ? dtz : -dtz;
+                                    { 0x71, 0xE8, 0x23, 0x5D } };
-  }
+    fname =  (e.key == pos.material_key() ? w + 'v' + b : b + 'v' + w)
+           + (IsWDL ? ".rtbw" : ".rtbz");
+    uint8_t* data = TBFile(fname).map(&e.baseAddress, &e.mapping, TB_MAGIC[IsWDL]);
-  if (wdl > 0) {
+    if (data)
+        e.hasPawns ? do_init(e, e.pawnTable, data) : do_init(e, e.pieceTable, data);
+    e.ready.store(true, std::memory_order_release);
+    return e.baseAddress;
+}
+template<typename E, typename T = typename Ret<E>::type>
+T probe_table(const Position& pos, ProbeState* result, WDLScore wdl = WDLDraw) {
+    if (!(pos.pieces() ^ pos.pieces(KING)))
+        return T(WDLDraw); // KvK
+    E* entry = EntryTable.get<E>(pos.material_key());
+    if (!entry || !init(*entry, pos))
+        return *result = FAIL, T();
+    return do_probe_table(pos, entry, wdl, result);
+}
+// For a position where the side to move has a winning capture it is not necessary
+// to store a winning value so the generator treats such positions as "don't cares"
+// and tries to assign to it a value that improves the compression ratio. Similarly,
+// if the side to move has a drawing capture, then the position is at least drawn.
+// If the position is won, then the TB needs to store a win value. But if the
+// position is drawn, the TB may store a loss value if that is better for compression.
+// All of this means that during probing, the engine must look at captures and probe
+// their results and must probe the position itself. The "best" result of these
+// probes is the correct result for the position.
+// DTZ table don't store values when a following move is a zeroing winning move
+// (winning capture or winning pawn move). Also DTZ store wrong values for positions
+// where the best move is an ep-move (even if losing). So in all these cases set
+// the state to ZEROING_BEST_MOVE.
+template<bool CheckZeroingMoves = false>
+WDLScore search(Position& pos, ProbeState* result) {
+    WDLScore value, bestValue = WDLLoss;
-    int best = 0xffff;
+    StateInfo st;
+    auto moveList = MoveList<LEGAL>(pos);
+    size_t totalCount = moveList.size(), moveCount = 0;
-    for (moves = stack; moves < end; moves++) {
+    for (const Move& move : moveList)
+    {
-      Move move = moves->move;
+        if (   !pos.capture(move)
-      if (pos.capture(move) || type_of(pos.moved_piece(move)) == PAWN
+            && (!CheckZeroingMoves || type_of(pos.moved_piece(move)) != PAWN))
-                || !pos.legal(move))
+            continue;
-        continue;
+        moveCount++;
-      pos.do_move(move, st, pos.gives_check(move));
+        pos.do_move(move, st);
-      int v = -Tablebases::probe_dtz(pos, success);
+        value = -search(pos, result);
-      pos.undo_move(move);
+        pos.undo_move(move);
-      if (*success == 0) return 0;
+        if (*result == FAIL)
+            return WDLDraw;
-      if (v > 0 && v + 1 < best)
+        if (value > bestValue)
+        {
-        best = v + 1;
+            bestValue = value;
+            if (value >= WDLWin)
+            {
+                *result = ZEROING_BEST_MOVE; // Winning DTZ-zeroing move
+                return value;
+            }
+        }
+    }
+    // In case we have already searched all the legal moves we don't have to probe
+    // the TB because the stored score could be wrong. For instance TB tables
+    // do not contain information on position with ep rights, so in this case
+    // the result of probe_wdl_table is wrong. Also in case of only capture
+    // moves, for instance here 4K3/4q3/6p1/2k5/6p1/8/8/8 w - - 0 7, we have to
-    return best;
+    // return with ZEROING_BEST_MOVE set.
