GPU-accelerated micropolar <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" altimg="si66.svg"> <mml:msup> <mml:mi>δ</mml:mi> <mml:mo>+</mml:mo> </mml:msup> </mml:math> -SPH framework with enhanced particle shifting for complex fluid dynamics

Smoothed Particle Hydrodynamics (SPH) faces persistent challenges in computational efficiency and numerical accuracy. This paper presents a GPU-accelerated framework integrating micropolar δ + -SPH with enhanced stabilization techniques. The implementation features four key components: cross-platform execution on both NVIDIA and AMD GPUs via OpenGL compute shaders; an in-out buffer system enabling fully parallel inter-voxel particle migration without global synchronization; an enhanced particle shifting technique maintaining uniform distributions without explicit free-surface detection; and a micropolar formulation preserving vortical structures through bidirectional coupling between velocity and micro-rotation fields. These design choices yield substantial improvements: (a) benchmarks demonstrate 17–18% faster execution than DualSPHysics v5.4 and 97 × speedup over server-class CPU implementations; (b) lid-driven cavity simulations achieve L 2 errors below 1.3% for Reynolds numbers 100–10,000, while NACA0015 airfoil simulations at R e = 180 , 000 show lift coefficient agreement within 7% of reference data; (c) the micropolar formulation reduces aerodynamic coefficient fluctuations by 20–22% at stall conditions compared to classical Navier–Stokes; and (d) the zero-copy architecture eliminates I/O bottlenecks, enabling real-time visualization and parameter adjustment. This framework establishes SPH as a practical tool for real-time fluid dynamics simulation, bridging the gap between academic research and industrial applications. Program Title: GPU-Micropolar-DeltaPlus-SPH Developer(s): Chen Yixuan Licensing provisions: MIT License Programming language: Python, GLSL (OpenGL Shading Language) Computer: Desktop workstation or HPC system with OpenGL 4.5+ support Operating system: Linux, Windows 10/11 RAM: Minimum 8 GB, recommended 32 GB GPU Memory: Minimum 4 GB, recommended 12 GB or more Keywords: SPH, GPU computing, Micropolar fluids, δ + -SPH, CFD Classification: 02.70.-c (Computational techniques; simulations), 07.05.Tp (Computer modeling and simulation), 02.70.Ns (Molecular dynamics and particle methods), 47.11.Kb (Spectral and particle methods) External routines: OpenGL 4.5+, GLFW 3.3+ Nature of problem: Simulation of incompressible viscous flows at high Reynolds numbers with complex vortical structures, free surfaces, and moving boundaries Solution method: GPU-accelerated SPH with micropolar formulation, δ + -SPH stabilization, and enhanced particle shifting Unusual features: Zero-copy visualization pipeline, cross-platform GPU compatibility via OpenGL, adaptive particle shifting without free-surface detection Running time: Real-time (300+ FPS) for 10 6 particles on NVIDIA RTX 5090