Hacker News Reader

> The Over-sampling Crisis (Computational Waste):

> To prevent "tunneling" (objects passing through each other) in high-dynamic scenes, developers are forced to infinitely shrink the time-step (Δt).

Surely this was avoided from decades ago by analytical solving.

View on HN

You are referring to Continuous Collision Detection (CCD), which has indeed existed for decades. However, CCD is a detection patch, not an integrator cure. 1. The Scaling Wall: While CCD avoids tunneling for a single pair of objects, solving it analytically for a system with thousands of constraints leads to a Non-linear Complementarity Problem (NCP) explosion. Most engines fallback to iterative solvers (like PGS or Jacobi), which brings us back to square one: high-frequency iterations to resolve 'shaking' constraints. 2. Integrator Drift: CCD finds the time of impact, but the integration still happens in discrete space. You still suffer from Numerical Dissipation (energy loss) because the state manifold is disconnected between steps. 3. The 'Why' of Octonions: Our approach isn't just 'detecting' the collision; it's about State Coupling. By using Non-associative algebra, we lock the causal dependency into the movement itself. We are replacing the O(n^2) geometric 'check-then-fix' loop with a single-pass O(n) algebraic update. In short: CCD tells you when you crashed; Octonions ensure the state update respects the causal sequence without the iterative overhead.

In short：CCD is a diagnostic patch; Octonions are an algebraic cure. One checks for crashes, the other makes physics causal by design.

Reply on news.ycombinator.com