Wgpu-rs: Documentation about aligning data before mapping it to a buffer

[yutannihilation]

(While I asked this on Matrix and my issue was solved, I'm filing this here as I feel it's worth a documentation)

After https://github.com/gfx-rs/wgpu/pull/758, buffers mapped at creation are required to be aligned to 4. But, iiuc (sorry, I'm still a newbie to computer graphics...), vertices are triangular in most cases, which means the length of index data is a multiple of 3. Besides, a index is u16, not a multiple of 4 byte. So, it's a common case the user needs to add some padding to the index data by themselves, especially when the index is generated by external libraries (I got this when I was playing with lyon).

Fortunately (or unfortunately?), current examples use only the even number of length of indices, so they don't hit this problem. I think it's good to show how the user can handle the alignment issue.

Not sure if this is useful, but this is what I'm doing:

        // record the range before extending
        let stroke_range = 0..(indices.len() as u32);

        // extend
        let alignment = wgpu::COPY_BUFFER_ALIGNMENT as usize / std::mem::size_of::<u16>();
        let fraction = indices.len() % alignment;
        if fraction > 0 {
            indices
                .extend(std::iter::repeat(0).take(alignment - fraction));
        }

[kvark]

Thank you for the feedback!
How about we do this in fn create_buffer_with_data? We know the size of an element there, and we know the alignment, so we could easily request the underlying buffer to be larger, and just handle the extra elements internally.

[yutannihilation]

Oh, nice. Sounds good to me!