/*
Stockfish, a UCI chess playing engine derived from Glaurung 2.1
- Copyright (C) 2004-2008 Tord Romstad (Glaurung author)
- Copyright (C) 2008-2010 Marco Costalba, Joona Kiiski, Tord Romstad
+ Copyright (C) 2004-2021 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
along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
-#if !defined(_MSC_VER)
-
-# include <sys/time.h>
-# include <sys/types.h>
-# include <unistd.h>
-# if defined(__hpux)
-# include <sys/pstat.h>
-# endif
-
-#else
-
-#define _CRT_SECURE_NO_DEPRECATE
-#define NOMINMAX // disable macros min() and max()
-#include <windows.h>
-#include <sys/timeb.h>
+#ifdef _WIN32
+#if _WIN32_WINNT < 0x0601
+#undef _WIN32_WINNT
+#define _WIN32_WINNT 0x0601 // Force to include needed API prototypes
+#endif
+#ifndef NOMINMAX
+#define NOMINMAX
#endif
-#if !defined(NO_PREFETCH)
-# include <xmmintrin.h>
+#include <windows.h>
+// 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
+// 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);
+}
#endif
-#include <cassert>
-#include <cstdio>
+#include <fstream>
#include <iomanip>
#include <iostream>
#include <sstream>
-#include <algorithm>
+#include <vector>
+#include <cstdlib>
+
+#if defined(__linux__) && !defined(__ANDROID__)
+#include <stdlib.h>
+#include <sys/mman.h>
+#endif
+
+#if defined(__APPLE__) || defined(__ANDROID__) || defined(__OpenBSD__) || (defined(__GLIBCXX__) && !defined(_GLIBCXX_HAVE_ALIGNED_ALLOC) && !defined(_WIN32)) || defined(__e2k__)
+#define POSIXALIGNEDALLOC
+#include <stdlib.h>
+#endif
-#include "bitcount.h"
#include "misc.h"
#include "thread.h"
using namespace std;
-/// Version number. If EngineVersion is left empty, then AppTag plus
-/// current date (in the format YYMMDD) is used as a version number.
+namespace Stockfish {
-static const string AppName = "Stockfish";
-static const string EngineVersion = "";
-static const string AppTag = "";
+namespace {
+/// Version number. If Version is left empty, then compile date in the format
+/// DD-MM-YY and show in engine_info.
+const string Version = "";
-/// engine_name() returns the full name of the current Stockfish version.
-/// This will be either "Stockfish YYMMDD" (where YYMMDD is the date when
-/// the program was compiled) or "Stockfish <version number>", depending
-/// on whether the constant EngineVersion is empty.
+/// 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
+/// can toggle the logging of std::cout and std:cin at runtime whilst preserving
+/// usual I/O functionality, all without changing a single line of code!
+/// Idea from http://groups.google.com/group/comp.lang.c++/msg/1d941c0f26ea0d81
-const string engine_name() {
+struct Tie: public streambuf { // MSVC requires split streambuf for cin and cout
+
+ Tie(streambuf* b, streambuf* l) : buf(b), logBuf(l) {}
+
+ 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;
+
+ int log(int c, const char* prefix) {
+
+ static int last = '\n'; // Single log file
+
+ if (last == '\n')
+ logBuf->sputn(prefix, 3);
+
+ return last = logBuf->sputc((char)c);
+ }
+};
+
+class Logger {
+
+ Logger() : in(cin.rdbuf(), file.rdbuf()), out(cout.rdbuf(), file.rdbuf()) {}
+ ~Logger() { start(""); }
+
+ ofstream file;
+ Tie in, out;
+
+public:
+ static void start(const std::string& fname) {
+
+ static Logger l;
+
+ 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);
+ }
+ else if (fname.empty() && l.file.is_open())
+ {
+ cout.rdbuf(l.out.buf);
+ cin.rdbuf(l.in.buf);
+ l.file.close();
+ }
+ }
+};
+
+} // namespace
- const string months("Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec");
- const string cpu64(CpuIs64Bit ? " 64bit" : "");
- if (!EngineVersion.empty())
- return AppName + " " + EngineVersion + cpu64;
+/// engine_info() returns the full name of the current Stockfish version. This
+/// will be either "Stockfish <Tag> DD-MM-YY" (where DD-MM-YY is the date when
+/// the program was compiled) or "Stockfish <Version>", depending on whether
+/// Version is empty.
