X-Git-Url: https://git.sesse.net/?p=stockfish;a=blobdiff_plain;f=src%2Fmisc.cpp;h=cf18d2e2fc56b826f3e627d712f02d8ff96e9f1d;hp=42dc0b06d59eb16f3119e4078f8647918fe9598c;hb=3b70932b0dee0cf1817baf0daa43ac92e18003c4;hpb=b48439e90643cb6f65f9e34d1421976883c12efc diff --git a/src/misc.cpp b/src/misc.cpp index 42dc0b06..cf18d2e2 100644 --- a/src/misc.cpp +++ b/src/misc.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) 2015-2017 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 @@ -23,6 +23,11 @@ #undef _WIN32_WINNT #define _WIN32_WINNT 0x0601 // Force to include needed API prototypes #endif + +#ifndef NOMINMAX +#define NOMINMAX +#endif + #include // The needed Windows API for processor groups could be missed from old Windows // versions, so instead of calling them directly (forcing the linker to resolve @@ -42,6 +47,11 @@ typedef bool(*fun3_t)(HANDLE, CONST GROUP_AFFINITY*, PGROUP_AFFINITY); #include #include +#if defined(__linux__) && !defined(__ANDROID__) +#include +#include +#endif + #include "misc.h" #include "thread.h" @@ -63,10 +73,10 @@ struct Tie: public streambuf { // MSVC requires split streambuf for cin and cout Tie(streambuf* b, streambuf* l) : buf(b), logBuf(l) {} - int sync() { return logBuf->pubsync(), buf->pubsync(); } - int overflow(int c) { return log(buf->sputc((char)c), "<< "); } - int underflow() { return buf->sgetc(); } - int uflow() { return log(buf->sbumpc(), ">> "); } + int sync() override { return logBuf->pubsync(), buf->pubsync(); } + int overflow(int c) override { return log(buf->sputc((char)c), "<< "); } + int underflow() override { return buf->sgetc(); } + int uflow() override { return log(buf->sbumpc(), ">> "); } streambuf *buf, *logBuf; @@ -97,6 +107,13 @@ public: if (!fname.empty() && !l.file.is_open()) { l.file.open(fname, ifstream::out); + + if (!l.file.is_open()) + { + cerr << "Unable to open debug log file " << fname << endl; + exit(EXIT_FAILURE); + } + cin.rdbuf(&l.in); cout.rdbuf(&l.out); } @@ -139,8 +156,79 @@ const string engine_info(bool to_uci) { } +/// compiler_info() returns a string trying to describe the compiler we use + +const 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: +/// +/// __GNUC__ Compiler is gcc, Clang or Intel on Linux +/// __INTEL_COMPILER Compiler is Intel +/// _MSC_VER Compiler is MSVC or Intel on Windows +/// _WIN32 Building on Windows (any) +/// _WIN64 Building on Windows 64 bit + + std::string compiler = "\nCompiled by "; + + #ifdef __clang__ + compiler += "clang++ "; + compiler += make_version_string(__clang_major__, __clang_minor__, __clang_patchlevel__); + #elif __INTEL_COMPILER + compiler += "Intel compiler "; + compiler += "(version "; + compiler += stringify(__INTEL_COMPILER) " update " stringify(__INTEL_COMPILER_UPDATE); + compiler += ")"; + #elif _MSC_VER + compiler += "MSVC "; + compiler += "(version "; + compiler += stringify(_MSC_FULL_VER) "." stringify(_MSC_BUILD); + compiler += ")"; + #elif __GNUC__ + compiler += "g++ (GNUC) "; + compiler += make_version_string(__GNUC__, __GNUC_MINOR__, __GNUC_PATCHLEVEL__); + #else + compiler += "Unknown compiler "; + compiler += "(unknown version)"; + #endif + + #if defined(__APPLE__) + compiler += " on Apple"; + #elif defined(__CYGWIN__) + compiler += " on Cygwin"; + #elif defined(__MINGW64__) + compiler += " on MinGW64"; + #elif defined(__MINGW32__) + compiler += " on MinGW32"; + #elif defined(__ANDROID__) + compiler += " on Android"; + #elif defined(__linux__) + compiler += " on Linux"; + #elif defined(_WIN64) + compiler += " on Microsoft Windows 64-bit"; + #elif defined(_WIN32) + compiler += " on Microsoft Windows 32-bit"; + #else + compiler += " on unknown system"; + #endif + + compiler += "\n __VERSION__ macro expands to: "; + #ifdef __VERSION__ + compiler += __VERSION__; + #else + compiler += "(undefined macro)"; + #endif + compiler += "\n"; + + return