Fluid simulation remains a core challenge in computer graphics, with ongoing progress in achieving realistic modeling, rendering, and dynamic behavior. This paper introduces a novel real-time fluid simulation framework grounded in the Lagrangian approach, leveraging Smoothed Particle Hydrodynamics (SPH) for flexible and efficient boundary condition handling. Our primary innovation lies in integrating GPU acceleration to address performance bottlenecks commonly seen in high-resolution simulations. By utilizing multi-texture operations and spatial hashing, we streamline neighbor searches and boost computation efficiency. This implementation, developed within the Unity engine, enhances real-time interactions through custom interactive components—such as collision handling, fluid collectors, and visual effects—enabled by Signed Distance Fields (SDF). Additionally, we improve upon conventional Position-Based Fluids (PBF) techniques by optimizing the algorithm for interactive applications. The paper details the theoretical foundations, data structures, and algorithmic processes, along with GPU-specific optimizations employed. We conclude by discussing potential future research directions, such as exploring alternative GPU acceleration techniques, introducing color blending for visual variety, and creating more advanced SDF-based interactions.