path: root/contrib/libs/apache/arrow_next/cpp/src/arrow/util/thread_pool.cc

// Licensed to the Apache Software Foundation (ASF) under one
// or more contributor license agreements.  See the NOTICE file
// distributed with this work for additional information
// regarding copyright ownership.  The ASF licenses this file
// to you under the Apache License, Version 2.0 (the
// "License"); you may not use this file except in compliance
// with the License.  You may obtain a copy of the License at
//
//   http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing,
// software distributed under the License is distributed on an
// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
// KIND, either express or implied.  See the License for the
// specific language governing permissions and limitations
// under the License.

#include "contrib/libs/apache/arrow_next/cpp/src/arrow/util/thread_pool.h"

#include <algorithm>
#include <condition_variable>
#include <deque>
#include <list>
#include <mutex>
#include <string>
#include <thread>
#include <vector>

#include "contrib/libs/apache/arrow_next/cpp/src/arrow/util/atfork_internal.h"
#include "contrib/libs/apache/arrow_next/src/arrow/util/config.h"
#include "contrib/libs/apache/arrow_next/cpp/src/arrow/util/io_util.h"
#include "contrib/libs/apache/arrow_next/cpp/src/arrow/util/logging.h"
#include "contrib/libs/apache/arrow_next/cpp/src/arrow/util/mutex.h"

#include "contrib/libs/apache/arrow_next/cpp/src/arrow/util/tracing_internal.h"

namespace arrow20 {
namespace internal {

Executor::~Executor() = default;

// By default we do nothing here.  Subclasses that expect to be allocated
// with static storage duration should override this and ensure any threads respect the
// lifetime of these resources.
void Executor::KeepAlive(std::shared_ptr<Resource> resource) {}

namespace {

struct Task {
  FnOnce<void()> callable;
  StopToken stop_token;
  Executor::StopCallback stop_callback;
};

struct QueuedTask {
  Task task;
  int32_t priority;
  uint64_t spawn_index;

  // Implement comparison so that std::priority_queue will pop the low priorities more
  // urgently.
  bool operator<(const QueuedTask& other) const {
    if (priority == other.priority) {
      // Maintain execution order for tasks with the same priority. Its preferable to keep
      // the execution order of tasks deterministic.
      return spawn_index > other.spawn_index;
    }
    return priority > other.priority;
  }
};

}  // namespace

struct SerialExecutor::State {
  std::priority_queue<QueuedTask> task_queue;
  uint64_t spawned_tasks_count_ = 0;
  std::mutex mutex;
  std::condition_variable wait_for_tasks;
  std::thread::id current_thread;
  bool paused{false};
  bool finished{false};
#ifndef ARROW_ENABLE_THREADING
  int max_tasks_running{1};
  int tasks_running{0};
#endif
};

#ifndef ARROW_ENABLE_THREADING
// list of all SerialExecutor objects - as we need to run tasks from all pools at once in
// Run()
struct SerialExecutorGlobalState {
  // a set containing all the executors that currently exist
  std::unordered_set<SerialExecutor*> all_executors;

  // this is the executor which is currently running a task
  SerialExecutor* current_executor = NULL;

  // in RunTasksOnAllExecutors we run tasks on executors in turn
  // this is used to keep track of the last fired task so that it
  // doesn't always run tasks on the first executor
  // in case of nested calls to RunTasksOnAllExecutors
  SerialExecutor* last_called_executor = NULL;
};

static SerialExecutorGlobalState* GetSerialExecutorGlobalState() {
  static SerialExecutorGlobalState state;
  return &state;
}

SerialExecutor* SerialExecutor::GetCurrentExecutor() {
  return GetSerialExecutorGlobalState()->current_executor;
}

bool SerialExecutor::IsCurrentExecutor() { return GetCurrentExecutor() == this; }

#endif

SerialExecutor::SerialExecutor() : state_(std::make_shared<State>()) {
#ifndef ARROW_ENABLE_THREADING
  GetSerialExecutorGlobalState()->all_executors.insert(this);
  state_->max_tasks_running = 1;
#endif
}

