PARALUTION  1.0.0
PARALUTION
cuda_kernels_vector.hpp File Reference

Go to the source code of this file.

Namespaces

 paralution
 

Functions

template<typename ValueType , typename IndexType >
__global__ void paralution::kernel_scaleadd (const IndexType n, const ValueType alpha, const ValueType *x, ValueType *out)
 
template<typename ValueType , typename IndexType >
__global__ void paralution::kernel_scaleaddscale (const IndexType n, const ValueType alpha, const ValueType beta, const ValueType *x, ValueType *out)
 
template<typename ValueType , typename IndexType >
__global__ void paralution::kernel_scaleaddscale_offset (const IndexType n, const IndexType src_offset, const IndexType dst_offset, const ValueType alpha, const ValueType beta, const ValueType *x, ValueType *out)
 
template<typename ValueType , typename IndexType >
__global__ void paralution::kernel_scaleadd2 (const IndexType n, const ValueType alpha, const ValueType beta, const ValueType gamma, const ValueType *x, const ValueType *y, ValueType *out)
 
template<typename ValueType , typename IndexType >
__global__ void paralution::kernel_pointwisemult (const IndexType n, const ValueType *x, ValueType *out)
 
template<typename ValueType , typename IndexType >
__global__ void paralution::kernel_pointwisemult2 (const IndexType n, const ValueType *x, const ValueType *y, ValueType *out)
 
template<typename ValueType , typename IndexType >
__global__ void paralution::kernel_copy_offset_from (const IndexType n, const IndexType src_offset, const IndexType dst_offset, const ValueType *in, ValueType *out)
 
template<typename ValueType , typename IndexType >
__global__ void paralution::kernel_permute (const IndexType n, const IndexType *permute, const ValueType *in, ValueType *out)
 
template<typename ValueType , typename IndexType >
__global__ void paralution::kernel_permute_backward (const IndexType n, const IndexType *permute, const ValueType *in, ValueType *out)
 
template<typename ValueType , typename IndexType , unsigned int BLOCK_SIZE>
__global__ void paralution::kernel_reduce (const IndexType n, const ValueType *data, ValueType *out, const IndexType GROUP_SIZE, const IndexType LOCAL_SIZE)
 
template<typename ValueType , typename IndexType , unsigned int BLOCK_SIZE>
__global__ void paralution::kernel_max (const IndexType n, const ValueType *data, ValueType *out, const IndexType GROUP_SIZE, const IndexType LOCAL_SIZE)
 
template<typename ValueType , typename IndexType , unsigned int BLOCK_SIZE>
__global__ void paralution::kernel_amax (const IndexType n, const ValueType *data, ValueType *out, const IndexType GROUP_SIZE, const IndexType LOCAL_SIZE)
 
template<typename IndexType >
__global__ void paralution::kernel_powerd (const IndexType n, const double power, double *out)
 
template<typename IndexType >
__global__ void paralution::kernel_powerf (const IndexType n, const double power, float *out)
 
template<typename ValueType , typename IndexType >
__global__ void paralution::kernel_copy_from_float (const IndexType n, const float *in, ValueType *out)
 
template<typename ValueType , typename IndexType >
__global__ void paralution::kernel_copy_from_double (const IndexType n, const double *in, ValueType *out)