A construct that maps parallel-reduce behavior to a MicroScheduler. More...

Public Member Functions
GTS_INLINE	ParallelReduce (MicroScheduler &scheduler, uint32_t priority=0)

template<typename TRange , typename TResultValue , typename TMapReduceFunc , typename TJoinFunc , typename TPartitioner = AdaptivePartitioner>
GTS_INLINE TResultValue	operator() (TRange const &range, TMapReduceFunc mapReduceFunc, TJoinFunc joinFunc, TResultValue identityValue, TPartitioner partitioner=AdaptivePartitioner(), void *userData=nullptr)

Detailed Description

A construct that maps parallel-reduce behavior to a MicroScheduler.

Constructor & Destructor Documentation

GTS_INLINE gts::ParallelReduce::ParallelReduce	(	MicroScheduler &	scheduler,
		uint32_t	priority = `0`
	)

inline

Creates a ParallelReduce object bound to the specified 'scheduler'. All parallel-reduce operations will be scheduled with the specified 'priority'.

template<typename TRange , typename TResultValue , typename TMapReduceFunc , typename TJoinFunc , typename TPartitioner = AdaptivePartitioner>

GTS_INLINE TResultValue gts::ParallelReduce::operator()	(	TRange const &	range,
		TMapReduceFunc	mapReduceFunc,
		TJoinFunc	joinFunc,
		TResultValue	identityValue,
		TPartitioner	partitioner = `AdaptivePartitioner()`,
		void *	userData = `nullptr`
	)

inline

Combines all the values in 'range' via 'reduceFunc'.

Parameters

range	The divisible range to iterate over.
mapReduceFunc	The function to execute on 'range'. Requires the signature: TResultValue mapReduceFunc(TRange range, void* userData, TaskContext const&). NOTE: This can be more that just a reduction operation. For instance one could fuse map-reduce here.
joinFunc	The function that joins the results of mapReduceFunc to produce the final reduction value. Requires the signature: TResultValue joinFunc(TResultValue const& lhs, TResultValue const& rhs, void* userData, TaskContext const&).
identityValue	An identity value for the join function.
partitioner	The partitioner object that determines when work is subdivided during scheduling.
userData	Optional data to be used in the passed in functions.