SerialExecutor::~SerialExecutor() {
#ifndef ARROW_ENABLE_THREADING
  GetSerialExecutorGlobalState()->all_executors.erase(this);
#endif
  auto state = state_;
  std::unique_lock<std::mutex> lk(state->mutex);
  if (!state->task_queue.empty()) {
    // We may have remaining tasks if the executor is being abandoned.  We could have
    // resource leakage in this case.  However, we can force the cleanup to happen now
    state->paused = false;
    lk.unlock();
    RunLoop();
    lk.lock();
  }
}

int SerialExecutor::GetNumTasks() {
  auto state = state_;
  return static_cast<int>(state_->task_queue.size());
}

#ifdef ARROW_ENABLE_THREADING
Status SerialExecutor::SpawnReal(TaskHints hints, FnOnce<void()> task,
                                 StopToken stop_token, StopCallback&& stop_callback) {
#  ifdef ARROW_WITH_OPENTELEMETRY
  // Wrap the task to propagate a parent tracing span to it
  // XXX should there be a generic utility in tracing_internal.h for this?
  task = [func = std::move(task),
          active_span =
              ::arrow20::internal::tracing::GetTracer()->GetCurrentSpan()]() mutable {
    auto scope = ::arrow20::internal::tracing::GetTracer()->WithActiveSpan(active_span);
    std::move(func)();
  };
#  endif
  // While the SerialExecutor runs tasks synchronously on its main thread,
  // SpawnReal may be called from external threads (e.g. when transferring back
  // from blocking I/O threads), so we need to keep the state alive *and* to
  // lock its contents.
  //
  // Note that holding the lock while notifying the condition variable may
  // not be sufficient, as some exit paths in the main thread are unlocked.
  auto state = state_;
  {
    std::lock_guard<std::mutex> lk(state->mutex);
    if (state_->finished) {
      return Status::Invalid(
          "Attempt to schedule a task on a serial executor that has already finished or "
          "been abandoned");
    }
    state->task_queue.push(QueuedTask{std::move(task), std::move(stop_token),
                                      std::move(stop_callback), hints.priority,
                                      state_->spawned_tasks_count_++});
  }
  state->wait_for_tasks.notify_one();
  return Status::OK();
}

void SerialExecutor::Finish() {
  auto state = state_;
  {
    std::lock_guard<std::mutex> lk(state->mutex);
    state->finished = true;
  }
  state->wait_for_tasks.notify_one();
}

#else  // ARROW_ENABLE_THREADING
Status SerialExecutor::SpawnReal(TaskHints hints, FnOnce<void()> task,
                                 StopToken stop_token, StopCallback&& stop_callback) {
#  ifdef ARROW_WITH_OPENTELEMETRY
  // Wrap the task to propagate a parent tracing span to it
  // XXX should there be a generic utility in tracing_internal.h for this?
  task = [func = std::move(task),
          active_span =
              ::arrow20::internal::tracing::GetTracer()->GetCurrentSpan()]() mutable {
    auto scope = ::arrow20::internal::tracing::GetTracer()->WithActiveSpan(active_span);
    std::move(func)();
  };
#  endif  // ARROW_WITH_OPENTELEMETRY

  if (state_->finished) {
    return Status::Invalid(
        "Attempt to schedule a task on a serial executor that has already finished or "
        "been abandoned");
  }

  state_->task_queue.push(QueuedTask{std::move(task), std::move(stop_token),
                                     std::move(stop_callback), hints.priority,
                                     state_->spawned_tasks_count_++});

  return Status::OK();
}

void SerialExecutor::Finish() {
  auto state = state_;
  { state->finished = true; }
  // empty any tasks from the loop on finish
  RunLoop();
}

#endif  // ARROW_ENABLE_THREADING
void SerialExecutor::Pause() {
  // Same comment as SpawnReal above
  auto state = state_;
  {
    std::lock_guard<std::mutex> lk(state->mutex);
    state->paused = true;
  }
  state->wait_for_tasks.notify_one();
}

bool SerialExecutor::IsFinished() {
  std::lock_guard<std::mutex> lk(state_->mutex);
  return state_->finished;
}

void SerialExecutor::Unpause() {
  auto state = state_;
  {
    std::lock_guard<std::mutex> lk(state->mutex);
    state->paused = false;
  }
}

bool SerialExecutor::OwnsThisThread() {
  std::lock_guard lk(state_->mutex);
  return std::this_thread::get_id() == state_->current_thread;
}
#ifdef ARROW_ENABLE_THREADING

