Hey guys!

I am going through Rendering in Call of Duty presentation.

I would like to double-check with you if I follow correctly their shadow caching scheme (slides 21, 22) .

As the first step, they cull light view frustums against camera frustum.

Then, they take from the stale cache all shadow maps which require update. Phrasing "DO NOT need update" sounds a bit odd to me in this context.

Among these shadow maps from the stale cache, they pick 4-8 of them based on importance (detected as visible on the first step?, distance, projection size, etc) and update them.

They are sort of trying to fill their stale cache as soon as possible.

If I understand correctly, the stale cache is the one which drives shadow map update as they want to populate it as soon as possible?

I guess, another strategy would be to pick 4-8 relevant shadow maps to the camera frustum.

For each shadow map, check if it is in the stale cache and update if needed. Otherwise, do nothing.

By doing this, stale cache gets populated not so aggressively.

BTW, what does keyword "FIN" mean in the slide?

Thanks!

@Hodgman Thanks for the input. They do use the first strategy I described trying to fill the stale cache asap, right?

@Hodgman Suppose, we have 128 shadow maps in the cache and 50 out of them require update. Say we can handle 8 shadow maps per frame and, currently, we have 6 lights visible in the camera with 1 up-to-date shadow map. So we need to update only 6-1=5 shadow maps. But since our budget for shadow map update is 8, I was under impression, they would pick 8-5=3 outdated shadow maps more from the cache so that to keep it up-to-date as much as possible. But if they do not want to update stale cache as soon as possible they would just update currently visible 5 shadow maps.