Thread Class Reference

Detailed Description

The Thread class represents an execution thread.

To create threads, derive a class from Thread and reimplement the run() function. That is what objects of the new class will execute in parallel when the start() function is called.

struct Worker : public Thread
{
    void run()
    {
        ... // what the function will do in a parallel thread
    }
};
 
Worker worker;
worker.start();
...
worker.join();

In compilers supporting lambdas threads can now be started with a lambda without defining a specific Thread subclass. These threads can get input data or keep state using variables captured by the lambda.

Thread thread([&]() {
    ... // what the thread does
});
...
thread.join();

The static functions Thread::parallel_for() and Thread::parallel_invoke() are deprecated. You can probably just use OpenMP.

See also Mutex, Semaphore and Lock.

#include <Thread.h>

Public Member Functions
virtual void	run ()
	The thread procedure.
void	start ()
	Starts a new thread by calling run() in parallel.
void	kill ()
	Cancels a thread, but you should not normally do this.
void	join ()
	Waits for this thread to end.
bool	finished () const
	Returns true if this thread has finished.
template<class F >
	Thread (const F &f, bool detached=false)
	Starts a thread given a function or lambda expression as the run() function.

Static Public Member Functions
static int	numProcessors ()
	Returns the number of logical processors or cores.
template<class F >
static void	parallel_for (int i0, int i1, const F &f, int nth=8)
	Emulates an OpenMP parallel for by running function f several times in parallel.
template<class F1 , class F2 >
static void	parallel_invoke (const F1 &f1, const F2 &f2)
	Runs the two functions/lambdas in parallel and returns when both are finished.
template<class F1 , class F2 , class F3 >
static void	parallel_invoke (const F1 &f1, const F2 &f2, const F3 &f3)
	Runs the 3 functions/lambdas in parallel and returns when all are finished.

Member Function Documentation

kill()

void kill ( )

inline

Cancels a thread, but you should not normally do this.

Deprecated:

Threads should end when their function ends

parallel_for()

template<class F >

static void parallel_for ( int  i0, int  i1, const F &  f, int  nth = 8  )

static

Emulates an OpenMP parallel for by running function f several times in parallel.

Function f must receive an integer argument which will get the values from i0 up to but not including i1.

Thread::parallel_for(i0, i1, [&](int i) {
    f(i);
});

Is equivalent to running:

for(int i=i0; i<i1; i++) {
    f(i);
}

but with iterations run in nth parallel threads.

Deprecated:

Use OpenMP

parallel_invoke() [1/2]

template<class F1 , class F2 >

static void parallel_invoke ( const F1 &  f1, const F2 &  f2  )

static

Runs the two functions/lambdas in parallel and returns when both are finished.

Deprecated:

Use OpenMP

parallel_invoke() [2/2]

template<class F1 , class F2 , class F3 >

static void parallel_invoke ( const F1 &  f1, const F2 &  f2, const F3 &  f3  )

static

Runs the 3 functions/lambdas in parallel and returns when all are finished.

Deprecated:

Use OpenMP

run()

virtual void run ( )

inlinevirtual

The thread procedure.

Reimplement this function to create new threads

---

The documentation for this class was generated from the following file:

• Thread.h