Problem
Research questions and friction points this paper is trying to address.
Lattice Boltzmann method
GPU computing
memory footprint
large-scale simulation
multi-GPU
LBFAST: A Lightweight Moment-Represented Lattice Boltzmann Solver for Multi-GPU Architectures
🤖 AI Summary
This study addresses the constraints of GPU memory capacity and bandwidth in large-scale three-dimensional lattice Boltzmann simulations by proposing a lightweight GPU solver based on a reduced moment set. The proposed method significantly decreases storage requirements through the online dynamic reconstruction of particle distribution functions, while achieving efficient computation via multi-node parallel scaling techniques. Experimental results demonstrate that the solver attains near-ideal weak scaling across 512 GPUs, effectively balancing numerical accuracy, stability, and high throughput. Consequently, this work provides an energy-efficient acceleration framework for large-scale fluid dynamics simulations.
📝 Abstract
We present LBFAST, a GPU-oriented lattice Boltzmann solver based on a lightweight moment-represented formulation, in which post-collision populations are reconstructed on the fly from a reduced set of moments rather than stored explicitly. This approach significantly lowers the memory footprint, enabling large three-dimensional simulations within the constraints of modern accelerator architectures, where VRAM capacity and bandwidth are critical resources. The method is assessed through standard single- and two-component benchmarks demonstrating good accuracy and stability. Extensive scaling experiments on multi-GPU systems show near-ideal weak scaling up to 512 GPUs and sustained performance across different velocity sets. The combination of reduced memory usage, competitive throughput, and stable energy efficiency makes the proposed formulation a practical route for large-scale lattice Boltzmann simulations on current and emerging HPC platforms.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Lattice Boltzmann Method
Moment-represented formulation
Multi-GPU
Memory footprint reduction
Weak scaling
🔎 Similar Papers
No similar papers found.
💼 Related Jobs
