Today’s update will be a quick one, as the task is still underway.
I decided to do some test to check the performance of the current system if we use larger terrain chunks. So far, all testing had been done on a 16×16 grid, and performance was fine.
So I decided to up the ante to 32 x 32, and the performance was bad…
There was a very notable dip below 60fps, and that’s not good. Not to mention this was just a barren terrain, and that is bad. I’m sure there could be some optimization done, but it still isn’t good enough.
So I decided to redo the tile system again (third time is the charm!), this way separating the terrain into chunks of 16×16 tiles, all contained inside a single mesh. (Chunks can be made up to near 64×64, due to Unity limitations regarding vertices in a mesh). As a test, I spawned 3 chunks of 32×64, and placed it all on the scene. This is roughly 6000 tiles on display, spread across 3 large meshes.
This is the result:
As you can see, the performance stayed at 70fps range, render thread was cut by 75%, cpu remained 4ms faster, much smaller number of shadow casters, batches, and a vastly greater number of verts and tris.
So I’ll need some time to adjusting the current system to match this. The generation itself is working fine, as is the UV mapping. But I’ll need to rework the shot system since it’s all under one big mesh collider. Some things will be vastly simplified, though. And the game will run better for it!
Stay tuned for more!