X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=src%2Fmisc.cpp;h=e36a04bccfbef712f07229e78417b252a68625ea;hb=66bf45b99e2061c1ba74f9975bc5059ac0121dfd;hp=c7fa0e9ae104927867675cf4dbc2698df6cb2bd3;hpb=e8caa6640df15b1823d5d4b94e759d52923769e6;p=stockfish diff --git a/src/misc.cpp b/src/misc.cpp index c7fa0e9a..e36a04bc 100644 --- a/src/misc.cpp +++ b/src/misc.cpp @@ -1,6 +1,6 @@ /* Stockfish, a UCI chess playing engine derived from Glaurung 2.1 - Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file) + Copyright (C) 2004-2023 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 @@ -32,21 +32,26 @@ // the calls at compile time), try to load them at runtime. To do this we need // first to define the corresponding function pointers. extern "C" { -typedef bool(*fun1_t)(LOGICAL_PROCESSOR_RELATIONSHIP, - PSYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX, PDWORD); -typedef bool(*fun2_t)(USHORT, PGROUP_AFFINITY); -typedef bool(*fun3_t)(HANDLE, CONST GROUP_AFFINITY*, PGROUP_AFFINITY); -typedef bool(*fun4_t)(USHORT, PGROUP_AFFINITY, USHORT, PUSHORT); -typedef WORD(*fun5_t)(); +using fun1_t = bool(*)(LOGICAL_PROCESSOR_RELATIONSHIP, + PSYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX, PDWORD); +using fun2_t = bool(*)(USHORT, PGROUP_AFFINITY); +using fun3_t = bool(*)(HANDLE, CONST GROUP_AFFINITY*, PGROUP_AFFINITY); +using fun4_t = bool(*)(USHORT, PGROUP_AFFINITY, USHORT, PUSHORT); +using fun5_t = WORD(*)(); +using fun6_t = bool(*)(HANDLE, DWORD, PHANDLE); +using fun7_t = bool(*)(LPCSTR, LPCSTR, PLUID); +using fun8_t = bool(*)(HANDLE, BOOL, PTOKEN_PRIVILEGES, DWORD, PTOKEN_PRIVILEGES, PDWORD); } #endif +#include +#include #include #include #include #include +#include #include -#include #if defined(__linux__) && !defined(__ANDROID__) #include @@ -68,7 +73,7 @@ namespace Stockfish { namespace { /// Version number or dev. -const string version = "dev"; +constexpr string_view version = "dev"; /// Our fancy logging facility. The trick here is to replace cin.rdbuf() and /// cout.rdbuf() with two Tie objects that tie cin and cout to a file stream. We @@ -151,13 +156,13 @@ string engine_info(bool to_uci) { stringstream ss; ss << "Stockfish " << version << setfill('0'); - if (version == "dev") + if constexpr (version == "dev") { ss << "-"; #ifdef GIT_DATE - ss << GIT_DATE; + ss << stringify(GIT_DATE); #else - const string months("Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec"); + constexpr string_view months("Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec"); string month, day, year; stringstream date(__DATE__); // From compiler, format is "Sep 21 2008" @@ -168,7 +173,7 @@ string engine_info(bool to_uci) { ss << "-"; #ifdef GIT_SHA - ss << GIT_SHA; + ss << stringify(GIT_SHA); #else ss << "nogit"; #endif @@ -185,8 +190,6 @@ string engine_info(bool to_uci) { std::string compiler_info() { - #define stringify2(x) #x - #define stringify(x) stringify2(x) #define make_version_string(major, minor, patch) stringify(major) "." stringify(minor) "." stringify(patch) /// Predefined macros hell: @@ -298,21 +301,94 @@ std::string compiler_info() { /// Debug functions used mainly to collect run-time statistics -static std::atomic hits[2], means[2]; +constexpr int MaxDebugSlots = 32; + +namespace { + +template +struct DebugInfo { + std::atomic data[N] = { 0 }; -void dbg_hit_on(bool b) { ++hits[0]; if (b) ++hits[1]; } -void dbg_hit_on(bool c, bool b) { if (c) dbg_hit_on(b); } -void dbg_mean_of(int v) { ++means[0]; means[1] += v; } + constexpr inline std::atomic& operator[](int index) { return data[index]; } +}; + +DebugInfo<2> hit[MaxDebugSlots]; +DebugInfo<2> mean[MaxDebugSlots]; +DebugInfo<3> stdev[MaxDebugSlots]; +DebugInfo<6> correl[MaxDebugSlots]; + +} // namespace + +void dbg_hit_on(bool cond, int slot) { + + ++hit[slot][0]; + if (cond) + ++hit[slot][1]; +} + +void dbg_mean_of(int64_t value, int slot) { + + ++mean[slot][0]; + mean[slot][1] += value; +} + +void dbg_stdev_of(int64_t value, int slot) { + + ++stdev[slot][0]; + stdev[slot][1] += value; + stdev[slot][2] += value * value; +} + +void dbg_correl_of(int64_t value1, int64_t value2, int slot) { + + ++correl[slot][0]; + correl[slot][1] += value1; + correl[slot][2] += value1 * value1; + correl[slot][3] += value2; + correl[slot][4] += value2 * value2; + correl[slot][5] += value1 * value2; +} void dbg_print() { - if (hits[0]) - cerr << "Total " << hits[0] << " Hits " << hits[1] - << " hit rate (%) " << 100 * hits[1] / hits[0] << endl; + int64_t n; + auto E = [&n](int64_t x) { return double(x) / n; }; + auto sqr = [](double x) { return x * x; }; + + for (int i = 0; i < MaxDebugSlots; ++i) + if ((n = hit[i][0])) + std::cerr << "Hit #" << i + << ": Total " << n << " Hits " << hit[i][1] + << " Hit Rate (%) " << 100.0 * E(hit[i][1]) + << std::endl; + + for (int i = 0; i < MaxDebugSlots; ++i) + if ((n = mean[i][0])) + { + std::cerr << "Mean #" << i + << ": Total " << n << " Mean " << E(mean[i][1]) + << std::endl; + } - if (means[0]) - cerr << "Total " << means[0] << " Mean " - << (double)means[1] / means[0] << endl; + for (int i = 0; i < MaxDebugSlots; ++i) + if ((n = stdev[i][0])) + { + double r = sqrtl(E(stdev[i][2]) - sqr(E(stdev[i][1]))); + std::cerr << "Stdev #" << i + << ": Total " << n << " Stdev " << r + << std::endl; + } + + for (int i = 0; i < MaxDebugSlots; ++i) + if ((n = correl[i][0])) + { + double r = (E(correl[i][5]) - E(correl[i][1]) * E(correl[i][3])) + / ( sqrtl(E(correl[i][2]) - sqr(E(correl[i][1]))) + * sqrtl(E(correl[i][4]) - sqr(E(correl[i][3])))); + std::cerr << "Correl. #" << i + << ": Total " << n << " Coefficient " << r + << std::endl; + } } @@ -373,8 +449,10 @@ void* std_aligned_alloc(size_t alignment, size_t size) { #if defined(POSIXALIGNEDALLOC) void *mem; return posix_memalign(&mem, alignment, size) ? nullptr : mem; -#elif defined(_WIN32) +#elif defined(_WIN32) && !defined(_M_ARM) && !defined(_M_ARM64) return _mm_malloc(size, alignment); +#elif defined(_WIN32) + return _aligned_malloc(size, alignment); #else return std::aligned_alloc(alignment, size); #endif @@ -384,8 +462,10 @@ void std_aligned_free(void* ptr) { #if defined(POSIXALIGNEDALLOC) free(ptr); -#elif defined(_WIN32) +#elif defined(_WIN32) && !defined(_M_ARM) && !defined(_M_ARM64) _mm_free(ptr); +#elif defined(_WIN32) + _aligned_free(ptr); #else free(ptr); #endif @@ -409,11 +489,26 @@ static void* aligned_large_pages_alloc_windows([[maybe_unused]] size_t allocSize if (!largePageSize) return nullptr; + // Dynamically link OpenProcessToken, LookupPrivilegeValue and AdjustTokenPrivileges + HMODULE k32 = GetModuleHandle("Advapi32.dll"); + auto fun6 = (fun6_t)(void(*)())GetProcAddress(k32, "OpenProcessToken"); + if (!fun6) + return nullptr; + auto fun7 = (fun7_t)(void(*)())GetProcAddress(k32, "LookupPrivilegeValueA"); + if (!fun7) + return nullptr; + auto fun8 = (fun8_t)(void(*)())GetProcAddress(k32, "AdjustTokenPrivileges"); + if (!fun8) + return nullptr; + + // We need SeLockMemoryPrivilege, so try to enable it for the process - if (!OpenProcessToken(GetCurrentProcess(), TOKEN_ADJUST_PRIVILEGES | TOKEN_QUERY, &hProcessToken)) + // OpenProcessToken() + if (!fun6(GetCurrentProcess(), TOKEN_ADJUST_PRIVILEGES | TOKEN_QUERY, &hProcessToken)) return nullptr; - if (LookupPrivilegeValue(NULL, SE_LOCK_MEMORY_NAME, &luid)) + // LookupPrivilegeValueA() + if (fun7(nullptr, SE_LOCK_MEMORY_NAME, &luid)) { TOKEN_PRIVILEGES tp { }; TOKEN_PRIVILEGES prevTp { }; @@ -425,17 +520,19 @@ static void* aligned_large_pages_alloc_windows([[maybe_unused]] size_t allocSize // Try to enable SeLockMemoryPrivilege. Note that even if AdjustTokenPrivileges() succeeds, // we still need to query GetLastError() to ensure that the privileges were actually obtained. - if (AdjustTokenPrivileges( + // AdjustTokenPrivileges() + if (fun8( hProcessToken, FALSE, &tp, sizeof(TOKEN_PRIVILEGES), &prevTp, &prevTpLen) && GetLastError() == ERROR_SUCCESS) { // Round up size to full pages and allocate allocSize = (allocSize + largePageSize - 1) & ~size_t(largePageSize - 1); mem = VirtualAlloc( - NULL, allocSize, MEM_RESERVE | MEM_COMMIT | MEM_LARGE_PAGES, PAGE_READWRITE); + nullptr, allocSize, MEM_RESERVE | MEM_COMMIT | MEM_LARGE_PAGES, PAGE_READWRITE); // Privilege no longer needed, restore previous state - AdjustTokenPrivileges(hProcessToken, FALSE, &prevTp, 0, NULL, NULL); + // AdjustTokenPrivileges () + fun8(hProcessToken, FALSE, &prevTp, 0, nullptr, nullptr); } } @@ -453,7 +550,7 @@ void* aligned_large_pages_alloc(size_t allocSize) { // Fall back to regular, page aligned, allocation if necessary if (!mem) - mem = VirtualAlloc(NULL, allocSize, MEM_RESERVE | MEM_COMMIT, PAGE_READWRITE); + mem = VirtualAlloc(nullptr, allocSize, MEM_RESERVE | MEM_COMMIT, PAGE_READWRITE); return mem; } @@ -517,7 +614,7 @@ void bindThisThread(size_t) {} /// API and returns the best node id for the thread with index idx. Original /// code from Texel by Peter Österlund. -int best_node(size_t idx) { +static int best_node(size_t idx) { int threads = 0; int nodes = 0; @@ -526,7 +623,7 @@ int best_node(size_t idx) { DWORD byteOffset = 0; // Early exit if the needed API is not available at runtime - HMODULE k32 = GetModuleHandle("Kernel32.dll"); + HMODULE k32 = GetModuleHandle(TEXT("Kernel32.dll")); auto fun1 = (fun1_t)(void(*)())GetProcAddress(k32, "GetLogicalProcessorInformationEx"); if (!fun1) return -1; @@ -596,7 +693,7 @@ void bindThisThread(size_t idx) { return; // Early exit if the needed API are not available at runtime - HMODULE k32 = GetModuleHandle("Kernel32.dll"); + HMODULE k32 = GetModuleHandle(TEXT("Kernel32.dll")); auto fun2 = (fun2_t)(void(*)())GetProcAddress(k32, "GetNumaNodeProcessorMaskEx"); auto fun3 = (fun3_t)(void(*)())GetProcAddress(k32, "SetThreadGroupAffinity"); auto fun4 = (fun4_t)(void(*)())GetProcAddress(k32, "GetNumaNodeProcessorMask2");