Select

Detailed Description

The functions in this module implement a pipeline stage in which, for all the active reads in the ScoringQueues, a group of seed hits is selected for a new round of Scoring. In terms of inputs and outputs, this stage takes the set of input ScoringQueues::active_reads, and fills the ScoringQueues::hits with a new set of hits, writing the HitQueues::read_id, HitQueues::seed, and HitQueues::loc fields. The hits' location specified by the HitQueues::loc field is expressed in Suffix Array coordinates. Finally, this stage is also responsible for producing a new set of ScoringQueues::active_reads.

inputs:

• SeedHitDequeArray
• ScoringQueues::active_reads

outputs:

• SeedHitDequeArray
• ScoringQueues::active_reads
• HitQueues::read_id
• HitQueues::seed
• HitQueues::loc

Modules
	SelectDetail

Classes
struct	nvbio::bowtie2::cuda::SelectBestApproxContext

Functions
void	nvbio::bowtie2::cuda::select_init (const uint32 count, const char *read_names, const uint32 *read_names_idx, const SeedHitDequeArrayDeviceView hits, uint32 *trys, uint32 *rseeds, const ParamsPOD params)
void	nvbio::bowtie2::cuda::select_init (BestApproxScoringPipelineState< EditDistanceScoringScheme > &pipeline, const ParamsPOD &params)
void	nvbio::bowtie2::cuda::select_init (BestApproxScoringPipelineState< SmithWatermanScoringScheme<> > &pipeline, const ParamsPOD &params)
__global__ void	nvbio::bowtie2::cuda::select_n_from_top_range_kernel (const uint32 begin, const uint32 count, const uint32 n_reads, const SeedHit *hit_data, const uint32 *hit_range_scan, uint32 *loc_queue, uint32 *seed_queue, uint32 *read_info)
void	nvbio::bowtie2::cuda::select (const SelectBestApproxContext context, const BestApproxScoringPipelineState< EditDistanceScoringScheme > &pipeline, const ParamsPOD params)
void	nvbio::bowtie2::cuda::select (const SelectBestApproxContext context, const BestApproxScoringPipelineState< SmithWatermanScoringScheme<> > &pipeline, const ParamsPOD params)
void	nvbio::bowtie2::cuda::select_all (const uint64 begin, const uint32 count, const uint32 n_reads, const uint32 n_hit_ranges, const uint64 n_hits, const SeedHitDequeArrayDeviceView hits, const uint32 *hit_count_scan, const uint64 *hit_range_scan, HitQueuesDeviceView scoring_queues)
template<typename ScoringScheme >
void	nvbio::bowtie2::cuda::select_init_t (BestApproxScoringPipelineState< ScoringScheme > &pipeline, const ParamsPOD &params)
template<typename BatchType , typename ContextType >
void	nvbio::bowtie2::cuda::select (const BatchType read_batch, SeedHitDequeArrayDeviceView hits, const ContextType context, ScoringQueuesDeviceView scoring_queues, const ParamsPOD params)
template<typename BatchType , typename ContextType >
void	nvbio::bowtie2::cuda::rand_select (const BatchType read_batch, SeedHitDequeArrayDeviceView hits, uint32 *rseeds, const ContextType context, ScoringQueuesDeviceView scoring_queues, const ParamsPOD params)
template<typename BatchType , typename ContextType >
void	nvbio::bowtie2::cuda::select_multi (const BatchType read_batch, SeedHitDequeArrayDeviceView hits, const ContextType context, ScoringQueuesDeviceView scoring_queues, const uint32 n_multi, const ParamsPOD params)
template<typename BatchType , typename ContextType >
void	nvbio::bowtie2::cuda::select (const BatchType read_batch, SeedHitDequeArrayDeviceView hits, uint32 *rseeds, const ContextType context, ScoringQueuesDeviceView scoring_queues, const uint32 n_multi, const ParamsPOD params)
template<typename ScoringScheme , typename ContextType >
void	nvbio::bowtie2::cuda::select_t (const ContextType context, const BestApproxScoringPipelineState< ScoringScheme > &pipeline, const ParamsPOD params)

Function Documentation

template<typename BatchType , typename ContextType >

void nvbio::bowtie2::cuda::rand_select	(	const BatchType	read_batch,
		SeedHitDequeArrayDeviceView	hits,
		uint32 *	rseeds,
		const ContextType	context,
		ScoringQueuesDeviceView	scoring_queues,
		const ParamsPOD	params
	)

Prepare for a round of seed extension by selecting the next SA row from each of the seed-hit deque arrays (SeedHitDequeArray) bound to the active-reads in the scoring queues (ScoringQueues::active_reads).

Definition at line 638 of file select_inl.h.

template<typename BatchType , typename ContextType >

void nvbio::bowtie2::cuda::select	(	const BatchType	read_batch,
		SeedHitDequeArrayDeviceView	hits,
		const ContextType	context,
		ScoringQueuesDeviceView	scoring_queues,
		const ParamsPOD	params
	)

Prepare for a round of seed extension by selecting the next SA row from each of the seed-hit deque arrays (SeedHitDequeArray) bound to the active-reads in the scoring queues (ScoringQueues::active_reads).

