Toward GPU-Resident Climate Models: A Feasibility Study on Lossy Compression for the Spherical Harmonic Transform's Communication Bottleneck
In the authors' words
Operational pseudospectral atmospheric models such as the ECMWF Integrated Forecasting System (IFS) run today almost exclusively on CPUs; GPU ports are under active development but not yet used in production. These models rely on the Spherical Harmonic Transform (SHT). Each time-step requires forward and inverse SHTs, and both passes depend on global pencil transposition that redistribute multi-dimensional arrays across compute nodes. At large node counts these global collectives dominate wall-clock time. We investigate GPU-resident lossy compression, using representative fields from the DYAMOND high-resolution operational dataset as input, and combining measured GPU compression throughput with SimGrid network simulation, we show that ZFP at 16 bits per value (rate-16), about the same storage budget as float16, matches the communication-time reduction of float16 truncation while delivering approximately lower mean relative error. ZFP at 8 bits per value (rate-8) achieves approximately the speedup of float16 while retaining lower mean relative error.
Appeared: Tuesday, September 22. arXiv. Preprint, not yet peer-reviewed.
Authors' comment: Accepted for publication in the 2026 IEEE International Conference for High Performance Computing, Networking, Storage and Analysis Workshops (SC-W) - High-Performance Computing for Environmental and Earth Sciences (HPC4EES). \textcopyright