Class ParallelBreadthFirstVisit
public class ParallelBreadthFirstVisit
extends java.lang.Object
To use this class you first create an instance, and then invoke visit(long). If you want perform
more visits preserving the marker state you can invoke visit(long) again.
By calling clear(), instead, you can reset marker (i.e., forget about visited nodes).
Alternatively, visitAll() will start a visit from all the nodes of the graph in a more efficient way.
After the visit, you can peek at the “big array“ marker field to discover details about the visit.
Depending on the parent value provided at construction time, marker
will be filled with parent information (e.g., with the index
of the parent node in the visit tree) or with a round number increased at each nonempty visit,
which act as a connected-component index if the graph is symmetric.
Observe that in the former case (if parent is true), marker will
contain the value -1 for the nodes that have not been reached by the visit, the parent of the node in the BFS tree
if the node was not the root, or the node itself for the root.
In the case of visit(long), queue and cutPoints, too, provide useful information. In
particular, the nodes in queue from the d-th to the (d +1)-th cutpoint
are exactly the nodes at distance d from the source.
Performance issues
This class needs three longs per node. If there are several available cores, breadth-first visits will be decomposed into relatively small tasks (small blocks of nodes in the queue at the same distance from the starting node) and each task will be assigned to the first available core. Since all tasks are completely independent, this ensures a very high degree of parallelism. However, on very sparse graphs the cost of keeping the threads synchronised can be extremely high, and even end up increasing the visit time.
Note that if the degree distribution is extremely skewed some cores might get stuck in the enumeration of the successors of some nodes with a very high degree.
-
Field Summary
Fields Modifier and Type Field Description LongBigArrayBigListcutPointsAt the end of a visit, the cutpoints ofqueue.ImmutableGraphgraphThe graph under examination.java.util.concurrent.atomic.AtomicLongArray[]markerbooleanparentWhethermarkercontains parent nodes or round numbers.LongBigArrayBigListqueueThe queue of visited nodes.longround -
Constructor Summary
Constructors Constructor Description ParallelBreadthFirstVisit(ImmutableGraph graph, int requestedThreads, boolean parent, ProgressLogger pl)Creates a new class for keeping track of the state of parallel breadth-first visits. -
Method Summary
Modifier and Type Method Description voidclear()longmaxDistance()Returns the maximum distance computed during the last visit (e.g., the eccentricity of the source).longnodeAtMaxDistance()Returns a node at maximum distance during the last visit (e.g., a node realising the positive eccentricity of the starting node).longvisit(long start)Performs a breadth-first visit of the given graph starting from the given node.longvisit(long start, long expectedSize)Performs a breadth-first visit of the given graph starting from the given node.voidvisitAll()Visits all nodes.
-
Field Details
-
graph
The graph under examination. -
queue
The queue of visited nodes. -
cutPoints
At the end of a visit, the cutpoints ofqueue. The d-th cutpoint is the first node in the queue at distance d. The last cutpoint is the queue size. -
parent
public final boolean parentWhethermarkercontains parent nodes or round numbers. -
marker
public final java.util.concurrent.atomic.AtomicLongArray[] markerThe marker “big” array; contains -1 for nodes that have not still been enqueued, the parent of the visit tree ifparentis true, or an index increased at each visit ifparentis false, which in the symmetric case is the index of the connected component of the node. It has the same form of a big array, but it is handled manually. -
round
public long round
-
-
Constructor Details
-
ParallelBreadthFirstVisit
public ParallelBreadthFirstVisit(ImmutableGraph graph, int requestedThreads, boolean parent, ProgressLogger pl)Creates a new class for keeping track of the state of parallel breadth-first visits.- Parameters:
graph- a graph.requestedThreads- the requested number of threads (0 forRuntime.availableProcessors()).parent- if true,markerwill contain parent nodes; otherwise, it will contain round numbers.pl- a progress logger, ornull.
-
-
Method Details
-
clear
public void clear() -
visit
public long visit(long start)Performs a breadth-first visit of the given graph starting from the given node.This method will increment
round.- Parameters:
start- the starting node.- Returns:
- the number of visited nodes.
- See Also:
visit(long,long)
-
visit
public long visit(long start, long expectedSize)Performs a breadth-first visit of the given graph starting from the given node.This method will increment
roundif at least one node is visited.- Parameters:
start- the starting node.expectedSize- the expected size (number of nodes) of the visit (for logging), or -1 to use the number of nodes of the graph.- Returns:
- the number of visited nodes.
-
visitAll
public void visitAll()Visits all nodes. Callsclear()initially.This method is more efficient than invoking
visit(long, long)on all nodes as threads are created just once. -
nodeAtMaxDistance
public long nodeAtMaxDistance()Returns a node at maximum distance during the last visit (e.g., a node realising the positive eccentricity of the starting node).- Returns:
- the maximum distance computed during the last visit.
-
maxDistance
public long maxDistance()Returns the maximum distance computed during the last visit (e.g., the eccentricity of the source).- Returns:
- the maximum distance computed during the last visit.
-