X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=src%2Fnnue%2Fnnue_common.h;h=d41e02377ac34b35378a969034682a9b3029190b;hb=58054fd0fa6294510fc8cf76b0ba9673d5094c10;hp=e7ce84f7b9f420eaaa3dbadb4302eeb242315c4f;hpb=875183b310a8249922c2155e82cb4cecfae2097e;p=stockfish

diff --git a/src/nnue/nnue_common.h b/src/nnue/nnue_common.h
index e7ce84f7..d41e0237 100644
--- a/src/nnue/nnue_common.h
+++ b/src/nnue/nnue_common.h
@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2020 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2021 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -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>
 
@@ -33,55 +36,40 @@
 #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 {
+namespace Stockfish::Eval::NNUE {
 
   // Version of the evaluation file
-  constexpr std::uint32_t kVersion = 0x7AF32F16u;
+  constexpr std::uint32_t Version = 0x7AF32F16u;
 
   // Constant used in evaluation value calculation
-  constexpr int FV_SCALE = 16;
-  constexpr int kWeightScaleBits = 6;
+  constexpr int OutputScale = 16;
+  constexpr int WeightScaleBits = 6;
 
   // Size of cache line (in bytes)
-  constexpr std::size_t kCacheLineSize = 64;
+  constexpr std::size_t CacheLineSize = 64;
 
   // SIMD width (in bytes)
   #if defined(USE_AVX2)
-  constexpr std::size_t kSimdWidth = 32;
+  constexpr std::size_t SimdWidth = 32;
 
   #elif defined(USE_SSE2)
-  constexpr std::size_t kSimdWidth = 16;
+  constexpr std::size_t SimdWidth = 16;
+
+  #elif defined(USE_MMX)
+  constexpr std::size_t SimdWidth = 8;
 
   #elif defined(USE_NEON)
-  constexpr std::size_t kSimdWidth = 16;
+  constexpr std::size_t SimdWidth = 16;
   #endif
 
-  constexpr std::size_t kMaxSimdWidth = 32;
+  constexpr std::size_t MaxSimdWidth = 32;
 
   // Type of input feature after conversion
   using TransformedFeatureType = std::uint8_t;
@@ -89,10 +77,46 @@ 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;
+  constexpr IntType ceil_to_multiple(IntType n, IntType base) {
+      return (n + base - 1) / base * base;
   }
 
-}  // namespace Eval::NNUE
+  // 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;
+  }
+
+  template <typename IntType>
+  inline void write_little_endian(std::ostream& stream, IntType value) {
+
+      std::uint8_t u[sizeof(IntType)];
+      typename std::make_unsigned<IntType>::type v = value;
+
+      std::size_t i = 0;
+      // if constexpr to silence the warning about shift by 8
+      if constexpr (sizeof(IntType) > 1) {
+        for (; i + 1 < sizeof(IntType); ++i) {
+            u[i] = v;
+            v >>= 8;
+        }
+      }
+      u[i] = v;
+
+      stream.write(reinterpret_cast<char*>(u), sizeof(IntType));
+  }
+}  // namespace Stockfish::Eval::NNUE
 
 #endif // #ifndef NNUE_COMMON_H_INCLUDED