X-Git-Url: https://git.sesse.net/?p=stockfish;a=blobdiff_plain;f=src%2Fnnue%2Fnnue_common.h;h=f9ff2bc81ee8f919903fef0f37217379a1304262;hp=e7ce84f7b9f420eaaa3dbadb4302eeb242315c4f;hb=9b7983a4521b66bf8d3c37ee58963d39deb2695c;hpb=875183b310a8249922c2155e82cb4cecfae2097e

diff --git a/src/nnue/nnue_common.h b/src/nnue/nnue_common.h
index e7ce84f7..f9ff2bc8 100644
--- a/src/nnue/nnue_common.h
+++ b/src/nnue/nnue_common.h
@@ -21,6 +21,9 @@
 #ifndef NNUE_COMMON_H_INCLUDED
 #define NNUE_COMMON_H_INCLUDED
 
+#include <cstring>
+#include <iostream>
+
 #if defined(USE_AVX2)
 #include <immintrin.h>
 
@@ -33,31 +36,13 @@
 #elif defined(USE_SSE2)
 #include <emmintrin.h>
 
+#elif defined(USE_MMX)
+#include <mmintrin.h>
+
 #elif defined(USE_NEON)
 #include <arm_neon.h>
 #endif
 
-// HACK: Use _mm256_loadu_si256() instead of _mm256_load_si256. Otherwise a binary
-//       compiled with older g++ crashes because the output memory is not aligned
-//       even though alignas is specified.
-#if defined(USE_AVX2)
-#if defined(__GNUC__ ) && (__GNUC__ < 9)
-#define _mm256_loadA_si256  _mm256_loadu_si256
-#define _mm256_storeA_si256 _mm256_storeu_si256
-#else
-#define _mm256_loadA_si256  _mm256_load_si256
-#define _mm256_storeA_si256 _mm256_store_si256
-#endif
-#endif
-
-#if defined(USE_AVX512)
-#if defined(__GNUC__ ) && (__GNUC__ < 9)
-#define _mm512_loadA_si512  _mm512_loadu_si512
-#else
-#define _mm512_loadA_si512  _mm512_load_si512
-#endif
-#endif
-
 namespace Eval::NNUE {
 
   // Version of the evaluation file
@@ -77,12 +62,43 @@ namespace Eval::NNUE {
   #elif defined(USE_SSE2)
   constexpr std::size_t kSimdWidth = 16;
 
+  #elif defined(USE_MMX)
+  constexpr std::size_t kSimdWidth = 8;
+
   #elif defined(USE_NEON)
   constexpr std::size_t kSimdWidth = 16;
   #endif
 
   constexpr std::size_t kMaxSimdWidth = 32;
 
+  // unique number for each piece type on each square
+  enum {
+    PS_NONE     =  0,
+    PS_W_PAWN   =  1,
+    PS_B_PAWN   =  1 * SQUARE_NB + 1,
+    PS_W_KNIGHT =  2 * SQUARE_NB + 1,
+    PS_B_KNIGHT =  3 * SQUARE_NB + 1,
+    PS_W_BISHOP =  4 * SQUARE_NB + 1,
+    PS_B_BISHOP =  5 * SQUARE_NB + 1,
+    PS_W_ROOK   =  6 * SQUARE_NB + 1,
+    PS_B_ROOK   =  7 * SQUARE_NB + 1,
+    PS_W_QUEEN  =  8 * SQUARE_NB + 1,
+    PS_B_QUEEN  =  9 * SQUARE_NB + 1,
+    PS_W_KING   = 10 * SQUARE_NB + 1,
+    PS_END      = PS_W_KING, // pieces without kings (pawns included)
+    PS_B_KING   = 11 * SQUARE_NB + 1,
+    PS_END2     = 12 * SQUARE_NB + 1
+  };
+
+  constexpr uint32_t kpp_board_index[COLOR_NB][PIECE_NB] = {
+    // convention: W - us, B - them
+    // viewed from other side, W and B are reversed
+    { PS_NONE, PS_W_PAWN, PS_W_KNIGHT, PS_W_BISHOP, PS_W_ROOK, PS_W_QUEEN, PS_W_KING, PS_NONE,
+      PS_NONE, PS_B_PAWN, PS_B_KNIGHT, PS_B_BISHOP, PS_B_ROOK, PS_B_QUEEN, PS_B_KING, PS_NONE },
+    { PS_NONE, PS_B_PAWN, PS_B_KNIGHT, PS_B_BISHOP, PS_B_ROOK, PS_B_QUEEN, PS_B_KING, PS_NONE,
+      PS_NONE, PS_W_PAWN, PS_W_KNIGHT, PS_W_BISHOP, PS_W_ROOK, PS_W_QUEEN, PS_W_KING, PS_NONE }
+  };
+
   // Type of input feature after conversion
   using TransformedFeatureType = std::uint8_t;
   using IndexType = std::uint32_t;
@@ -90,7 +106,25 @@ namespace Eval::NNUE {
   // Round n up to be a multiple of base
   template <typename IntType>
   constexpr IntType CeilToMultiple(IntType n, IntType base) {
-    return (n + base - 1) / base * base;
+      return (n + base - 1) / base * base;
+  }
+
+  // read_little_endian() is our utility to read an integer (signed or unsigned, any size)
+  // from a stream in little-endian order. We swap the byte order after the read if
+  // necessary to return a result with the byte ordering of the compiling machine.
+  template <typename IntType>
+  inline IntType read_little_endian(std::istream& stream) {
+
+      IntType result;
+      std::uint8_t u[sizeof(IntType)];
+      typename std::make_unsigned<IntType>::type v = 0;
+
+      stream.read(reinterpret_cast<char*>(u), sizeof(IntType));
+      for (std::size_t i = 0; i < sizeof(IntType); ++i)
+          v = (v << 8) | u[sizeof(IntType) - i - 1];
+
+      std::memcpy(&result, &v, sizeof(IntType));
+      return result;
   }
 
 }  // namespace Eval::NNUE