Periodic Reporting for period 1 - ESiWACE3 (Center of excellence for weather and climate phase 3)
Período documentado: 2023-01-01 hasta 2023-12-31
A new generation of efficient, optimised weather and climate models running on supercomputers is needed for reliable predictions of climate change and the weather and extreme weather events in a changing climate. The Centre of Excellence in Simulation of Weather and Climate in Europe, ESiWACE, was created to serve the weather and climate modelling community with innovative technologies and tools to improve the model performance and provide enhanced support and training. Nowadays, the path towards exascale computing holds enormous challenges regarding portability, scalability and data management, which individual institutes can hardly face. In this direction, the third phase of the Centre of Excellence in Simulation of Weather and Climate in Europe, ESiWACE3, aims to link, organise and enhance Europe’s excellence in weather and climate modelling to enable more detailed resolving weather and climate simulations on the upcoming exascale supercomputers and provide the required technology for better and more detailed climate-related risk assessments on a local level. To achieve this goal, ESiWACE3 focus on three main goals that are essential to prepare existing operational weather and climate prediction systems for the exascale era: 1) the transfer and establishment of knowledge and technology for efficient and scalable simulations of weather and climate across the Earth System modelling community in Europe; 2) closing common technology gaps in the knowledge and toolbox for high-resolution Earth System modelling via joint developments across the European community; and 3) serving as a sustainable community hub for training, communication, and dissemination of high-performance computing for weather and climate modelling in Europe. ESiWACE3 will bring the various approaches to address these challenges from the different modelling groups together to transfer knowledge across the weather and climate domain, generate synergies between the local efforts, provide targeted support to modelling groups via customised high-performance computing services, and provide training to educate the next generation of researchers.
To facilitate co-design efforts and furnish HPC vendors with meaningful benchmarks, the High-Performance Climate and Weather (HPCW) benchmark has been improved. Some of the core applications (dwarfs) have been used to enhance the LLVM compiler of RISC-V in collaboration with the EUPILOT project or employed to assess the new High Bandwidth Memory (HBM) with the EPI and EUPEX projects. Selected dwarfs have been successfully deployed on two pre-exascale machines, LUMI and LEONARDO. Developments have been made on two Domain Specific Languages (DSL) for porting weather and climate models to heterogeneous supercomputers: the development of FVM_LAM in Python using the GT4Py DSL and of NEMO using the Psyclone DSL, with the GPUs porting of two NEMO configurations. Significant data volumes generated by weather and climate codes have been achieved. The generation of a dataset of model output from IFS without any data compression has been initiated. An "Online Field Compression Laboratory'' library has been developed permitting the loading of model output fields into Jupyter notebooks for further data manipulation. Various data compression approaches have been studied together with diagnostics to check for data quality. A portfolio and list of service topics have been addressed to bridge the broader community with HPC services. Three proposal have been granted from the two calls for HPC service opened: DALES for GPU porting with openACC; GLOBO to explore Multicore HPC profiling and IO optimisation; SMValTool for code optimisations and data load speed. ESiWACE3 has been organized the following training activities: The first hackathon was organised by CSC in Espoo, Finland, on October 2023; An online training events calendar has been developed, including a comprehensive list of events related to Earth System Modeling (ESM), High-Performance Computing (HPC), and their intersection also from other European projects.
ESiWACE3 has played a pivotal role in ushering in a notable transformation in the capabilities of European climate and weather models. It developed a set of applications to encapsulate the complexity of weather and climate models to furnish HPC vendors with meaningful benchmarks employed for evaluating new architectures through co-design. ESiWACE3 has made substantial contributions mainly through the High-Performance Climate and Weather (HPCW) benchmark. This includes formulating a governance plan for the coming years and the promotion of its utilisation for diverse purposes such as deployment on EuroHPC machines, collaboration with other projects like HANAMI, and serving as test cases for emerging architectures such as RISC-V or de GPU porting through two leading initiatives dedicated to domain-specific languages (DSLs) for weather and climate models. ESiWACE3 is tackling the entire data challenge, achieving notable improvements with novel data storage and cataloguing approaches. These advancements are propelling weather and climate models to a new level, preparing them for efficient adaptation to exascale machines with faster and more accurate codes for the new generation of EuroHPC machines and the efficient handling and analysis of model output. This plays a pivotal role in preparing society for extreme events and mitigating the effects of climate change. We aim to contribute to initiating follow-on projects and initiatives that will further exploit, consolidate, and advance these applications.