Browsing by Author "Ledoux, Franck"
Now showing 1 - 3 of 3
Results Per Page
Sort Options
Item A GPU-based Out-of-core Architecture for Interactive Visualization of AMR Time Series Data(The Eurographics Association, 2023) Alexandre-Barff, Welcome; Deleau, Hervé; Sarton, Jonathan; Ledoux, Franck; Lucas, Laurent; Bujack, Roxana; Pugmire, David; Reina, GuidoThis paper presents a scalable approach for large-scale Adaptive Mesh Refinement (AMR) time series interactive visualization. We can define AMR data as a dynamic gridding format of cells hierarchically refined from a computational domain described in this study as a regular Cartesian grid. This adaptive feature is essential for tracking time-dependent evolutionary phenomena and makes the AMR format an essential representation for 3D numerical simulations. However, the visualization of numerical simulation data highlights one critical issue: the significant increases in generated data memory footprint reaching petabytes, thus greatly exceeding the memory capabilities of the most recent graphics hardware. Therefore, the question is how to access this massive data - AMR time series in particular - for interactive visualization on a simple workstation. To overcome this main problem, we present an out-of-core GPU-based architecture. Our proposal is a cache system based on an ad-hoc bricking identified by a Space-Filling Curve (SFC) indexing and managed by a GPU-based page table that loads required AMR data on-the-fly from disk to GPU memory.Item Hex Me If You Can(The Eurographics Association and John Wiley & Sons Ltd., 2022) Beaufort, Pierre-Alexandre; Reberol, Maxence; Kalmykov, Denis; Liu, Heng; Ledoux, Franck; Bommes, David; Campen, Marcel; Spagnuolo, MichelaHexMe consists of 189 tetrahedral meshes with tagged features and a workflow to generate them. The primary purpose of HexMe meshes is to enable consistent and practically meaningful evaluation of hexahedral meshing algorithms and related techniques, specifically regarding the correct meshing of specified feature points, curves, and surfaces. The tetrahedral meshes have been generated with Gmsh, starting from 63 computer-aided design (CAD) models from various databases. To highlight and label the diverse and challenging aspects of hexahedral mesh generation, the CAD models are classified into three categories: simple, nasty, and industrial. For each CAD model, we provide three kinds of tetrahedral meshes (uniform, curvature-adapted, and box-embedded). The mesh generation pipeline is defined with the help of Snakemake, a modern workflow management system, which allows us to specify a fully automated, extensible, and sustainable workflow. It is possible to download the whole dataset or select individual meshes by browsing the online catalog. The HexMe dataset is built with evolution in mind and prepared for future developments. A public GitHub repository hosts the HexMe workflow, where external contributions and future releases are possible and encouraged. We demonstrate the value of HexMe by exploring the robustness limitations of state-of-the-art frame-field-based hexahedral meshing algorithm. Only for 19 of 189 tagged tetrahedral inputs all feature entities are meshed correctly, while the average success rates are 70.9% / 48.5% / 34.6% for feature points/curves/surfaces.Item Time Series AMR Data Representation for Out-of-core Interactive Visualization(The Eurographics Association, 2022) Alexandre-Barff, Welcome; Deleau, Hervé; Sarton, Jonathan; Ledoux, Franck; Lucas, Laurent; Sauvage, Basile; Hasic-Telalovic, JasminkaTime-varying Adaptive Mesh Refinement (AMR) data have become an essential representation for 3D numerical simulations in many scientific fields. This observation is even more relevant considering that the data volumetry has increased significantly, reaching petabytes, hence largely exceeding the memory capacities of the most recent graphics hardware. Therefore, the question is how to access these massive data - AMR time series in particular - for interactive visualization purposes, without cracks, artifacts or latency. In this paper, we present a time-varying AMR data representation to enable a possible fully GPU-based out-of-core approach. We propose to convert the input data initially expressed as regular voxel grids into a set of AMR bricks uniquely identified by a 3D Hilbert's curve and store them in mass storage.