How to Efficiently Handle Tick Batching for Input-Dependent Systems ?

Hello, Everyone!

What is the best approach for batching ticks in systems that depend on individual tick inputs ?
EG: We need to re-simulate the last 10 ticks. If we set MaxPredictionStepBatchSizeRepeatedTick to 2, those 10 ticks will be reduced to 5, and each tick’s deltatime will be multiplied by 2. However, this approach presents a challenge, especially for systems like the MovementSystem, which require simulating each tick with some different input data. the simulation is likely to result in frequent mispredictions.

Thanks !

Hey Opeth! Our solution to this is that we don’t batch client prediction ‘rollback & resimulation ticks’ when inputs change (i.e. we only batch when inputs don’t change between frames) for this reason. I’ve noted that the XML comment does not communicate this.

1 Like

As always, thank your for the detailed response.

1 Like