Client side prediction rework
Network game code, particularly the client side prediction, has been completely rewritten.. again. This was probably the third time, but I have a strong feeling that this time I got it right. At least it looks very promising.
I have currently no motion filtering or smoothing implemented and client side movement already looks a lot smoother than it ever did before, even with heavy filtering applied. The code is also lot less complicated and much more maintainable.
The downside to all this is that system requirements on the client went up. I'm now doing a full "rewind and replay" type of prediction using the physics engine, which is really heavy on the CPU, especially with high network latency. There are still a few optimizations to be made, but I'm not expecting radical improvements.
There is still a lot of work to do until everything I broke works again and a few (hopefully) minor issues to solve. I'll report back later when I have done some more testing.
Posted on 2011-11-06