The latter. Alternatively you can use the vertex shader to for example change the color and position of vertices. Often shader-based programs (especially fragment shaders) move everything to the shader so the application won't need to get back any information, since CPU-GPU interactions are a bottleneck.

How the shaders operate under Processing isn't different from their regular task. So Processing just allows you to use shaders, the complexity of shaders themselves remains unchanged. See the world wide web for examples and information on GLSL shaders as such.

Thank you very much.