Functional Definition of

Collaboration diagram for Functional Definition of:

Functions
template<typename T , typename Tag = stapl::use_default, typename Span = stapl::use_default>
skeletons::result_of::alltoall< T, Tag, Span >	stapl::skeletons::alltoall (void)
	This alltoall skeleton is a recursive-doubling alltoall and works only for power-of-two sizes. More...
template<bool is_pos_aware = false, typename Flows = stapl::use_default, typename Span = stapl::use_default, typename Op , typename Filter = skeletons::no_filter>
result_of::butterfly< is_pos_aware, Flows, Span, Op, Filter >	stapl::skeletons::butterfly (Op &&op, Filter &&filter=Filter())
	A butterfly skeleton is used in various applications including but not limited to a type of allreduce, FFT DIT. More...
template<bool is_pos_aware = false, typename Flows = stapl::use_default, typename Span = stapl::use_default, typename Op , typename Filter = skeletons::no_filter>
result_of::reverse_butterfly< is_pos_aware, Flows, Span, Op, Filter >	stapl::skeletons::reverse_butterfly (Op &&op, Filter &&filter=Filter())
	A reverse butterfly skeleton is used in various applications including but not limited to a type of allreduce, FFT DIT. More...
template<typename Flows = stapl::use_default, typename Op >
result_of::pointer_jumping< Flows, Op >	stapl::skeletons::pointer_jumping (Op &&op)
	A pointer jumping skeleton consists of levels of combining values of a given input with an element with the index . More...

Detailed Description

Skeletons Exchanging Data

Function Documentation

alltoall()

template<typename T , typename Tag = stapl::use_default, typename Span = stapl::use_default>

skeletons::result_of::alltoall<T, Tag, Span> stapl::skeletons::alltoall

(

void

)

This alltoall skeleton is a recursive-doubling alltoall and works only for power-of-two sizes.

In a recursive doubling alltoall each node sends its lower and upper half chunks to its butterflied pair in the parametric dependency. Each chunk is of size num_elem_per_node/2. The communication is therefore:

Note

The current default implementation can only be used for power-of-two sizes, since it uses a butterfly skeleton.

Template Parameters

T	the type of elements to be communicated
Span	the span to be used for butterfly inside
Tag	determines which type of alltoall to use

butterfly()

template<bool is_pos_aware = false, typename Flows = stapl::use_default, typename Span = stapl::use_default, typename Op , typename Filter = skeletons::no_filter>

result_of::butterfly<is_pos_aware, Flows, Span, Op, Filter> stapl::skeletons::butterfly	(	Op &&	op,
		Filter &&	filter = Filter()
	)

A butterfly skeleton is used in various applications including but not limited to a type of allreduce, FFT DIT.

Template Parameters

is_pos_aware	in some algorithms that use butterfly skeleton knowing the position of the element in the overall butterfly dependence graph is necessary, e.g., in FFT DIT. If you set this template parameter to true, this information will be sent to your workfunction by invoking its set_position
Flows	the flow to be used between the levels of the butterfly
Span	the iteration space for elements on each level of the butterfly skeleton

Parameters

op	the workfunction to be used in each butterfly parametric dependency.

Returns

a butterfly skeleton

See also

butterfly_pd

log_lazysize

reverse_butterfly()

template<bool is_pos_aware = false, typename Flows = stapl::use_default, typename Span = stapl::use_default, typename Op , typename Filter = skeletons::no_filter>

result_of::reverse_butterfly<is_pos_aware, Flows, Span, Op, Filter> stapl::skeletons::reverse_butterfly	(	Op &&	op,
		Filter &&	filter = Filter()
	)

A reverse butterfly skeleton is used in various applications including but not limited to a type of allreduce, FFT DIT.

Template Parameters

is_pos_aware	in some algorithms that use reverse butterfly skeleton knowing the position of the element in the overall reverse butterfly dependence graph is necessary, e.g., in FFT DIT. If you set this template parameter to true, this information will be sent to your workfunction by invoking its set_position
Flows	the flow to be used between the levels of the reverse butterfly
Span	the iteration space for elements on each level of the reverse butterfly skeleton

Parameters

op	the workfunction to be used in each reverse butterfly parametric dependency.

Returns

a reverse butterfly skeleton

See also

butterfly_pd

log_lazysize

pointer_jumping()

template<typename Flows = stapl::use_default, typename Op >

result_of::pointer_jumping<Flows, Op> stapl::skeletons::pointer_jumping

(

Op &&

op

)

A pointer jumping skeleton consists of levels of combining values of a given input with an element with the index .

Pointer jumping is used in various algorithm such as Hillis-Steele scan.

Parameters

op	the operation to be used in each level

Template Parameters

Flows

the flows to be used in repeat