// group nvBowtie

Definition at line 617 of file select_inl.h.

template<typename BatchType , typename ContextType >

void nvbio::bowtie2::cuda::select	(	const BatchType	read_batch,
		SeedHitDequeArrayDeviceView	hits,
		uint32 *	rseeds,
		const ContextType	context,
		ScoringQueuesDeviceView	scoring_queues,
		const uint32	n_multi,
		const ParamsPOD	params
	)

Prepare for a round of seed extension by selecting the next SA row from each of the seed-hit deque arrays (SeedHitDequeArray) bound to the active-reads in the scoring queues (ScoringQueues::active_reads).

Definition at line 726 of file select_inl.h.

void nvbio::bowtie2::cuda::select	(	const SelectBestApproxContext	context,
		const BestApproxScoringPipelineState< EditDistanceScoringScheme > &	pipeline,
		const ParamsPOD	params
	)

Prepare for a round of seed extension by selecting the next SA row from each of the seed-hit deque arrays (SeedHitDequeArray) bound to the active-reads in the scoring queues (ScoringQueues::active_reads).

void nvbio::bowtie2::cuda::select	(	const SelectBestApproxContext	context,
		const BestApproxScoringPipelineState< SmithWatermanScoringScheme<> > &	pipeline,
		const ParamsPOD	params
	)

Prepare for a round of seed extension by selecting the next SA row from each of the seed-hit deque arrays (SeedHitDequeArray) bound to the active-reads in the scoring queues (ScoringQueues::active_reads).

void nvbio::bowtie2::cuda::select_all	(	const uint64	begin,
		const uint32	count,
		const uint32	n_reads,
		const uint32	n_hit_ranges,
		const uint64	n_hits,
		const SeedHitDequeArrayDeviceView	hits,
		const uint32 *	hit_count_scan,
		const uint64 *	hit_range_scan,
		HitQueuesDeviceView	scoring_queues
	)

Select next hit extensions for all-mapping

void nvbio::bowtie2::cuda::select_init	(	const uint32	count,
		const char *	read_names,
		const uint32 *	read_names_idx,
		const SeedHitDequeArrayDeviceView	hits,
		uint32 *	trys,
		uint32 *	rseeds,
		const ParamsPOD	params
	)

Initialize the hit-selection pipeline

void nvbio::bowtie2::cuda::select_init	(	BestApproxScoringPipelineState< EditDistanceScoringScheme > &	pipeline,
		const ParamsPOD &	params
	)

Initialize the hit-selection pipeline

void nvbio::bowtie2::cuda::select_init	(	BestApproxScoringPipelineState< SmithWatermanScoringScheme<> > &	pipeline,
		const ParamsPOD &	params
	)

Initialize the hit-selection pipeline

template<typename ScoringScheme >

void nvbio::bowtie2::cuda::select_init_t	(	BestApproxScoringPipelineState< ScoringScheme > &	pipeline,
		const ParamsPOD &	params
	)

Initialize the hit-selection pipeline

Definition at line 50 of file select_inl.h.

template<typename BatchType , typename ContextType >

void nvbio::bowtie2::cuda::select_multi	(	const BatchType	read_batch,
		SeedHitDequeArrayDeviceView	hits,
		const ContextType	context,
		ScoringQueuesDeviceView	scoring_queues,
		const uint32	n_multi,
		const ParamsPOD	params
	)

Prepare for a round of seed extension by selecting a set of up to 'n_multi' SA rows from each of the seed-hit deque arrays (SeedHitDequeArray) bound to the active-reads in the scoring queues (ScoringQueues::active_reads).

For each read in the input queue, this kernel generates:

1. one or zero output reads, in the main output read queue,
2. zero to 'n_multi' SA rows. These are made of three entries, one in 'loc_queue', identifying the corresponding SA index, one in 'seed_queue', storing information about the seed hit, and one in 'parent_queue', storing the index of the "parent" read in the output queue (i.e. the slot where the read is is being stored)

Definition at line 670 of file select_inl.h.

__global__ void nvbio::bowtie2::cuda::select_n_from_top_range_kernel	(	const uint32	begin,
		const uint32	count,
		const uint32	n_reads,
		const SeedHit *	hit_data,
		const uint32 *	hit_range_scan,
		uint32 *	loc_queue,
		uint32 *	seed_queue,
		uint32 *	read_info
	)

select next hit extensions from the top seed ranges

template<typename ScoringScheme , typename ContextType >

void nvbio::bowtie2::cuda::select_t	(	const ContextType	context,
		const BestApproxScoringPipelineState< ScoringScheme > &	pipeline,
		const ParamsPOD	params
	)

Prepare for a round of seed extension by selecting the next SA row from each of the seed-hit deque arrays (SeedHitDequeArray) bound to the active-reads in the scoring queues (ScoringQueues::active_reads).

Definition at line 793 of file select_inl.h.