Checkpoint — groups, fixed step & the spiral

01why

Why must the frame split into a variable outer loop containing a fixed inner loop?

1. aTo reduce CPU usage
2. bRendering wants a variable rate to show the freshest state at whatever speed the hardware allows, while simulation needs a fixed rate or the same inputs give different results on different machines — incompatible requirements that one loop cannot serve
3. cBecause the job system requires nested groups
4. dTo separate managed from unmanaged code

02why

A frame took 33 ms with a 16.67 ms step. Why loop twice instead of running one step with dt = 33 ms?

1. aTwo smaller steps are more numerically stable
2. bScaling the step makes step size a function of frame rate, so a 30 fps machine computes different physics than a 60 fps one — exactly the variable-delta problem the fixed step existed to eliminate. Determinism requires the step never varies, so the only free variable is how many steps you take
3. cThe physics engine can't handle large steps
4. dLooping is faster

03how

Why does the accumulator carry the remainder forward instead of discarding it?

1. aTo improve precision of the next step
2. bDiscarding it makes the simulation clock drift from wall time — a few discarded milliseconds per frame becomes seconds of desync between simulated and experienced time over minutes
3. cThe remainder is needed for interpolation
4. dIt prevents the spiral of death

04why

Describe the spiral of death and why it is inherent rather than a bug.

1. aA memory leak in the fixed-step group
2. bExpensive steps make the frame longer, which accumulates more time, which demands more steps, which makes the frame longer still — positive feedback that falls out of catch-up itself and doesn't recover on its own
3. cA deadlock between simulation and presentation
4. dFloating-point error accumulating over many steps

05why

The step cap discards time under load, making the game run in slow motion. Why is that the right choice?

1. aIt isn't; it's a known flaw
2. bThe alternatives are hanging forever (spiral) or varying the step size (losing determinism) — running slow is the only option that keeps the simulation both responsive and reproducible. When overloaded, a fixed-step simulation trades time fidelity to preserve determinism
3. cIt reduces power consumption
4. dSlow motion is more forgiving for players

06how

Why does raising the step cap make things worse rather than better?

1. aIt uses more memory
2. bIt trades a slow-motion symptom for a spiral — more permitted steps means more time spent per frame, which accumulates more time, pushing you toward the runaway the cap existed to prevent
3. cThe cap is hardcoded and cannot be raised
4. dIt breaks interpolation

07why

Why does presentation interpolate between simulated states, and which earlier decision does that justify?

1. aFor anti-aliasing; it justifies the LocalToWorld matrix
2. bThe simulation sits at a discrete tick that rarely aligns with the display moment, so presentation blends the last two states for smooth motion — which is exactly why Module 6 stores LocalTransform as decomposed TRS with a quaternion, since blending matrices produces garbage
3. cTo reduce GPU load; it justifies uniform scale
4. dFor network prediction; it justifies the Parent component

08how

A system samples player input and is placed in the fixed-step group. What goes wrong?

1. aNothing; that's the correct placement
2. bIn a frame with zero fixed steps input is never sampled, and in a frame with three it's sampled three times — the standard shape is to sample at variable rate into a buffer that fixed-step systems consume
3. cIt causes a race condition with rendering
4. dInput becomes nondeterministic