简体   繁体   English

如何使用 pybind11 在 C++ 线程中调用 Python 函数作为回调

[英]How to invoke Python function as a callback inside C++ thread using pybind11

I designed a C++ system that invokes user defined callbacks from procedure running in a separate thread.我设计了一个 C++ 系统,它从在单独线程中运行的过程调用用户定义的回调。 Simplified system.hpp looks like this:简化后的system.hpp如下所示:

#pragma once

#include <atomic>
#include <chrono>
#include <functional>
#include <thread>

class System
{
public:
  using Callback = std::function<void(int)>;
  System(): t_(), cb_(), stop_(true) {}
  ~System()
  {
    stop();
  }
  bool start()
  {
    if (t_.joinable()) return false;
    stop_ = false;
    t_ = std::thread([this]()
    {
      while (!stop_)
      {
        std::this_thread::sleep_for(std::chrono::milliseconds(100));
        if (cb_) cb_(1234);
      }
    });
    return true;
  }
  bool stop()
  {
    if (!t_.joinable()) return false;
    stop_ = true;
    t_.join();
    return true;
  }
  bool registerCallback(Callback cb)
  {
    if (t_.joinable()) return false;
    cb_ = cb;
    return true;
  }

private:
  std::thread t_;
  Callback cb_;
  std::atomic_bool stop_;
};

It works just fine and can be tested with this short example main.cpp :它工作得很好,可以用这个简短的例子main.cpp进行测试:

#include <iostream>
#include "system.hpp"

int g_counter = 0;

void foo(int i)
{
  std::cout << i << std::endl;
  g_counter++;
}

int main()
{
  System s;
  s.registerCallback(foo);
  s.start();
  while (g_counter < 3)
  {
    std::this_thread::sleep_for(std::chrono::milliseconds(1));
  }
  s.stop();
  return 0;
}

which will output 1234 a few times and then will stop.这将输出1234几次,然后将停止。 However I encountered a problem trying to create python bindings for my System .但是,我在尝试为我的System创建 python 绑定时遇到了问题。 If I register a python function as a callback, my program will deadlock after calling System::stop .如果我注册一个 python 函数作为回调,我的程序将在调用System::stop后死锁。 I investigated the topic a bit and it seems that I face the issue with GIL .我调查了这个话题,似乎我面临着GIL的问题。 Reproducible example:可重现的例子:

binding.cpp : binding.cpp

#include "pybind11/functional.h"
#include "pybind11/pybind11.h"

#include "system.hpp"

namespace py = pybind11;

PYBIND11_MODULE(mysystembinding, m) {
  py::class_<System>(m, "System")
    .def(py::init<>())
    .def("start", &System::start)
    .def("stop", &System::stop)
    .def("registerCallback", &System::registerCallback);
}

python script:蟒蛇脚本:

#!/usr/bin/env python

import mysystembinding
import time

g_counter = 0

def foo(i):
  global g_counter
  print(i)
  g_counter = g_counter + 1

s = mysystembinding.System()
s.registerCallback(foo)
s.start()
while g_counter < 3:
  time.sleep(1)
s.stop()

I have read the pybind11 docs section about the possibility to acquire or release GIL on the C++ side.我已阅读有关在 C++ 端获取或释放 GIL 的可能性的pybind11 文档部分。 However I did not manage to get rid of the deadlock that occurs in my case:但是我没有设法摆脱在我的情况下发生的僵局:

PYBIND11_MODULE(mysystembinding, m) {
  py::class_<System>(m, "System")
    .def(py::init<>())
    .def("start", &System::start)
    .def("stop", &System::stop)
    .def("registerCallback", [](System* s, System::Callback cb)
      {
        s->registerCallback([cb](int i)
        {
          // py::gil_scoped_acquire acquire;
          // py::gil_scoped_release release;
          cb(i);
        });
      });
}

If I call py::gil_scoped_acquire acquire;如果我调用py::gil_scoped_acquire acquire; before calling the callback, deadlock occurs anyway.在调用回调之前,无论如何都会发生死锁。 If I call py::gil_scoped_release release;如果我调用py::gil_scoped_release release; before calling the callback, I get在调用回调之前,我得到

Fatal Python error: PyEval_SaveThread: NULL tstate致命的 Python 错误:PyEval_SaveThread:NULL tstate

What should I do to register python functions as callbacks and avoid deadlocks?我应该怎么做才能将 python 函数注册为回调并避免死锁?

Thanks to this discussion and many other resources ( 1 , 2 , 3 ) I figured out that guarding the functions that start and join the C++ thread with gil_scoped_release seems to solve the problem:由于这个讨论和许多其他资源( 123 ),我发现用gil_scoped_release保护启动和加入 C++ 线程的gil_scoped_release似乎可以解决问题:

PYBIND11_MODULE(mysystembinding, m) {
  py::class_<System>(m, "System")
    .def(py::init<>())
    .def("start", &System::start, py::call_guard<py::gil_scoped_release>())
    .def("stop", &System::stop, py::call_guard<py::gil_scoped_release>())
    .def("registerCallback", &System::registerCallback);
}

Apparently deadlocks occurred because python was holding a lock while invoking a binding responsible for C++ thread manipulation.显然,死锁的发生是因为 python 在调用负责 C++ 线程操作的绑定时持有锁。 I am still not really sure if my reasoning is correct, so I would appreciate any expert's comments.我仍然不确定我的推理是否正确,因此我将不胜感激任何专家的评论。

Call gil_scoped_release before join() will get rid of the deadlock in my case.join()之前调用gil_scoped_release将摆脱我的情况下的僵局。

void Tick::WaitLifeOver() {
  if (thread_.joinable()) {
    thread_.join();
  }
}
PYBIND11_MODULE(tick_pb, m) {
  py::class_<Tick, std::shared_ptr<Tick>>(m, "Tick")
    // ...
    .def("wait_life_over", &Tick::WaitLifeOver,
        py::call_guard<py::gil_scoped_release>());
}

Here are the codes: C++ Thread Callback Python Function以下是代码: C++ 线程回调 Python 函数

声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.

 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM