organ
/
Sanmill
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

refactor: 从 AIAlgorithm 类中将 MTDF 函数和最底层的 search 移动到类外部

This commit is contained in:
Calcitem 2020-07-12 12:06:53 +08:00
parent 0d24cb9969
commit ab716466dd
4 changed files with 270 additions and 278 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
src/position.cpp
View File

@ -973,6 +973,12 @@ bool Position::undo_move(Move m)
return ret;
}
void Position::undo_move(Stack<Position> &ss)
{
memcpy(this, ss.top(), sizeof(Position));
ss.pop();
}
Color Position::get_winner() const
{
return winner;

1
src/position.h
View File

@ -101,6 +101,7 @@ public:
// Doing and undoing moves
bool do_move(Move m);
bool undo_move(Move m);
void undo_move(Stack<Position> &ss);
bool undo_null_move();
bool do_null_move();

524
src/search.cpp
View File

@ -52,6 +52,8 @@ LimitsType Limits;
using namespace Search;
Value MTDF(Position *pos, Stack<Position> &ss, Value firstguess, Depth depth, Depth originDepth, Move &bestMove);
vector<Key> moveHistory;
AIAlgorithm::AIAlgorithm()
@ -178,7 +180,7 @@ void AIAlgorithm::setPosition(Position *p)
pos = p;
// position = pos;
requiredQuit = false;
//requiredQuit = false;
}
#ifdef ALPHABETA_AI
@ -245,9 +247,9 @@ int AIAlgorithm::search()
#endif
#ifdef MTDF_AI
value = MTDF(value, i);
value = MTDF(pos, ss, value, i, originDepth, bestMove);
#else
value = search(i, alpha, beta);
value = search(pos, ss, i, originDepth, alpha, beta, bestMove);
#endif
loggerDebug("%d(%d) ", value, value - lastValue);
@ -275,9 +277,9 @@ int AIAlgorithm::search()
originDepth = d;
#ifdef MTDF_AI
value = MTDF(value, d);
value = MTDF(pos, ss, value, d, originDepth, bestMove);
#else
value = search(d, alpha, beta);
value = search(pos, ss, d, originDepth, alpha, beta, bestMove);
#endif
#ifdef TIME_STAT
@ -290,275 +292,13 @@ int AIAlgorithm::search()
return 0;
}
Value AIAlgorithm::MTDF(Value firstguess, Depth depth)
{
Value g = firstguess;
Value lowerbound = -VALUE_INFINITE;
Value upperbound = VALUE_INFINITE;
Value beta;
while (lowerbound < upperbound) {
if (g == lowerbound) {
beta = g + VALUE_MTDF_WINDOW;
} else {
beta = g;
}
g = search(depth, beta - VALUE_MTDF_WINDOW, beta);
if (g < beta) {
upperbound = g; // fail low
} else {
lowerbound = g; // fail high
}
}
return g;
}
Value AIAlgorithm::search(Depth depth, Value alpha, Value beta)
{
Value value;
Value bestValue = -VALUE_INFINITE;
Depth epsilon;
#ifdef TT_MOVE_ENABLE
Move ttMove = MOVE_NONE;
#endif // TT_MOVE_ENABLE
#if defined (TRANSPOSITION_TABLE_ENABLE) || defined(ENDGAME_LEARNING)
Key posKey = pos->key();
#endif
#ifdef ENDGAME_LEARNING
// 检索残局库
Endgame endgame;
if (gameOptions.getLearnEndgameEnabled() &&
findEndgameHash(posKey, endgame)) {
switch (endgame.type) {
case ENDGAME_PLAYER_BLACK_WIN:
bestValue = VALUE_MATE;
bestValue += depth;
break;
case ENDGAME_PLAYER_WHITE_WIN:
bestValue = -VALUE_MATE;
bestValue -= depth;
break;
default:
break;
}
return bestValue;
}
#endif /* ENDGAME_LEARNING */
#ifdef TRANSPOSITION_TABLE_ENABLE
Bound type = BOUND_NONE;
Value probeVal = TranspositionTable::probe(posKey, depth, alpha, beta, type
#ifdef TT_MOVE_ENABLE
, ttMove
#endif // TT_MOVE_ENABLE
);
if (probeVal != VALUE_UNKNOWN) {
#ifdef TRANSPOSITION_TABLE_DEBUG
ttHitCount++;
#endif
bestValue = probeVal;
#if 0
// TODO: 有必要针对深度微调 value?
if (position->turn == BLACK)
bestValue += tte.depth - depth;
else
bestValue -= tte.depth - depth;
#endif
#ifdef TT_MOVE_ENABLE
// if (ttMove != MOVE_NONE) {
// bestMove = ttMove;
// }
#endif // TT_MOVE_ENABLE
return bestValue;
}
#ifdef TRANSPOSITION_TABLE_DEBUG
else {
ttMissCount++;
}
#endif
//hashMapMutex.unlock();
#endif /* TRANSPOSITION_TABLE_ENABLE */
#if 0
if (position->phase == PHASE_PLACING && depth == 1 && pos->nPiecesNeedRemove > 0) {
depth--;
}
#endif
if (unlikely(pos->phase == PHASE_GAMEOVER) || // TODO: Deal with hash
depth <= 0 ||
unlikely(requiredQuit)) {
bestValue = Eval::evaluate(pos);
// For win quickly
if (bestValue > 0) {
