Expand description
Reusable ping-pong offscreen field for stateful feedback scenes (ADR-0012).
Two same-format textures a simulation swaps each sub-step: the sim samples the previous state (the read view) and writes the next (the write view), then the field swaps so the fresh state becomes the next read. This is the engine’s first feedback path; the reaction-diffusion scene is its first user and future warp/feedback variants reuse it (ADR-0002 named it a deferred follow-up).
Composition, not engine machinery. The field owns only the texture pair
and the read/write selector; the scene owns its sim/present pipelines and
the shader that steps the field. That keeps the Scene seam thin — ADR-0012
rejected an engine-managed multi-pass pipeline for exactly this reason.
Structs§
- Feedback
Config - A preset’s
[feedback]table: the two load-time choices about how an accumulation reads its own past (ADR-0048).