The idea is to be able to replace multithreaded code with boost::asio and a thread pool, on a consumer/producer problem. Currently, each consumer thread waits on a boost::condition_variable - when a producer adds something to the queue, it calls notify_one/notify_all to notify all the consumers. Now what happens when you (potentially) have 1k+ consumers? Threads won't scale!
I decided to use boost::asio, but then I ran into the fact that it doesn't have condition variables. And then async_condition_variable was born:
class async_condition_variable
{
private:
boost::asio::io_service& service_;
typedef boost::function<void ()> async_handler;
std::queue<async_handler> waiters_;
public:
async_condition_variable(boost::asio::io_service& service) : service_(service)
{
}
void async_wait(async_handler handler)
{
waiters_.push(handler);
}
void notify_one()
{
service_.post(waiters_.front());
waiters_.pop();
}
void notify_all()
{
while (!waiters_.empty()) {
notify_one();
}
}
};
Basically, each consumer would call async_condition_variable::wait(...). Then, a producer would eventually call async_condition_variable::notify_one() or async_condition_variable::notify_all(). Each consumer's handle would be called, and would either act on the condition or call async_condition_variable::wait(...) again. Is this feasible or am I being crazy here? What kind of locking (mutexes) should be performed, given the fact that this would be run on a thread pool?
P.S.: Yes, this is more a RFC (Request for Comments) than a question :).
Have a list of things that need to be done when an event occurs. Have a function to add something to that list and a function to remove something from that list. Then, when the event occurs, have a pool of threads work on the list of jobs that now need to be done. You don't need threads specifically waiting for the event.
Boost::asio can be kind of hard to wrap your head around. At least, I have difficult time doing it.
You don't need to have the threads wait on anything. They do that on their own when they don't have any work to do. The examples that seemed to look like what you wanted to do had work posted to the io_service for each item.
The following code was inspired from this link. It actually open my eyes to how you could use it do a lot of things.
I'm sure this isn't perfect, but I think it gives the general idea. I hope this helps.
Code
#include <iostream>
#include <boost/thread.hpp>
#include <boost/bind.hpp>
#include <boost/asio.hpp>
class ServerProcessor
{
protected:
void handleWork1(WorkObject1* work)
{
//The code to do task 1 goes in here
}
void handleWork2(WorkObject2* work)
{
//The code to do task 2 goes in here
}
boost::thread_group worker_threads_;
boost::asio::io_service io_service_;
//This is used to keep io_service from running out of work and exiting to soon.
boost::shared_ptr<boost::asio::io_service::work> work_;
public:
void start(int numberOfThreads)
{
boost::shared_ptr<boost::asio::io_service::work> myWork(new boost::asio::io_service::work(io_service_));
work_=myWork;
for (int x=0; x < numberOfThreads; ++x)
worker_threads_.create_thread( boost::bind( &ServerProcessor::threadAction, this ) );
}
void doWork1(WorkObject1* work)
{
io_service_.post(boost::bind(&ServerProcessor::handleWork1, this, work));
}
void doWork2(WorkObject2* work)
{
io_service_.post(boost::bind(&ServerProcessor::handleWork2, this, work));
}
void threadAction()
{
io_service_.run();
}
void stop()
{
work_.reset();
io_service_.stop();
worker_threads_.join_all();
}
};
int main()
{
ServerProcessor s;
std::string input;
std::cout<<"Press f to stop"<<std::endl;
s.start(8);
std::cin>>input;
s.stop();
return 0;
}
How about using boost::signals2?
It is a thread safe spinoff of boost::signals that lets your clients subscribe a callback to a signal to be emitted.
Then, when the signal is emitted asynchronously in an io_service dispatched job all the registered callbacks will be executed (on the same thread that emitted the signal).
Related
Implementing a timer class on a UI application.
Basically the problem I have is that calling io.run() would block, rendering the async_wait call useless. From reading other posts I got the impression that somehow Timer, or at least the code that calls startCountdown, should be on another thread.
