Namespace Tensor.Cuda

Cannot register host memory with CUDA, maybe because it is not properly aligned.

Options for configuring operations performed on CUDA tensors.

Methods for locking a TensorHostStorage into memory and registering the memory with CUDA for fast data transfers with GPU device.

CUDA registered memory for fast data transfer. Dispose to unregister memory with CUDA.

C++ NativeIdx

C++ NativeIdx template info

C++ tensor marshaling

Out of CUDA memory.

CUDA backend for tensors.

Creates Tensors on a CUDA device.

Tensor storage on a CUDA device.

type-neutral interface to CUDA backend for tensors

type neutral interface to a CudaStorageT