CLIP: A CUDA-Accelerated Lattice Boltzmann Framework for Interfacial Phenomena

CLIP is a high-performance, extensible Lattice Boltzmann Method (LBM) library developed in C++/CUDA, designed to simulate interfacial phenomena such as bubbles, drops, and jets in both 2D and 3D domains.

It features a modular, GPU-accelerated architecture that enables easy integration of additional physical models and equations. Thanks to its flexible design, adding new models or equations requires minimal effort, making it ideal for extending to novel research problems.

Moreover, CLIP is highly optimized, allowing users to run many simulations — including 3D multiphase problems — on a personal computer without requiring a high-end GPU. For memory-constrained or low-end GPUs, users can enable single precision to significantly reduce memory usage and improve performance. With its fully configurable input system, CLIP is ideal for rapid prototyping, educational use, and scalable research.

🔧 Key Features

CUDA-Accelerated Kernels
Two-Phase Interfacial Flow
Versatile Boundary Conditions
Checkpointing Support
Config-Driven Simulation
WMRT Collision Stability
Allen–Cahn Phase Field
2D/3D Lattice Support
Single/Double Precision Toggle

🧪 Examples

CLIP supports the following types of simulations:

Rayleigh–Taylor Instability (2D & 3D)
Bubble Dynamics (2D & 3D)
Drop Dynamics (2D & 3D)
Jet Breakup & Pinch-Off (2D & 3D)

Animations are available in the GitHub version of the README.

⚙️ Build Instructions

Prerequisites

CUDA 8.0+
CMake ≥ 3.18
GCC ≥ 9
NVIDIA GPU (e.g., V100, A100, RTX, or even entry-level GPUs)
Python (optional, for post-processing)

Steps

Clone the repository

git clone https://github.com/your-org/CLIP.git

cd CLIP

Create a build directory

mkdir build && cd build

Configure with CMake

cmake .. -DCMAKE_BUILD_TYPE=Release -DENABLE_2D=ON                 # 2D double precision
cmake .. -DCMAKE_BUILD_TYPE=Release -DENABLE_3D=ON                 # 3D double precision
cmake .. -DCMAKE_BUILD_TYPE=Release -DENABLE_2D=ON -DUSE_SINGLE_PRECISION=ON  # 2D single precision

Build

make -j

Contact