RooFit::Detail::CudaKernels Namespace Reference

Namespaces
namespace	Reducers
	Dedicated namespace for reduction kernels.

Typedefs
using	Size_t = int
	The type for array size parameters.

Functions
template<class Num_t , class Den_t >
__global__ void	DivideBy (Size_t arraySize, Den_t *__restrict__ num, Num_t const *__restrict__ den)
	Divides elements of the num array in-place by corresponding elements of the den array.
template<class Idx_t , class Elem_t >
__global__ void	Lookup2D (Size_t arraySize, Idx_t n2, Idx_t const *__restrict__ x1, Idx_t const *__restrict__ x2, Elem_t const *__restrict__ lut, Elem_t *__restrict__ output)
	Fills an output array 'output' with values from a row-wise flattened two-dimensional lookup table lut based on input arrays x1 and x2:

Typedef Documentation

Size_t

using RooFit::Detail::CudaKernels::Size_t = typedef int

The type for array size parameters.

Definition at line 26 of file CudaKernels.cuh.

Function Documentation

DivideBy()

template<class Num_t , class Den_t >

__global__ void RooFit::Detail::CudaKernels::DivideBy	(	Size_t	arraySize,
		Den_t *__restrict__	num,
		Num_t const *__restrict__	den
	)

Divides elements of the num array in-place by corresponding elements of the den array.

Template Parameters

Num_t	Data type of the 'num' array elements.
Den_t	Data type of the 'den' array elements.

Parameters

[in]	arraySize	Size of the input arrays.
[in,out]	num	Numerator array containing values to be divided.
[in]	den	Denominator array containing divisor values.

Definition at line 229 of file CudaKernels.cuh.

Lookup2D()

template<class Idx_t , class Elem_t >

__global__ void RooFit::Detail::CudaKernels::Lookup2D	(	Size_t	arraySize,
		Idx_t	n2,
		Idx_t const *__restrict__	x1,
		Idx_t const *__restrict__	x2,
		Elem_t const *__restrict__	lut,
		Elem_t *__restrict__	output
	)

Fills an output array 'output' with values from a row-wise flattened two-dimensional lookup table lut based on input arrays x1 and x2:

output[ i ] = lut[ n2 * x1[i] + x2[i] ].

Each thread processes a portion of the input arrays using grid-stride looping.

Template Parameters

Idx_t	Data type of the indices.
Elem_t	Data type of the lookup table and output array elements.

Parameters

[in]	arraySize	Size of the input and output arrays.
[in]	n2	Size of the second dimension of the lookup table.
[in]	x1	Input array containing indices for the first dimension of the lookup table.
[in]	x2	Input array containing indices for the second dimension of the lookup table.
[in]	lut	Lookup table containing values.
[out]	output	Output array to be filled with values from the lookup table.

Definition at line 258 of file CudaKernels.cuh.