Below is my code. How do I manage this in a way that is considered correct in Boost?
class Timer
{
public:
Timer() : countdownTimer(io) { }
void startCountdown(int seconds)
{
countdownTimer.expires_from_now(boost::posix_time::seconds(seconds));
countdownTimer.async_wait(boost::bind(&Timer::on_timeout, this, _1));
io.run(); // this blocks
}
void on_timeout(const boost::system::error_code& e)
{
if (e != boost::asio::error::operation_aborted) {
cout << "Timer expired!";
}
}
private:
boost::asio::io_service io;
boost::asio::deadline_timer countdownTimer;
}
You shouldn't call the run member function of io_service but rahter one of :
run_one
poll
poll_one
Depending on your design goal.
Also note that boost Asio is not really designed to be a secondary library to use for some features but rather as a core element of your program, it is more a spine than an arm.
A thread dedicated to handle everything linked to the io_service could also work, but good luck synchronizing shared data between a synchronous and an asynchronous world :)
I have a program starting an std::thread doing the following: sleep X, execute a function, terminate.
create std::thread(Xms, &func)
wait Xms
then do func()
end
I was wondering if I could for example send a signal to my std::thread in order to instantly break the sleep and do func, then quit.
Do I need to send the signal to std::thread::id in order to perform this?
my thread is launched this way, with a lambda function:
template<typename T, typename U>
void execAfter(T func, U params, const int ms)
{
std::thread thread([=](){
std::this_thread::sleep_for(std::chrono::milliseconds(ms));
func(params);
});
thread.detach();
}
Using wait_for of std::condition_variable would be the way to go, if the thread model can't be changed. In the code snippet below, the use of the condition_variable is wrapped into a class of which objects have to be shared across the threads.
#include <iostream>
#include <atomic>
#include <condition_variable>
#include <thread>
#include <chrono>
class BlockCondition
{
private:
mutable std::mutex m;
std::atomic<bool> done;
mutable std::condition_variable cv;
public:
BlockCondition()
:
m(),
done(false),
cv()
{
}
void wait_for(int duration_ms)
{
std::unique_lock<std::mutex> l(m);
int ms_waited(0);
while ( !done.load() && ms_waited < duration_ms )
{
auto t_0(std::chrono::high_resolution_clock::now());
cv.wait_for(l, std::chrono::milliseconds(duration_ms - ms_waited));
auto t_1(std::chrono::high_resolution_clock::now());
ms_waited += std::chrono::duration_cast<std::chrono::milliseconds>(t_1 - t_0).count();
}
}
void release()
{
std::lock_guard<std::mutex> l(m);
done.store(true);
cv.notify_one();
}
};
void delayed_func(BlockCondition* block)
{
block->wait_for(1000);
std::cout << "Hello actual work\n";
}
void abortSleepyFunction(BlockCondition* block)
{
block->release();
}
void test_aborted()
{
BlockCondition b();
std::thread delayed_thread(delayed_func, &b);
abortSleepyFunction(&b);
delayed_thread.join();
}
void test_unaborted()
{
BlockCondition b();
std::thread delayed_thread(delayed_func, &b);
delayed_thread.join();
}
int main()
{
test_aborted();
test_unaborted();
}
Note that there might be spurious wakeups that abort the wait call prematurely. To account for that, we count the milliseconds actually waited and continue waiting until the done flag is set.
As was pointed out in the comments, this wasn't the smartest approach for solving your problem in the first place. As implementing a proper interruption mechanism is quite complex and extremely easy to get wrong, here are suggestions for a workaround:
Instead of sleeping for the whole timeout, simply loop over a sleep of fixed small size (e.g. 10 milliseconds) until the desired duration has elapsed. After each sleep you check an atomic flag whether interruption was requested. This is a dirty solution, but is the quickest to pull of.
Alternatively, supply each thread with a condition_variable and do a wait on it instead of doing the this_thread::sleep. Notify the condition variable to indicate the request for interruption. You will probably still want an additional flag to protect against spurious wakeups so you don't accidentally return too early.
Ok, to figure this out I found a new implementation, it's inspired by all your answers so thanks a lot.
First I am gonna do a BombHandler item, in the main Game item. It will have a an attribute containing all the Bomb items.
This BombHandler will be a singleton, containing a timerLoop() function who will execute in a thread (This way I only use ONE thread for xxx bombs, way more effective)
The timerLoop() will usleep(50) then pass through the whole std::list elements and call Bomb::incrTimer() who will increment their internal _timer attribute by 10ms indefinitely, and check bombs who have to explode.
