thorn socket
So my engine pipeline is very dumb right now, I have a scene and a camera in it. To display something on screen I have a "renderer" that update every frame with the input scene and camera to render. This means the renderer own all the render target necessary to be displayed on screen later on, it manages which passes to do on its own based on what is present/enabled in the scene.
Until now that was working great, but arise the question of rendering cubemaps, and I'm struggling a bit on what should be the best course of action. I'm thinking of adding the concept of "views", basically a pipeline on its own inside a view with a specific camera attached to it which would render something. This means this view would own its own buffers to render something. I don't feel this is practical in term of memory consumption. However this feels simpler than trying to set up a render graph. I don't want to over engineer and keep things straightforward.
What do you folks think ?
What are the best practices for this stuff ?
sonic cosmos