compact_kernel.cu File Reference

CUDPP kernel-level compact routines. More...

#include <cudpp_globals.h>
#include "sharedmem.h"

Functions

Compact Functions

template<bool isBackward>

__global__ void computeNumValidElements (size_t *d_numValidElements, const unsigned int *d_indices, const unsigned int *d_isValid, size_t numElements)

Compute the number of valid flags in an array given the array and its exclusive sum scan.

template<class T, bool isBackward>

__global__ void compactData (T *d_out, size_t *d_numValidElements, const unsigned int *d_indices, const unsigned int *d_isValid, const T *d_in, unsigned int numElements)

Consolidate non-null elements - for each non-null element in d_in write it to d_out, in the position specified by d_isValid. Called by compactArray().

Detailed Description

CUDPP kernel-level compact routines.

compact_kernel.cu


Functions
Compact Functions
template<bool isBackward>
__global__ void	computeNumValidElements (size_t d_numValidElements, const unsigned int d_indices, const unsigned int *d_isValid, size_t numElements)
	Compute the number of valid flags in an array given the array and its exclusive sum scan.
template<class T, bool isBackward>
__global__ void	compactData (T d_out, size_t d_numValidElements, const unsigned int d_indices, const unsigned int d_isValid, const T *d_in, unsigned int numElements)
	Consolidate non-null elements - for each non-null element in d_in write it to d_out, in the position specified by d_isValid. Called by compactArray().