X-Git-Url: https://git.sesse.net/?p=stockfish;a=blobdiff_plain;f=src%2Ftt.cpp;h=6ee63138d15497851843ce114108cddba262151e;hp=26791809f8890bb6aed2e492838b9a9473b2aaf2;hb=d4763424d2728fe2dfd0a6fe747666feb6a2fdbb;hpb=e7cfa5d020efb5a0ad2521afc7b886f3b2d3e6b3 diff --git a/src/tt.cpp b/src/tt.cpp index 26791809..6ee63138 100644 --- a/src/tt.cpp +++ b/src/tt.cpp @@ -2,7 +2,7 @@ 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) 2015-2020 Marco Costalba, Joona Kiiski, Gary Linscott, Tord Romstad Stockfish is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by @@ -24,15 +24,16 @@ #include "bitboard.h" #include "misc.h" +#include "thread.h" #include "tt.h" #include "uci.h" TranspositionTable TT; // Our global transposition table -/// TTEntry::save saves a TTEntry -void TTEntry::save(Key k, Value v, Bound b, Depth d, Move m, Value ev) { +/// TTEntry::save populates the TTEntry with a new node's data, possibly +/// overwriting an old position. Update is not atomic and can be racy. - assert(d / ONE_PLY * ONE_PLY == d); +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 || (k >> 48) != key16) @@ -40,14 +41,16 @@ void TTEntry::save(Key k, Value v, Bound b, Depth d, Move m, Value ev) { // Overwrite less valuable entries if ( (k >> 48) != key16 - || d / ONE_PLY > depth8 - 4 + || d - DEPTH_OFFSET > depth8 - 4 || b == BOUND_EXACT) { + assert(d >= DEPTH_OFFSET); + key16 = (uint16_t)(k >> 48); value16 = (int16_t)v; eval16 = (int16_t)ev; - genBound8 = (uint8_t)(TT.generation8 | b); - depth8 = (int8_t)(d / ONE_PLY); + genBound8 = (uint8_t)(TT.generation8 | uint8_t(pv) << 2 | b); + depth8 = (uint8_t)(d - DEPTH_OFFSET); } } @@ -58,11 +61,19 @@ void TTEntry::save(Key k, Value v, Bound b, Depth d, Move m, Value ev) { void TranspositionTable::resize(size_t mbSize) { - clusterCount = mbSize * 1024 * 1024 / sizeof(Cluster); + Threads.main()->wait_for_search_finished(); + + if (mem) + aligned_ttmem_free(mem); - free(mem); - mem = malloc(clusterCount * sizeof(Cluster) + CacheLineSize - 1); + if (!mbSize) + { + mem = nullptr; + return; + } + clusterCount = mbSize * 1024 * 1024 / sizeof(Cluster); + table = static_cast(aligned_ttmem_alloc(clusterCount * sizeof(Cluster), mem)); if (!mem) { std::cerr << "Failed to allocate " << mbSize @@ -70,7 +81,6 @@ void TranspositionTable::resize(size_t mbSize) { exit(EXIT_FAILURE); } - table = (Cluster*)((uintptr_t(mem) + CacheLineSize - 1) & ~(CacheLineSize - 1)); clear(); } @@ -82,12 +92,12 @@ 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 @@ -97,7 +107,7 @@ void TranspositionTable::clear() { stride : clusterCount - start; std::memset(&table[start], 0, len * sizeof(Cluster)); - })); + }); } for (std::thread& th: threads) @@ -119,8 +129,7 @@ TTEntry* TranspositionTable::probe(const Key key, bool& found) const { 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]; } @@ -129,11 +138,11 @@ 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; @@ -146,12 +155,9 @@ 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; }