Originially a Direct2D term, since diluted into “keep all the state on your GPU”. Contrast with:
It is possible to load everything upfront and only use “get drawing” call of your API. For most applications, however, it turns into long fat-tailed distribution of how much stuff do you want to update and how often.
There’s a natural range of update intensity:
- Loaded at start and not changing ever: fonts, UI elements, models for simple games, camera projection (unless the window is resizeable).
- Loaded when needed, but then sit in memory until you get far away from that house/planet/level.
- Changes quite frequently, but not every frame: camera view and some model transforms.
- Does not make sense to retain for more than one draw call: particles, animations and effects, time.
OpenGL has buffer usage hints to reflect update patterns. Alas, they are only hints:
Is it better to use STATIC for buffers that are updated very infrequently? Is it better to use DYNAMIC for buffers that get updated frequently, but not at STREAM speed? Is it better to use DYNAMIC for buffers that get partially updated? These are questions that can only be answered with careful profiling. And even then, the answer will only be accurate for that particular driver version from that particular hardware vendor.
One can go even further with indirect drawing buffers, where a buffer hold a batch of arguments for draw calls.
- You, when doing things right.