ecl_threads Documentation

ecl_threads: c++ extensions for threaded programming

This package provides the c++ extensions for a variety of threaded programming tools. These are usually different on different platforms, so the architecture for a cross-platform framework is also implemented.


Threads are not defined in standard C++ and as a result are operating system and platform dependant. Linux generally uses POSIX threads, which are also available on windows, but the WIN32 API is usually preferred. Because of this, some standard cross-platform interfaces are developed here.



Include the following at the top of any translation unit that uses these container classes.

        #include <ecl/threads.hpp>

        // The thread classes
        using ecl::Mutex;
        using ecl::Thread;
        using ecl::Threadable;
        // Priorities
        using ecl::set_priority;
        using ecl::get_priority;
        using ecl::print_priority_diagnostics;
        using ecl::BackgroundPriority;
        using ecl::LowPriority;
        using ecl::NormalPriority;      
        using ecl::HighPriority;        
        using ecl::CriticalPriority;
        using ecl::RealTimePriority1;   
        using ecl::RealTimePriority2;   
        using ecl::RealTimePriority3;   
        using ecl::RealTimePriority4;   

You will also need to link to -lecl_threads.



The mutex is fairly standard, and on some systems can be more complex, but here the wrapper simply acts as a closed door with the usual lock, trylock and unlock features. The behaviour is also somewhat different depending on the platform.

On posix, the class is set up to run in two modes. When NDEBUG is not defined, it will do exception handling (via ecl's StandardException) for posix errors as well as configuring the mutexes for deadlock checking (see below for example code). Exception handling and deadlock checking are disabled when NDEBUG is defined.

                Mutex mutex;

                mutex.lock();     // Locks
                // do work here
                mutex.trylock();  // Locks
                mutex.trylock();  // Fails to lock and returns immediately.

                // mutex.lock();     // The DEADLOCK! Like this, the program will usually halt forever.

                // If NDEBUG is not defined, then on posix systems, you can catch deadlocks like this.
                try {
                } catch ( StandardException &e ) {
                   std::cout << e.what() << std::endl;

                // Typical output from a caught deadlock:
                // Location : /mnt/froody/work/code/cpp/projects/ecl/modules/core/src/lib/threads/mutex_pos.cpp:57
                // Flag     : The object was used incorrectly.
                // Detail   : DEADLOCK! The mutex has already been locked by this thread, it now has to wait on itself.


These can be configured via the ecl::set_priority() function using the ecl::Priority enum values as an abstraction to a platform's implementation (which varies quite significantly from platform to platform). Check the documentation's api for further details on its usage, particularly for posix which complicates things with both userland priorities and real time priorities.


The Thread class is a raii style object which initialises and automatically starts a thread when constructed and manages the thread cleanly when the thread object goes out of scope.


Construction can be done directly through free and member function handles or via function objects. Refer to the documentation on 'Function Objects' and 'Reference Wrappers' in ecl_utilities for more details about creating and using function objects/reference wrappers (note, use a reference wrapper if you want to pass a 'heavy' function object!).

                using ecl::utilities::generateFunctionObject;

                   int f() {}
                   int g(int i) {}
                   class A {
                           void f() {}
                           void g(int i) {}
                   class FunctionObject {
                           typedef void result_type;
                           void operator()() { //
                   // ...
                   A a;
                   FunctionObject function_object;

                   Thread thread1(f));                                   // Thread a nullary global function.
                   Thread thread2(generateFunctionObject(g, 3));         // Thread a bound unary global function.
                   Thread thread3(&A::f, a);                                                     // Thread a nullary member function.
                   Thread thread4(generateFunctionObject(&A::g, a, 2));  // Thread a bound unary member function.
                   Thread thread5(function_object);                      // Thread a nullary function object.
                   Thread thread6(ref(function_object));                 // Thread a reference to a nullary function object.


Also, this object is permitted to go out of scope without affecting the thread that it started (it may very well still be running!). When it goes out of scope, it simply detaches it and lets it clean itself up. At this point you only lose control of administration of the thread (joining, checking if it is running, cancelling etc).

                void g() {
                        for (int i = 0; i < 10; ++i ) {
                                cout << i << endl;

                void create_out_of_scope_thread() {
                        Thread thread(g);
                } // thread will go out of scope here.
                // ...
                // Cannot manage the thread from here, but it will continue running.
                sleep(10); // Note that we have no way of joining with it.

Thread Priority:

Thread priorities can be specified at construction (ecl::Thread) or at the call to execute a Threadable (Threadable::start()). This will impose the specified priority for scheduling for the lifetime of the thread.

If you wish to configure the thread's priority dynamically, you'll have to fall back to using the ecl::set_priority() function directly from inside the worker function.

Stack Size:

On embedded systems with no swap its important to watch how much stack memory you supply to the thread. This can be manually specified in the thread constructor.

                Thread thread(g,1024*1024); // allocates 1M to the thread instead of the system default which is usually 8M.


Other member functions include

Error handling is done in debug mode (i.e. -DNDEBUG is not set) via exceptions. These will throw and report any information on the resulting posix errors should they occur.


The threadable class implements a concept is for worker threads that want to retain state information (possibly for use by other parts of the program) in a class. Think of it as a threaded function object rather than a regular c-style threaded function.

Implementation wise, it provides an inheritable interface for your threading class. All the class needs to do is:

Note that it will not spawn multiple threads - it has a check that ensures it will only execute one thread at any point in time. It is designed to be something more akin to a thread function object rather than a thread factory.

                class A : Threadable {
                        void runnable() {
                        // thread work here

                int main() {
                        A a;
                        return 0;



Author(s): Daniel Stonier (
autogenerated on Thu Jan 2 2014 11:12:47