support camera frame data pipeline ?

[lix19937]

It does not support data-driven pipeline or not ? like camera frame data pipeline, camera keep a fixed frame rate, see #148.
If Graph Processing Pipeline support, how to use ? Thanks !

Like OpenVX graph model.

Graph pipelining allow users to schedule a graph multiple times, without having to wait for a graph execution to complete. Each such execution of the graph operates on different input or output references.

// follow is my code, run the taskflow iteratively with changing frame data

I think follow is not right.

#include <taskflow/taskflow.hpp>

int main(){

  tf::Executor executor;
  tf::Taskflow taskflow("camera pipeline");
  
  // camera frame data   
  void *data_in{nullptr}; 
  
  // out of node(task) 
  A_OUT a_out;
  B_OUT b_out;
  C_OUT c_out;

  // create a task and attach it the camera data
  auto NodeA = taskflow.placeholder();
  NodeA.data(&data_in).work([NodeA, &a_out](){
    // task impl api
    // ...
   }); 
  
  auto NodeB = taskflow.placeholder();
  NodeB.data(&a_out).work([NodeB, &b_out](){
    // task impl api
    // ...
  });

  auto NodeC = taskflow.placeholder();
  NodeB.data(&b_out).work([NodeC, &c_out](){
    // task impl api
    // ...
  });

  NodeA.precede(NodeB);
  NodeB.precede(NodeC);

  // iteratively with changing camera data
  while(1){

    // get camera data here, define in other    
    GetCameraData(data_in);
    
    // here maybe wrong, here next frame`s should wait last frame process done   !!!   
    executor.run(taskflow).wait();

  }

  return 0;
}

[Fang Chao]

DFS Scheduler always process full Node.
A fixed Process is below：

// define a frame processing ITERATION ie. 10Hz
define ITERATION 100
while(True){
  
  while(pipline.alive() || pipline.ready()){
    auto start = std::chrono::high_resolution_clock::now();
    auto iteration = std::chrono::high_resolution_clock::now();
    if(pipline.timeout()){
      break;
    }
    if(data.error() || process.error()){
      break;
    }
    
    auto duration_ = std::chrono::duration_cast<std::chrono::microseconds>(
                         std::chrono::high_resolution_clock::now() - iteration)
                         .count();
    // Calculate the remaining time to wait
    auto remainingTime = std::chrono::microseconds(ITERATION) -
                         std::chrono::microseconds(duration_);

    if (remainingTime.count() > 0)
    {
      std::this_thread::sleep_for(remainingTime);
    }
  }

}

You should split your pipline, example creating a chain for three pipline. Giving each pipline a timeout, or node timeout.

[lix19937]

@fangchaooo thanks, can you show a sample about pipeline you defined ?