A High-Quality Workflow for Multi-Resolution Scientific Data Reduction and Visualization

TL;DR

The paper tackles the challenge of reducing I/O and storage for large-scale multi-resolution scientific data by presenting a workflow that converts uniform data to multi-resolution form using ROI-based extraction, optimizes a global compressor SZ3 for multi-resolution data (SZ3MR) with dynamic padding and per-level error bounds, and augments block-wise compressors with an adaptive, error-bounded post-processing stage. It also introduces an uncertainty-visualization framework based on probabilistic marching cubes to quantify the impact of compression on isosurface features. Across in-situ and offline experiments with AMR and adaptive data from WarpX and Nyx, the approach achieves significant rate-distortion improvements, reduces visualization artifacts, and maintains low processing overhead, enabling practical, high-fidelity in-situ compression. The work advances storage efficiency for exascale computing and provides a robust methodology for assessing compression-induced uncertainty in scientific visualization and analysis.

Abstract

Multi-resolution methods such as Adaptive Mesh Refinement (AMR) can enhance storage efficiency for HPC applications generating vast volumes of data. However, their applicability is limited and cannot be universally deployed across all applications. Furthermore, integrating lossy compression with multi-resolution techniques to further boost storage efficiency encounters significant barriers. To this end, we introduce an innovative workflow that facilitates high-quality multi-resolution data compression for both uniform and AMR simulations. Initially, to extend the usability of multi-resolution techniques, our workflow employs a compression-oriented Region of Interest (ROI) extraction method, transforming uniform data into a multi-resolution format. Subsequently, to bridge the gap between multi-resolution techniques and lossy compressors, we optimize three distinct compressors, ensuring their optimal performance on multi-resolution data. Lastly, we incorporate an advanced uncertainty visualization method into our workflow to understand the potential impacts of lossy compression. Experimental evaluation demonstrates that our workflow achieves significant compression quality improvements.

Figures (18)

• Figure 1: Example of an AMR dataset of Rayleigh–Taylor instability.
• Figure 2: Vis of data distributions for different level for Fig. \ref{['rt-vis']}.
• Figure 3: Overview of our proposed workflow for multi-resolution scientific data compression.
• Figure 4: Visualization of the original Nyx cosmology dataset (left) and the ROI (right, 15% of the dataset) extracted using our approach, the SSIM of the two pictures is 0.99995.
• Figure 5: Vis comparison (one $1.5\times$ zoom in 2D slice) of original data and decompressed data produced by TAC's SZ3, AMRIC's SZ3 and our SZ3MR on Nyx's "baryon density" field (fine level). Warmer colors indicate higher values. The CR of TAC, AMRIC, and ours is the same, 163.