- stringstream s, date(__DATE__); // From compiler, format is "Sep 21 2008"
+string engine_info(bool to_uci) {
+
+ const string months("Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec");
string month, day, year;
+ stringstream ss, date(__DATE__); // From compiler, format is "Sep 21 2008"
+
+ ss << "Stockfish " << Version << setfill('0');
+
+ if (Version.empty())
+ {
+ date >> month >> day >> year;
+ ss << setw(2) << day << setw(2) << (1 + months.find(month) / 4) << year.substr(2);
+ }
- date >> month >> day >> year;
+ ss << (to_uci ? "\nid author ": " by ")
+ << "the Stockfish developers (see AUTHORS file)";
+
+ return ss.str();
+}
- s << setfill('0') << AppName + " " + AppTag + " "
- << year.substr(2, 2) << setw(2)
- << (1 + months.find(month) / 4) << setw(2)
- << day << cpu64;
- return s.str();
+/// compiler_info() returns a string trying to describe the compiler we use
+
+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 defined(__e2k__) && defined(__LCC__)
+ #define dot_ver2(n) \
+ compiler += (char)'.'; \
+ compiler += (char)('0' + (n) / 10); \
+ compiler += (char)('0' + (n) % 10);
+
+ compiler += "MCST LCC ";
+ compiler += "(version ";
+ compiler += std::to_string(__LCC__ / 100);
+ dot_ver2(__LCC__ % 100)
+ dot_ver2(__LCC_MINOR__)
+ 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 += "\nCompilation settings include: ";
+ compiler += (Is64Bit ? " 64bit" : " 32bit");
+ #if defined(USE_VNNI)
+ compiler += " VNNI";
+ #endif
+ #if defined(USE_AVX512)
+ compiler += " AVX512";
+ #endif
+ compiler += (HasPext ? " BMI2" : "");
+ #if defined(USE_AVX2)
+ compiler += " AVX2";
+ #endif
+ #if defined(USE_SSE41)
+ compiler += " SSE41";
+ #endif
+ #if defined(USE_SSSE3)
+ compiler += " SSSE3";
+ #endif
+ #if defined(USE_SSE2)
+ compiler += " SSE2";
+ #endif
+ compiler += (HasPopCnt ? " POPCNT" : "");
+ #if defined(USE_MMX)
+ compiler += " MMX";
+ #endif
+ #if defined(USE_NEON)
+ compiler += " NEON";
+ #endif
+
+ #if !defined(NDEBUG)
+ compiler += " DEBUG";
+ #endif
+
+ compiler += "\n__VERSION__ macro expands to: ";
+ #ifdef __VERSION__
+ compiler += __VERSION__;
+ #else
+ compiler += "(undefined macro)";
+ #endif
+ compiler += "\n";
+
+ return compiler;
}
-/// Our brave developers! Required by UCI
+/// Debug functions used mainly to collect run-time statistics
+static std::atomic<int64_t> hits[2], means[2];
-const string engine_authors() {
+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; }
- return "Tord Romstad, Marco Costalba and Joona Kiiski";
+void dbg_print() {
+
+ if (hits[0])
+ cerr << "Total " << hits[0] << " Hits " << hits[1]
+ << " hit rate (%) " << 100 * hits[1] / hits[0] << endl;
+
+ if (means[0])
+ cerr << "Total " << means[0] << " Mean "
+ << (double)means[1] / means[0] << endl;
}
-/// Debug stuff. Helper functions used mainly for debugging purposes
+/// Used to serialize access to std::cout to avoid multiple threads writing at
+/// the same time.
