TopoSZp: Lightweight Topology-Aware Error-controlled Compression for Scientific Data
New compressor achieves 3-100x better topology preservation with 100-10000x speed gains over rivals.
Deep Dive
Researchers led by Tripti Agarwal developed TopoSZp, a lightweight, topology-aware lossy compressor for HPC simulation data. Built on SZp, it preserves critical points (minima, maxima, saddles) within strict error bounds. It delivers 100-10000x faster compression and 10-500x faster decompression than existing topology-aware methods while maintaining competitive ratios. Scientists can now compress massive datasets without losing structural features essential for analysis.
Why It Matters
Enables efficient storage of exascale simulation data while preserving the topological features researchers actually need for discovery.