Fixed compile errors.

src/eval/nnue/evaluate_nnue.cpp

@@ -1,10 +1,12 @@
 // NNUE評価関数の計算に関するコード
 
 #include <fstream>
+#include <iostream>
 
 #include "../../evaluate.h"
 #include "../../position.h"
 #include "../../misc.h"
+#include "../../uci.h"
 
 #include "evaluate_nnue.h"
 
@@ -263,7 +265,7 @@ Value compute_eval(const Position& pos) {
 }
 
 // 評価関数
-Value evaluate(const Position& pos) {
+Value NNUE::evaluate(const Position& pos) {
   const auto& accumulator = pos.state()->accumulator;
   if (accumulator.computed_score) {
     return accumulator.score;

src/eval/nnue/evaluate_nnue.h

@@ -55,6 +55,8 @@ bool ReadParameters(std::istream& stream);
 // 評価関数パラメータを書き込む
 bool WriteParameters(std::ostream& stream);
 
+Value evaluate(const Position& pos);
+
 }  // namespace NNUE
 
 }  // namespace Eval

src/eval/nnue/features/feature_set.h

@@ -71,7 +71,7 @@ class FeatureSetBase {
   template <typename IndexListType>
   static void AppendActiveIndices(
       const Position& pos, TriggerEvent trigger, IndexListType active[2]) {
-    for (const auto perspective : COLOR) {
+    for (const auto perspective : Colors) {
       Derived::CollectActiveIndices(
           pos, trigger, perspective, &active[perspective]);
     }
@@ -85,7 +85,7 @@ class FeatureSetBase {
     const auto& dp = pos.state()->dirtyPiece;
     if (dp.dirty_num == 0) return;
 
-    for (const auto perspective : COLOR) {
+    for (const auto perspective : Colors) {
       reset[perspective] = false;
       switch (trigger) {
         case TriggerEvent::kNone:
@@ -105,7 +105,7 @@ class FeatureSetBase {
           reset[perspective] = true;
           break;
         default:
-          ASSERT_LV5(false);
+          assert(false);
           break;
       }
       if (reset[perspective]) {

src/eval/nnue/features/half_kp.cpp

@@ -28,7 +28,7 @@ inline void HalfKP<AssociatedKing>::GetPieces(
   const PieceNumber target = (AssociatedKing == Side::kFriend) ?
       static_cast<PieceNumber>(PIECE_NUMBER_KING + perspective) :
       static_cast<PieceNumber>(PIECE_NUMBER_KING + ~perspective);
-  *sq_target_k = static_cast<Square>(((*pieces)[target] - f_king) % SQ_NB);
+  *sq_target_k = static_cast<Square>(((*pieces)[target] - f_king) % SQUARE_NB);
 }
 
 // 特徴量のうち、値が1であるインデックスのリストを取得する

src/eval/nnue/features/half_kp.h

@@ -26,7 +26,7 @@ class HalfKP {
       0x5D69D5B9u ^ (AssociatedKing == Side::kFriend);
   // 特徴量の次元数
   static constexpr IndexType kDimensions =
-      static_cast<IndexType>(SQ_NB) * static_cast<IndexType>(fe_end);
+      static_cast<IndexType>(SQUARE_NB) * static_cast<IndexType>(fe_end);
   // 特徴量のうち、同時に値が1となるインデックスの数の最大値
   static constexpr IndexType kMaxActiveDimensions = PIECE_NUMBER_KING;
   // 差分計算の代わりに全計算を行うタイミング

src/eval/nnue/features/half_relative_kp.cpp

@@ -17,8 +17,8 @@ inline IndexType HalfRelativeKP<AssociatedKing>::MakeIndex(
     Square sq_k, BonaPiece p) {
   constexpr IndexType W = kBoardWidth;
   constexpr IndexType H = kBoardHeight;
-  const IndexType piece_index = (p - fe_hand_end) / SQ_NB;
-  const Square sq_p = static_cast<Square>((p - fe_hand_end) % SQ_NB);
+  const IndexType piece_index = (p - fe_hand_end) / SQUARE_NB;
+  const Square sq_p = static_cast<Square>((p - fe_hand_end) % SQUARE_NB);
   const IndexType relative_file = file_of(sq_p) - file_of(sq_k) + (W / 2);
   const IndexType relative_rank = rank_of(sq_p) - rank_of(sq_k) + (H / 2);
   return H * W * piece_index + H * relative_file + relative_rank;
@@ -35,7 +35,7 @@ inline void HalfRelativeKP<AssociatedKing>::GetPieces(
   const PieceNumber target = (AssociatedKing == Side::kFriend) ?
       static_cast<PieceNumber>(PIECE_NUMBER_KING + perspective) :
       static_cast<PieceNumber>(PIECE_NUMBER_KING + ~perspective);
-  *sq_target_k = static_cast<Square>(((*pieces)[target] - f_king) % SQ_NB);
+  *sq_target_k = static_cast<Square>(((*pieces)[target] - f_king) % SQUARE_NB);
 }
 