bestValue += depth;
} else {
bestValue -= depth;
}
#ifdef TRANSPOSITION_TABLE_ENABLE
TranspositionTable::save(bestValue,
depth,
BOUND_EXACT,
posKey
#ifdef TT_MOVE_ENABLE
, MOVE_NONE
#endif // TT_MOVE_ENABLE
);
#endif
return bestValue;
}
MovePicker mp(pos);
mp.endMoves = generate(pos, mp.moves);
mp.score();
partial_insertion_sort(mp.moves, mp.endMoves, -100);
ExtMove *cur = mp.moves;
int nchild = mp.endMoves - cur;
if (nchild == 1 && depth == originDepth) {
bestMove = mp.moves[0].move;
bestValue = VALUE_UNIQUE;
return bestValue;
}
#ifdef TRANSPOSITION_TABLE_ENABLE
#ifdef PREFETCH_SUPPORT
for (int i = 0; i < nchild; i++) {
TranspositionTable::prefetch(pos->next_primary_key(mp.moves[i].move));
}
#ifdef PREFETCH_DEBUG
if (posKey << 8 >> 8 == 0x0)
{
int pause = 1;
}
#endif // PREFETCH_DEBUG
#endif // PREFETCH_SUPPORT
#endif // TRANSPOSITION_TABLE_ENABLE
for (int i = 0; i < nchild; i++) {
stashPosition();
Color before = pos->sideToMove;
Move move = mp.moves[i].move;
do_move(move);
Color after = pos->sideToMove;
if (gameOptions.getDepthExtension() == true && nchild == 1) {
epsilon = 1;
} else {
epsilon = 0;
}
#ifdef PVS_AI
if (i == 0) {
if (after != before) {
value = -search(depth - 1 + epsilon, -beta, -alpha);
} else {
value = search(depth - 1 + epsilon, alpha, beta);
}
} else {
if (after != before) {
value = -search(depth - 1 + epsilon, -alpha - VALUE_PVS_WINDOW, -alpha);
if (value > alpha && value < beta) {
value = -search(depth - 1 + epsilon, -beta, -alpha);
//assert(value >= alpha && value <= beta);
}
} else {
value = search(depth - 1 + epsilon, alpha, alpha + VALUE_PVS_WINDOW);
if (value > alpha && value < beta) {
value = search(depth - 1 + epsilon, alpha, beta);
//assert(value >= alpha && value <= beta);
}
}
}
#else
if (after != before) {
value = -search(depth - 1 + epsilon, -beta, -alpha);
} else {
value = search(depth - 1 + epsilon, alpha, beta);
}
#endif // PVS_AI
undo_move();
assert(value > -VALUE_INFINITE && value < VALUE_INFINITE);
if (value >= bestValue) {
bestValue = value;
if (value > alpha) {
if (depth == originDepth) {
bestMove = move;
}
break;
}
}
}
#ifdef TRANSPOSITION_TABLE_ENABLE
TranspositionTable::save(bestValue,
depth,
bestValue >= beta ? BOUND_LOWER :
BOUND_UPPER,
posKey
#ifdef TT_MOVE_ENABLE
, bestMove
#endif // TT_MOVE_ENABLE
);
#endif /* TRANSPOSITION_TABLE_ENABLE */
#ifdef HOSTORY_HEURISTIC
movePicker->setHistoryScore(bestMove, depth);
#endif
assert(bestValue > -VALUE_INFINITE && bestValue < VALUE_INFINITE);
return bestValue;
}
#endif // ALPHABETA_AI
void AIAlgorithm::stashPosition()
{
positionStack.push(*(pos));
}
void AIAlgorithm::do_move(Move move)
{
pos->do_move(move);
}
void AIAlgorithm::undo_move()
{
memcpy(pos, positionStack.top(), sizeof(Position));
//tmppos = positionStack.top();
positionStack.pop();
}
#ifdef ALPHABETA_AI
const char* AIAlgorithm::nextMove()
{
@ -860,3 +600,253 @@ void MainThread::check_time()
|| (Limits.nodes && Threads.nodes_searched() >= (uint64_t)Limits.nodes))
Threads.stop = true;
}
// search<>() is the main search function for both PV and non-PV nodes
Value search(Position *pos, Stack<Position> &ss, Depth depth, Depth originDepth, Value alpha, Value beta, Move &bestMove)
{
Value value;
Value bestValue = -VALUE_INFINITE;
Depth epsilon;
#ifdef TT_MOVE_ENABLE
Move ttMove = MOVE_NONE;
#endif // TT_MOVE_ENABLE
#if defined (TRANSPOSITION_TABLE_ENABLE) || defined(ENDGAME_LEARNING)
Key posKey = pos->key();
#endif
#ifdef ENDGAME_LEARNING
// 检索残局库
Endgame endgame;
if (gameOptions.getLearnEndgameEnabled() &&
findEndgameHash(posKey, endgame)) {
switch (endgame.type) {
case ENDGAME_PLAYER_BLACK_WIN:
bestValue = VALUE_MATE;
bestValue += depth;
break;
case ENDGAME_PLAYER_WHITE_WIN:
bestValue = -VALUE_MATE;
bestValue -= depth;
break;
default:
break;
}
return bestValue;
}
#endif /* ENDGAME_LEARNING */
#ifdef TRANSPOSITION_TABLE_ENABLE
Bound type = BOUND_NONE;
Value probeVal = TranspositionTable::probe(posKey, depth, alpha, beta, type