When they reach 2000ms for instance, BombHandler.explode() will be called, exploding the bomb and deleting it.
If another bomb is in range Bomb::touchByFire() will be called, and set the internal attribute of Bomb, _timer, to TIME_TO_EXPLODE (1950ms).
Then it will be explode 50ms later by BombHandler::explode().
Isn't this a nice solution?
Again, thanks for your answers! Hope this can help.
I'm playing with boost library and C++. I want to create a multithreaded program that contains a producer, conumer, and a stack. The procuder fills the stack, the consumer remove items (int) from the stack. everything work (pop, push, mutex) But when i call the pop/push winthin a thread, i don't get any effect
i made this simple code :
#include "stdafx.h"
#include <stack>
#include <iostream>
#include <algorithm>
#include <boost/shared_ptr.hpp>
#include <boost/thread.hpp>
#include <boost/date_time.hpp>
#include <boost/signals2/mutex.hpp>
#include <ctime>
using namespace std;
/ *
* this class reprents a stack which is proteced by mutex
* Pop and push are executed by one thread each time.
*/
class ProtectedStack{
private :
stack<int> m_Stack;
boost::signals2::mutex m;
public :
ProtectedStack(){
}
ProtectedStack(const ProtectedStack & p){
}
void push(int x){
m.lock();
m_Stack.push(x);
m.unlock();
}
void pop(){
m.lock();
//return m_Stack.top();
if(!m_Stack.empty())
m_Stack.pop();
m.unlock();
}
int size(){
return m_Stack.size();
}
bool isEmpty(){
return m_Stack.empty();
}
int top(){
return m_Stack.top();
}
};
/*
*The producer is the class that fills the stack. It encapsulate the thread object
*/
class Producer{
public:
Producer(int number ){
//create thread here but don't start here
m_Number=number;
}
void fillStack (ProtectedStack& s ) {
int object = 3; //random value
s.push(object);
//cout<<"push object\n";
}
void produce (ProtectedStack & s){
//call fill within a thread
m_Thread = boost::thread(&Producer::fillStack,this, s);
}
private :
int m_Number;
boost::thread m_Thread;
};
/* The consumer will consume the products produced by the producer */
class Consumer {
private :
int m_Number;
boost::thread m_Thread;
public:
Consumer(int n){
m_Number = n;
}
void remove(ProtectedStack &s ) {
if(s.isEmpty()){ // if the stack is empty sleep and wait for the producer to fill the stack
//cout<<"stack is empty\n";
boost::posix_time::seconds workTime(1);
boost::this_thread::sleep(workTime);
}
else{
s.pop(); //pop it
//cout<<"pop object\n";
}
}
void consume (ProtectedStack & s){
//call remove within a thread
m_Thread = boost::thread(&Consumer::remove, this, s);
}
};
int main(int argc, char* argv[])
{
ProtectedStack s;
Producer p(0);
p.produce(s);
Producer p2(1);
p2.produce(s);
cout<<"size after production "<<s.size()<<endl;
Consumer c(0);
c.consume(s);
Consumer c2(1);
c2.consume(s);
cout<<"size after consumption "<<s.size()<<endl;
getchar();
return 0;
}
After i run that in VC++ 2010 / win7
i got :
0
0
Could you please help me understand why when i call fillStack function from the main i got an effect but when i call it from a thread nothing happens?
Thank you
Your example code suffers from a couple synchronization issues as noted by others:
Missing locks on calls to some of the members of ProtectedStack.
Main thread could exit without allowing worker threads to join.
The producer and consumer do not loop as you would expect. Producers should always (when they can) be producing, and consumers should keep consuming as new elements are pushed onto the stack.
cout's on the main thread may very well be performed before the producers or consumers have had a chance to work yet.
I would recommend looking at using a condition variable for synchronization between your producers and consumers. Take a look at the producer/consumer example here: http://en.cppreference.com/w/cpp/thread/condition_variable
It is a rather new feature in the standard library as of C++11 and supported as of VS2012. Before VS2012, you would either need boost or to use Win32 calls.
