Thread: my questions on Boost

Hello all .
Im starting to learn boost , and now im on threads.
and for the time being , im trying to use two threads to sum lets say 1000000 numbers as an example , i want these threads iterate through a loop , one after another , i mean if thread one counted to 3 , thread 2 continues with 4 and so on ,how can i share that variable with these threads ? how can i do that ?
i already tried to use mutex .

Code:

//in the name of GOD
#define BOOST_THREAD_USE_LIB
#include <iostream>
#include <boost/date_time/posix_time/posix_time.hpp>
#include <boost/thread.hpp>


boost::uint64_t  i = 0;
boost::uint64_t  sum=0;
boost::mutex mutex;

void IteratorFunc()
{
     mutex.lock();
    for (i ; i<1000000000 ; i++)
        {
            sum+=i;
        }
        mutex.unlock();
}

using namespace std;

int main()
{
    boost::posix_time::ptime start = boost::posix_time::microsec_clock::local_time();
    boost::thread thread(IteratorFunc);
    boost::thread thread2(IteratorFunc);

    boost::posix_time::ptime end = boost::posix_time::microsec_clock::local_time();
     cout << "sum =\t" << sum<< "\t"<<end-start<<endl;
     thread.join();
     thread2.join();

    return 0;
}

but it seems im doing it wrong! , i also tried using the condition variable (condition_variable_any) so that threads can notify eachother and thus speed up the process but im stuck how to use wait(mutex) in first place cause there is only one function!
i dont want to separete the loop into two functions! i just want that one function to be used by two threads in the specified manner
can someone help get this going ?
------
how can i get all threads ID issued by me in my app?
how can i iterate through running threads in my app ?
is there any kind of means to get all the running threads using boost library? if it does whats the calss? if it doesnt how can i do that?
can i resume a thread after pausing it ? ( how can i pause a thread rather than using sleep? )

thank you in advance

and also why is this not working ?
im trying to get track of all threads i create and then whenever i wish , terminate them or query them .
what am i missing here?

Code:

    vector<thread> threadStore;
    
    threadStore.push_back(thrd);
    threadStore.push_back(thrd2);
    
    BOOST_FOREACH(thread th,threadStore)
    {
        cout<<th.get_id();
    }

and i have difficulties using the timed_join() method as in

Code:

boost::thread mythread(myfunction);
mythread.timed_join(sth should be here!);

i dont know what im supposed to send as an argument so that it works , i checked the documentation , it says either TimeDuration ? ! or system_time?
im clueless , can some one please do me a favor and enlight me on this too ?

Haven't used boost threads but this:

Code:

void IteratorFunc()
{
     mutex.lock();
    for (i ; i<1000000000 ; i++)
        {
            sum+=i;
        }
        mutex.unlock();
}

Certainly won't do this:

im trying to use two threads to sum lets say 1000000 numbers as an example , i want these threads iterate through a loop , one after another , i mean if thread one counted to 3 , thread 2 continues with 4 and so on ,how can i share that variable with these threads ?

Instead, the first thread to acquire the lock will increment i to 1000000 then release it, then next thread to acquire the lock will have nothing to do. You want something more like:

Code:

void IteratorFunc()
{

    for (i ; i<1000000000 ; i++)
        {
            mutex.lock();
            sum+=i;
            mutex.unlock();
        }

}

But even in this case, one thread will possibly do that in one pass anyway. Put a short sleep or something in there if you are trying to figure out how this all hangs together.

Originally Posted by MK27

Haven't used boost threads but this:

Code:

void IteratorFunc()
{
     mutex.lock();
    for (i ; i<1000000000 ; i++)
        {
            sum+=i;
        }
        mutex.unlock();
}

Certainly won't do this:



Instead, the first thread to acquire the lock will increment i to 1000000 then release it, then next thread to acquire the lock will have nothing to do. You want something more like:

Code:

void IteratorFunc()
{

    for (i ; i<1000000000 ; i++)
        {
            mutex.lock();
            sum+=i;
            mutex.unlock();
        }

}

But even in this case, one thread will possibly do that in one pass anyway. Put a short sleep or something in there if you are trying to figure out how this all hangs together.

thank you very much , i've already tried the second bit of code which you too suggested as well , as you mentioned again , the first thread just does the whole chore , is that because the given chore is not that time-consuming ? :-/ ?
do we have sth , liked timed_mutex? i mean so that a thread releases a mutex in a period of time specified by locking method ? will the timed_lock() method do that ?
by the way why doesnt this work ? :

Code:

boost::thread::sleep(boost::posix_time::microseconds(200));

edited :
got it , i should have used "this_thread instead" of" thread."
and i also tried using yield() , so that when each one of those two threads gets one count done , they give out their remaining time slice , so that the other one could get inside .! seems not working though! what am i missing here again ? ( in know thats not the correct way , but just wana know what is happening here )
and also for that

Code:

boost::thread mythread(myfunction);

mythread.timed_join(sth should be here!);

issue , i found

Code:

boost::posix_time

, seems to be working , if im wrong please sb enlighten me

i managed to get two threads iterate through the loop by adding a sleep() with 200 microseconds delay for each threads .
i still want to know if we have a timed_lock so that any thread releases its mutex in the given timeslice .
i also would appreciate any help on this ?
i'm trying to get the tack of all created treads by me , by using thread_groups , i can add them to the group , but i cant use foreach to traverse inside that group .

