1-Bit Fluids

Deterministic checkpointing of fluid-mediated artificial life simulations can be prohibitive. For example, a single checkpoint of a 256^3 fluid velocity field occupies 100 MB at half precision. This paper presents a lossless compression scheme that reduces this to under 1 bit per velocity component (under 6.3 MB) on typical simulation data. A hierarchical non-causal predictor exploits spatial correlation across octave spatial scales, using up to 26 neighbors per sample. ECDQ-style dither synchronization accommodates upstream subtractive dither at near-zero compression cost, reducing dither’s penalty from over 70% to under 9% by aligning encoder and decoder rounding thresholds. The regime in which this penalty vanishes with increasing quantization resolution is characterized. Prediction residuals are split into homogeneous streams for entropy coding, achieving file sizes within 3% of the residual entropy. Extended version of a late breaking abstract accepted at ALife 2026.