/*
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) 2004-2024 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
#define THREAD_H_INCLUDED
#include <atomic>
-#include <bitset>
#include <condition_variable>
+#include <cstddef>
+#include <cstdint>
+#include <functional>
+#include <memory>
#include <mutex>
-#include <thread>
#include <vector>
-#include "material.h"
-#include "movepick.h"
-#include "pawns.h"
+#include "numa.h"
#include "position.h"
#include "search.h"
-#include "thread_win32.h"
-
-struct Thread;
-
-const size_t MAX_THREADS = 128;
-const size_t MAX_SPLITPOINTS_PER_THREAD = 8;
-const size_t MAX_SLAVES_PER_SPLITPOINT = 4;
-
-
-/// Spinlock class wraps low level atomic operations to provide a spin lock
-
-class Spinlock {
-
- Mutex m; // WARNING: Diasabled spinlocks to test on fishtest
-
-public:
- void acquire() { m.lock(); }
- void release() { m.unlock(); }
-};
-
-
-/// SplitPoint struct stores information shared by the threads searching in
-/// parallel below the same split point. It is populated at splitting time.
-
-struct SplitPoint {
-
- // Const data after split point has been setup
- const Position* pos;
- Search::Stack* ss;
- Thread* master;
- Depth depth;
- Value beta;
- int nodeType;
- bool cutNode;
-
- // Const pointers to shared data
- MovePicker* movePicker;
- SplitPoint* parentSplitPoint;
-
- // Shared variable data
- Spinlock spinlock;
- std::bitset<MAX_THREADS> slavesMask;
- volatile bool allSlavesSearching;
- volatile uint64_t nodes;
- volatile Value alpha;
- volatile Value bestValue;
- volatile Move bestMove;
- volatile int moveCount;
- volatile bool cutoff;
-};
-
-
-/// ThreadBase struct is the base of the hierarchy from where we derive all the
-/// specialized thread classes.
-
-struct ThreadBase {
-
- virtual ~ThreadBase() = default;
- virtual void idle_loop() = 0;
- void notify_one();
- void wait_for(volatile const bool& b);
-
- std::thread nativeThread;
- Mutex mutex;
- ConditionVariable sleepCondition;
- volatile bool exit = false;
-};
-
-
-/// Thread struct keeps together all the thread related stuff like locks, state
-/// and especially split points. We also use per-thread pawn and material hash
-/// tables so that once we get a pointer to an entry its life time is unlimited
-/// and we don't have to care about someone changing the entry under our feet.
-
-struct Thread : public ThreadBase {
-
- Thread();
- virtual void idle_loop();
- bool cutoff_occurred() const;
- bool can_join(const SplitPoint* sp) const;
-
- void split(Position& pos, Search::Stack* ss, Value alpha, Value beta, Value* bestValue, Move* bestMove,
- Depth depth, int moveCount, MovePicker* movePicker, int nodeType, bool cutNode);
-
- SplitPoint splitPoints[MAX_SPLITPOINTS_PER_THREAD];
- Pawns::Table pawnsTable;
- Material::Table materialTable;
- Endgames endgames;
- Position* activePosition;
- size_t idx;
- int maxPly;
- SplitPoint* volatile activeSplitPoint;
- volatile size_t splitPointsSize;
- volatile bool searching;
+#include "thread_win32_osx.h"
+
+namespace Stockfish {
+
+
+class OptionsMap;
+using Value = int;
+
+// Sometimes we don't want to actually bind the threads, but the recipient still
+// needs to think it runs on *some* NUMA node, such that it can access structures
+// that rely on NUMA node knowledge. This class encapsulates this optional process
+// such that the recipient does not need to know whether the binding happened or not.
+class OptionalThreadToNumaNodeBinder {
+ public:
+ OptionalThreadToNumaNodeBinder(NumaIndex n) :
+ numaConfig(nullptr),
+ numaId(n) {}
+
+ OptionalThreadToNumaNodeBinder(const NumaConfig& cfg, NumaIndex n) :
+ numaConfig(&cfg),
+ numaId(n) {}
+
+ NumaReplicatedAccessToken operator()() const {
+ if (numaConfig != nullptr)
+ return numaConfig->bind_current_thread_to_numa_node(numaId);
+ else
+ return NumaReplicatedAccessToken(numaId);
+ }
+
+ private:
+ const NumaConfig* numaConfig;
+ NumaIndex numaId;
};
-
-/// MainThread and TimerThread are derived classes used to characterize the two
-/// special threads: the main one and the recurring timer.
-
-struct MainThread : public Thread {
- virtual void idle_loop();
- volatile bool thinking = true; // Avoid a race with start_thinking()
+// Abstraction of a thread. It contains a pointer to the worker and a native thread.
+// After construction, the native thread is started with idle_loop()
+// waiting for a signal to start searching.