void SerialExecutor::RunLoop() {
  // This is called from the SerialExecutor's main thread, so the
  // state is guaranteed to be kept alive.
  std::unique_lock<std::mutex> lk(state_->mutex);
  state_->current_thread = std::this_thread::get_id();
  // If paused we break out immediately.  If finished we only break out
  // when all work is done.
  while (!state_->paused && !(state_->finished && state_->task_queue.empty())) {
    // The inner loop is to check if we need to sleep (e.g. while waiting on some
    // async task to finish from another thread pool).  We still need to check paused
    // because sometimes we will pause even with work leftover when processing
    // an async generator
    while (!state_->paused && !state_->task_queue.empty()) {
      Task task = std::move(const_cast<Task&>(state_->task_queue.top().task));
      state_->task_queue.pop();
      lk.unlock();
      if (!task.stop_token.IsStopRequested()) {
        std::move(task.callable)();
      } else {
        if (task.stop_callback) {
          std::move(task.stop_callback)(task.stop_token.Poll());
        }
        // Can't break here because there may be cleanup tasks down the chain we still
        // need to run.
      }
      lk.lock();
    }
    // In this case we must be waiting on work from external (e.g. I/O) executors.  Wait
    // for tasks to arrive (typically via transferred futures).
    state_->wait_for_tasks.wait(lk, [&] {
      return state_->paused || state_->finished || !state_->task_queue.empty();
    });
  }
  state_->current_thread = {};
}
#else   // ARROW_ENABLE_THREADING
bool SerialExecutor::RunTasksOnAllExecutors() {
  auto globalState = GetSerialExecutorGlobalState();
  // if the previously called executor was deleted, ignore last_called_executor
  if (globalState->last_called_executor != NULL &&
      globalState->all_executors.count(globalState->last_called_executor) == 0) {
    globalState->last_called_executor = NULL;
  }
  bool run_task = true;
  bool keep_going = true;
  while (keep_going) {
    run_task = false;
    keep_going = false;
    for (auto it = globalState->all_executors.begin();
         it != globalState->all_executors.end(); ++it) {
      if (globalState->last_called_executor != NULL) {
        // always rerun loop if we have a last_called_executor, otherwise
        // we may drop out before every executor has been checked
        keep_going = true;
        if (globalState->all_executors.count(globalState->last_called_executor) == 0 ||
            globalState->last_called_executor == *it) {
          // found the last one (or it doesn't exist ih the set any more)
          // now we can start running things
          globalState->last_called_executor = NULL;
        }
        // skip until after we have seen the last executor we called
        // so that we do things nicely in turn
        continue;
      }
      auto exe = *it;
      // don't make more reentrant calls inside an
      // executor than the number of concurrent tasks set on a threadpool, or
      // 1 in the case of a serialexecutor -
      // this is because users will expect a serial executor not to be able to
      // run the next task until the current one is finished (and a threadpool
      // only to be able to run a certain number of tasks concurrently)
      if (exe->state_->tasks_running >= exe->state_->max_tasks_running) {
        continue;
      }
      if (exe->state_->paused == false && exe->state_->task_queue.empty() == false) {
        SerialExecutor* old_exe = globalState->current_executor;
        globalState->current_executor = exe;
        Task task = std::move(const_cast<Task&>(exe->state_->task_queue.top().task));
        exe->state_->task_queue.pop();
        run_task = true;
        exe->state_->tasks_running += 1;
        if (!task.stop_token.IsStopRequested()) {
          std::move(task.callable)();
        } else {
          if (task.stop_callback) {
            std::move(task.stop_callback)(task.stop_token.Poll());
          }
        }
        exe->state_->tasks_running -= 1;
        globalState->current_executor = old_exe;

        globalState->last_called_executor = exe;
        keep_going = false;
        break;
      }
    }
  }
  return run_task;
}

