pp Namespace Reference

Parallel Processing primitives. More...

Classes
class	Controller
	An optional thread pool controller, which can be used to pause/resume/cancel the queued work items. More...
class	ThreadPool
	A basic thread pool (managing a fixed number of threads) More...
struct	WorkBatch
	A collection of work items to be processed in a fork/join fashion. More...
class	WorkItem
	Generic work item interface. More...

Functions
template<class T , class Body >
void	for_each (T &array, const Body &loop_body)
	Iterates over an array, with support for parallel execution. More...

Detailed Description

Parallel Processing primitives.

Function Documentation

for_each()

template<class T , class Body >

void pp::for_each	(	T &	array,
		const Body &	loop_body
	)

Iterates over an array, with support for parallel execution.

pp::for_each() offers an easy way to parallelize the processing of the elements in an array (the array must support size() and operator[]):

vector<int> array;

// parallel-for-loop
pp::for_each(array, [](int index, int& value) {
  ... read and/or mutate the value ...
});

Note

pp::for_each() loops can't be nested

Warning

Iterations will likely happen on different threads, so the access to any shared state must be properly synchronized:

int counter = 0;
pp::for_each(array, [&](int index, int& value) {
  ...
  // BUG: counter is captured by reference and shared between multiple
  // iterations, so the increment below introduces a data race
  ++counter;
});