Code:

    boost::thread_group threadstore;

    threadstore.add_thread(&thread);
    threadstore.add_thread(&thread2);
    BOOST_FOREACH(boost::thread t ,threadstore.threads)
    {
        cout<<t.get_id();
    }

and the threadstore has a threads member , which i dont know how to work around it , there is also an m member , which i have no clue of what it is ! and where it came form! and what to do with it! i couldnt find any information on these members on the documentation either !
any help ?

What is this lock for? Why won't you just split this one big set of numbers into smaller subsets and then let each thread sum one set, finally adding computed sums in the main thread?

Originally Posted by kmdv

What is this lock for? Why won't you just split this one big set of numbers into smaller subsets and then let each thread sum one set, finally adding computed sums in the main thread?

because im trying to teach myself this way , so that in area and situations where the data are not separable like this simple loop , i have a solution for that beforehand . thats why im trying to get this thing done this way .

Originally Posted by Masterx

thank you very much , i've already tried the second bit of code which you too suggested as well , as you mentioned again , the first thread just does the whole chore , is that because the given chore is not that time-consuming ? :-/ ?

Yes. The computer can count to a million pretty quickly. On a multi-core system, there is some chance that you might get the two threads running in parallel and trading off a bit, although, using your code, how would you know? Throw some stderr reporting in there, that will also slow things down, and pass each thread a unique id to report with.

and i also tried using yield() , so that when each one of those two threads gets one count done , they give out their remaining time slice , so that the other one could get inside .! seems not working though!

Yeah, I noticed that yield() think looking at the docs. Did you unlock the mutex before you yielded?

I think a possible complication with that will be the OS scheduler. I don't think you can tell it what to do in this regard, so if Boost yield() just causes the thread to cease execution in hopes that some other thread will then get picked up, that "other thread" might be the same thread again.

You should really post the code you are using. Eventually someone who uses Boost::Threads will come along, but it might be while you are off-line, etc, and plain language descriptions of "what I did next" are very ambiguous.

Also keep in mind that using some kind of sleep is fine for experimenting with how threads work, but don't get too into it as it is not reliable as a means of synchronization (even if it seems to work sometimes) and tends to negate the purpose of using threads.

thank you very much Mk27 ,
here is the code :

Code:

#define BOOST_THREAD_USE_LIB
#include <iostream>
#include <boost/date_time/posix_time/posix_time.hpp>
#include <boost/thread.hpp>
using namespace std;

boost::uint64_t  i = 0;
boost::uint64_t  sum=0;
boost::mutex mutex;

void IteratorFunc()
{

    for (i ; i<100000000000; i++)
    {
        mutex.lock();
        sum+=i;
       cout<<i<<"\t"<<boost::this_thread::get_id()<<endl;
        mutex.unlock();
        boost::this_thread::sleep(boost::posix_time::microseconds(200));
       //boost::this_thread::yield();
    }

}


int main()
{
    boost::posix_time::ptime start = boost::posix_time::microsec_clock::local_time();
    boost::thread thread(IteratorFunc);
    boost::thread thread2(IteratorFunc);

    boost::thread_group threadstore;

    threadstore.add_thread(&thread);
    threadstore.add_thread(&thread2);

    BOOST_FOREACH(boost::thread t ,threadstore.threads)
    {
        cout<<t.get_id();
    }
    thread.join();
    thread2.join();
    boost::posix_time::ptime end = boost::posix_time::microsec_clock::local_time();



    cout << "sum =\t" << sum<< "\t"<<end-start<<endl;
    return 0;
}

i cant get much out of boost::threading features , it doesnt have really much to offer .
that yield() method , you are right , the scheduler takes the control and it is not guaranteed that my expected thread gets the chance to own the mutex! ( actually despite this very situation which it actually works just fine!!) so i cant rely on that any more.

and about using thread_groups , it seems its kinda useless , you cant iterate through threads , or even rely on that , cause i heard that in the next version its going to be changed and the verion after that!
there are no means in boost::threading to get the running threads nor is there any means to get the status of threads , nor is there any means to pause/resume threads and alot more .
and according to my research , alot of functionalities such as threads priority needs me to implement the needed feature using the host OS API , what i was trying to avoid using boost!
so the answers to my questions :

how can i get all threads ID issued by me in my app? NO
how can i iterate through running threads in my app ? NO
is there any kind of means to get all the running threads using boost library? NO
if it does whats the calss? if it doesnt how can i do that? use the OS API to do that!
can i resume a thread after pausing it ? NO

i'm really disapointed , there is'nt any kind of threading samples on boost either and this just makes it worst . all in all i think for a novice like me , with a C# background on threads , boost is not definitely an answer .Wxwidgets may be

