com.ibm.cuda

Class CudaGrid

java.lang.Object

com.ibm.cuda.CudaGrid

public final class CudaGrid
extends Object

The CudaGrid class represents a kernel launch configuration.

Field Summary

Fields

Modifier and Type	Field and Description
int	blockDimX The size of the thread block in the x dimension.
int	blockDimY The size of the thread block in the y dimension.
int	blockDimZ The size of the thread block in the z dimension.
int	gridDimX The size of the grid in the x dimension.
int	gridDimY The size of the grid in the y dimension.
int	gridDimZ The size of the grid in the z dimension.
int	sharedMemBytes The number of bytes of shared memory to allocate to each thread block.
CudaStream	stream The stream on which the kernel should be queued (or null for the default stream).

Constructor Summary

Constructors

Constructor and Description
CudaGrid(Dim3 gridDim, Dim3 blockDim) Creates a grid with the specified dimensions, with no shared memory on the default stream.
CudaGrid(Dim3 gridDim, Dim3 blockDim, CudaStream stream) Creates a grid with the specified dimensions with no shared memory on the specified stream.
CudaGrid(Dim3 gridDim, Dim3 blockDim, int sharedMemBytes) Creates a grid with the specified dimensions and shared memory size on the default stream.
CudaGrid(Dim3 gridDim, Dim3 blockDim, int sharedMemBytes, CudaStream stream) Creates a grid with the specified dimensions and shared memory size on the specified stream.
CudaGrid(int gridDim, int blockDim) Creates a grid with the specified x dimensions with no shared memory on the default stream.
CudaGrid(int gridDim, int blockDim, CudaStream stream) Creates a grid with the specified x dimensions with no shared memory on the specified stream.
CudaGrid(int gridDim, int blockDim, int sharedMemBytes) Creates a grid with the specified x dimensions and shared memory size on the default stream.
CudaGrid(int gridDim, int blockDim, int sharedMemBytes, CudaStream stream) Creates a grid with the specified x dimensions and shared memory size on the specified stream.

Method Summary

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

blockDimX

public final int blockDimX

The size of the thread block in the x dimension.

blockDimY

public final int blockDimY

The size of the thread block in the y dimension.

blockDimZ

public final int blockDimZ

The size of the thread block in the z dimension.

gridDimX

public final int gridDimX

The size of the grid in the x dimension.

gridDimY

public final int gridDimY

The size of the grid in the y dimension.

gridDimZ

public final int gridDimZ

The size of the grid in the z dimension.

sharedMemBytes

public final int sharedMemBytes

The number of bytes of shared memory to allocate to each thread block.

stream

public final CudaStream stream

The stream on which the kernel should be queued (or null for the default stream).

Constructor Detail

CudaGrid

public CudaGrid(Dim3 gridDim,
                Dim3 blockDim)

Creates a grid with the specified dimensions, with no shared memory on the default stream.

Parameters:

gridDim - the dimensions of the grid

blockDim - the dimensions of the thread block

CudaGrid

public CudaGrid(Dim3 gridDim,
                Dim3 blockDim,
                CudaStream stream)

Creates a grid with the specified dimensions with no shared memory on the specified stream.

Parameters:

gridDim - the dimensions of the grid

blockDim - the dimensions of the thread block

stream - the stream on which the kernel should be queued (or null for the default stream)

CudaGrid

public CudaGrid(Dim3 gridDim,
                Dim3 blockDim,
                int sharedMemBytes)

Creates a grid with the specified dimensions and shared memory size on the default stream.

Parameters:

gridDim - the dimensions of the grid

blockDim - the dimensions of the thread block

sharedMemBytes - the number of bytes of shared memory to allocate to each thread block

CudaGrid

public CudaGrid(Dim3 gridDim,
                Dim3 blockDim,
                int sharedMemBytes,
                CudaStream stream)

Creates a grid with the specified dimensions and shared memory size on the specified stream.

Parameters:

gridDim - the dimensions of the grid

blockDim - the dimensions of the thread block

sharedMemBytes - the number of bytes of shared memory to allocate to each thread block

stream - the stream on which the kernel should be queued (or null for the default stream)

CudaGrid

public CudaGrid(int gridDim,
                int blockDim)

Creates a grid with the specified x dimensions with no shared memory on the default stream. The y and z dimensions are set to 1.

Parameters:

gridDim - the x dimension of the grid

blockDim - the x dimension of the thread block

CudaGrid

public CudaGrid(int gridDim,
                int blockDim,
                CudaStream stream)

Creates a grid with the specified x dimensions with no shared memory on the specified stream. The y and z dimensions are set to 1.

Parameters:

gridDim - the x dimension of the grid

blockDim - the x dimension of the thread block

stream - the stream on which the kernel should be queued (or null for the default stream)

CudaGrid

public CudaGrid(int gridDim,
                int blockDim,
                int sharedMemBytes)

Creates a grid with the specified x dimensions and shared memory size on the default stream. The y and z dimensions are set to 1.

Parameters:

gridDim - the x dimension of the grid

blockDim - the x dimension of the thread block

sharedMemBytes - the number of bytes of shared memory to allocate to each thread block

CudaGrid

public CudaGrid(int gridDim,
                int blockDim,
                int sharedMemBytes,
                CudaStream stream)

Creates a grid with the specified x dimensions and shared memory size on the specified stream. The y and z dimensions are set to 1.

Parameters:

gridDim - the x dimension of the grid

blockDim - the x dimension of the thread block

sharedMemBytes - the number of bytes of shared memory to allocate to each thread block

stream - the stream on which the kernel should be queued (or null for the default stream)