-  } else {
-    int best = -1;
+    bool noMoreMoves = (moveCount && moveCount == totalCount);
-    if (!pos.checkers())
+    if (noMoreMoves)
-      end = generate<NON_EVASIONS>(pos, stack);
+        value = bestValue;
     else
+    {
-      end = generate<EVASIONS>(pos, stack);
+        value = probe_table<WDLEntry>(pos, result);
+        if (*result == FAIL)
+            return WDLDraw;
+    }
+    // DTZ stores a "don't care" value if bestValue is a win
+    if (bestValue >= value)
+        return *result = (   bestValue > WDLDraw
+                          || noMoreMoves ? ZEROING_BEST_MOVE : OK), bestValue;
+    return *result = OK, value;
+}
+} // namespace
+void Tablebases::init(const std::string& paths) {
+    EntryTable.clear();
+    MaxCardinality = 0;
+    TBFile::Paths = paths;
+    if (paths.empty() || paths == "<empty>")
+        return;
+    // MapB1H1H7[] encodes a square below a1-h8 diagonal to 0..27
+    int code = 0;
-    for (moves = stack; moves < end; moves++) {
+    for (Square s = SQ_A1; s <= SQ_H8; ++s)
+        if (off_A1H8(s) < 0)
+            MapB1H1H7[s] = code++;
+    // MapA1D1D4[] encodes a square in the a1-d1-d4 triangle to 0..9
+    std::vector<Square> diagonal;
-      int v;
+    code = 0;
+    for (Square s = SQ_A1; s <= SQ_D4; ++s)
+        if (off_A1H8(s) < 0 && file_of(s) <= FILE_D)
-      Move move = moves->move;
+            MapA1D1D4[s] = code++;
+        else if (!off_A1H8(s) && file_of(s) <= FILE_D)
+            diagonal.push_back(s);
+    // Diagonal squares are encoded as last ones
-      if (!pos.legal(move))
+    for (auto s : diagonal)
+        MapA1D1D4[s] = code++;
+    // MapKK[] encodes all the 461 possible legal positions of two kings where
+    // the first is in the a1-d1-d4 triangle. If the first king is on the a1-d4
+    // diagonal, the other one shall not to be above the a1-h8 diagonal.
+    std::vector<std::pair<int, Square>> bothOnDiagonal;
+    code = 0;
+    for (int idx = 0; idx < 10; idx++)
+        for (Square s1 = SQ_A1; s1 <= SQ_D4; ++s1)
+            if (MapA1D1D4[s1] == idx && (idx || s1 == SQ_B1)) // SQ_B1 is mapped to 0
-        continue;
+            {
+                for (Square s2 = SQ_A1; s2 <= SQ_H8; ++s2)
+                    if ((PseudoAttacks[KING][s1] | s1) & s2)
+                        continue; // Illegal position
+                    else if (!off_A1H8(s1) && off_A1H8(s2) > 0)
+                        continue; // First on diagonal, second above
+                    else if (!off_A1H8(s1) && !off_A1H8(s2))
+                        bothOnDiagonal.push_back(std::make_pair(idx, s2));
+                    else
-      pos.do_move(move, st, pos.gives_check(move));
+                        MapKK[idx][s2] = code++;
+            }
+    // Legal positions with both kings on diagonal are encoded as last ones
+    for (auto p : bothOnDiagonal)
+        MapKK[p.first][p.second] = code++;
+    // Binomial[] stores the Binomial Coefficents using Pascal rule. There
+    // are Binomial[k][n] ways to choose k elements from a set of n elements.
-      if (st.rule50 == 0) {
+    Binomial[0][0] = 1;
+    for (int n = 1; n < 64; n++) // Squares
+        for (int k = 0; k < 6 && k <= n; ++k) // Pieces
+            Binomial[k][n] =  (k > 0 ? Binomial[k - 1][n - 1] : 0)
+                            + (k < n ? Binomial[k    ][n - 1] : 0);
+    // MapPawns[s] encodes squares a2-h7 to 0..47. This is the number of possible
+    // available squares when the leading one is in 's'. Moreover the pawn with
+    // highest MapPawns[] is the leading pawn, the one nearest the edge and,
+    // among pawns with same file, the one with lowest rank.
+    int availableSquares = 47; // Available squares when lead pawn is in a2
+    // Init the tables for the encoding of leading pawns group: with 6-men TB we
+    // can have up to 4 leading pawns (KPPPPK).
