tf::DataPipe< Input, Output, C > Class Template Reference

class to create a stage in a data-parallel pipeline More...

#include <taskflow/algorithm/data_pipeline.hpp>

Public Types
using	callable_t = C
	callable type of the data pipe
using	input_t = Input
	input type of the data pipe
using	output_t = Output
	output type of the data pipe

Public Member Functions
	DataPipe ()=default
	default constructor
	DataPipe (PipeType d, callable_t &&callable)
	constructs a data pipe
PipeType	type () const
	queries the type of the data pipe
void	type (PipeType type)
	assigns a new type to the data pipe
template<typename U >
void	callable (U &&callable)
	assigns a new callable to the data pipe

Friends
template<typename... Ps>
class	DataPipeline

Detailed Description

template<typename Input, typename Output, typename C>
class tf::DataPipe< Input, Output, C >

class to create a stage in a data-parallel pipeline

A data pipe represents a stage of a data-parallel pipeline. A data pipe can be either parallel direction or serial direction (specified by tf::PipeType) and is associated with a callable to invoke by the pipeline scheduler.

You need to use the template function, tf::make_data_pipe, to create a data pipe. The input and output types of a tf::DataPipe should be decayed types (though the library will always decay them for you using std::decay) to allow internal storage to work. The data will be passed by reference to your callable, at which you can take it by copy or reference.

tf::make_data_pipe<int, std::string>(
  tf::PipeType::SERIAL, 
  [](int& input) {return std::to_string(input + 100);}
);

In addition to the data, you callable can take an additional reference of tf::Pipeflow in the second argument to probe the runtime information for a stage task, such as its line number and token number:

tf::make_data_pipe<int, std::string>(
  tf::PipeType::SERIAL, 
  [](int& input, tf::Pipeflow& pf) {
    printf("token=%lu, line=%lu\n", pf.token(), pf.line());
    return std::to_string(input + 100);
  }
);

Constructor & Destructor Documentation

DataPipe()

template<typename Input , typename Output , typename C >

tf::DataPipe< Input, Output, C >::DataPipe ( PipeType d, callable_t && callable )

inline

constructs a data pipe

You should use the helper function, tf::make_data_pipe, to create a DataPipe object, especially when you need tf::DataPipe to automatically deduct the lambda type.

Member Function Documentation

callable()

template<typename Input , typename Output , typename C >

template<typename U >

void tf::DataPipe< Input, Output, C >::callable ( U && callable )

inline

assigns a new callable to the data pipe

Template Parameters

U	callable type

Parameters

callable

a callable object constructible from the callable type of this data pipe

Assigns a new callable to the pipe using universal forwarding.

type()

template<typename Input , typename Output , typename C >

PipeType tf::DataPipe< Input, Output, C >::type ( ) const

inline

queries the type of the data pipe

A data pipe can be either parallel (tf::PipeType::PARALLEL) or serial (tf::PipeType::SERIAL).

---

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

• taskflow/algorithm/data_pipeline.hpp