Using a condition variable to tackle a producer/consumer problem is nice because it almost enforces the use of a mutex to lock shared data and it provides a signaling mechanism to let consumers know something is ready to be consumed so they don't have so spin (which is always a trade off between the responsiveness of the consumer and CPU usage polling the queue). It also does so being atomic itself which prevents the possibility of threads missing a signal that there is something to consume as explained here: https://en.wikipedia.org/wiki/Sleeping_barber_problem
To give a brief run-down of how a condition variable takes care of this...
A producer does all time consuming activities on its thread without the owning the mutex.
The producer locks the mutex, adds the item it produced to a global data structure (probably a queue of some sort), lets go of the mutex and signals a single consumer to go -- in that order.
A consumer that is waiting on the condition variable re-acquires the mutex automatically, removes the item out of the queue and does some processing on it. During this time, the producer is already working on producing a new item but has to wait until the consumer is done before it can queue the item up.
This would have the following impact on your code:
No more need for ProtectedStack, a normal stack/queue data structure will do.
No need for boost if you are using a new enough compiler - removing build dependencies is always a nice thing.
I get the feeling that threading is rather new to you so I can only offer the advice to look at how others have solved synchronization issues as it is very difficult to wrap your mind around. Confusion about what is going on in an environment with multiple threads and shared data typically leads to issues like deadlocks down the road.
The major problem with your code is that your threads are not synchronized.
Remember that by default threads execution isn't ordered and isn't sequenced, so consumer threads actually can be (and in your particular case are) finished before any producer thread produces any data.
To make sure consumers will be run after producers finished its work you need to use thread::join() function on producer threads, it will stop main thread execution until producers exit:
// Start producers
...
p.m_Thread.join(); // Wait p to complete
p2.m_Thread.join(); // Wait p2 to complete
// Start consumers
...
This will do the trick, but probably this is not good for typical producer-consumer use case.
To achieve more useful case you need to fix consumer function.
Your consumer function actually doesn't wait for produced data, it will just exit if stack is empty and never consume any data if no data were produced yet.
It shall be like this:
void remove(ProtectedStack &s)
{
// Place your actual exit condition here,
// e.g. count of consumed elements or some event
// raised by producers meaning no more data available etc.
// For testing/educational purpose it can be just while(true)
while(!_some_exit_condition_)
{
if(s.isEmpty())
{
// Second sleeping is too big, use milliseconds instead
boost::posix_time::milliseconds workTime(1);
boost::this_thread::sleep(workTime);
}
else
{
s.pop();
}
}
}
Another problem is wrong thread constructor usage:
m_Thread = boost::thread(&Producer::fillStack, this, s);
Quote from Boost.Thread documentation:
Thread Constructor with arguments
template <class F,class A1,class A2,...>
thread(F f,A1 a1,A2 a2,...);
Preconditions:
F and each An must by copyable or movable.
Effects:
As if thread(boost::bind(f,a1,a2,...)). Consequently, f and each an are copied into
internal storage for access by the new thread.
This means that each your thread receives its own copy of s and all modifications aren't applied to s but to local thread copies. It's the same case when you pass object to function argument by value. You need to pass s object by reference instead - using boost::ref:
void produce(ProtectedStack& s)
{
m_Thread = boost::thread(&Producer::fillStack, this, boost::ref(s));
}
void consume(ProtectedStack& s)
{
m_Thread = boost::thread(&Consumer::remove, this, boost::ref(s));
}
Another issues is about your mutex usage. It's not the best possible.
Why do you use mutex from Signals2 library? Just use boost::mutex from Boost.Thread and remove uneeded dependency to Signals2 library.
Use RAII wrapper boost::lock_guard instead of direct lock/unlock calls.
As other people mentioned, you shall protect with lock all members of ProtectedStack.
Sample:
boost::mutex m;
void push(int x)
{
boost::lock_guard<boost::mutex> lock(m);
m_Stack.push(x);
}
void pop()
{
boost::lock_guard<boost::mutex> lock(m);
if(!m_Stack.empty()) m_Stack.pop();
}
int size()
{
boost::lock_guard<boost::mutex> lock(m);
return m_Stack.size();
}
bool isEmpty()
{
boost::lock_guard<boost::mutex> lock(m);
return m_Stack.empty();
}
int top()
{
boost::lock_guard<boost::mutex> lock(m);
return m_Stack.top();
}
You're not checking that the producing thread has executed before you try to consume. You're also not locking around size/empty/top... that's not safe if the container's being updated.