Originally Posted by Masterx

i'm really disapointed , there is'nt any kind of threading samples on boost either and this just makes it worst . all in all i think for a novice like me , with aC# background on threads , boost is not definitely an answer .Wxwidgets may be

Try std::thread s from C++11. They are similar to boost's but differ in some places. I find it somewhat easier to comprehend that boost threads.
If you need a more high level api, try Intel's tbb. As far as I've learnt, it is quite easy.

For std::threads, you can pust the threads onto a vector and iterate throught them later to ..say..call join.
There exists a get_id function of the thread in C++11.
There also exists a way to get a native handle to the thread...but I'm not sure what it can be used for.

>is there any kind of means to get all the running threads using boost library? NO
That is the C++ way of thinking... you get only what you specify. For example, you can't get to iterate over all member functions of a class.

>can i resume a thread after pausing it ? NO
Look into sleep_for and sleep_until (again.. in std::thread.. no idea about boost's). Beyond that, it is the OS's scheduler's job, not your.

Originally Posted by phantomotap

O_o

Okay.

Let's just back this "rollicoster world" up! (Vincent Price was awesome.)

The "answer" to those questions is: why on earth do you need those features?

Your idea of threads seems to be poorly realized.

I'm totally not trying to be rude here; you really don't seem to know the purpose and use of the threading and serialization primitives "Boost" provides.

This "it is not guaranteed that my expected thread gets the chance to own the mutex" craziness for example. The job of a mutual exclusive lock is strictly ordering access of a facility. A mutual exclusive lock doesn't guarantee strict "fair" sharing. That is simply not the tool for that job. You can make such a facility, yes even using primitives provided by "Boost", but you aren't going to find such a beast ready "out of the box" from most libraries simply because it is rarely a necessary thing.

Soma

I agree with you on the part that says i dont know much about threading , thats right , and im just starting to learn about all of these .
actually i was trying to experiment all the possibilities with the wrong example .
there is a reason behind all of those craziness and sure the reason is my lack of knowledge on the field .
and for the part why i would need such features :
i imagined a situation in which there are messages going to a message dispatcher , message dispatcher gets one message at the time and process it and then tries to do what it was ordered to , so it fires a worker thread , the worker thread again
either deals with reading or writing on a buffer ( suppose the message is to turn a light bulb on ! ) if i make the dispatcher multi treaded , instead of one message at a time it will get several messages at the time and will process them , then each of these threads will execute workers to do the job and so on .
i needed a mechanism to make sure how a specific message is dealt with , was it successful or not ? how could i know about the status of threads being created in the process ? how could i impose some priority on threads , so that for example a writer thread has a more priority towards the reader threads , and how could i terminate a thread if the need arises ? how could i manage the concurrency of the application if i didnt know how many threads are working or blocked or what ever .
i know i myself could implement them using windows api , or maybe some of these questions are just irrelavent and are because of poor planning or the scenario , but my first reason to start threading with boost was that i dont implement much of stuff, that some standard gets me in the correct way , i had a look at wxwidgets threading samples and its thread class , most of the features i requested here or asked here , are present in wxwidgets(http://docs.wxwidgets.org/trunk/classwx_thread.html) ,
i think one of the most crucial reasons for my misplanning or mis understanding of threading capabilities , is just because of the lack of samples on threading itself , if there were practical examples demonstrating common scenarios , people like me would have a greater chance of success and not getting lost like this .!

Originally Posted by manasij7479

Try std::thread s from C++11. They are similar to boost's but differ in some places. I find it somewhat easier to comprehend that boost threads.
If you need a more high level api, try Intel's tbb. As far as I've learnt, it is quite easy.

For std::threads, you can pust the threads onto a vector and iterate throught them later to ..say..call join.
There exists a get_id function of the thread in C++11.
There also exists a way to get a native handle to the thread...but I'm not sure what it can be used for.

>is there any kind of means to get all the running threads using boost library? NO
That is the C++ way of thinking... you get only what you specify. For example, you can't get to iterate over all member functions of a class.

>can i resume a thread after pausing it ? NO
Look into sleep_for and sleep_until (again.. in std::thread.. no idea about boost's). Beyond that, it is the OS's scheduler's job, not your.

thank you very much . but the problem is , no compiler supports all threading capabilities of C++11 yet. the gcc 4.7.0 has a partial implementation of C++11 threading but not much ( http://gcc.gnu.org/gcc-4.7/cxx0x_status.html ) , and those parts are of no use really .i need to wait for the 4.8.0 or 4.9.0 to see if they get implemented .