Per-mesh vertex counts currently show up in the CLI, but there are a lot of different ways to compute a "vertex count" with different practical uses. Consider cases like:
if a mesh is drawn N times, how do we count its vertices?
if a mesh shares a position buffer with another mesh?
if a mesh shares a position buffer with another mesh, but not a color buffer?
We might want to know how many vertices are represented in the file conceptually, or how many will be uploaded to the GPU, or how many will be drawn on each render pass. These may all be different numbers. Consider providing tools to disambiguate between:
total interchange vertices (total vertices in all unique position attributes)
total GPU vertices uploaded, expected
total GPU vertices uploaded, minimum
total GPU vertices drawn
...
unused vertices
This would be helpful for writing unit tests to ensure functions like join() do not increase the relevant counts.
Per-mesh vertex counts currently show up in the CLI, but there are a lot of different ways to compute a "vertex count" with different practical uses. Consider cases like:
We might want to know how many vertices are represented in the file conceptually, or how many will be uploaded to the GPU, or how many will be drawn on each render pass. These may all be different numbers. Consider providing tools to disambiguate between:
This would be helpful for writing unit tests to ensure functions like join() do not increase the relevant counts.
Resources, references, and explainers: