A parallel tasks computing system (with DAG dependency), implemented via sleeping-thread-pool.
Usage
Add itask.h and ParallelTasks.hpp into project, and include ParallelTasks.hpp.
Getting Start
It's very easy to use ParallelTasks class to parallelize multi-tasks.
First, define a customed task class MyTask (must inherit the interface ITask), and implement the runTask method.
class MyTask : public ITask
{
public:
void runTask(TaskID task_id, int job_idx, int num_jobs)
{
printf("TaskID = %u (idx = %d, total jobs = %d) MyRunner \n", task_id, job_idx, num_jobs);
}
};
class AnotherTask : public ITask
{
// ...
};The TaskID is assigned by class ParallelTasks (for debug purpose), in most cases, we do not need to care about it.
And we can write main program like this:
int main()
{
ParallelTasks parallel(4); // 4 threads to execute the tasks
MyTask *task = new MyTask();
AnotherTask *another = new AnotherTask();
TaskID A = parallel.addTaskWithDeps(task, 1, {}); // task has only 1 job
TaskID B = parallel.addTaskWithDeps(another, 1, {A}); // B must be executed after A
parallel.sync(); // execute the tasks we added above
}Simple Example
See example-simple.cpp, and we can type make simple to build it.
In this program, we construct a task graph like this:
And task A, B, C, D have {5, 4, 1, 2} sub-jobs, respectively. This example shows how does ParallelTasks system works.
Merge Sort Example
See example-merge-sort.cpp, and we can type make sorting to build it.
This program implements a multi-threads merge-sorting algorithm by ParallelTasks system.