I have an object that once created executes many tasks in the background, but should block untill /all/ posted tasks are finished. I.e.:
struct run_many{
boost::asio::io_service m_io_service;
boost::thread_group m_threads;
boost::asio::signal_set m_signals;
void evaluate(std::string work, int i){ /*...*/ }
void run_tasks(int tasks, std::string work){
{
boost::asio::io_service::work w(m_io_service); //
for(int i=0;i<tasks;i++)
m_io_service.post(boost::bind(&run_many::evaluate, this, work, i));
}
//m_io_service.run(); // blocks forever
m_io_service.stop(); // seems to cut off queued jobs
m_threads.join_all(); // works only after m_io_service.stop()
}
run_many(int n_workers)
{
m_threads.create_thread(boost::bind(&boost::asio::io_service::run,m_io_service);
}
};
So I am stuck... it seems that I can either wait forever or cut off the queue after the currently running job in each thread. There must be something I'm missing in the docs?
According to the documentation this idea should work(pseudocode):
...
// put a few handlers into io_service
...
// don't forget to destroy all boost::asio::io_service::work objects
while (m_io_service.stopped())
{
m_io_service.run();
}
// when code reaches this line m_io_service will be in stopped state and all handlers will be executed
// this code can be called without any doubts about being locked
m_threads.join_all();
Is it possible to perform an asynchronous wait (read : non-blocking) on a conditional variable in boost::asio ? if it isn't directly supported any hints on implementing it would be appreciated.
I could implement a timer and fire a wakeup even every few ms, but this is approach is vastly inferior, I find it hard to believe that condition variable synchronization is not implemented / documented.
If I understand the intent correctly, you want to launch an event handler, when some condition variable is signaled, in context of asio thread pool? I think it would be sufficient to wait on the condition variable in the beginning of the handler, and io_service::post() itself back in the pool in the end, something of this sort:
#include <iostream>
#include <boost/asio.hpp>
#include <boost/thread.hpp>
boost::asio::io_service io;
boost::mutex mx;
boost::condition_variable cv;
void handler()
{
boost::unique_lock<boost::mutex> lk(mx);
cv.wait(lk);
std::cout << "handler awakened\n";
io.post(handler);
}
void buzzer()
{
for(;;)
{
boost::this_thread::sleep(boost::posix_time::seconds(1));
boost::lock_guard<boost::mutex> lk(mx);
cv.notify_all();
}
}
int main()
{
io.post(handler);
boost::thread bt(buzzer);
io.run();
}
I can suggest solution based on boost::asio::deadline_timer which works fine for me. This is kind of async event in boost::asio environment.
One very important thing is that the 'handler' must be serialised through the same 'strand_' as 'cancel', because using 'boost::asio::deadline_timer' from multiple threads is not thread safe.
class async_event
{
public:
async_event(
boost::asio::io_service& io_service,
boost::asio::strand<boost::asio::io_context::executor_type>& strand)
: strand_(strand)
, deadline_timer_(io_service, boost::posix_time::ptime(boost::posix_time::pos_infin))
{}
// 'handler' must be serialised through the same 'strand_' as 'cancel' or 'cancel_one'
// because using 'boost::asio::deadline_timer' from multiple threads is not thread safe
template<class WaitHandler>
void async_wait(WaitHandler&& handler) {
deadline_timer_.async_wait(handler);
}
void async_notify_one() {
boost::asio::post(strand_, boost::bind(&async_event::async_notify_one_serialized, this));
}
void async_notify_all() {
boost::asio::post(strand_, boost::bind(&async_event::async_notify_all_serialized, this));
}
private:
void async_notify_one_serialized() {
deadline_timer_.cancel_one();
}
void async_notify_all_serialized() {
deadline_timer_.cancel();
}
boost::asio::strand<boost::asio::io_context::executor_type>& strand_;
boost::asio::deadline_timer deadline_timer_;
};
Unfortunately, Boost ASIO doesn't have an async_wait_for_condvar() method.
