I’m afraid I don’t have much to add. I’ve used KernelAbstractions.jl a few times, but I don’t have much experience with it. More often, I’ve used AcceleratedKernels.jl, which, as far as I understand, is a higher level abstraction built on top of KernelAbstractions.jl. As for support for various hardware architectures, I agree with you. Some of the packages I mentioned earlier can even run on TPUs. I’m not sure if it’s technically possible to program FPGAs using Julia, but that would be really cool. Regarding GPU array support, I’ve often thought it would be great if package maintainers always included a section about multithreading or parallel processing on the front page of their GitHub repositories.This would probably make the package ecosystem more coherent and, overall, be beneficial for the community.