Why dont we have stable diffusion using OpenCL yet?

[SeaN0X]

I'm quite a novice to this kind of AI technology, i've used lots of different WebUI, but none of them seems to use OpenCL when generating images, Only CUDA & cuDNN, which is basically limited to only NVidia cards. we can still run them using CPU, but it's a lot slower, with this in mind, why dont we use OpenCL along with oneDNN? it supports most of the graphics cards today, even including iGPUs, this could make image generation a lot more accessible for everyone.