C# Class ManagedCuda.CudaOccupancy

Cuda occupancy from CudaOccupancy.h
Datei anzeigen Open project: kunzmi/managedCuda Class Usage Examples

Public Methods

Method Description
cudaOccMaxActiveBlocksPerMultiprocessor ( cudaOccResult result, cudaOccDeviceProp properties, cudaOccFuncAttributes attributes, cudaOccDeviceState state, int blockSize, ManagedCuda.BasicTypes.SizeT dynamicSmemSize ) : void

Determine the maximum number of CTAs that can be run simultaneously per SM. This is equivalent to the calculation done in the CUDA Occupancy Calculator spreadsheet

cudaOccMaxPotentialOccupancyBlockSize ( int &minGridSize, int &blockSize, cudaOccDeviceProp properties, cudaOccFuncAttributes attributes, cudaOccDeviceState state, ManagedCuda.BasicTypes.SizeT dynamicSMemSize ) : void

cudaOccMaxPotentialOccupancyBlockSize ( int &minGridSize, int &blockSize, cudaOccDeviceProp properties, cudaOccFuncAttributes attributes, cudaOccDeviceState state, del_blockSizeToDynamicSMemSize blockSizeToDynamicSMemSize, ManagedCuda.BasicTypes.SizeT dynamicSMemSize ) : void

Private Methods

Method Description
__occDivideRoundUp ( int x, int y ) : int
__occMin ( int lhs, int rhs ) : int
__occRoundUp ( int x, int y ) : int
cudaOccDevicePropCheck ( cudaOccDeviceProp properties ) : cudaOccError
cudaOccDeviceStateCheck ( cudaOccDeviceState state ) : cudaOccError
cudaOccFuncAttributesCheck ( cudaOccFuncAttributes attributes ) : cudaOccError
cudaOccInputCheck ( cudaOccDeviceProp properties, cudaOccFuncAttributes attributes, cudaOccDeviceState state ) : void
cudaOccMaxBlocksPerMultiprocessor ( cudaOccDeviceProp properties ) : int
cudaOccMaxBlocksPerSMRegsLimit ( cudaOccPartitionedGCConfig &gcConfig, cudaOccResult result, cudaOccDeviceProp properties, cudaOccFuncAttributes attributes, int blockSize ) : int
cudaOccMaxBlocksPerSMSmemLimit ( cudaOccResult result, cudaOccDeviceProp properties, cudaOccFuncAttributes attributes, cudaOccDeviceState state, int blockSize, ManagedCuda.BasicTypes.SizeT dynamicSmemSize ) : int
cudaOccMaxBlocksPerSMWarpsLimit ( cudaOccPartitionedGCConfig gcConfig, cudaOccDeviceProp properties, cudaOccFuncAttributes attributes, int blockSize ) : int
cudaOccPartitionedGCExpected ( cudaOccDeviceProp properties, cudaOccFuncAttributes attributes ) : cudaOccPartitionedGCConfig
cudaOccPartitionedGCForced ( cudaOccDeviceProp properties ) : bool
cudaOccPartitionedGlobalCachingModeSupport ( cudaOccDeviceProp properties ) : cudaOccPartitionedGCSupport
cudaOccRegAllocationGranularity ( cudaOccDeviceProp properties, int regsPerThread ) : int
cudaOccSMemAllocationGranularity ( cudaOccDeviceProp properties ) : int
cudaOccSMemPerMultiprocessor ( cudaOccDeviceProp properties, cudaOccCacheConfig cacheConfig ) : ManagedCuda.BasicTypes.SizeT
cudaOccSubPartitionsPerMultiprocessor ( cudaOccDeviceProp properties ) : int

Method Details

cudaOccMaxActiveBlocksPerMultiprocessor() public static method

Determine the maximum number of CTAs that can be run simultaneously per SM. This is equivalent to the calculation done in the CUDA Occupancy Calculator spreadsheet
public static cudaOccMaxActiveBlocksPerMultiprocessor ( cudaOccResult result, cudaOccDeviceProp properties, cudaOccFuncAttributes attributes, cudaOccDeviceState state, int blockSize, ManagedCuda.BasicTypes.SizeT dynamicSmemSize ) : void
result cudaOccResult
properties cudaOccDeviceProp
attributes cudaOccFuncAttributes
state cudaOccDeviceState
blockSize int
dynamicSmemSize ManagedCuda.BasicTypes.SizeT
return void

cudaOccMaxPotentialOccupancyBlockSize() public static method

public static cudaOccMaxPotentialOccupancyBlockSize ( int &minGridSize, int &blockSize, cudaOccDeviceProp properties, cudaOccFuncAttributes attributes, cudaOccDeviceState state, ManagedCuda.BasicTypes.SizeT dynamicSMemSize ) : void
minGridSize int
blockSize int
properties cudaOccDeviceProp
attributes cudaOccFuncAttributes
state cudaOccDeviceState
dynamicSMemSize ManagedCuda.BasicTypes.SizeT
return void

cudaOccMaxPotentialOccupancyBlockSize() public static method

public static cudaOccMaxPotentialOccupancyBlockSize ( int &minGridSize, int &blockSize, cudaOccDeviceProp properties, cudaOccFuncAttributes attributes, cudaOccDeviceState state, del_blockSizeToDynamicSMemSize blockSizeToDynamicSMemSize, ManagedCuda.BasicTypes.SizeT dynamicSMemSize ) : void
minGridSize int
blockSize int
properties cudaOccDeviceProp
attributes cudaOccFuncAttributes
state cudaOccDeviceState
blockSizeToDynamicSMemSize del_blockSizeToDynamicSMemSize
dynamicSMemSize ManagedCuda.BasicTypes.SizeT
return void