@rapidsai/cuda

node-cuda (npm install @rapidsai/cuda)

A node native addon that provides bindings to the CUDA driver and runtime APIs.
These bindings allow calling the CUDA device management, memory, stream, event, ipc, and
DX/GL interop APIs from JS. These APIs allow node (v8 or chakracore) applications to read,
write, and share memory via zero-copy CUDA IPC with external processes that also use the
CUDA, OpenGL, and RAPIDS libraries.

Device management:

cudaChooseDevice, cudaGetDeviceCount, cuDeviceGet, cudaDeviceGetPCIBusId, cudaDeviceGetByPCIBusId, cudaGetDevice, cudaGetDeviceFlags, cudaGetDeviceProperties, cudaSetDevice, cudaSetDeviceFlags, cudaDeviceReset, cudaDeviceSynchronize, cudaDeviceCanAccessPeer, cudaDeviceEnablePeerAccess, cudaDeviceDisablePeerAccess

Memory:

• CUDADevice: A class to wrap and manage a CUDA device.
• CUDABuffer: A class to wrap and manage device memory allocations (similar to ArrayBuffer).
• CUDAArray: A class to wrap operations like read/write/share on a CUDABuffer (similar to TypedArray).

• cuPointerGetAttribute, cudaMalloc, cudaFree, cudaMallocHost, cudaFreeHost, cudaHostRegister, cudaHostUnregister, cudaMemcpy, cudaMemset, cudaMemcpyAsync, cudaMemsetAsync, cudaMemGetInfo
  

IPC:

cudaIpcGetMemHandle, cudaIpcOpenMemHandle, cudaIpcCloseMemHandle

Stream:

cudaStreamCreate, cudaStreamDestroy, cudaStreamSynchronize

OpenGL:

cuGraphicsGLRegisterBuffer, cuGraphicsGLRegisterImage, cuGraphicsUnregisterResource, cuGraphicsMapResources, cuGraphicsUnapResources, cuGraphicsResourceGetMappedPointer