culite::Grid2D Class Reference

2D CUDA kernel grid configuration. More...

Inheritance diagram for culite::Grid2D:

Public Member Functions
	Grid2D ()
	Default constructor.
	Grid2D (std::size_t m, std::size_t n, std::size_t blockSizeX=16, std::size_t blockSizeY=16)
	Constructs a 2D grid configuration with explicit block dimensions.
	~Grid2D ()
	Default destructor.
Public Member Functions inherited from culite::GridXD
	GridXD ()
	Default constructor.
	GridXD (dim3 numBlocks, dim3 threadsPerBlock)
	Constructs a GridXD with explicit grid and block dimensions.
	~GridXD ()
	Default destructor.
const dim3 &	numBlocks () const
	Gets the number of blocks in the grid.
const dim3 &	threadsPerBlock () const
	Gets the number of threads per block.

Additional Inherited Members
Protected Types inherited from culite::GridXD
using	cudaInt = unsigned int
	Type used by CUDA dim3 dimensions.

Detailed Description

2D CUDA kernel grid configuration.

Grid2D simplifies configuration for 2D kernels (e.g., matrix operations) by automatically calculating the number of blocks needed to cover m×n elements with the specified block dimensions. Optimized for column-major matrices.

Constructor & Destructor Documentation

Grid2D() [1/2]

culite::Grid2D::Grid2D

(

)

Default constructor.

Creates a Grid2D with default-initialized dimensions.

Grid2D() [2/2]

culite::Grid2D::Grid2D	(	std::size_t	m,
		std::size_t	n,
		std::size_t	blockSizeX = 16,
		std::size_t	blockSizeY = 16 )

Constructs a 2D grid configuration with explicit block dimensions.

Automatically calculates the number of blocks needed in X and Y dimensions to cover m×n elements with the given block sizes. Performs overflow and validation checks.

Parameters

m	Number of elements in the X dimension (rows).
n	Number of elements in the Y dimension (columns).
blockSizeX	Number of threads per block in X dimension (default: 16).
blockSizeY	Number of threads per block in Y dimension (default: 16).

Exceptions

err::CudaException

if block sizes are zero, if arithmetic overflow occurs, or if calculated values exceed unsigned int maximum.