// run tasks in this thread and queue things from other executors if required
// (e.g. when a compute task depends on an IO request)
void SerialExecutor::RunLoop() {
  auto globalState = GetSerialExecutorGlobalState();
  // If paused we break out immediately.  If finished we only break out
  // when all work is done.
  while (!state_->paused && !(state_->finished && state_->task_queue.empty())) {
    // first empty us until paused or empty
    // if we're already running as many tasks as possible then
    // we can't run any more until something else drops off the queue
    if (state_->tasks_running <= state_->max_tasks_running) {
      while (!state_->paused && !state_->task_queue.empty()) {
        Task task = std::move(const_cast<Task&>(state_->task_queue.top().task));
        state_->task_queue.pop();
        auto last_executor = globalState->current_executor;
        globalState->current_executor = this;
        state_->tasks_running += 1;
        if (!task.stop_token.IsStopRequested()) {
          std::move(task.callable)();
        } else {
          if (task.stop_callback) {
            std::move(task.stop_callback)(task.stop_token.Poll());
          }
        }
        state_->tasks_running -= 1;
        globalState->current_executor = last_executor;
      }
      if (state_->paused || (state_->finished && state_->task_queue.empty())) {
        break;
      }
    }
    // now wait for anything on other executors (unless we're finished in which case it
    // will drop out of the outer loop
    RunTasksOnAllExecutors();
  }
}
#endif  // ARROW_ENABLE_THREADING

#ifdef ARROW_ENABLE_THREADING

struct ThreadPool::State {
  State() = default;

  // NOTE: in case locking becomes too expensive, we can investigate lock-free FIFOs
  // such as https://github.com/cameron314/concurrentqueue

  std::mutex mutex_;
  std::condition_variable cv_;
  std::condition_variable cv_shutdown_;
  std::condition_variable cv_idle_;

  std::list<std::thread> workers_;
  // Trashcan for finished threads
  std::vector<std::thread> finished_workers_;
  std::priority_queue<QueuedTask> pending_tasks_;
  uint64_t spawned_tasks_count_ = 0;

  // Desired number of threads
  int desired_capacity_ = 0;

  // Total number of tasks that are either queued or running
  int tasks_queued_or_running_ = 0;

  // Are we shutting down?
  bool please_shutdown_ = false;
  bool quick_shutdown_ = false;

  std::vector<std::shared_ptr<Resource>> kept_alive_resources_;

  // At-fork machinery

  void BeforeFork() { mutex_.lock(); }

  void ParentAfterFork() { mutex_.unlock(); }

  void ChildAfterFork() {
    int desired_capacity = desired_capacity_;
    bool please_shutdown = please_shutdown_;
    bool quick_shutdown = quick_shutdown_;
    new (this) State;  // force-reinitialize, including synchronization primitives
    desired_capacity_ = desired_capacity;
    please_shutdown_ = please_shutdown;
    quick_shutdown_ = quick_shutdown;
  }

  std::shared_ptr<AtForkHandler> atfork_handler_;
};

// The worker loop is an independent function so that it can keep running
// after the ThreadPool is destroyed.
static void WorkerLoop(std::shared_ptr<ThreadPool::State> state,
                       std::list<std::thread>::iterator it) {
  std::unique_lock<std::mutex> lock(state->mutex_);

  // Since we hold the lock, `it` now points to the correct thread object
  // (LaunchWorkersUnlocked has exited)
  DCHECK_EQ(std::this_thread::get_id(), it->get_id());

  // If too many threads, we should secede from the pool
  const auto should_secede = [&]() -> bool {
    return state->workers_.size() > static_cast<size_t>(state->desired_capacity_);
  };

  while (true) {
    // By the time this thread is started, some tasks may have been pushed
    // or shutdown could even have been requested.  So we only wait on the
    // condition variable at the end of the loop.

    // Execute pending tasks if any
    while (!state->pending_tasks_.empty() && !state->quick_shutdown_) {
      // We check this opportunistically at each loop iteration since
      // it releases the lock below.
      if (should_secede()) {
        break;
      }

      DCHECK_GE(state->tasks_queued_or_running_, 0);
      {
        Task task = std::move(const_cast<Task&>(state->pending_tasks_.top().task));
        state->pending_tasks_.pop();
        StopToken* stop_token = &task.stop_token;
        lock.unlock();
        if (!stop_token->IsStopRequested()) {
          std::move(task.callable)();
        } else {
          if (task.stop_callback) {
            std::move(task.stop_callback)(stop_token->Poll());
          }
        }
        {
          auto tmp_task = std::move(task);  // release resources before waiting for lock
          ARROW_UNUSED(tmp_task);
        }
        lock.lock();
      }
      if (ARROW_PREDICT_FALSE(--state->tasks_queued_or_running_ == 0)) {
        state->cv_idle_.notify_all();
      }
    }
    // Now either the queue is empty *or* a quick shutdown was requested
    if (state->please_shutdown_ || should_secede()) {
      break;
    }
    // Wait for next wakeup
    state->cv_.wait(lock);
  }
  DCHECK_GE(state->tasks_queued_or_running_, 0);

