#include <ring_graph_builder.h>

Inheritance diagram for fpmas::graph::RingGraphBuilder< T >:

Collaboration diagram for fpmas::graph::RingGraphBuilder< T >:

Public Member Functions
	RingGraphBuilder (std::size_t k, RingType ring_type=CYCLE)

std::vector< api::graph::DistributedNode< T > * >	build (api::graph::DistributedNodeBuilder< T > &node_builder, api::graph::LayerId layer, api::graph::DistributedGraph< T > &graph) override

virtual std::vector< api::graph::DistributedNode< T > * >	build (DistributedNodeBuilder< T > &node_builder, LayerId layer, DistributedGraph< T > &graph)=0

Detailed Description

template<typename T>
class fpmas::graph::RingGraphBuilder< T >

Implements an api::graph::DistributedGraphBuilder algorithm to build ring shaped graphs.

The ring can be linked in one direction (LOOP) or in both directions (CYCLE).

The number of closest neighbors to which each node is connected can be controlled with the parameter k.

Examples

N_processes	LOOP, k=1		LOOP, k=2
1
4
4

N_processes	CYCLE, k=1		CYCLE, k=2
1
4
4

Constructor & Destructor Documentation

◆ RingGraphBuilder()

template<typename T >

fpmas::graph::RingGraphBuilder< T >::RingGraphBuilder	(	std::size_t	k,
		RingType	ring_type = `CYCLE`
	)

inline

RingGraphBuilder constructor.

Parameters

k	the connected nearest neighbors count of each node in each direction. So, if `k=2` in a LOOP, each node has 2 outgoing neighbors. If `k=2` in a CYCLE, each node has 2 outgoing neighbors in each direction, so 4 in total.
ring_type	ring type: CYCLE or LOOP

Member Function Documentation

◆ build()

template<typename T >

std::vector< api::graph::DistributedNode< T > * > fpmas::graph::RingGraphBuilder< T >::build	(	api::graph::DistributedNodeBuilder< T > &	node_builder,
		api::graph::LayerId	layer,
		api::graph::DistributedGraph< T > &	graph
	)

overridevirtual

Automatically builds the specified graph according to the current implementation.

Nodes are generated using the specified DistributedNodeBuilder. More precisely, DistributedNodeBuilder::localNodeCount() nodes will be inserted in the graph. Generated nodes can then be linked on the specified layer, according to rules defined by the implemented algorithm.

Notice that the specified graph is not required to be empty.

Contrary to the GraphBuilder, this process is synchronous and should be called from all processes.