How should I implement multi-pass rendering in a game engine?

Question

I have done multi-pass rendering, before, and understand how it works. I made a simple example, which rendered a basic scene with shadows. This was all part of one file. Now, I am trying to figure out is how to put it into my game engine.

Currently, my game engine uses a single pass. It is in a hierarchical structure, and uses Direct3D 9. I have a graphics component, which will load and draw a 3D model. In my game loop, I update all of the entities in the world, then I call the draw function for each one. This draw function gets the vertex buffer, index buffer and texture or material, and draws the 3D model using a shader. This works fine.

To do multi-pass rendering, to allow for shadows, I will need to draw each model multiple times. It doesn't seem right, to me, that in each models draw function I should put the second pass code; this will then be completed before the next models first pass.

How should I implement multi-pass rendering in a game engine?

Answer 1

The entities should not be responsible for drawing themselves. They (or just their renderable components) should be passed to the rendering engine and the rendering engine should have the responsibility of rendering them.

Answer 2

With your current structure you'll need a second API for your models to handle the shadow pass.

If you wanted to invert some of your logic then let the models store their data and a friendly description of how they want to be rendered, and let the graphics engine pull data from the models to render according to those descriptions. This will let the graphics code handle the shadows (or post processing FX, etc...) in its own way with the second pass.