Adapting Atmospheric Chemistry Components for Efficient GPU Accelerators

TL;DR

This paper enhances atmospheric chemistry modeling by adapting the MONARCH system for GPU acceleration, achieving significant speedups through novel multi-instance solving strategies and optimized GPU memory access.

Contribution

It introduces a new GPU-based approach and multi-instance solving method for atmospheric chemistry models, significantly improving computational efficiency.

Findings

Up to 9x speedup with multi-instance solving.

GPU implementation achieves up to 1.7x speedup over CPU.

Optimized memory access further increases GPU speedup.

Abstract

Atmospheric models demand a lot of computational power and solving the chemical processes is one of its most computationally intensive components. This work shows how to improve the computational performance of the Multiscale Online Nonhydrostatic AtmospheRe CHemistry model (MONARCH), a chemical weather prediction system developed by the Barcelona Supercomputing Center. The model implements the new flexible external package Chemistry Across Multiple Phases (CAMP) for the solving of gas- and aerosol-phase chemical processes, that allows multiple chemical processes to be solved simultaneously as a single system. We introduce a novel strategy to simultaneously solve multiple instances of a chemical mechanism, represented in the model as grid-cells, obtaining a speedup up to 9x using thousands of cells. In addition, we present a GPU strategy for the most time-consuming function of CAMP. The GPU version achieves up to 1.2x speedup compared to CPU. Also, we optimize the memory access in the GPU to increase its speedup up to 1.7x.