Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Parallel Adaptive Anisotropic Meshing on cc-NUMA Machines

Christos Tsolakis, Nikos Chrisochoides

TL;DR

The paper presents CDT3D, a parallel speculative framework for metric-based anisotropic 3D mesh adaptation on shared-memory cc-NUMA systems. By decomposing mesh operations into topological and geometrical components and integrating a multiscale metric ${oldsymbol{ m M}}$ with CAD geometry via the EGADS kernel, the approach achieves high strong scaling (over 90% efficiency) and supports end-to-end CFD pipelines. Evaluation across analytic fields and public CFD test cases (Delta Wing, ONERA M6, JSM) demonstrates accurate solution-driven mesh refinement and competitive agreement with literature, while CAD handling enhances geometric fidelity and deviation control. The work demonstrates a practical, scalable pathway for integrating advanced anisotropic mesh adaptation into production CFD workflows, with clear implications for exascale-ready meshing on shared-memory nodes. Practical impact includes robust, efficient adaptive pipelines that can handle complex geometries and high-fidelity simulations with reduced mesh counts and improved accuracy.

Abstract

Efficient and robust anisotropic mesh adaptation is crucial for Computational Fluid Dynamics (CFD) simulations. The CFD Vision 2030 Study highlights the pressing need for this technology, particularly for simulations targeting supercomputers. This work applies a fine-grained speculative approach to anisotropic mesh operations. Our implementation exhibits more than 90% parallel efficiency on a multi-core node. Additionally, we evaluate our method within an adaptive pipeline for a spectrum of publicly available test-cases that includes both analytically derived and error-based fields. For all test-cases, our results are in accordance with published results in the literature. Support for CAD-based data is introduced, and its effectiveness is demonstrated on one of NASA's High-Lift prediction workshop cases.

Parallel Adaptive Anisotropic Meshing on cc-NUMA Machines

TL;DR

The paper presents CDT3D, a parallel speculative framework for metric-based anisotropic 3D mesh adaptation on shared-memory cc-NUMA systems. By decomposing mesh operations into topological and geometrical components and integrating a multiscale metric with CAD geometry via the EGADS kernel, the approach achieves high strong scaling (over 90% efficiency) and supports end-to-end CFD pipelines. Evaluation across analytic fields and public CFD test cases (Delta Wing, ONERA M6, JSM) demonstrates accurate solution-driven mesh refinement and competitive agreement with literature, while CAD handling enhances geometric fidelity and deviation control. The work demonstrates a practical, scalable pathway for integrating advanced anisotropic mesh adaptation into production CFD workflows, with clear implications for exascale-ready meshing on shared-memory nodes. Practical impact includes robust, efficient adaptive pipelines that can handle complex geometries and high-fidelity simulations with reduced mesh counts and improved accuracy.

Abstract

Efficient and robust anisotropic mesh adaptation is crucial for Computational Fluid Dynamics (CFD) simulations. The CFD Vision 2030 Study highlights the pressing need for this technology, particularly for simulations targeting supercomputers. This work applies a fine-grained speculative approach to anisotropic mesh operations. Our implementation exhibits more than 90% parallel efficiency on a multi-core node. Additionally, we evaluate our method within an adaptive pipeline for a spectrum of publicly available test-cases that includes both analytically derived and error-based fields. For all test-cases, our results are in accordance with published results in the literature. Support for CAD-based data is introduced, and its effectiveness is demonstrated on one of NASA's High-Lift prediction workshop cases.
Paper Structure (20 sections, 10 equations, 26 figures, 1 table)

This paper contains 20 sections, 10 equations, 26 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related work
  3. Metric-based Adaptation within the CDT3D library
  4. Parallel Speculative Local Reconnection in Metric Spaces
  5. Parallel Point Creation for Anisotropic Mesh Adaptation
  6. Speculative Edge Collapse for Metric-based Adaptation
  7. Vertex Smoothing
  8. Effect of the complete set of operations
  9. Handling Geometry through metric spaces
  10. Evaluation
  11. Experimental Setup
  12. Analytic scalar fields
  13. Laminar Subsonic Flow over a Delta Wing
  14. Inviscid Onera M6 case
  15. Inviscid flow over the JAXA Standard Model
  16. ...and 5 more sections

Figures (26)

  • Figure 1: Pipeline of the presented approach. Dotted modules are utilized only when CAD data are available.
  • Figure 2: Topological Flips utilized by CDT3D for local reconnection.
  • Figure 3: Steps of speculative edge collapse.
  • Figure 4: Search Space for Smoothing Operation.
  • Figure 5: Left: The effect of adding more metric-aware operations. Right: The improved approach versus our previous results presented in tsolakis_parallel_2019.
  • ...and 21 more figures