-static uint64_t dbg_hit_cnt0;
-static uint64_t dbg_hit_cnt1;
-static uint64_t dbg_mean_cnt0;
-static uint64_t dbg_mean_cnt1;
+std::ostream& operator<<(std::ostream& os, SyncCout sc) {
-void dbg_print_hit_rate() {
+ static std::mutex m;
- if (dbg_hit_cnt0)
- cerr << "Total " << dbg_hit_cnt0 << " Hit " << dbg_hit_cnt1
- << " hit rate (%) " << 100 * dbg_hit_cnt1 / dbg_hit_cnt0 << endl;
-}
+ if (sc == IO_LOCK)
+ m.lock();
-void dbg_print_mean() {
+ if (sc == IO_UNLOCK)
+ m.unlock();
- if (dbg_mean_cnt0)
- cerr << "Total " << dbg_mean_cnt0 << " Mean "
- << (float)dbg_mean_cnt1 / dbg_mean_cnt0 << endl;
+ return os;
}
-void dbg_mean_of(int v) {
- dbg_mean_cnt0++;
- dbg_mean_cnt1 += v;
-}
+/// Trampoline helper to avoid moving Logger to misc.h
+void start_logger(const std::string& fname) { Logger::start(fname); }
+
+
+/// prefetch() preloads the given address in L1/L2 cache. This is a non-blocking
+/// function that doesn't stall the CPU waiting for data to be loaded from memory,
+/// which can be quite slow.
+#ifdef NO_PREFETCH
+
+void prefetch(void*) {}
+
+#else
+
+void prefetch(void* addr) {
-void dbg_hit_on(bool b) {
+# if defined(__INTEL_COMPILER)
+ // This hack prevents prefetches from being optimized away by
+ // Intel compiler. Both MSVC and gcc seem not be affected by this.
+ __asm__ ("");
+# endif
- dbg_hit_cnt0++;
- if (b)
- dbg_hit_cnt1++;
+# if defined(__INTEL_COMPILER) || defined(_MSC_VER)
+ _mm_prefetch((char*)addr, _MM_HINT_T0);
+# else
+ __builtin_prefetch(addr);
+# endif
}
-void dbg_hit_on_c(bool c, bool b) { if (c) dbg_hit_on(b); }
-void dbg_before() { dbg_hit_on(false); }
-void dbg_after() { dbg_hit_on(true); dbg_hit_cnt0--; }
+#endif
+
+/// std_aligned_alloc() is our wrapper for systems where the c++17 implementation
+/// does not guarantee the availability of aligned_alloc(). Memory allocated with
+/// std_aligned_alloc() must be freed with std_aligned_free().
-/// get_system_time() returns the current system time, measured in milliseconds
+void* std_aligned_alloc(size_t alignment, size_t size) {
-int get_system_time() {
+#if defined(POSIXALIGNEDALLOC)
+ void *mem;
+ return posix_memalign(&mem, alignment, size) ? nullptr : mem;
+#elif defined(_WIN32)
+ return _mm_malloc(size, alignment);
+#else
+ return std::aligned_alloc(alignment, size);
+#endif
+}
+
+void std_aligned_free(void* ptr) {
-#if defined(_MSC_VER)
- struct _timeb t;
- _ftime(&t);
- return int(t.time * 1000 + t.millitm);
+#if defined(POSIXALIGNEDALLOC)
+ free(ptr);
+#elif defined(_WIN32)
+ _mm_free(ptr);
#else
- struct timeval t;
- gettimeofday(&t, NULL);
- return t.tv_sec * 1000 + t.tv_usec / 1000;
+ free(ptr);
#endif
}
+/// aligned_large_pages_alloc() will return suitably aligned memory, if possible using large pages.
