Not just NVIDIA: GPU programming that runs everywhere

If you’re doing computations on a GPU, NVIDIA is the default, alongside its CUDA libraries.
Some libraries like PyTorch support do support AMD GPUs and Macs.
But from the re-implementations of NumPy, SciPy, and Pandas in the RAPIDS project, to Numba’s GPU support, NVIDIA has best software support in the Python world.

Sticking to NVIDIA-specific software has some downsides, however:

  • It won’t run on modern Mac laptops.
  • Testing in CI is more difficult: you need custom runners that have NVIDIA GPUs.
  • You can’t use any other GPUs you might have access to, like AMD GPUs.

What can you do if you want to use GPUs in a portable manner?
In this article we’ll cover one option, the wgpu-py library.



To finish reading, please visit source site