compiler; +} + + /// Debug functions used mainly to collect run-time statistics -static int64_t hits[2], means[2]; +static std::atomic hits[2], means[2]; 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); } @@ -163,7 +251,7 @@ void dbg_print() { std::ostream& operator<<(std::ostream& os, SyncCout sc) { - static Mutex m; + static std::mutex m; if (sc == IO_LOCK) m.lock(); @@ -205,12 +293,35 @@ void prefetch(void* addr) { #endif -void prefetch2(void* addr) { - prefetch(addr); - prefetch((uint8_t*)addr + 64); +/// aligned_ttmem_alloc will return suitably aligned memory, and if possible use large pages. +/// The returned pointer is the aligned one, while the mem argument is the one that needs to be passed to free. +/// With c++17 some of this functionality can be simplified. +#if defined(__linux__) && !defined(__ANDROID__) + +void* aligned_ttmem_alloc(size_t allocSize, void** mem) { + + constexpr size_t alignment = 2 * 1024 * 1024; // assumed 2MB page sizes + size_t size = ((allocSize + alignment - 1) / alignment) * alignment; // multiple of alignment + *mem = aligned_alloc(alignment, size); + madvise(*mem, allocSize, MADV_HUGEPAGE); + return *mem; +} + +#else + +void* aligned_ttmem_alloc(size_t allocSize, void** mem) { + + constexpr size_t alignment = 64; // assumed cache line size + size_t size = allocSize + alignment - 1; // allocate some extra space + *mem = malloc(size); + void* ret = reinterpret_cast((uintptr_t(*mem) + alignment - 1) & ~uintptr_t(alignment - 1)); + return ret; } +#endif + + namespace WinProcGroup { #ifndef _WIN32 @@ -219,11 +330,11 @@ void bindThisThread(size_t) {} #else -/// get_group() retrieves logical processor information using Windows specific +/// best_group() retrieves logical processor information using Windows specific /// API and returns the best group id for the thread with index idx. Original /// code from Texel by Peter Österlund. -int get_group(size_t idx) { +int best_group(size_t idx) { int threads = 0; int nodes = 0; @@ -233,7 +344,7 @@ int get_group(size_t idx) { // Early exit if the needed API is not available at runtime HMODULE k32 = GetModuleHandle("Kernel32.dll"); - auto fun1 = (fun1_t)GetProcAddress(k32, "GetLogicalProcessorInformationEx"); + auto fun1 = (fun1_t)(void(*)())GetProcAddress(k32, "GetLogicalProcessorInformationEx"); if (!fun1) return -1; @@ -252,7 +363,7 @@ int get_group(size_t idx) { return -1; } - while (ptr->Size > 0 && byteOffset + ptr->Size <= returnLength) + while (byteOffset < returnLength) { if (ptr->Relationship == RelationNumaNode) nodes++; @@ -263,6 +374,7 @@ int get_group(size_t idx) { threads += (ptr->Processor.Flags == LTP_PC_SMT) ? 2 : 1; } + assert(ptr->Size); byteOffset += ptr->Size; ptr = (SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX*)(((char*)ptr) + ptr->Size); } @@ -293,24 +405,16 @@ int get_group(size_t idx) { void bindThisThread(size_t idx) { - // If OS already scheduled us on a different group than 0 then don't overwrite - // the choice, eventually we are one of many one-threaded processes running on - // some Windows NUMA hardware, for instance in fishtest. To make it simple, - // just check if running threads are below a threshold, in this case all this - // NUMA machinery is not needed. - if (Threads.size() < 8) - return; - // Use only local variables to be thread-safe - int group = get_group(idx); + int group = best_group(idx); if (group == -1) return; // Early exit if the needed API are not available at runtime HMODULE k32 = GetModuleHandle("Kernel32.dll"); - auto fun2 = (fun2_t)GetProcAddress(k32, "GetNumaNodeProcessorMaskEx"); - auto fun3 = (fun3_t)GetProcAddress(k32, "SetThreadGroupAffinity"); + auto fun2 = (fun2_t)(void(*)())GetProcAddress(k32, "GetNumaNodeProcessorMaskEx"); + auto fun3 = (fun3_t)(void(*)())GetProcAddress(k32, "SetThreadGroupAffinity"); if (!fun2 || !fun3) return;