There are definitely still quirks and glitches that need to be hammered out over time, and I need to stop adding new stuff soon and go into a performance analysis mode so I can refine and boost performance, but that damn statue business was a giant pain. A down side of the new approach is no custom thug models until a way is found to dynamically change the model of an existing entity. Another down side is the code now directly depends on existing entities, so machines with low resources or low settings have less available entities to draw from, which is directly proportional to how many things AW can "create" obviously. In the original approach I could create whatever I needed regardless, but the game as we've seen has a huge problem in dealing with that over time.
And I think I know why. The game is constantly recycling peds and cars as part of keeping itself within a 2gig footprint. And as you'll remember from one of my posts what we're doing is injecting content into the game on rogue threads the game is not aware of in the normal sense. It processes the thread because it's in the thread pool but it is clueless as to the nature of what it's doing or creating outside the control of it's normal collection system. So none of what is created in this manner falls within the games normal garbage collection and recycle except the fact it would be seeing higher numbers than it can account for, because it didn't create them, so over a long period of time it is probably litterally breaking its garbage collection. The stuff gets orphaned and stranded because the game isn't treating them the same as within its own collection system since it's not aware of or accounted for in that manner.
Switching to using what the game itself provides eliminates that problem at the cost of entity model customization. To get a stable long term war I'll definitely pay that price if no way is found to dynamically change models.
Last edited by Ironhide; 08-16-2010 at 08:03 PM..