Difficulty understanding on rendered_opacity or depth

[Reed-yang]

Thank you for your wonderful work!
But I have a little confusion when reading the code.

After rasterizer, we get the S-dimensional rendered_feature. Why can we get rendered_opacity and rendered_depth just by
rendered_other, rendered_depth, rendered_opacity = rendered_feature.split([S_other, 1, 1], dim=0)?
I read the relevant CUDA implementation, and it seems that there is no special processing for the last two dimensions of rendered_feature.

Thanks in advance!