  // We're done.  Move our thread object to the trashcan of finished
  // workers.  This has two motivations:
  // 1) the thread object doesn't get destroyed before this function finishes
  //    (but we could call thread::detach() instead)
  // 2) we can explicitly join() the trashcan threads to make sure all OS threads
  //    are exited before the ThreadPool is destroyed.  Otherwise subtle
  //    timing conditions can lead to false positives with Valgrind.
  DCHECK_EQ(std::this_thread::get_id(), it->get_id());
  state->finished_workers_.push_back(std::move(*it));
  state->workers_.erase(it);
  if (state->please_shutdown_) {
    // Notify the function waiting in Shutdown().
    state->cv_shutdown_.notify_one();
  }
}

void ThreadPool::WaitForIdle() {
  std::unique_lock<std::mutex> lk(state_->mutex_);
  state_->cv_idle_.wait(lk, [this] { return state_->tasks_queued_or_running_ == 0; });
}

ThreadPool::ThreadPool()
    : sp_state_(std::make_shared<ThreadPool::State>()),
      state_(sp_state_.get()),
      shutdown_on_destroy_(true) {
  // Eternal thread pools would produce false leak reports in the vector of
  // atfork handlers.
#  if !(defined(_WIN32) || defined(ADDRESS_SANITIZER) || defined(ARROW_VALGRIND))
  state_->atfork_handler_ = std::make_shared<AtForkHandler>(
      /*before=*/
      [weak_state = std::weak_ptr<ThreadPool::State>(sp_state_)]() {
        auto state = weak_state.lock();
        if (state) {
          state->BeforeFork();
        }
        return state;  // passed to after-forkers
      },
      /*parent_after=*/
      [](std::any token) {
        auto state = std::any_cast<std::shared_ptr<ThreadPool::State>>(token);
        if (state) {
          state->ParentAfterFork();
        }
      },
      /*child_after=*/
      [](std::any token) {
        auto state = std::any_cast<std::shared_ptr<ThreadPool::State>>(token);
        if (state) {
          state->ChildAfterFork();
        }
      });
  RegisterAtFork(state_->atfork_handler_);
#  endif
}

ThreadPool::~ThreadPool() {
  if (shutdown_on_destroy_) {
    ARROW_UNUSED(Shutdown(false /* wait */));
  }
}

Status ThreadPool::SetCapacity(int threads) {
  std::unique_lock<std::mutex> lock(state_->mutex_);
  if (state_->please_shutdown_) {
    return Status::Invalid("operation forbidden during or after shutdown");
  }
  if (threads <= 0) {
    return Status::Invalid("ThreadPool capacity must be > 0");
  }
  CollectFinishedWorkersUnlocked();

  state_->desired_capacity_ = threads;
  // See if we need to increase or decrease the number of running threads
  const int required = std::min(static_cast<int>(state_->pending_tasks_.size()),
                                threads - static_cast<int>(state_->workers_.size()));
  if (required > 0) {
    // Some tasks are pending, spawn the number of needed threads immediately
    LaunchWorkersUnlocked(required);
  } else if (required < 0) {
    // Excess threads are running, wake them so that they stop
    state_->cv_.notify_all();
  }
  return Status::OK();
}

int ThreadPool::GetCapacity() {
  std::unique_lock<std::mutex> lock(state_->mutex_);
  return state_->desired_capacity_;
}

int ThreadPool::GetNumTasks() {
  std::unique_lock<std::mutex> lock(state_->mutex_);
  return state_->tasks_queued_or_running_;
}

int ThreadPool::GetActualCapacity() {
  std::unique_lock<std::mutex> lock(state_->mutex_);
  return static_cast<int>(state_->workers_.size());
}