+
+#if defined(_WIN32)
+
+static void* aligned_large_pages_alloc_windows(size_t allocSize) {
+
+ #if !defined(_WIN64)
+ (void)allocSize; // suppress unused-parameter compiler warning
+ return nullptr;
+ #else
+
+ HANDLE hProcessToken { };
+ LUID luid { };
+ void* mem = nullptr;
-/// cpu_count() tries to detect the number of CPU cores
+ const size_t largePageSize = GetLargePageMinimum();
+ if (!largePageSize)
+ return nullptr;
-int cpu_count() {
+ // We need SeLockMemoryPrivilege, so try to enable it for the process
+ if (!OpenProcessToken(GetCurrentProcess(), TOKEN_ADJUST_PRIVILEGES | TOKEN_QUERY, &hProcessToken))
+ return nullptr;
+
+ if (LookupPrivilegeValue(NULL, SE_LOCK_MEMORY_NAME, &luid))
+ {
+ TOKEN_PRIVILEGES tp { };
+ TOKEN_PRIVILEGES prevTp { };
+ DWORD prevTpLen = 0;
+
+ tp.PrivilegeCount = 1;
+ tp.Privileges[0].Luid = luid;
+ tp.Privileges[0].Attributes = SE_PRIVILEGE_ENABLED;
+
+ // 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(
+ 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);
+
+ // Privilege no longer needed, restore previous state
+ AdjustTokenPrivileges(hProcessToken, FALSE, &prevTp, 0, NULL, NULL);
+ }
+ }
+
+ CloseHandle(hProcessToken);
+
+ return mem;
+
+ #endif
+}
+
+void* aligned_large_pages_alloc(size_t allocSize) {
+
+ // Try to allocate large pages
+ void* mem = aligned_large_pages_alloc_windows(allocSize);
+
+ // Fall back to regular, page aligned, allocation if necessary
+ if (!mem)
+ mem = VirtualAlloc(NULL, allocSize, MEM_RESERVE | MEM_COMMIT, PAGE_READWRITE);
+
+ return mem;
+}
-#if defined(_MSC_VER)
- SYSTEM_INFO s;
- GetSystemInfo(&s);
- return std::min(int(s.dwNumberOfProcessors), MAX_THREADS);
#else
-# if defined(_SC_NPROCESSORS_ONLN)
- return std::min(sysconf(_SC_NPROCESSORS_ONLN), MAX_THREADS);
-# elif defined(__hpux)
- struct pst_dynamic psd;
- if (pstat_getdynamic(&psd, sizeof(psd), (size_t)1, 0) == -1)
- return 1;
- return std::min(psd.psd_proc_cnt, MAX_THREADS);
-# else
- return 1;
-# endif
+void* aligned_large_pages_alloc(size_t allocSize) {
+
+#if defined(__linux__)
+ constexpr size_t alignment = 2 * 1024 * 1024; // assumed 2MB page size
+#else
+ constexpr size_t alignment = 4096; // assumed small page size
+#endif
+ // round up to multiples of alignment
+ size_t size = ((allocSize + alignment - 1) / alignment) * alignment;
+ void *mem = std_aligned_alloc(alignment, size);
+#if defined(MADV_HUGEPAGE)
+ madvise(mem, size, MADV_HUGEPAGE);
#endif
+ return mem;
}
+#endif
-/// Check for console input. Original code from Beowulf, Olithink and Greko
-#ifndef _WIN32
+/// aligned_large_pages_free() will free the previously allocated ttmem
-int input_available() {
+#if defined(_WIN32)
- fd_set readfds;
- struct timeval timeout;
+void aligned_large_pages_free(void* mem) {
+
+ if (mem && !VirtualFree(mem, 0, MEM_RELEASE))
+ {
+ DWORD err = GetLastError();
+ std::cerr << "Failed to free large page memory. Error code: 0x"
+ << std::hex << err
+ << std::dec << std::endl;
+ exit(EXIT_FAILURE);
+ }
+}
- FD_ZERO(&readfds);
- FD_SET(fileno(stdin), &readfds);
- timeout.tv_sec = 0; // Set to timeout immediately
- timeout.tv_usec = 0;
- select(16, &readfds, 0, 0, &timeout);
+#else
- return (FD_ISSET(fileno(stdin), &readfds));
+void aligned_large_pages_free(void *mem) {
+ std_aligned_free(mem);
}
+#endif
+
+
+namespace WinProcGroup {
+
+#ifndef _WIN32
+
+void bindThisThread(size_t) {}
+
#else
-int input_available() {
+/// 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.