+    for (int leadPawnsCnt = 1; leadPawnsCnt <= 4; ++leadPawnsCnt)
+        for (File f = FILE_A; f <= FILE_D; ++f)
+        {
+            // Restart the index at every file because TB table is splitted
+            // by file, so we can reuse the same index for different files.
-        if (wdl == -2) v = -1;
+            int idx = 0;
+            // Sum all possible combinations for a given file, starting with
+            // the leading pawn on rank 2 and increasing the rank.
+            for (Rank r = RANK_2; r <= RANK_7; ++r)
-        else {
+            {
-          v = probe_ab(pos, 1, 2, success);
+                Square sq = make_square(f, r);
+                // Compute MapPawns[] at first pass.
+                // If sq is the leading pawn square, any other pawn cannot be
+                // below or more toward the edge of sq. There are 47 available
+                // squares when sq = a2 and reduced by 2 for any rank increase
+                // due to mirroring: sq == a3 -> no a2, h2, so MapPawns[a3] = 45
-          v = (v == 2) ? 0 : -101;
+                if (leadPawnsCnt == 1)
+                {
+                    MapPawns[sq] = availableSquares--;
+                    MapPawns[sq ^ 7] = availableSquares--; // Horizontal flip
+                }
+                LeadPawnIdx[leadPawnsCnt][sq] = idx;
+                idx += Binomial[leadPawnsCnt - 1][MapPawns[sq]];
+            }
+            // After a file is traversed, store the cumulated per-file index
+            LeadPawnsSize[leadPawnsCnt][f] = idx;
+        }
+    for (PieceType p1 = PAWN; p1 < KING; ++p1) {
+        EntryTable.insert({KING, p1, KING});
+        for (PieceType p2 = PAWN; p2 <= p1; ++p2) {
+            EntryTable.insert({KING, p1, p2, KING});
+            EntryTable.insert({KING, p1, KING, p2});
+            for (PieceType p3 = PAWN; p3 < KING; ++p3)
+                EntryTable.insert({KING, p1, p2, KING, p3});
+            for (PieceType p3 = PAWN; p3 <= p2; ++p3) {
+                EntryTable.insert({KING, p1, p2, p3, KING});
+                for (PieceType p4 = PAWN; p4 <= p3; ++p4)
+                    EntryTable.insert({KING, p1, p2, p3, p4, KING});
+                for (PieceType p4 = PAWN; p4 < KING; ++p4)
+                    EntryTable.insert({KING, p1, p2, p3, KING, p4});
+            }
+            for (PieceType p3 = PAWN; p3 <= p1; ++p3)
+                for (PieceType p4 = PAWN; p4 <= (p1 == p3 ? p2 : p3); ++p4)
+                    EntryTable.insert({KING, p1, p2, KING, p3, p4});
+        }
-      } else {
-        v = -Tablebases::probe_dtz(pos, success) - 1;
-      }
-      pos.undo_move(move);
-      if (*success == 0) return 0;
-      if (v < best)
-        best = v;
+    }
-    return best;
-  }
+    sync_cout << "info string Found " << EntryTable.size() << " tablebases" << sync_endl;
+}
+// Probe the WDL table for a particular position.
+// If *result != FAIL, the probe was successful.
+// The return value is from the point of view of the side to move:
+// -2 : loss
+// -1 : loss, but draw under 50-move rule
+//  0 : draw
-static int wdl_to_dtz[] = {
+//  1 : win, but draw under 50-move rule
+//  2 : win
+WDLScore Tablebases::probe_wdl(Position& pos, ProbeState* result) {
-  -1, -101, 0, 101, 1
+    *result = OK;
+    return search(pos, result);
-};
+}
 // Probe the DTZ table for a particular position.
-// If *success != 0, the probe was successful.
+// If *result != FAIL, the probe was successful.
 // The return value is from the point of view of the side to move:
 //         n < -100 : loss, but draw under 50-move rule
 // -100 <= n < -1   : loss in n ply (assuming 50-move counter == 0)
 //         0        : draw
 //     1 < n <= 100 : win in n ply (assuming 50-move counter == 0)
 // capture or pawn move, the inequality to be preserved is
 // dtz + 50-movecounter <= 100.
