Module openj9.cuda

Package com.ibm.cuda

Class CudaKernel

java.lang.Object

com.ibm.cuda.CudaKernel

public class CudaKernel
extends Object

The CudaKernel class represents a kernel function in a loaded CudaModule.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	CudaKernel.Parameters	The Parameters class represents the actual parameters in a kernel launch.

Constructor Summary

Constructors

Constructor	Description
CudaKernel(CudaModule module, CudaFunction function)	Creates a new kernel object in the given module whose entry point is the specified function.
CudaKernel(CudaModule module, String functionName)	Creates a new kernel object in the given module whose entry point is the function with the specified name.

Method Summary

Modifier and Type	Method	Description
int	getAttribute(int attribute)	Returns the value of the specified @{code attribute} for the function associated with this kernel.
void	launch(CudaGrid grid, CudaKernel.Parameters parameters)	Launches this kernel.
void	launch(CudaGrid grid, Object... parameters)	Launches this kernel.
void	setCacheConfig(CudaDevice.CacheConfig config)	Configures the cache for the function associated with this kernel.
void	setSharedMemConfig(CudaDevice.SharedMemConfig config)	Configures the shared memory of the function associated with this kernel.

Methods declared in class java.lang.Object

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

Constructor Detail

CudaKernel

public CudaKernel(CudaModule module,
                  CudaFunction function)

Creates a new kernel object in the given module whose entry point is the specified function.

Parameters:

module - the module containing the kernel code

function - the entry point of the kernel

CudaKernel

public CudaKernel(CudaModule module,
                  String functionName)
           throws CudaException

Creates a new kernel object in the given module whose entry point is the function with the specified name.

Parameters:

module - the module containing the kernel code

functionName - the name of the entry point of the kernel

Throws:

CudaException - if a CUDA exception occurs

Method Detail

getAttribute

public final int getAttribute(int attribute)
                       throws CudaException

Returns the value of the specified @{code attribute} for the function associated with this kernel.

Parameters:

attribute - the attribute to be queried (see CudaFunction.ATTRIBUTE_XXX)

Returns:

the attribute value

Throws:

CudaException - if a CUDA exception occurs

launch

public final void launch(CudaGrid grid,
                         Object... parameters)
                  throws CudaException

Launches this kernel. The launch configuration is given by grid and the actual parameter values are specified by parameters.

Each parameter value must be one of the following:

• a boxed primitive value
• a CudaBuffer object
• null

Parameters:

grid - the launch configuration

parameters - the actual parameter values

Throws:

CudaException - if a CUDA exception occurs

IllegalArgumentException - if parameters contains any unsupported types

launch

public final void launch(CudaGrid grid,
                         CudaKernel.Parameters parameters)
                  throws CudaException

Launches this kernel. The launch configuration is given by grid and the actual parameter values are specified by parameters.

Parameters:

grid - the launch configuration

parameters - the actual parameter values

Throws:

CudaException - if a CUDA exception occurs

IllegalArgumentException - if parameters does not contain the correct number of values

setCacheConfig

public final void setCacheConfig(CudaDevice.CacheConfig config)
                          throws CudaException

Configures the cache for the function associated with this kernel.

Parameters:

config - the desired cache configuration

Throws:

CudaException - if a CUDA exception occurs

setSharedMemConfig

public final void setSharedMemConfig(CudaDevice.SharedMemConfig config)
                              throws CudaException

Configures the shared memory of the function associated with this kernel.

Parameters:

config - the desired shared memory configuration

Throws:

CudaException - if a CUDA exception occurs