Status ThreadPool::Shutdown(bool wait) {
  std::unique_lock<std::mutex> lock(state_->mutex_);

  if (state_->please_shutdown_) {
    return Status::Invalid("Shutdown() already called");
  }
  state_->please_shutdown_ = true;
  state_->quick_shutdown_ = !wait;
  state_->cv_.notify_all();
  state_->cv_shutdown_.wait(lock, [this] { return state_->workers_.empty(); });
  if (!state_->quick_shutdown_) {
    DCHECK_EQ(state_->pending_tasks_.size(), 0);
  } else {
    std::priority_queue<QueuedTask> empty;
    std::swap(state_->pending_tasks_, empty);
  }
  CollectFinishedWorkersUnlocked();
  return Status::OK();
}

void ThreadPool::CollectFinishedWorkersUnlocked() {
  for (auto& thread : state_->finished_workers_) {
    // Make sure OS thread has exited
    thread.join();
  }
  state_->finished_workers_.clear();
}

thread_local ThreadPool* current_thread_pool_ = nullptr;

bool ThreadPool::OwnsThisThread() { return current_thread_pool_ == this; }

void ThreadPool::LaunchWorkersUnlocked(int threads) {
  std::shared_ptr<State> state = sp_state_;

  for (int i = 0; i < threads; i++) {
    state_->workers_.emplace_back();
    auto it = --(state_->workers_.end());
    *it = std::thread([this, state, it] {
      current_thread_pool_ = this;
      WorkerLoop(state, it);
    });
  }
}

Status ThreadPool::SpawnReal(TaskHints hints, FnOnce<void()> task, StopToken stop_token,
                             StopCallback&& stop_callback) {
  {
#  ifdef ARROW_WITH_OPENTELEMETRY
    // Wrap the task to propagate a parent tracing span to it
    // This task-wrapping needs to be done before we grab the mutex because the
    // first call to OT (whatever that happens to be) will attempt to grab this mutex
    // when calling KeepAlive to keep the OT infrastructure alive.
    struct {
      void operator()() {
        auto scope = ::arrow20::internal::tracing::GetTracer()->WithActiveSpan(activeSpan);
        std::move(func)();
      }
      FnOnce<void()> func;
      opentelemetry::nostd::shared_ptr<opentelemetry::trace::Span> activeSpan;
    } wrapper{std::forward<FnOnce<void()>>(task),
              ::arrow20::internal::tracing::GetTracer()->GetCurrentSpan()};
    task = std::move(wrapper);
#  endif
    std::lock_guard<std::mutex> lock(state_->mutex_);
    if (state_->please_shutdown_) {
      return Status::Invalid("operation forbidden during or after shutdown");
    }
    CollectFinishedWorkersUnlocked();
    state_->tasks_queued_or_running_++;
    if (static_cast<int>(state_->workers_.size()) < state_->tasks_queued_or_running_ &&
        state_->desired_capacity_ > static_cast<int>(state_->workers_.size())) {
      // We can still spin up more workers so spin up a new worker
      LaunchWorkersUnlocked(/*threads=*/1);
    }
    state_->pending_tasks_.push(
        QueuedTask{{std::move(task), std::move(stop_token), std::move(stop_callback)},
                   hints.priority,
                   state_->spawned_tasks_count_++});
  }
  state_->cv_.notify_one();
  return Status::OK();
}

void ThreadPool::KeepAlive(std::shared_ptr<Executor::Resource> resource) {
  // Seems unlikely but we might as well guard against concurrent calls to KeepAlive
  std::lock_guard<std::mutex> lk(state_->mutex_);
  state_->kept_alive_resources_.push_back(std::move(resource));
}

Result<std::shared_ptr<ThreadPool>> ThreadPool::Make(int threads) {
  auto pool = std::shared_ptr<ThreadPool>(new ThreadPool());
  RETURN_NOT_OK(pool->SetCapacity(threads));
  return pool;
}

Result<std::shared_ptr<ThreadPool>> ThreadPool::MakeEternal(int threads) {
  ARROW_ASSIGN_OR_RAISE(auto pool, Make(threads));
  // On Windows, the ThreadPool destructor may be called after non-main threads
  // have been killed by the OS, and hang in a condition variable.
  // On Unix, we want to avoid leak reports by Valgrind.
#  ifdef _WIN32
  pool->shutdown_on_destroy_ = false;
#  endif
  return pool;
}

