Best practice for per draw call uniform data

[junglie85]

In #733 it’s suggested to use dynamic uniform offsets as an approach to provide per draw call data. However, I have 32 bytes that I need to offset by and my limits have a minimum offset size of 256.

Am I going to have to allocate 256 bytes per uniform data to make this approach work? I assume that would need to be determined at runtime based on the actual limit?

I’d like to avoid duplicating this data per vertex (although it’s probably not loads in the grand scheme of things). I don’t want to use instancing but I would like to limit the number of buffers that I’m using and ideally keep the draw call count lower than the number of objects I’m drawing (but not imperative). What are some other options I have?

[junglie85]

For now, I've moved the uniform data to a vertex buffer, which means I don't have to deal with the uniform buffer size or offset limits. Not that I'm actually drawing that many objects, but it's nice to be able to.

I'm using the vertex in instance step mode and binding a slice of each buffer for every object to render:

pass.set_pipeline(&primitives_pipeline);
for (bind_group, vbo_range, ibo_range, transform_range, amount) in render_groups {
    pass.set_bind_group(1, bind_group, &[]);
    pass.set_vertex_buffer(0, primitives_vbo.slice(vbo_range));
    pass.set_vertex_buffer(1, transforms_vbo.slice(transform_range));
    pass.set_index_buffer(primitives_ibo.slice(ibo_range), IndexFormat::Uint32);
    pass.draw_indexed(amount, 0, 0..1);
}

Are there any reasons why this approach would be a bad idea?