10

Guess who has to finally swallow his pride and implement traditional deferred rendering with a traditional gbuffer even though he swore to never do that

This guy right here

Comments
  • 1
    And why you're doing that now?
  • 4
    @retoor Turns out I actually dont! Haha, my stubbornness strikes again!

    No actually the problem with visibility rendering is that you can't keep vertex positions across your passes meaning that for each pixel you have to transform 3 vertices twice (once for a "prepass" generating normals and stuff, and again for the actual shade pass), plus calculating barycentric coordinates twice plus transforming all vertices again in the vis pass

    With a gbuffer you would only need to do it once, since your shading only cares about the gbuffer parameters, also good for dynamic geometry since you can render that directly into your gbuffer with traditional vertex/fragment shaders...

    Well whatever, as far as I can tell doing the vertex transforms per pixel is more or less free (yeah really) taking only about 15 micros, most of which is the overhead of just launching the compute shader invocations themselves lol
Add Comment