High-Performance Moment-Encoded Lattice Boltzmann Method with Stability-Guided Quantization

TL;DR

This paper introduces a GPU-optimized, stable, and memory-efficient lattice Boltzmann method using moment encoding and stability-guided quantization, achieving significant speedups and reduced memory usage for large-scale fluid simulations.

Contribution

It presents a novel split-kernel GPU framework for LBM, performs the first von Neumann stability analysis of high-order moment-encoded LBM, and develops a stable 16-bit quantization scheme.

Findings

Up to 6x speedup over existing solvers

Memory footprint reduced by up to 50%

Maintains physical accuracy in large-scale benchmarks

Abstract

In this work, we present a memory-efficient, high-performance GPU framework for moment-based lattice Boltzmann methods (LBM) with fluid-solid coupling. We introduce a split-kernel scheme that decouples fluid updates from solid boundary handling, substantially reducing warp divergence and improving utilization on GPUs. We further perform the first von Neumann stability analysis of the high-order moment-encoded LBM (HOME-LBM) formulation, characterizing its spectral behavior and deriving stability bounds for individual moment components. These theoretical insights directly guide a practical 16-bit moment quantization without compromising numerical stability. Our framework achieves up to 6x speedup and reduces GPU memory footprint by up to 50% in fluid-only scenarios and 25% in scenes with complex solid boundaries compared to the state-of-the-art LBM solver, while preserving physical fidelity across a range of large-scale benchmarks and real-time demonstrations. The proposed approach enables scalable, stable, and high-resolution LBM simulation on a single GPU, bridging theoretical stability analysis with practical GPU algorithm design.