Reproduce “PeriPy -- A High Performance OpenCL Peridynamics Package” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
B. Boys, T. J. Dodwell, M. Hobbs, M. Girolami
PeriPy, an open-source Python package, accelerates peridynamic simulations by leveraging OpenCL for GPU/CPU acceleration, achieving up to 10x faster performance than prior implementations for small problems and 1.4-2.0x for larger ones. This enables complex "outer-loop" applications and large-scale fracture modeling in engineering.
Key insights
- Kernel fusing, which combines bond force calculation and bond breaking checks, significantly reduces overhead and redundant computations.
- A binary parallel reduction algorithm, executed within OpenCL work-groups and utilizing fast local memory, efficiently sums bond force contributions to nodal forces.
- Minimizing thread divergence by parallelizing computations over individual bonds rather than nodes improves GPU utilization, especially for irregular meshes with varying bond counts per node.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.