Modifier and Type | Method and Description |
---|---|
Configuration |
Allocator.getConfiguration()
Returns current Allocator configuration
|
Modifier and Type | Method and Description |
---|---|
void |
Allocator.applyConfiguration(Configuration configuration)
Consume and apply configuration passed in as argument
|
Constructor and Description |
---|
DeviceAllocationsTracker(Configuration configuration) |
Modifier and Type | Method and Description |
---|---|
Configuration |
AtomicAllocator.getConfiguration()
Returns current Allocator configuration
|
Modifier and Type | Method and Description |
---|---|
void |
AtomicAllocator.applyConfiguration(Configuration configuration)
Consume and apply configuration passed in as argument
PLEASE NOTE: This method should only be used BEFORE any calculations were started.
|
Modifier and Type | Method and Description |
---|---|
void |
Balancer.init(Configuration configuration)
Deprecated.
This method initializes this Balancer instance
|
Modifier and Type | Method and Description |
---|---|
Configuration |
Configuration.allowCrossDeviceAccess(boolean reallyAllow)
Enables/disables P2P memory access for multi-gpu
|
Configuration |
Configuration.allowFallbackFromDevice(boolean reallyAllow)
NOT IMPLEMENTED YET
|
Configuration |
Configuration.allowMultiGPU(boolean reallyAllow)
This method allows you to enable or disable multi-GPU mode.
|
Configuration |
Configuration.allowPreallocation(boolean reallyAllow)
If set to true, each non-cached allocation request will cause few additional allocations,
Default value: true
|
Configuration |
Configuration.banDevice(Integer deviceId)
This method allows you to ban specific device.
|
Configuration |
Configuration.enableDebug(boolean debug)
With debug enabled all CUDA launches will become synchronous, with forced stream synchronizations after calls.
|
Configuration |
Configuration.enableStatisticsGathering(boolean reallyEnable)
This method enables/disables
|
Configuration |
CudaEnvironment.getConfiguration() |
Configuration |
Configuration.setAllocationModel(Configuration.AllocationModel allocationModel)
This method allows to specify allocation model for memory.
|
Configuration |
Configuration.setCommandLanesNumber(int length)
This method allows you to specify maximum number of probable parallel cuda processes
Default value: 4
PLEASE NOTE: This parameter has effect only for ASYNCHRONOUS execution model
|
Configuration |
Configuration.setCommandQueueLength(int length)
This method allows you to specify command queue length, as primary argument for asynchronous execution controller
Default value: 3
|
Configuration |
Configuration.setExecutionModel(Configuration.ExecutionModel executionModel)
This method allows to specify execution model for matrix/blas operations
SEQUENTIAL: Issue commands in order Java compiler sees them.
|
Configuration |
Configuration.setFirstMemory(AllocationStatus initialMemory)
This method allows to specify initial memory to be used within system.
|
Configuration |
Configuration.setMaximumBlockSize(int blockDim)
This methos allows to specify max blockSize for kernel launches
Default value: -1 (that means pick value automatically, device occupancy dependent)
|
Configuration |
Configuration.setMaximumDeviceAllocation(long max)
This method allows you to set maximum device allocation.
|
Configuration |
Configuration.setMaximumDeviceCache(long maxCache)
This method allows you to specify maximum memory cache per device
|
Configuration |
Configuration.setMaximumDeviceCacheableLength(long maxLen)
This method allows to specify maximum length of single memory chunk that's allowed to be cached.
|
Configuration |
Configuration.setMaximumDeviceMemoryUsed(double percentage)
This method allows you to specify max per-device memory use.
|
Configuration |
Configuration.setMaximumGridSize(int gridDim)
This method allows to specify max gridDim for kernel launches.
|
Configuration |
Configuration.setMaximumHostCache(long maxCache)
This method allows you to specify maximum memory cache for host memory
|
Configuration |
Configuration.setMaximumHostCacheableLength(long maxLen)
This method allows to specify maximum length of single memory chunk that's allowed to be cached.
|
Configuration |
Configuration.setMaximumSingleDeviceAllocation(long max)
This method allows to specify maximum single allocation on device.
|
Configuration |
Configuration.setMaximumSingleHostAllocation(long max)
This method allows to specify maximum single allocation on host.
|
Configuration |
Configuration.setMaximumZeroAllocation(long max)
This method allows you to set maximum host allocation.
|
Configuration |
Configuration.setMemoryModel(Configuration.MemoryModel model) |
Configuration |
Configuration.setMinimumBlockSize(int blockDim) |
Configuration |
Configuration.setMinimumRelocationThreshold(int threshold) |
Configuration |
Configuration.setNumberOfGcThreads(int numThreads)
This method allows you to set number of threads that'll handle memory releases on native side.
|
Configuration |
Configuration.setPoolSize(int poolSize)
Per-device resources pool size.
|
Configuration |
Configuration.setPreallocationCalls(int numCalls)
This method allows to specify number of preallocation calls done by cache subsystem in parallel, to serve later requests.
|
Configuration |
Configuration.setVerbose(boolean verbose) |
Configuration |
Configuration.triggerDebug(int code) |
Configuration |
Configuration.useDevice(Integer deviceId)
This method forces one specific device to be used.
|
Modifier and Type | Field and Description |
---|---|
protected Configuration |
SynchronousFlowController.configuration |
Modifier and Type | Method and Description |
---|---|
void |
MemoryHandler.init(Configuration configuration,
Allocator allocator)
This method gets called from Allocator, during Allocator/MemoryHandler initialization
|
Modifier and Type | Method and Description |
---|---|
void |
CudaZeroHandler.init(Configuration configuration,
Allocator allocator)
This method gets called from Allocator, during Allocator/MemoryHandler initialization
|
Modifier and Type | Field and Description |
---|---|
protected Configuration |
CudaCachingZeroProvider.configuration |
Copyright © 2016. All Rights Reserved.