org.apache.sysml.runtime.instructions.gpu.context

Class JCudaContext

java.lang.Object

org.apache.sysml.runtime.instructions.gpu.context.GPUContext

org.apache.sysml.runtime.instructions.gpu.context.JCudaContext

public class JCudaContext
extends GPUContext

Field Summary

Fields

Modifier and Type	Field and Description
static ArrayList<GPUObject>	allocatedPointers Global list of allocated GPUObject instances.
static boolean	DEBUG enable this to print debug information before code pertaining to the GPU is executed
static int	deviceCount The total number of cuda devices on this machine
double	GPU_MEMORY_UTILIZATION_FACTOR
boolean	REFRESH_AVAILABLE_MEMORY_EVERY_TIME
static Object	syncObj Synchronization object to make sure no allocations happen when something is being evicted from memory

Fields inherited from class org.apache.sysml.runtime.instructions.gpu.context.GPUContext

currContext, isGPUContextCreated

Constructor Summary

Constructors

Constructor and Description
JCudaContext()

Method Summary

Modifier and Type	Method and Description
void	destroy()
void	ensureComputeCapability() Ensures that all the CUDA cards on the current system are of the minimum required compute capability.
long	getAndAddAvailableMemory(long v) Gets the available memory and then adds value to it
long	getAvailableMemory() Gets device memory available for SystemML operations
static jcuda.runtime.cudaDeviceProp	getGPUProperties() Gets the device properties for the active GPU (set with cudaSetDevice())
static jcuda.runtime.cudaDeviceProp	getGPUProperties(int device) Gets the device properties
static int	getMaxBlocks() Gets the maximum number of blocks supported by the active cuda device
static long	getMaxSharedMemory() Gets the shared memory per block supported by the active cuda device
static int	getMaxThreadsPerBlock() Gets the maximum number of threads per block for "active" GPU
static int	getWarpSize() Gets the warp size supported by the active cuda device

Methods inherited from class org.apache.sysml.runtime.instructions.gpu.context.GPUContext

createGPUObject, getGPUContext

Methods inherited from class java.lang.Object

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

Field Detail

syncObj

public static final Object syncObj

Synchronization object to make sure no allocations happen when something is being evicted from memory

allocatedPointers

public static ArrayList<GPUObject> allocatedPointers

Global list of allocated GPUObject instances. This list must be accessed in a synchronized way

deviceCount

public static int deviceCount

The total number of cuda devices on this machine

DEBUG

public static boolean DEBUG

enable this to print debug information before code pertaining to the GPU is executed

GPU_MEMORY_UTILIZATION_FACTOR

public double GPU_MEMORY_UTILIZATION_FACTOR

REFRESH_AVAILABLE_MEMORY_EVERY_TIME

public boolean REFRESH_AVAILABLE_MEMORY_EVERY_TIME

Constructor Detail

JCudaContext

public JCudaContext()
             throws DMLRuntimeException

Throws:

DMLRuntimeException

Method Detail

getAvailableMemory

public long getAvailableMemory()

Description copied from class: GPUContext

Gets device memory available for SystemML operations

Specified by:

getAvailableMemory in class GPUContext

Returns:

available memory

ensureComputeCapability

public void ensureComputeCapability()
                             throws DMLRuntimeException

Description copied from class: GPUContext

Ensures that all the CUDA cards on the current system are of the minimum required compute capability. (The minimum required compute capability is hard coded in JCudaContext.

Specified by:

ensureComputeCapability in class GPUContext

Throws:

DMLRuntimeException - if DMLRuntimeException occurs

getGPUProperties

public static jcuda.runtime.cudaDeviceProp getGPUProperties()

Gets the device properties for the active GPU (set with cudaSetDevice())

Returns:

the device properties

getGPUProperties

public static jcuda.runtime.cudaDeviceProp getGPUProperties(int device)

Gets the device properties

Parameters:

device - the device number (on a machine with more than 1 GPU)

Returns:

the device properties

getMaxThreadsPerBlock

public static int getMaxThreadsPerBlock()

Gets the maximum number of threads per block for "active" GPU

Returns:

the maximum number of threads per block

getMaxBlocks

public static int getMaxBlocks()

Gets the maximum number of blocks supported by the active cuda device

Returns:

the maximum number of blocks supported

getMaxSharedMemory

public static long getMaxSharedMemory()

Gets the shared memory per block supported by the active cuda device

Returns:

the shared memory per block

getWarpSize

public static int getWarpSize()

Gets the warp size supported by the active cuda device

Returns:

the warp size

getAndAddAvailableMemory

public long getAndAddAvailableMemory(long v)

Gets the available memory and then adds value to it

Parameters:

v - the value to add

Returns:

the current available memory before adding value to it

destroy

public void destroy()
             throws DMLRuntimeException

Specified by:

destroy in class GPUContext

Throws:

DMLRuntimeException