cuda_utilities::AutomaticLaunchParams< T > Struct Template Reference

Struct to determine the optimal number of blocks and threads per block to use when launching a kernel. The member variables are threadsPerBlock and numBlocks which are chosen with the occupancy API. More...

#include <cuda_utilities.h>

Public Member Functions
	AutomaticLaunchParams (T &kernel, size_t numElements=0)
	Construct a new AutomaticLaunchParams object. By default it generates values of numBlocks and threadsPerBlock suitable for a kernel with a grid-stride loop. For a kernel with one thread per element set the optional numElements argument to the number of elements.
	~AutomaticLaunchParams ()=default
	Defaulted Destructor.
int	get_threadsPerBlock () const
	Getter for threadsPerBlock.
int	get_numBlocks () const
	Getter for numBlocks.

Detailed Description

template<typename T>
struct cuda_utilities::AutomaticLaunchParams< T >

Struct to determine the optimal number of blocks and threads per block to use when launching a kernel. The member variables are threadsPerBlock and numBlocks which are chosen with the occupancy API.

Constructor & Destructor Documentation

AutomaticLaunchParams()

template<typename T >

cuda_utilities::AutomaticLaunchParams< T >::AutomaticLaunchParams ( T &  kernel, size_t  numElements = 0  )

inline

Construct a new AutomaticLaunchParams object. By default it generates values of numBlocks and threadsPerBlock suitable for a kernel with a grid-stride loop. For a kernel with one thread per element set the optional numElements argument to the number of elements.

Parameters

[in]	kernel	The kernel to determine the launch parameters for
[in]	numElements	The number of elements in the array that the kernel operates on

---

The documentation for this struct was generated from the following file:

• cuda_utilities.h