- static HANDLE inh = NULL;
- static bool usePipe = false;
- INPUT_RECORD rec[256];
- DWORD nchars, recCnt;
+int best_group(size_t idx) {
- if (!inh)
+ int threads = 0;
+ int nodes = 0;
+ int cores = 0;
+ DWORD returnLength = 0;
+ DWORD byteOffset = 0;
+
+ // Early exit if the needed API is not available at runtime
+ HMODULE k32 = GetModuleHandle("Kernel32.dll");
+ auto fun1 = (fun1_t)(void(*)())GetProcAddress(k32, "GetLogicalProcessorInformationEx");
+ if (!fun1)
+ return -1;
+
+ // First call to get returnLength. We expect it to fail due to null buffer
+ if (fun1(RelationAll, nullptr, &returnLength))
+ return -1;
+
+ // Once we know returnLength, allocate the buffer
+ SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX *buffer, *ptr;
+ ptr = buffer = (SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX*)malloc(returnLength);
+
+ // Second call, now we expect to succeed
+ if (!fun1(RelationAll, buffer, &returnLength))
{
- inh = GetStdHandle(STD_INPUT_HANDLE);
- if (GetConsoleMode(inh, &nchars))
+ free(buffer);
+ return -1;
+ }
+
+ while (byteOffset < returnLength)
+ {
+ if (ptr->Relationship == RelationNumaNode)
+ nodes++;
+
+ else if (ptr->Relationship == RelationProcessorCore)
{
- SetConsoleMode(inh, nchars & ~(ENABLE_MOUSE_INPUT | ENABLE_WINDOW_INPUT));
- FlushConsoleInputBuffer(inh);
- } else
- usePipe = true;
+ cores++;
+ 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);
}
- // When using Standard C input functions, also check if there
- // is anything in the buffer. After a call to such functions,
- // the input waiting in the pipe will be copied to the buffer,
- // and the call to PeekNamedPipe can indicate no input available.
- // Setting stdin to unbuffered was not enough. [from Greko]
- if (stdin->_cnt > 0)
- return 1;
-
- // When running under a GUI the input commands are sent to us
- // directly over the internal pipe. If PeekNamedPipe() returns 0
- // then something went wrong. Probably the parent program exited.
- // Returning 1 will make the next call to the input function
- // return EOF, where this should be catched then.
- if (usePipe)
- return PeekNamedPipe(inh, NULL, 0, NULL, &nchars, NULL) ? nchars : 1;
-
- // Count the number of unread input records, including keyboard,
- // mouse, and window-resizing input records.
- GetNumberOfConsoleInputEvents(inh, &nchars);
-
- // Read data from console without removing it from the buffer
- if (nchars <= 0 || !PeekConsoleInput(inh, rec, std::min(int(nchars), 256), &recCnt))
- return 0;
-
- // Search for at least one keyboard event
- for (DWORD i = 0; i < recCnt; i++)
- if (rec[i].EventType == KEY_EVENT)
- return 1;
-
- return 0;
+ free(buffer);
+
+ std::vector<int> groups;
+
+ // Run as many threads as possible on the same node until core limit is
+ // reached, then move on filling the next node.
