Safari supports WebGPU experimentally with WSL kernels. I wrote a simple tuner that tries to optimize matrix multiplication. If you have Safari, you can try it here. (You’ll need to enable WebGPU in Develop > Experimental Features.)
My M1 MacBook Air achieves 900GFlops after a couple seconds of tuning. My Intel MacBook Pro (16-inch, 2019, i9) only hits 100GFlops with the same exhaustive search.
https://jott.live/markdown/m1_webgpu_perf
This thread was posted by one of our members via one of our news source trackers.