cudaDeviceProp

canMapHostMemory clockRate computeMode concurrentKernels deviceOverlap ECCEnabled integrated kernelExecTimeoutEnabled major maxGridSize maxTexture1D maxTexture2D maxTexture2DArray maxTexture3D maxThreadsDim maxThreadsPerBlock memPitch minor multiProcessorCount name pciBusID pciDeviceID regsPerBlock sharedMemPerBlock surfaceAlignment tccDriver textureAlignment totalConstMem totalGlobalMem warpSize

int cudaDeviceProp::maxThreadsDim[3] Maximum size of each dimension of a block