Jeffrey Camlin
Numerical blow-up near peak vortex stretching has long been attributed to incipient singularity or fundamental limits of spatial resolution. We show it is neither. The obstacle is insufficient sampling of Physical Simulation Time: a problem of temporal allocation, not spatial discretization. We introduce iDNS (intelligent Direct Numerical Simulation), a stiffness-aware integration framework based on bounded deterministic temporal lifting: a diffeomorphic reparameterization t = phi(tau) that separates Physical Simulation Time t from Lifted Computational Time tau. The lifting is governed by a bounded sigmoid controller where s is a real-time vorticity stiffness indicator. The controller parameters (A, k, c) are fixed from Re = 1600 to Re = 10^8, spanning five orders of magnitude in Reynolds number. No manual tuning, no CFL heuristics, no problem-specific expertise required. iDNS achieves R_epsilon = 1.000 +/- 0.001 at N = 64^3 on a consumer laptop. The NASA Glenn WRLES benchmark requires N = 512^3 on 368 CPU cores to approach the same dissipation balance. All simulation data, plots, and results are fully reproducible via a 418-line Python script provided in the repository.