In most cases, you also won't need it. Programming the ASIO way usually means, that you use strands, not mutexes or condition variables, to protect shared resources. Except for rare cases, which usually focus around correct construction or destruction order at startup and exit, you won't need mutexes or condition variables at all.
When modifying a shared resource, the classic, partially synchronous threaded way is as follows:
Lock the mutex protecting the resource
Update whatever needs to be updated
Signal a condition variable, if further processing by a waiting thread is required
Unlock the mutex
The fully asynchronous ASIO way is though:
Generate a message, that contains everything, that is needed to update the resource
Post a call to an update handler with that message to the resource's strand
If further processing is needed, let that update handler create further message(s) and post them to the apropriate resources' strands.
If jobs can be executed on fully private data, then post them directly to the io-context instead.
Here is an example of a class some_shared_resource, that receives a string state and triggers some further processing depending on the state received. Please note, that all processing in the private method some_shared_resource::receive_state() is fully thread-safe, as the strand serializes all calls.
Of course, the example is not complete; some_other_resource needs a similiar send_code_red() method as some_shared_ressource::send_state().
#include <boost/asio>
#include <memory>
using asio_context = boost::asio::io_context;
using asio_executor_type = asio_context::executor_type;
using asio_strand = boost::asio::strand<asio_executor_type>;
class some_other_resource;
class some_shared_resource : public std::enable_shared_from_this<some_shared_resource> {
asio_strand strand;
std::shared_ptr<some_other_resource> other;
std::string state;
void receive_state(std::string&& new_state) {
std::string oldstate = std::exchange(state, new_state);
if(state == "red" && oldstate != "red") {
// state transition to "red":
other.send_code_red(true);
} else if(state != "red" && oldstate == "red") {
// state transition from "red":
other.send_code_red(false);
}
}
public:
some_shared_resource(asio_context& ctx, const std::shared_ptr<some_other_resource>& other)
: strand(ctx.get_executor()), other(other) {}
void send_state(std::string&& new_state) {
boost::asio::post(strand, [me = weak_from_this(), new_state = std::move(new_state)]() mutable {
if(auto self = me.lock(); self) {
self->receive_state(std::move(new_state));
}
});
}
};
As you see, posting always into ASIO's strands can be a bit tedious at first. But you can move most of that "equip a class with a strand" code into a template.
The good thing about message passing: As you are not using mutexes, you cannot deadlock yourself anymore, even in extreme situations. Also, using message passing, it is often easier to create a high level of parallelity than with classical multithreading. On the downside, moving and copying around all these message objects is time consuming, which can slow down your application.
A last note: Using the weak pointer in the message formed by send_state() facilitates the reliable destruction of some_shared_resource objects: Otherwise, if A calls B and B calls C and C calls A (possibly only after a timeout or similiar), using shared pointers instead of weak pointers in the messages would create cyclic references, which then prevents object destruction. If you are sure, that you never will have cycles, and that processing messages from to-be-deleted objects doesn't pose a problem, you can use shared_from_this() instead of weak_from_this(), of course. If you are sure, that objects won't get deleted before ASIO has been stopped (and all working threads been joined back to the main thread), then you can also directly capture the this pointer instead.
FWIW, I implemented an asynchronous mutex using the rather good continuable library:
class async_mutex
{
cti::continuable<> tail_{cti::make_ready_continuable()};
std::mutex mutex_;
public:
async_mutex() = default;
async_mutex(const async_mutex&) = delete;
const async_mutex& operator=(const async_mutex&) = delete;
[[nodiscard]] cti::continuable<std::shared_ptr<int>> lock()
{
std::shared_ptr<int> result;
cti::continuable<> tail = cti::make_continuable<void>(
[&result](auto&& promise) {
result = std::shared_ptr<int>((int*)1,
[promise = std::move(promise)](auto) mutable {
promise.set_value();
}
);
}
);
{
std::lock_guard _{mutex_};
std::swap(tail, tail_);
}
co_await std::move(tail);
co_return result;
}
};
usage eg:
async_mutex mutex;
...
{
const auto _ = co_await mutex.lock();
// only one lock per mutex-instance
}