I know that this is not on top of the to-do list at this moment, but it could be a powerful unifying way of developing GPGPU code on embedded systems. Setting aside a wide variety of NPUs present on many modern SoCs, both Intel with UHD Graphics and ARM64 with Mali or Adreno offer opportunities for acceleration of image processing and computer vision among others. OpenCL seems to be the only somewhat universal way to do GPGPU computing on these systems today. Is there any chance of Mojo competing in this space any time soon?
3 Likes