Open bkmgit opened 3 years ago
It is a good idea, especially because the OpenCL part would be almost trivial to derive from the CUDA one (moving from CuPy to PyOpenCL and update the kernel code, plus remove the CUDA specific parts).
For now I would focus on finish the CUDA version though, and then we can measure the interest in the OpenCL one and take action.
CuPy has some experimental support for AMD GPUs.
Would there be a possibility of adding OpenCL and/or Sycl tracks so that the lesson can be used on a wider variety of devices?