parallel::internal Namespace Reference

Classes
struct	Body
struct	EnableOpenMPSimdFor
struct	ParallelForWrapper
struct	ReductionOnSubranges
class	TBBPartitioner

Functions
template<typename F >
Body< F >	make_body (const F &f)
template<typename Iterator , typename Functor >
void	parallel_for (Iterator x_begin, Iterator x_end, const Functor &functor, const unsigned int grainsize)
template<typename Iterator , typename Functor >
void	parallel_for (Iterator x_begin, Iterator x_end, const Functor &functor, const unsigned int grainsize, const std::shared_ptr< tbb::affinity_partitioner > &partitioner)
template<typename RangeType , typename Function >
void	apply_to_subranges (const tbb::blocked_range< RangeType > &range, const Function &f)

Detailed Description

A class that wraps a TBB affinity partitioner in a thread-safe way. In Vector, we use a shared pointer to share an affinity partitioner between different vectors of the same size for improving data (and NUMA) locality. However, when an outer task does multiple vector operations, the shared pointer could lead to race conditions. This class only allows one instance to get a partitioner. The other objects cannot use that object and need to create their own copy.

Function Documentation

make_body()

template<typename F >

Body<F> parallel::internal::make_body

(

const F &

f

)

Take a function object and create a Body object from it. We do this in this helper function since alternatively we would have to specify the actual data type of F – which for function objects is often extraordinarily complicated.

Definition at line 143 of file parallel.h.

parallel_for() [1/2]

template<typename Iterator , typename Functor >

void parallel::internal::parallel_for	(	Iterator	x_begin,
		Iterator	x_end,
		const Functor &	functor,
		const unsigned int	grainsize
	)

Encapsulate tbb::parallel_for.

Definition at line 156 of file parallel.h.

parallel_for() [2/2]

template<typename Iterator , typename Functor >

void parallel::internal::parallel_for	(	Iterator	x_begin,
		Iterator	x_end,
		const Functor &	functor,
		const unsigned int	grainsize,
		const std::shared_ptr< tbb::affinity_partitioner > &	partitioner
	)

Encapsulate tbb::parallel_for when an affinite_partitioner is provided.

Definition at line 173 of file parallel.h.

apply_to_subranges()

template<typename RangeType , typename Function >

void parallel::internal::apply_to_subranges	(	const tbb::blocked_range< RangeType > &	range,
		const Function &	f
	)

Take a range argument and call the given function with its begin and end.

Definition at line 365 of file parallel.h.