Work

Deliverable 2.3: Report on DSL development in NEMO and IFS-FVM

Year: 2024

Type: article

Abstract: Heterogeneous supercomputer architectures have advanced significantly in recent years, incorporatingmulti-node parallelism and graphics processing units (GPUs) to create larger and more energy-efficientsystems. At the same time, climate models are growing increasingly complex, computationally demanding, andtime-intensive. GPUs are becoming crucial for speeding up scientific applications, and by leveraging theirprocessing power, climate researchers can accelerate simulations and more quickly explore a wide range ofclimate scenarios.This work shows the enhancements made so far on DSL development in NEMO and IFS-FVM models with theaim of allowing the targeting of these models on GPUs. In particular, the document summarizes the progressmade when PSyclone DSL in NEMO and GT4Py DSL in IFS-FVM are used.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.14807847

Authors: Epicoco, Italo, Mele, Francesca, Kühnlein, Christian, Faghih-Naini, Sara, Dueben, Peter

Institutions: University of Salento, CMCC Foundation - Euro-Mediterranean Center on Climate Change, European Centre for Medium-Range Weather Forecasts, University of Oxford

Language: English

Topic: Parallel Computing and Optimization Techniques

Subfield: Hardware and Architecture

Field: Computer Science

Domain: Physical Sciences

SDG: Climate action

Open Access status: green

Funder: European Commission