#ifdef TT_MOVE_ENABLE
, ttMove
#endif // TT_MOVE_ENABLE
);
if (probeVal != VALUE_UNKNOWN) {
#ifdef TRANSPOSITION_TABLE_DEBUG
ttHitCount++;
#endif
bestValue = probeVal;
#if 0
// TODO: 有必要针对深度微调 value?
if (position->turn == BLACK)
bestValue += tte.depth - depth;
else
bestValue -= tte.depth - depth;
#endif
#ifdef TT_MOVE_ENABLE
// if (ttMove != MOVE_NONE) {
// bestMove = ttMove;
// }
#endif // TT_MOVE_ENABLE
return bestValue;
}
#ifdef TRANSPOSITION_TABLE_DEBUG
else {
ttMissCount++;
}
#endif
//hashMapMutex.unlock();
#endif /* TRANSPOSITION_TABLE_ENABLE */
#if 0
if (position->phase == PHASE_PLACING && depth == 1 && pos->nPiecesNeedRemove > 0) {
depth--;
}
#endif
if (unlikely(pos->phase == PHASE_GAMEOVER) || // TODO: Deal with hash and requiredQuit
depth <= 0) {
bestValue = Eval::evaluate(pos);
// For win quickly
if (bestValue > 0) {
bestValue += depth;
} else {
bestValue -= depth;
}
#ifdef TRANSPOSITION_TABLE_ENABLE
TranspositionTable::save(bestValue,
depth,
BOUND_EXACT,
posKey
#ifdef TT_MOVE_ENABLE
, MOVE_NONE
#endif // TT_MOVE_ENABLE
);
#endif
return bestValue;
}
MovePicker mp(pos);
mp.endMoves = generate(pos, mp.moves);
mp.score();
partial_insertion_sort(mp.moves, mp.endMoves, -100);
ExtMove *cur = mp.moves;
int nchild = mp.endMoves - cur;
if (nchild == 1 && depth == originDepth) {
bestMove = mp.moves[0].move;
bestValue = VALUE_UNIQUE;
return bestValue;
}
#ifdef TRANSPOSITION_TABLE_ENABLE
#ifdef PREFETCH_SUPPORT
for (int i = 0; i < nchild; i++) {
TranspositionTable::prefetch(pos->next_primary_key(mp.moves[i].move));
}
#ifdef PREFETCH_DEBUG
if (posKey << 8 >> 8 == 0x0) {
int pause = 1;
}
#endif // PREFETCH_DEBUG
#endif // PREFETCH_SUPPORT
#endif // TRANSPOSITION_TABLE_ENABLE
for (int i = 0; i < nchild; i++) {
ss.push(*(pos));
Color before = pos->sideToMove;
Move move = mp.moves[i].move;
pos->do_move(move);
Color after = pos->sideToMove;
if (gameOptions.getDepthExtension() == true && nchild == 1) {
epsilon = 1;
} else {
epsilon = 0;
}
#ifdef PVS_AI
if (i == 0) {
if (after != before) {
value = -search(depth - 1 + epsilon, -beta, -alpha);
} else {
value = search(depth - 1 + epsilon, alpha, beta);
}
} else {
if (after != before) {
value = -search(depth - 1 + epsilon, -alpha - VALUE_PVS_WINDOW, -alpha);
if (value > alpha && value < beta) {
value = -search(depth - 1 + epsilon, -beta, -alpha);
//assert(value >= alpha && value <= beta);
}
} else {
value = search(depth - 1 + epsilon, alpha, alpha + VALUE_PVS_WINDOW);
if (value > alpha && value < beta) {
value = search(depth - 1 + epsilon, alpha, beta);
//assert(value >= alpha && value <= beta);
}
}
}
#else
if (after != before) {
value = -search(pos, ss, depth - 1 + epsilon, originDepth, -beta, -alpha, bestMove);
} else {
value = search(pos, ss, depth - 1 + epsilon, originDepth, alpha, beta, bestMove);
}
#endif // PVS_AI
pos->undo_move(ss);
assert(value > -VALUE_INFINITE && value < VALUE_INFINITE);
if (value >= bestValue) {
bestValue = value;
if (value > alpha) {
if (depth == originDepth) {
bestMove = move;
}
break;
}
}
}
#ifdef TRANSPOSITION_TABLE_ENABLE
TranspositionTable::save(bestValue,
depth,
bestValue >= beta ? BOUND_LOWER :
BOUND_UPPER,
posKey
#ifdef TT_MOVE_ENABLE
, bestMove
#endif // TT_MOVE_ENABLE
);
#endif /* TRANSPOSITION_TABLE_ENABLE */
#ifdef HOSTORY_HEURISTIC
movePicker->setHistoryScore(bestMove, depth);
#endif
assert(bestValue > -VALUE_INFINITE && bestValue < VALUE_INFINITE);
return bestValue;
}
Value MTDF(Position *pos, Stack<Position> &ss, Value firstguess, Depth depth, Depth originDepth, Move &bestMove)
{
Value g = firstguess;
Value lowerbound = -VALUE_INFINITE;
Value upperbound = VALUE_INFINITE;
Value beta;
while (lowerbound < upperbound) {
if (g == lowerbound) {
beta = g + VALUE_MTDF_WINDOW;
} else {
beta = g;
}
g = search(pos, ss, depth, originDepth, beta - VALUE_MTDF_WINDOW, beta, bestMove);
if (g < beta) {
upperbound = g; // fail low
} else {
lowerbound = g; // fail high
}
}
return g;
}