// ----------------------------------------------------------------------
// Global thread pool

static int ParseOMPEnvVar(const char* name) {
  // OMP_NUM_THREADS is a comma-separated list of positive integers.
  // We are only interested in the first (top-level) number.
  auto result = GetEnvVar(name);
  if (!result.ok()) {
    return 0;
  }
  auto str = *std::move(result);
  auto first_comma = str.find_first_of(',');
  if (first_comma != std::string::npos) {
    str = str.substr(0, first_comma);
  }
  try {
    return std::max(0, std::stoi(str));
  } catch (...) {
    return 0;
  }
}

int ThreadPool::DefaultCapacity() {
  int capacity, limit;
  capacity = ParseOMPEnvVar("OMP_NUM_THREADS");
  if (capacity == 0) {
    capacity = std::thread::hardware_concurrency();
  }
  limit = ParseOMPEnvVar("OMP_THREAD_LIMIT");
  if (limit > 0) {
    capacity = std::min(limit, capacity);
  }
  if (capacity == 0) {
    ARROW_LOG(WARNING) << "Failed to determine the number of available threads, "
                          "using a hardcoded arbitrary value";
    capacity = 4;
  }
  return capacity;
}

#else  // ARROW_ENABLE_THREADING
ThreadPool::ThreadPool() {
  // default to max 'concurrency' of 8
  // if threading is disabled
  state_->max_tasks_running = 8;
}

Status ThreadPool::Shutdown(bool wait) {
  state_->finished = true;
  if (wait) {
    RunLoop();
  } else {
    // clear any pending tasks so that we behave
    // the same as threadpool on fast shutdown
    std::priority_queue<QueuedTask> empty;
    std::swap(state_->task_queue, empty);
  }
  return Status::OK();
}

// Wait for the 'thread pool' to become idle
// including running tasks from other pools if
// needed
void ThreadPool::WaitForIdle() {
  while (!state_->task_queue.empty()) {
    RunTasksOnAllExecutors();
  }
}

Status ThreadPool::SetCapacity(int threads) {
  state_->max_tasks_running = threads;
  return Status::OK();
}

int ThreadPool::GetCapacity() { return state_->max_tasks_running; }

int ThreadPool::GetActualCapacity() { return state_->max_tasks_running; }

Result<std::shared_ptr<ThreadPool>> ThreadPool::Make(int threads) {
  auto pool = std::shared_ptr<ThreadPool>(new ThreadPool());
  RETURN_NOT_OK(pool->SetCapacity(threads));
  return pool;
}

Result<std::shared_ptr<ThreadPool>> ThreadPool::MakeEternal(int threads) {
  ARROW_ASSIGN_OR_RAISE(auto pool, Make(threads));
  // On Windows, the ThreadPool destructor may be called after non-main threads
  // have been killed by the OS, and hang in a condition variable.
  // On Unix, we want to avoid leak reports by Valgrind.
  return pool;
}

ThreadPool::~ThreadPool() {
  // clear threadpool, otherwise ~SerialExecutor will
  // run any tasks left (which isn't threadpool behaviour)
  std::priority_queue<QueuedTask> empty;
  std::swap(state_->task_queue, empty);
}

#endif  // ARROW_ENABLE_THREADING

// Helper for the singleton pattern
std::shared_ptr<ThreadPool> ThreadPool::MakeCpuThreadPool() {
  auto maybe_pool = ThreadPool::MakeEternal(ThreadPool::DefaultCapacity());
  if (!maybe_pool.ok()) {
    maybe_pool.status().Abort("Failed to create global CPU thread pool");
  }
  return *std::move(maybe_pool);
}

ThreadPool* GetCpuThreadPool() {
  // Avoid using a global variable because of initialization order issues (ARROW-18383)
  static std::shared_ptr<ThreadPool> singleton = ThreadPool::MakeCpuThreadPool();
  return singleton.get();
}

}  // namespace internal

int GetCpuThreadPoolCapacity() { return internal::GetCpuThreadPool()->GetCapacity(); }

Status SetCpuThreadPoolCapacity(int threads) {
  return internal::GetCpuThreadPool()->SetCapacity(threads);
}

}  // namespace arrow20