GPUTopOpt: GPU-Accelerated Topology Optimization Pipeline for Large-Scale 3D Design

A GPU-accelerated topology optimization pipeline written in Julia and CUDA for large-scale 3D structural design. The code implements the SIMP formulation with Heaviside projection and a matrix-free multigrid-preconditioned conjugate gradient solver, and includes an optional porous infill extension validated on a femur reconstruction case at approximately 77 million degrees of freedom. It was authored by Vadillo Morillas, Abraham, and last updated on June 21, 2026.

Use Cases

• Optimizing large-scale 3D structural components for additive manufacturing based on the described voxel-based pipeline.
• Designing bio-inspired or lattice structures using the optional porous/local-volume infill extension based on the Wu–Aage–Sigmund p-norm formulation.
• Benchmarking GPU-accelerated, matrix-free multigrid solvers for topology optimization problems mentioned in the description.
• Teaching or replicating topology optimization methods like SIMP and the Method of Moving Asymptotes (MMA) in a Julia/CUDA environment.

Strengths

• Validated on a large-scale problem with approximately 77 million degrees of freedom for a femur reconstruction case.
• Includes CPU/GPU tiled assembly to minimize GPU memory pressure, enabling problems that would otherwise exceed VRAM.
• Distributed as a self-contained 61 MB package with 15 Julia source files, example cases, and user guides.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Requires specific software dependencies: Julia 1.10+, an NVIDIA GPU with CUDA 11+, and gmsh on PATH.

Provenance

Source

Vadillo Morillas, Abraham via e-cienciaDatos Harvested Dataverse

Collection Method

Forked and extended from the upstream FSI-Topology-Optimisation-Portable codebase.

Freshness

Last updated 2026-06-21 04:10:12