17
src/search.h
View File

@ -51,6 +51,7 @@ namespace Search
constexpr int CounterMovePruneThreshold = 0;
#if 0
/// Stack struct keeps track of the information we need to remember from nodes
/// shallower and deeper in the tree during the search. Each search thread has
/// its own array of Stack objects, indexed by the current ply.
@ -67,6 +68,7 @@ struct Stack
int moveCount;
bool inCheck;
};
#endif
/// RootMove struct is used for moves at the root of the tree. For each root move
@ -151,7 +153,7 @@ public:
void quit()
{
loggerDebug("Timeout\n");
requiredQuit = true;
//requiredQuit = true; // TODO
#ifdef HOSTORY_HEURISTIC
movePicker->clearHistoryScore();
#endif
@ -162,8 +164,6 @@ public:
const char *nextMove();
#endif // ALPHABETA_AI
void stashPosition();
void do_move(Move move);
void undo_move();
@ -181,13 +181,8 @@ public:
#endif // ENDGAME_LEARNING
public: /* TODO: Move to private or protected */
Value search(Depth depth, Value alpha, Value beta);
Value MTDF(Value firstguess, Depth depth);
public:
const char *moveToCommand(Move move);
protected:
Depth changeDepth();
@ -202,9 +197,9 @@ public:
private:
Position *pos { nullptr };
Stack<Position> positionStack;
Stack<Position> ss;
bool requiredQuit {false};
// bool requiredQuit {false}; // TODO
Move bestMove { MOVE_NONE };