TYPE

是啊，二维码又不是一定得以图片显示，黑白的方块拼出来就行了

Thanks for your reply. I meet this problem on the same OS version(12.5.1). And I tried to use `rfd` in a simple terminal project, `pick_file` works well without such problem.

Maybe we could decompose one concave polygon into some convex polygons, and store them. And I can implement it if possible.

Thanks for your reply. You are right, this workgroup_size is not for the best performance, it just an example to show the difference between M1 MacBook Air and Nvidia 2080ti...

Thank you for your contribution to this project! And I try to use naga to translate wgsl to msl, it does looks fine.

How about`arrow2`，which uses std::Vec as backend. https://crates.io/crates/arrow2

Hey, I've made a crate provides N-Dimension FFT acceleration convolution for ndarray. Maybe ndarray-vision could depend on my crate. [https://crates.io/crates/ndarray-conv](url) And there is the benchmark result:

And ndarray-conv also provides normal convolution and a lot of padding and convolution mode.

Of course, but would you like to expose ndarray-conv's interface directly?

OK, I will try to make a PR recently.