 //
 // In short, if a move is available resulting in dtz + 50-move-counter <= 99,
 // then do not accept moves leading to dtz + 50-move-counter == 100.
-//
-int Tablebases::probe_dtz(Position& pos, int *success)
+int Tablebases::probe_dtz(Position& pos, ProbeState* result) {
-{
-  *success = 1;
-  int v = probe_dtz_no_ep(pos, success);
-  if (pos.ep_square() == SQ_NONE)
-    return v;
+    *result = OK;
-  if (*success == 0) return 0;
+    WDLScore wdl = search<true>(pos, result);
-  // Now handle en passant.
+    if (*result == FAIL || wdl == WDLDraw) // DTZ tables don't store draws
-  int v1 = -3;
+        return 0;
+    // DTZ stores a 'don't care' value in this case, or even a plain wrong
-  ExtMove stack[192];
+    // one as in case the best move is a losing ep, so it cannot be probed.
-  ExtMove *moves, *end;
+    if (*result == ZEROING_BEST_MOVE)
-  StateInfo st;
+        return dtz_before_zeroing(wdl);
-  if (!pos.checkers())
-    end = generate<CAPTURES>(pos, stack);
-  else
-    end = generate<EVASIONS>(pos, stack);
+    int dtz = probe_table<DTZEntry>(pos, result, wdl);
-  for (moves = stack; moves < end; moves++) {
+    if (*result == FAIL)
-    Move capture = moves->move;
+        return 0;
-    if (type_of(capture) != ENPASSANT
+    if (*result != CHANGE_STM)
-                || !pos.legal(capture))
+        return (dtz + 100 * (wdl == WDLBlessedLoss || wdl == WDLCursedWin)) * sign_of(wdl);
-      continue;
-    pos.do_move(capture, st, pos.gives_check(capture));
+    // DTZ stores results for the other side, so we need to do a 1-ply search and
-    int v0 = -probe_ab(pos, -2, 2, success);
+    // find the winning move that minimizes DTZ.
-    pos.undo_move(capture);
+    StateInfo st;
-    if (*success == 0) return 0;
-    if (v0 > v1) v1 = v0;
+    int minDTZ = 0xFFFF;
-  }
-  if (v1 > -3) {
-    v1 = wdl_to_dtz[v1 + 2];
+    for (const Move& move : MoveList<LEGAL>(pos))
-    if (v < -100) {
-      if (v1 >= 0)
-        v = v1;
+    {
-    } else if (v < 0) {
-      if (v1 >= 0 || v1 < -100)
+        bool zeroing = pos.capture(move) || type_of(pos.moved_piece(move)) == PAWN;
-        v = v1;
+        pos.do_move(move, st);
-    } else if (v > 100) {
+        // For zeroing moves we want the dtz of the move _before_ doing it,
-      if (v1 > 0)
-        v = v1;
-    } else if (v > 0) {
+        // otherwise we will get the dtz of the next move sequence. Search the
-      if (v1 == 1)
-        v = v1;
-    } else if (v1 >= 0) {
+        // position after the move to get the score sign (because even in a
-      v = v1;
-    } else {
-      for (moves = stack; moves < end; moves++) {
+        // winning position we could make a losing capture or going for a draw).
-        Move move = moves->move;
+        dtz = zeroing ? -dtz_before_zeroing(search(pos, result))
-        if (type_of(move) == ENPASSANT) continue;
+                      : -probe_dtz(pos, result);
-        if (pos.legal(move)) break;
+        pos.undo_move(move);
-      }
-      if (moves == end && !pos.checkers()) {
+        if (*result == FAIL)
-        end = generate<QUIETS>(pos, end);
+            return 0;
-        for (; moves < end; moves++) {
+        // Convert result from 1-ply search. Zeroing moves are already accounted
-          Move move = moves->move;
+        // by dtz_before_zeroing() that returns the DTZ of the previous move.