+ for (int n = 0; n < nodes; n++)
+ for (int i = 0; i < cores / nodes; i++)
+ groups.push_back(n);
+
+ // In case a core has more than one logical processor (we assume 2) and we
+ // have still threads to allocate, then spread them evenly across available
+ // nodes.
+ for (int t = 0; t < threads - cores; t++)
+ groups.push_back(t % nodes);
+
+ // If we still have more threads than the total number of logical processors
+ // then return -1 and let the OS to decide what to do.
+ return idx < groups.size() ? groups[idx] : -1;
}
-#endif
+/// bindThisThread() set the group affinity of the current thread
+
+void bindThisThread(size_t idx) {
+
+ // Use only local variables to be thread-safe
+ int group = best_group(idx);
-/// prefetch() preloads the given address in L1/L2 cache. This is a non
-/// blocking function and do not stalls the CPU waiting for data to be
-/// loaded from memory, that can be quite slow.
-#if defined(NO_PREFETCH)
+ if (group == -1)
+ return;
-void prefetch(char*) {}
+ // Early exit if the needed API are not available at runtime
+ HMODULE k32 = GetModuleHandle("Kernel32.dll");
+ auto fun2 = (fun2_t)(void(*)())GetProcAddress(k32, "GetNumaNodeProcessorMaskEx");
+ auto fun3 = (fun3_t)(void(*)())GetProcAddress(k32, "SetThreadGroupAffinity");
+ if (!fun2 || !fun3)
+ return;
+
+ GROUP_AFFINITY affinity;
+ if (fun2(group, &affinity))
+ fun3(GetCurrentThread(), &affinity, nullptr);
+}
+
+#endif
+
+} // namespace WinProcGroup
+
+#ifdef _WIN32
+#include <direct.h>
+#define GETCWD _getcwd
#else
+#include <unistd.h>
+#define GETCWD getcwd
+#endif
-void prefetch(char* addr) {
+namespace CommandLine {
-#if defined(__INTEL_COMPILER) || defined(__ICL)
- // This hack prevents prefetches to be optimized away by
- // Intel compiler. Both MSVC and gcc seems not affected.
- __asm__ ("");
+string argv0; // path+name of the executable binary, as given by argv[0]
+string binaryDirectory; // path of the executable directory
+string workingDirectory; // path of the working directory
+
+void init(int argc, char* argv[]) {
+ (void)argc;
+ string pathSeparator;
+
+ // extract the path+name of the executable binary
+ argv0 = argv[0];
+
+#ifdef _WIN32
+ pathSeparator = "\\";
+ #ifdef _MSC_VER
+ // Under windows argv[0] may not have the extension. Also _get_pgmptr() had
+ // issues in some windows 10 versions, so check returned values carefully.
+ char* pgmptr = nullptr;
+ if (!_get_pgmptr(&pgmptr) && pgmptr != nullptr && *pgmptr)
+ argv0 = pgmptr;
+ #endif
+#else
+ pathSeparator = "/";
#endif
- _mm_prefetch(addr, _MM_HINT_T2);
- _mm_prefetch(addr+64, _MM_HINT_T2); // 64 bytes ahead
+ // extract the working directory
+ workingDirectory = "";
+ char buff[40000];
+ char* cwd = GETCWD(buff, 40000);
+ if (cwd)
+ workingDirectory = cwd;
+
+ // extract the binary directory path from argv0
+ binaryDirectory = argv0;
+ size_t pos = binaryDirectory.find_last_of("\\/");
+ if (pos == std::string::npos)
+ binaryDirectory = "." + pathSeparator;
+ else
+ binaryDirectory.resize(pos + 1);
+
+ // pattern replacement: "./" at the start of path is replaced by the working directory
+ if (binaryDirectory.find("." + pathSeparator) == 0)
+ binaryDirectory.replace(0, 1, workingDirectory);
}
-#endif
+
+} // namespace CommandLine
+
+} // namespace Stockfish
projects / stockfish / blobdiff