Swift: GPU-Accelerated Spectral Solver for the MOOSE Framework

Efficient 3D Mesoscale Simulations with GPU-Enabled Spectral Solvers
Technology No. CW-25-07 SWIFT

The Challenge

Simulating mesoscale microstructure evolution and micro-mechanics often requires solving large-scale partial differential equations (PDEs) in 3D. Traditional solvers, while effective, can be computationally intensive and time-consuming, especially for high-resolution simulations. Researchers need faster, more efficient methods to accelerate these simulations and take advantage of modern GPU hardware.

How It Works

Swift introduces a GPU-accelerated spectral solver to the MOOSE ecosystem, enabling faster, more efficient simulations for mesoscale microstructure evolution and micro-mechanics.

• Fast Fourier Transform (FFT)-Based Spectral Solver: Swift uses FFT techniques to efficiently solve PDEs on regular orthogonal grids, often applied to representative volume elements (RVEs).

• GPU Acceleration: Designed to utilize GPU hardware, Swift accelerates semi-implicit solutions of PDEs, achieving high computational efficiency.

• Seamless Coupling with MOOSE Models: The solver integrates directly into the MOOSE framework, coupling with existing MOOSE-based models to expand their simulation capabilities.


Key Advantages

• High Computational Efficiency: Accelerates large-scale 3D simulations of phase field mesoscale microstructure evolution and micro-mechanics using GPUs.

• Seamless Integration: Adds a new spectral solution method to the MOOSE framework while maintaining compatibility with existing models.

• Optimized for Orthogonal Grids: Operates on regular orthogonal grids, ideal for representative volume elements (RVEs).

• Enhanced Scalability: Unlocks GPU resources to handle large, complex simulations with greater speed and precision.


Market Applications

• Materials Science: Enables researchers to simulate mesoscale phase field evolution for materials design and analysis.

• Computational Mechanics: Supports micro-mechanics simulations that require solving PDEs efficiently at scale.

• High-Performance Computing (HPC): Leverages GPU acceleration to improve simulation efficiency for computationally demanding problems.


This software is open source and available at no cost. Download now by visiting the product's GitHub page.

  • swap_vertical_circlecloud_downloadSupporting documents (1)
    Product brochure
    Swift: GPU-Accelerated Spectral Solver for the MOOSE Framework.pdf
Questions about this technology?