-          if (pos.legal(move))
+        if (!zeroing)
-            break;
+            dtz += sign_of(dtz);
-        }
-      }
+        // Skip the draws and if we are winning only pick positive dtz
-      if (moves == end)
+        if (dtz < minDTZ && sign_of(dtz) == sign_of(wdl))
-        v = v1;
+            minDTZ = dtz;
+    }
-  }
+    // Special handle a mate position, when there are no legal moves, in this
+    // case return value is somewhat arbitrary, so stick to the original TB code
-  return v;
+    // that returns -1 in this case.
+    return minDTZ == 0xFFFF ? -1 : minDTZ;
+}
 // Check whether there has been at least one repetition of positions
 // since the last capture or pawn move.
 static int has_repeated(StateInfo *st)
+{
-  while (1) {
+    while (1) {
-    int i = 4, e = std::min(st->rule50, st->pliesFromNull);
+        int i = 4, e = std::min(st->rule50, st->pliesFromNull);
-    if (e < i)
-      return 0;
-    StateInfo *stp = st->previous->previous;
-    do {
-      stp = stp->previous->previous;
-      if (stp->key == st->key)
-        return 1;
-      i += 2;
-    } while (i <= e);
-    st = st->previous;
-  }
-}
-static Value wdl_to_Value[5] = {
+        if (e < i)
-  -VALUE_MATE + MAX_PLY + 1,
+            return 0;
+        StateInfo *stp = st->previous->previous;
-  VALUE_DRAW - 2,
+        do {
+            stp = stp->previous->previous;
+            if (stp->key == st->key)
-  VALUE_DRAW,
+                return 1;
-  VALUE_DRAW + 2,
+            i += 2;
+        } while (i <= e);
-  VALUE_MATE - MAX_PLY - 1
+        st = st->previous;
+    }
-};
+}
 // Use the DTZ tables to filter out moves that don't preserve the win or draw.
 // If the position is lost, but DTZ is fairly high, only keep moves that
 // maximise DTZ.
 //
 // A return value false indicates that not all probes were successful and that
 // no moves were filtered out.
 bool Tablebases::root_probe(Position& pos, Search::RootMoves& rootMoves, Value& score)
+{
-  int success;
+    assert(rootMoves.size());
-  int dtz = probe_dtz(pos, &success);
+    ProbeState result;
-  if (!success) return false;
+    int dtz = probe_dtz(pos, &result);
+    if (result == FAIL)
-  StateInfo st;
+        return false;
+    StateInfo st;
-  // Probe each move.
+    // Probe each move
-  for (size_t i = 0; i < rootMoves.size(); i++) {
+    for (size_t i = 0; i < rootMoves.size(); ++i) {
-    Move move = rootMoves[i].pv[0];
+        Move move = rootMoves[i].pv[0];
-    pos.do_move(move, st, pos.gives_check(move));
+        pos.do_move(move, st);
-    int v = 0;
+        int v = 0;
-    if (pos.checkers() && dtz > 0) {
+        if (pos.checkers() && dtz > 0) {
-      ExtMove s[192];
+            ExtMove s[MAX_MOVES];
-      if (generate<LEGAL>(pos, s) == s)
+            if (generate<LEGAL>(pos, s) == s)
+                v = 1;
+        }
+        if (!v) {
+            if (st.rule50 != 0) {
+                v = -probe_dtz(pos, &result);
+                if (v > 0)
+                    ++v;
+                else if (v < 0)
+                    --v;
+            } else {
+                v = -probe_wdl(pos, &result);
+                v = dtz_before_zeroing(WDLScore(v));
-        v = 1;
+            }
+        }
+        pos.undo_move(move);
+        if (result == FAIL)
+            return false;
+        rootMoves[i].score = (Value)v;
+    }
-    if (!v) {
-      if (st.rule50 != 0) {
-        v = -Tablebases::probe_dtz(pos, &success);
-        if (v > 0) v++;
-        else if (v < 0) v--;
-      } else {
-        v = -Tablebases::probe_wdl(pos, &success);
-        v = wdl_to_dtz[v + 2];
-      }
-    }
-    pos.undo_move(move);
-    if (!success) return false;
-    rootMoves[i].score = (Value)v;
-  }
-  // Obtain 50-move counter for the root position.
