Psi Lambda LLC ψ(λκ) Kappa Library User Guide

CUDA GPU Properties

The properties and attributes of the GPUs can be retrieved using the Kappa API access to the cudaGPU class. These properties are also made available to use for dynamic sizing as either:

/Kappa/CUDA/GPU/Current#<Property>

or

%/Kappa/CUDA/GPU/<GPU#>{<Property>}

where <GPU#> is the GPU associated with kappa::Process and <Property> is one of the following CUDA properties:

CUDA <Property>

Description

Name

Device name string

Major

Major revision number

Minor

Minor revision number

GlobalMemory

Total amount of global memory in bytes

ConstantMemory

Total amount of constant memory in bytes

SharedMemoryPerBlock

Total amount of shared memory per block in bytes

RegistersPerBlock

Total number of registers available per block

WarpSize

Warp size (currently always 32)

MaxThreadsPerBlock

Maximum number of threads per block

MaxThreads

Maximum sizes of each dimension of a block (vector integer Value of three numbers)

MaxGridSize

Maximum sizes of each dimension of a grid (vector integer Value of three numbers)

MemoryPitch

Maximum memory pitch in bytes

TextureAlignment

Texture alignment in bytes

ClockRate

Clock Rate (in hertz)

MultiProcessorCount

Multiprocessor Count

ConcurrentCopyExecute

Concurrent copy and execution (boolean)

KernelTimeout

Kernel timeout (boolean)

Integrated

Integrated (boolean)

CanMapHostMemory

Can Map Host Memory (boolean)

ComputeMode

Compute Mode (enumeration)

MaximumTexture1DWidth

Maximum texture one dimensional width

MaximumTexture2DWidth

Maximum texture two dimensional width

MaximumTexture2DHeight

Maximum texture two dimensional height

MaximumTexture3DWidth

Maximum texture three dimensional width

MaximumTexture3DHeight

Maximum texture three dimensional height

MaximumTexture3DDepth

Maximum texture three dimensional depth

MaximumArrayWidth

Maximum texture array width

MaximumArrayHeight

Maximum texture array height

MaximumArraySlices

Maximum texture array slices

SurfaceAlignment

Surface alignment in bytes

ConcurrentKernels

Whether concurrent kernel execution is supported (boolean)

ECCSupported

Whether ECC is supported (boolean)

Please refer to the CUDA Reference Manual or Programming Guide for further details on these properties and their uses.

Page