/*
Stockfish, a UCI chess playing engine derived from Glaurung 2.1
Copyright (C) 2004-2008 Tord Romstad (Glaurung author)
- Copyright (C) 2008-2014 Marco Costalba, Joona Kiiski, Tord Romstad
+ Copyright (C) 2008-2015 Marco Costalba, Joona Kiiski, Tord Romstad
+ Copyright (C) 2015-2019 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
along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
+#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
+
+#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 <fstream>
#include <iomanip>
#include <iostream>
#include <sstream>
+#include <vector>
#include "misc.h"
#include "thread.h"
using namespace std;
+namespace {
+
/// Version number. If Version is left empty, then compile date in the format
/// DD-MM-YY and show in engine_info.
-static const string Version = "";
+const string Version = "";
+
+/// 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
+
+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);
+ 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
/// 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
}
ss << (Is64Bit ? " 64" : "")
- << (HasPext ? " BMI2" : (HasPopCnt ? " SSE4.2" : ""))
+ << (HasPext ? " BMI2" : (HasPopCnt ? " POPCNT" : ""))
<< (to_uci ? "\nid author ": " by ")
- << "Tord Romstad, Marco Costalba and Joona Kiiski";
+ << "T. Romstad, M. Costalba, J. Kiiski, G. Linscott";
return ss.str();
}
/// Debug functions used mainly to collect run-time statistics
-
static int64_t hits[2], means[2];
void dbg_hit_on(bool b) { ++hits[0]; if (b) ++hits[1]; }
-void dbg_hit_on_c(bool c, bool b) { if (c) dbg_hit_on(b); }
+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; }
void dbg_print() {
}
-/// 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
-
-struct Tie: public streambuf { // MSVC requires splitted streambuf for cin and cout
-
- Tie(streambuf* b, ofstream* f) : buf(b), file(f) {}
-
- int sync() { return file->rdbuf()->pubsync(), buf->pubsync(); }
- int overflow(int c) { return log(buf->sputc((char)c), "<< "); }
- int underflow() { return buf->sgetc(); }
- int uflow() { return log(buf->sbumpc(), ">> "); }
-
- streambuf* buf;
- ofstream* file;
-
- int log(int c, const char* prefix) {
-
- static int last = '\n';
-
- if (last == '\n')
- file->rdbuf()->sputn(prefix, 3);
-
- return last = file->rdbuf()->sputc((char)c);
- }
-};
-
-class Logger {
-
- Logger() : in(cin.rdbuf(), &file), out(cout.rdbuf(), &file) {}
- ~Logger() { start(false); }
-
- ofstream file;
- Tie in, out;
-
-public:
- static void start(bool b) {
-
- static Logger l;
-
- if (b && !l.file.is_open())
- {
- l.file.open("io_log.txt", ifstream::out | ifstream::app);
- cin.rdbuf(&l.in);
- cout.rdbuf(&l.out);
- }
- else if (!b && l.file.is_open())
- {
- cout.rdbuf(l.out.buf);
- cin.rdbuf(l.in.buf);
- l.file.close();
- }
- }
-};
-
-
/// Used to serialize access to std::cout to avoid multiple threads writing at
/// the same time.
/// Trampoline helper to avoid moving Logger to misc.h
-void start_logger(bool b) { Logger::start(b); }
-
-
-/// timed_wait() waits for msec milliseconds. It is mainly a helper to wrap
-/// the conversion from milliseconds to struct timespec, as used by pthreads.
-
-void timed_wait(WaitCondition& sleepCond, Lock& sleepLock, int msec) {
-
-#ifdef _WIN32
- int tm = msec;
-#else
- timespec ts, *tm = &ts;
- uint64_t ms = Time::now() + msec;
-
- ts.tv_sec = ms / 1000;
- ts.tv_nsec = (ms % 1000) * 1000000LL;
-#endif
-
- cond_timedwait(sleepCond, sleepLock, tm);
-}
+void start_logger(const std::string& fname) { Logger::start(fname); }
/// prefetch() preloads the given address in L1/L2 cache. This is a non-blocking
/// which can be quite slow.
#ifdef NO_PREFETCH
-void prefetch(char*) {}
+void prefetch(void*) {}
#else
-void prefetch(char* addr) {
+void prefetch(void* addr) {
# if defined(__INTEL_COMPILER)
// This hack prevents prefetches from being optimized away by
# endif
# if defined(__INTEL_COMPILER) || defined(_MSC_VER)
- _mm_prefetch(addr, _MM_HINT_T0);
+ _mm_prefetch((char*)addr, _MM_HINT_T0);
# else
__builtin_prefetch(addr);
# endif
}
#endif
+
+void prefetch2(void* addr) {
+
+ prefetch(addr);
+ prefetch((uint8_t*)addr + 64);
+}
+
+namespace WinProcGroup {
+
+#ifndef _WIN32
+
+void bindThisThread(size_t) {}
+
+#else
+
+/// 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 Ă–sterlund.
+
+int best_group(size_t idx) {
+
+ 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))
+ {
+ free(buffer);
+ return -1;
+ }
+
+ while (ptr->Size > 0 && byteOffset + ptr->Size <= returnLength)
+ {
+ if (ptr->Relationship == RelationNumaNode)
+ nodes++;
+
+ else if (ptr->Relationship == RelationProcessorCore)
+ {
+ cores++;
+ threads += (ptr->Processor.Flags == LTP_PC_SMT) ? 2 : 1;
+ }
+
+ byteOffset += ptr->Size;
+ ptr = (SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX*)(((char*)ptr) + ptr->Size);
+ }
+
+ 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;
+}
+
+
+/// 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);
+
+ if (group == -1)
+ return;
+
+ // 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