+    // Obtain 50-move counter for the root position.
-  // In Stockfish there seems to be no clean way, so we do it like this:
+    // In Stockfish there seems to be no clean way, so we do it like this:
-  int cnt50 = st.previous->rule50;
+    int cnt50 = st.previous ? st.previous->rule50 : 0;
-  // Use 50-move counter to determine whether the root position is
+    // Use 50-move counter to determine whether the root position is
-  // won, lost or drawn.
+    // won, lost or drawn.
-  int wdl = 0;
+    WDLScore wdl = WDLDraw;
-  if (dtz > 0)
-    wdl = (dtz + cnt50 <= 100) ? 2 : 1;
-  else if (dtz < 0)
-    wdl = (-dtz + cnt50 <= 100) ? -2 : -1;
-  // Determine the score to report to the user.
-  score = wdl_to_Value[wdl + 2];
-  // If the position is winning or losing, but too few moves left, adjust the
-  // score to show how close it is to winning or losing.
-  // NOTE: int(PawnValueEg) is used as scaling factor in score_to_uci().
-  if (wdl == 1 && dtz <= 100)
+    if (dtz > 0)
-    score = (Value)(((200 - dtz - cnt50) * int(PawnValueEg)) / 200);
+        wdl = (dtz + cnt50 <= 100) ? WDLWin : WDLCursedWin;
-  else if (wdl == -1 && dtz >= -100)
+    else if (dtz < 0)
-    score = -(Value)(((200 + dtz - cnt50) * int(PawnValueEg)) / 200);
+        wdl = (-dtz + cnt50 <= 100) ? WDLLoss : WDLBlessedLoss;
+    // Determine the score to report to the user.
+    score = WDL_to_value[wdl + 2];
+    // If the position is winning or losing, but too few moves left, adjust the
+    // score to show how close it is to winning or losing.
+    // NOTE: int(PawnValueEg) is used as scaling factor in score_to_uci().
+    if (wdl == WDLCursedWin && dtz <= 100)
+        score = (Value)(((200 - dtz - cnt50) * int(PawnValueEg)) / 200);
+    else if (wdl == WDLBlessedLoss && dtz >= -100)
+        score = -(Value)(((200 + dtz - cnt50) * int(PawnValueEg)) / 200);
-  // Now be a bit smart about filtering out moves.
+    // Now be a bit smart about filtering out moves.
-  size_t j = 0;
+    size_t j = 0;
-  if (dtz > 0) { // winning (or 50-move rule draw)
+    if (dtz > 0) { // winning (or 50-move rule draw)
-    int best = 0xffff;
+        int best = 0xffff;
-    for (size_t i = 0; i < rootMoves.size(); i++) {
+        for (size_t i = 0; i < rootMoves.size(); ++i) {
-      int v = rootMoves[i].score;
+            int v = rootMoves[i].score;
-      if (v > 0 && v < best)
+            if (v > 0 && v < best)
-        best = v;
+                best = v;
-    }
+        }
-    int max = best;
+        int max = best;
-    // If the current phase has not seen repetitions, then try all moves
+        // If the current phase has not seen repetitions, then try all moves
-    // that stay safely within the 50-move budget, if there are any.
+        // that stay safely within the 50-move budget, if there are any.
-    if (!has_repeated(st.previous) && best + cnt50 <= 99)
+        if (!has_repeated(st.previous) && best + cnt50 <= 99)
-      max = 99 - cnt50;
+            max = 99 - cnt50;
-    for (size_t i = 0; i < rootMoves.size(); i++) {
+        for (size_t i = 0; i < rootMoves.size(); ++i) {
-      int v = rootMoves[i].score;
+            int v = rootMoves[i].score;
-      if (v > 0 && v <= max)
+            if (v > 0 && v <= max)
-        rootMoves[j++] = rootMoves[i];
+                rootMoves[j++] = rootMoves[i];
-    }
+        }
-  } else if (dtz < 0) { // losing (or 50-move rule draw)
+    } else if (dtz < 0) { // losing (or 50-move rule draw)
-    int best = 0;
+        int best = 0;
-    for (size_t i = 0; i < rootMoves.size(); i++) {
+        for (size_t i = 0; i < rootMoves.size(); ++i) {
-      int v = rootMoves[i].score;
+            int v = rootMoves[i].score;
-      if (v < best)
+            if (v < best)
-        best = v;
+                best = v;
-    }
+        }
-    // Try all moves, unless we approach or have a 50-move rule draw.