 // 特徴量のうち、値が1であるインデックスのリストを取得する

src/eval/nnue/features/half_relative_kp.h

@@ -25,7 +25,7 @@ class HalfRelativeKP {
   static constexpr std::uint32_t kHashValue =
       0xF9180919u ^ (AssociatedKing == Side::kFriend);
   // 玉を除いた駒種
-  static constexpr IndexType kNumPieceKinds = (fe_end - fe_hand_end) / SQ_NB;
+  static constexpr IndexType kNumPieceKinds = (fe_end - fe_hand_end) / SQUARE_NB;
   // 玉を中央に置いた仮想的な盤の幅
   static constexpr IndexType kBoardWidth = FILE_NB * 2 - 1;
   // 玉を中央に置いた仮想的な盤の高さ

src/eval/nnue/features/k.cpp

@@ -20,8 +20,8 @@ void K::AppendActiveIndices(
   const BonaPiece* pieces = (perspective == BLACK) ?
       pos.eval_list()->piece_list_fb() :
       pos.eval_list()->piece_list_fw();
-  ASSERT_LV5(pieces[PIECE_NUMBER_BKING] != BONA_PIECE_ZERO);
-  ASSERT_LV5(pieces[PIECE_NUMBER_WKING] != BONA_PIECE_ZERO);
+  assert(pieces[PIECE_NUMBER_BKING] != BONA_PIECE_ZERO);
+  assert(pieces[PIECE_NUMBER_WKING] != BONA_PIECE_ZERO);
   for (PieceNumber i = PIECE_NUMBER_KING; i < PIECE_NUMBER_NB; ++i) {
     active->push_back(pieces[i] - fe_end);
   }

src/eval/nnue/features/k.h

@@ -22,7 +22,7 @@ class K {
   // 評価関数ファイルに埋め込むハッシュ値
   static constexpr std::uint32_t kHashValue = 0xD3CEE169u;
   // 特徴量の次元数
-  static constexpr IndexType kDimensions = SQ_NB * 2;
+  static constexpr IndexType kDimensions = SQUARE_NB * 2;
   // 特徴量のうち、同時に値が1となるインデックスの数の最大値
   static constexpr IndexType kMaxActiveDimensions = 2;
   // 差分計算の代わりに全計算を行うタイミング

src/eval/nnue/nnue_accumulator.h

@@ -3,8 +3,6 @@
 #ifndef _NNUE_ACCUMULATOR_H_
 #define _NNUE_ACCUMULATOR_H_
 
-#include "../../config.h"
-
 #if defined(EVAL_NNUE)
 
 #include "nnue_architecture.h"

src/eval/nnue/nnue_architecture.h

@@ -8,6 +8,7 @@
 // 入力特徴量とネットワーク構造が定義されたヘッダをincludeする
 
 // KP256型を使いたいときは、これを事前にdefineする。
+#define EVAL_NNUE_KP256
 #if defined(EVAL_NNUE_KP256)
 #include "architectures/k-p_256x2-32-32.h"
 #else // #if defined(EVAL_NNUE_HALFKP256)

src/eval/nnue/nnue_feature_transformer.h

@@ -142,9 +142,9 @@ class FeatureTransformer {
       }
 #else
       for (IndexType j = 0; j < kHalfDimensions; ++j) {
-        BiasType sum = accumulation[perspectives[p]][0][j];
+        BiasType sum = accumulation[static_cast<int>(perspectives[p])][0][j];
         for (IndexType i = 1; i < kRefreshTriggers.size(); ++i) {
-          sum += accumulation[perspectives[p]][i][j];
+          sum += accumulation[static_cast<int>(perspectives[p])][i][j];
         }
         output[offset + j] = static_cast<OutputType>(
             std::max<int>(0, std::min<int>(127, sum)));
@@ -161,7 +161,7 @@ class FeatureTransformer {
       Features::IndexList active_indices[2];
       RawFeatures::AppendActiveIndices(pos, kRefreshTriggers[i],
                                        active_indices);
-      for (const auto perspective : COLOR) {
+      for (const auto perspective : Colors) {
         if (i == 0) {
           std::memcpy(accumulator.accumulation[perspective][i], biases_,
                       kHalfDimensions * sizeof(BiasType));
@@ -217,7 +217,7 @@ class FeatureTransformer {
       bool reset[2];
       RawFeatures::AppendChangedIndices(pos, kRefreshTriggers[i],
                                         removed_indices, added_indices, reset);
-      for (const auto perspective : COLOR) {
+      for (const auto perspective : Colors) {
 #if defined(USE_AVX2)
         constexpr IndexType kNumChunks = kHalfDimensions / (kSimdWidth / 2);
         auto accumulation = reinterpret_cast<__m256i*>(

src/eval/nnue/trainer/features/factorizer_half_kp.h

@@ -43,7 +43,7 @@ class Factorizer<HalfKP<AssociatedKing>> {
     // kFeaturesHalfKP
     {true, FeatureType::kDimensions},
     // kFeaturesHalfK
-    {true, SQ_NB},
+    {true, SQUARE_NB},
     // kFeaturesP
     {true, Factorizer<P>::GetDimensions()},
     // kFeaturesHalfRelativeKP