I am working on [dcompute](https://github.com/libmir/dcompute https://github.com/libmir/dcompute) which I hope will bring an equivalent experience when running

I am working on [dcompute](https://github.com/libmir/dcompute https://github.com/libmir/dcompute) which I hope will bring an equivalent experience when running on OpenCL and CUDA, and being written in D will provide all the advantages it does to writing kernels as it does with writing host code (with of course the ability to share compile time information across host and device). I believe that this levelling of the playing field will let the best hardware win. If you have any question, please ask!