X-Git-Url: https://git.sesse.net/?p=stockfish;a=blobdiff_plain;f=src%2Ftt.cpp;h=d494c27d4ec654fc705d1f45b11fb844cc47314d;hp=04cd7357c336af960f6b8ae8ee26bec2db79cd79;hb=8b8412ef87825d8e341e160585307dc89843b7f6;hpb=8c4f0ffa1d1cc6952874ee30fff9f826646b1aeb diff --git a/src/tt.cpp b/src/tt.cpp index 04cd7357..d494c27d 100644 --- a/src/tt.cpp +++ b/src/tt.cpp @@ -1,8 +1,6 @@ /* Stockfish, a UCI chess playing engine derived from Glaurung 2.1 - Copyright (C) 2004-2008 Tord Romstad (Glaurung author) - Copyright (C) 2008-2015 Marco Costalba, Joona Kiiski, Tord Romstad - Copyright (C) 2015-2018 Marco Costalba, Joona Kiiski, Gary Linscott, Tord Romstad + Copyright (C) 2004-2020 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 @@ -24,11 +22,36 @@ #include "bitboard.h" #include "misc.h" +#include "thread.h" #include "tt.h" #include "uci.h" TranspositionTable TT; // Our global transposition table +/// TTEntry::save() populates the TTEntry with a new node's data, possibly +/// overwriting an old position. Update is not atomic and can be racy. + +void TTEntry::save(Key k, Value v, bool pv, Bound b, Depth d, Move m, Value ev) { + + // Preserve any existing move for the same position + if (m || (uint16_t)k != key16) + move16 = (uint16_t)m; + + // Overwrite less valuable entries + if ((uint16_t)k != key16 + || d - DEPTH_OFFSET > depth8 - 4 + || b == BOUND_EXACT) + { + assert(d >= DEPTH_OFFSET); + + key16 = (uint16_t)k; + value16 = (int16_t)v; + eval16 = (int16_t)ev; + genBound8 = (uint8_t)(TT.generation8 | uint8_t(pv) << 2 | b); + depth8 = (uint8_t)(d - DEPTH_OFFSET); + } +} + /// TranspositionTable::resize() sets the size of the transposition table, /// measured in megabytes. Transposition table consists of a power of 2 number @@ -36,11 +59,12 @@ TranspositionTable TT; // Our global transposition table void TranspositionTable::resize(size_t mbSize) { - clusterCount = mbSize * 1024 * 1024 / sizeof(Cluster); + Threads.main()->wait_for_search_finished(); - free(mem); - mem = malloc(clusterCount * sizeof(Cluster) + CacheLineSize - 1); + aligned_ttmem_free(mem); + clusterCount = mbSize * 1024 * 1024 / sizeof(Cluster); + table = static_cast(aligned_ttmem_alloc(clusterCount * sizeof(Cluster), mem)); if (!mem) { std::cerr << "Failed to allocate " << mbSize @@ -48,7 +72,6 @@ void TranspositionTable::resize(size_t mbSize) { exit(EXIT_FAILURE); } - table = (Cluster*)((uintptr_t(mem) + CacheLineSize - 1) & ~(CacheLineSize - 1)); clear(); } @@ -60,28 +83,29 @@ void TranspositionTable::clear() { std::vector threads; - for (size_t idx = 0; idx < Options["Threads"]; idx++) + for (size_t idx = 0; idx < Options["Threads"]; ++idx) { - threads.push_back(std::thread([this, idx]() { + threads.emplace_back([this, idx]() { // Thread binding gives faster search on systems with a first-touch policy - if (Options["Threads"] >= 8) + if (Options["Threads"] > 8) WinProcGroup::bindThisThread(idx); // Each thread will zero its part of the hash table - const size_t stride = clusterCount / Options["Threads"], - start = stride * idx, + const size_t stride = size_t(clusterCount / Options["Threads"]), + start = size_t(stride * idx), len = idx != Options["Threads"] - 1 ? stride : clusterCount - start; std::memset(&table[start], 0, len * sizeof(Cluster)); - })); + }); } - for (std::thread& th: threads) + for (std::thread& th : threads) th.join(); } + /// TranspositionTable::probe() looks up the current position in the transposition /// table. It returns true and a pointer to the TTEntry if the position is found. /// Otherwise, it returns false and a pointer to an empty or least valuable TTEntry @@ -92,13 +116,12 @@ void TranspositionTable::clear() { TTEntry* TranspositionTable::probe(const Key key, bool& found) const { TTEntry* const tte = first_entry(key); - const uint16_t key16 = key >> 48; // Use the high 16 bits as key inside the cluster + const uint16_t key16 = (uint16_t)key; // Use the low 16 bits as key inside the cluster for (int i = 0; i < ClusterSize; ++i) if (!tte[i].key16 || tte[i].key16 == key16) { - if ((tte[i].genBound8 & 0xFC) != generation8 && tte[i].key16) - tte[i].genBound8 = uint8_t(generation8 | tte[i].bound()); // Refresh + tte[i].genBound8 = uint8_t(generation8 | (tte[i].genBound8 & 0x7)); // Refresh return found = (bool)tte[i].key16, &tte[i]; } @@ -107,11 +130,11 @@ TTEntry* TranspositionTable::probe(const Key key, bool& found) const { TTEntry* replace = tte; for (int i = 1; i < ClusterSize; ++i) // Due to our packed storage format for generation and its cyclic - // nature we add 259 (256 is the modulus plus 3 to keep the lowest - // two bound bits from affecting the result) to calculate the entry + // nature we add 263 (256 is the modulus plus 7 to keep the unrelated + // lowest three bits from affecting the result) to calculate the entry // age correctly even after generation8 overflows into the next cycle. - if ( replace->depth8 - ((259 + generation8 - replace->genBound8) & 0xFC) * 2 - > tte[i].depth8 - ((259 + generation8 - tte[i].genBound8) & 0xFC) * 2) + if ( replace->depth8 - ((263 + generation8 - replace->genBound8) & 0xF8) + > tte[i].depth8 - ((263 + generation8 - tte[i].genBound8) & 0xF8)) replace = &tte[i]; return found = false, replace; @@ -124,12 +147,9 @@ TTEntry* TranspositionTable::probe(const Key key, bool& found) const { int TranspositionTable::hashfull() const { int cnt = 0; - for (int i = 0; i < 1000 / ClusterSize; i++) - { - const TTEntry* tte = &table[i].entry[0]; - for (int j = 0; j < ClusterSize; j++) - if ((tte[j].genBound8 & 0xFC) == generation8) - cnt++; - } - return cnt; + for (int i = 0; i < 1000; ++i) + for (int j = 0; j < ClusterSize; ++j) + cnt += (table[i].entry[j].genBound8 & 0xF8) == generation8; + + return cnt / ClusterSize; }