+// When the signal is received, the thread starts searching and when
+// the search is finished, it goes back to idle_loop() waiting for a new signal.
+class Thread {
+ public:
+ Thread(Search::SharedState&,
+ std::unique_ptr<Search::ISearchManager>,
+ size_t,
+ OptionalThreadToNumaNodeBinder);
+ virtual ~Thread();
+
+ void idle_loop();
+ void start_searching();
+ void clear_worker();
+ void run_custom_job(std::function<void()> f);
+
+ void ensure_network_replicated();
+
+ // Thread has been slightly altered to allow running custom jobs, so
+ // this name is no longer correct. However, this class (and ThreadPool)
+ // require further work to make them properly generic while maintaining
+ // appropriate specificity regarding search, from the point of view of an
+ // outside user, so renaming of this function is left for whenever that happens.
+ void wait_for_search_finished();
+ size_t id() const { return idx; }
+
+ std::unique_ptr<Search::Worker> worker;
+ std::function<void()> jobFunc;
+
+ private:
+ std::mutex mutex;
+ std::condition_variable cv;
+ size_t idx, nthreads;
+ bool exit = false, searching = true; // Set before starting std::thread
+ NativeThread stdThread;
+ NumaReplicatedAccessToken numaAccessToken;
};
-struct TimerThread : public ThreadBase {
-
- static const int Resolution = 5; // Millisec between two check_time() calls
-
- virtual void idle_loop();
-
- bool run = false;
-};
-
-
-/// ThreadPool struct handles all the threads related stuff like init, starting,
-/// parking and, most importantly, launching a slave thread at a split point.
-/// All the access to shared thread data is done through this class.
-
-struct ThreadPool : public std::vector<Thread*> {
-
- void init(); // No c'tor and d'tor, threads rely on globals that should be
- void exit(); // initialized and are valid during the whole thread lifetime.
-
- MainThread* main() { return static_cast<MainThread*>(at(0)); }
- void read_uci_options();
- Thread* available_slave(const SplitPoint* sp) const;
- void wait_for_think_finished();
- void start_thinking(const Position&, const Search::LimitsType&, Search::StateStackPtr&);
- Depth minimumSplitDepth;
- Spinlock spinlock;
- ConditionVariable sleepCondition;
- TimerThread* timer;
+// ThreadPool struct handles all the threads-related stuff like init, starting,
+// parking and, most importantly, launching a thread. All the access to threads
+// is done through this class.
+class ThreadPool {
+ public:
+ ThreadPool() {}
+
+ ~ThreadPool() {
+ // destroy any existing thread(s)
+ if (threads.size() > 0)
+ {
+ main_thread()->wait_for_search_finished();
+
+ threads.clear();
+ }
+ }
+
+ ThreadPool(const ThreadPool&) = delete;
+ ThreadPool(ThreadPool&&) = delete;
+
+ ThreadPool& operator=(const ThreadPool&) = delete;
+ ThreadPool& operator=(ThreadPool&&) = delete;
+
+ void start_thinking(const OptionsMap&, Position&, StateListPtr&, Search::LimitsType);
+ void run_on_thread(size_t threadId, std::function<void()> f);
+ void wait_on_thread(size_t threadId);
+ size_t num_threads() const;
+ void clear();
+ void set(const NumaConfig& numaConfig,
+ Search::SharedState,
+ const Search::SearchManager::UpdateContext&);
+
+ Search::SearchManager* main_manager();
+ Thread* main_thread() const { return threads.front().get(); }
+ uint64_t nodes_searched() const;
+ uint64_t tb_hits() const;
+ Thread* get_best_thread() const;
+ void start_searching();
+ void wait_for_search_finished() const;
+
+ std::vector<size_t> get_bound_thread_count_by_numa_node() const;
+
+ void ensure_network_replicated();
+
+ std::atomic_bool stop, abortedSearch, increaseDepth;
+
+ auto cbegin() const noexcept { return threads.cbegin(); }
+ auto begin() noexcept { return threads.begin(); }
+ auto end() noexcept { return threads.end(); }
+ auto cend() const noexcept { return threads.cend(); }
+ auto size() const noexcept { return threads.size(); }
+ auto empty() const noexcept { return threads.empty(); }
+
+ private:
+ StateListPtr setupStates;
+ std::vector<std::unique_ptr<Thread>> threads;
+ std::vector<NumaIndex> boundThreadToNumaNode;
+
+ uint64_t accumulate(std::atomic<uint64_t> Search::Worker::*member) const {
+
+ uint64_t sum = 0;
+ for (auto&& th : threads)
+ sum += (th->worker.get()->*member).load(std::memory_order_relaxed);
+ return sum;
+ }
};
-extern ThreadPool Threads;
+} // namespace Stockfish
-#endif // #ifndef THREAD_H_INCLUDED
+#endif // #ifndef THREAD_H_INCLUDED