+        // Try all moves, unless we approach or have a 50-move rule draw.
-    if (-best * 2 + cnt50 < 100)
+        if (-best * 2 + cnt50 < 100)
-      return true;
+            return true;
-    for (size_t i = 0; i < rootMoves.size(); i++) {
+        for (size_t i = 0; i < rootMoves.size(); ++i) {
-      if (rootMoves[i].score == best)
+            if (rootMoves[i].score == best)
+                rootMoves[j++] = rootMoves[i];
+        }
+    } else { // drawing
+        // Try all moves that preserve the draw.
+        for (size_t i = 0; i < rootMoves.size(); ++i) {
+            if (rootMoves[i].score == 0)
-        rootMoves[j++] = rootMoves[i];
+                rootMoves[j++] = rootMoves[i];
+        }
+    }
-  } else { // drawing
-    // Try all moves that preserve the draw.
-    for (size_t i = 0; i < rootMoves.size(); i++) {
-      if (rootMoves[i].score == 0)
-        rootMoves[j++] = rootMoves[i];
-    }
-  }
-  rootMoves.resize(j, Search::RootMove(MOVE_NONE));
+    rootMoves.resize(j, Search::RootMove(MOVE_NONE));
-  return true;
+    return true;
+}
 // Use the WDL tables to filter out moves that don't preserve the win or draw.
 // This is a fallback for the case that some or all DTZ tables are missing.
 //
 // A return value false indicates that not all probes were successful and that
 // no moves were filtered out.
 bool Tablebases::root_probe_wdl(Position& pos, Search::RootMoves& rootMoves, Value& score)
+{
-  int success;
+    ProbeState result;
-  int wdl = Tablebases::probe_wdl(pos, &success);
+    WDLScore wdl = Tablebases::probe_wdl(pos, &result);
-  if (!success) return false;
-  score = wdl_to_Value[wdl + 2];
+    if (result == FAIL)
-  StateInfo st;
+        return false;
-  int best = -2;
+    score = WDL_to_value[wdl + 2];
-  // Probe each move.
-  for (size_t i = 0; i < rootMoves.size(); i++) {
-    Move move = rootMoves[i].pv[0];
-    pos.do_move(move, st, pos.gives_check(move));
-    int v = -Tablebases::probe_wdl(pos, &success);
-    pos.undo_move(move);
-    if (!success) return false;
-    rootMoves[i].score = (Value)v;
-    if (v > best)
-      best = v;
+    StateInfo st;
-  }
-  size_t j = 0;
+    int best = WDLLoss;
-  for (size_t i = 0; i < rootMoves.size(); i++) {
-    if (rootMoves[i].score == best)
-      rootMoves[j++] = rootMoves[i];
-  }
-  rootMoves.resize(j, Search::RootMove(MOVE_NONE));
-  return true;
+    // Probe each move
-}
+    for (size_t i = 0; i < rootMoves.size(); ++i) {
+        Move move = rootMoves[i].pv[0];
+        pos.do_move(move, st);
+        WDLScore v = -Tablebases::probe_wdl(pos, &result);
+        pos.undo_move(move);
+        if (result == FAIL)
+            return false;
+        rootMoves[i].score = (Value)v;
+        if (v > best)
+            best = v;
+    }
+    size_t j = 0;
+    for (size_t i = 0; i < rootMoves.size(); ++i) {
+        if (rootMoves[i].score == best)
+            rootMoves[j++] = rootMoves[i];
+    }
+    rootMoves.resize(j, Search::RootMove(MOVE_NONE));
+    return true;
+}

Subversion Repositories Games.Chess Giants

Games.Chess Giants/engine-stockfish/syzygy/tbprobe.cpp – Rev 154 → 169