Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Support __cuda_array_interface__ on GPU

See original GitHub issue

https://numba.pydata.org/numba-doc/dev/cuda/cuda_array_interface.html

It would not be hard to make DeviceArray implement this interface on GPU.

It would be slightly harder to support wrapping a DeviceArray around an existing CUDA array, but not that hard.

Issue Analytics

State:
Created 4 years ago
Reactions:15
Comments:12 (6 by maintainers)

Top GitHub Comments

4reactions

leofangcommented, Aug 20, 2020

Thanks John! Yeah we just finished a GPU Hackathon, and a few of our teams evaluating JAX asked us why JAX can’t work with other libraries like CuPy and PyTorch bidirectionally. It’d be very useful, say, to do autograd in JAX, postprocess in CuPy, then bring it back to JAX.

2reactions

hawkinspcommented, Aug 20, 2020

Also: I haven’t tried this, but since CuPy supports both __cuda_array_interface__ and DLPack, you can most likely “launder” an array via CuPy into JAX:

export the array via __cuda_array_interface__ to CuPy.
export the array via DLPack from CuPy.
import the DLPack into JAX.

(Obviously this isn’t ideal, but it might unblock you.)

Top Results From Across the Web

CUDA Array Interface (Version 2) - Numba

The cuda array interface is created for interoperability between different implementation of GPU array-like objects in various projects. The idea is borrowed ...

CUDA Array Interface (Version 3) - Numba documentation

The CUDA Array Interface (or CAI) is created for interoperability between different implementations of CUDA array-like objects in various projects. The idea is ......

Support the __cuda_array_interface__ protocol #29039 - GitHub

A number of GPU array computing libraries in Python (Numba, CuPy, PyTorch, RAPIDS) support the __cuda_array_interface__ , protocol as ...

CUDA C++ Programming Guide - NVIDIA Documentation Center

The programming guide to the CUDA model and interface. ... CUDA is designed to support various languages and application programming interfaces. ......

Interoperability — CuPy 11.4.0 documentation

Numba is a Python JIT compiler with NumPy support. cupy.ndarray implements __cuda_array_interface__ , which is the CUDA array interchange interface compatible ...