One of the forms in which ESiWACE3 supports the European modelling community is by offering tailored advice, engineering time and/or help with implementing code modifications – free of charge – to weather and climate modelling groups. Below are examples of engineering and software development services that ESiWACE3 might offer throughout the project.  In some cases, they build on cutting-edge development work being undertaken within ESiWACE3, giving you rapid and easy access to the latest tools and techniques.

Opportunities to apply for support will be made available through a series of calls for applications to be held during the ESiWACE3 project period. (More details about eligibility will be found in the respective calls.) If you would welcome help in any of the areas listed here that is not the subject of a current call, you are encouraged to contact us to discuss whether it could be included in a future call. 

  • Optimising weather and climate model codes
    Support to prepare ESM simulations to (pre-)exascale machines by helping to port and optimise model code to new machines. Analysis of the computational performance and scalability and optimisation of efficiency, targeting accelerator hardware as well as CPU codes (particularly for many-core machines). Also, the assessment of energy efficiency of model codes can be supported.
  • Reduced precision for ESM
    Support to improve the computational performance of weather and climate models by exploiting reduced precision calculations. This includes advice on implementation options (such as libraries or automated code transition) and help to profile and monitor the performance.
  • ESM containerisation
    Support the use of containerised versions of climate models. This includes the adoption of containerised model versions for specific hardware and workflows. 
  • Data compression
    Support to improve the performance of model experiments and data workflow with the help of data compression schemes. This activity builds upon developments in ESiWACE3 WP3, which will use (on-the-fly) compression methods well adapted for weather and climate data to reduce the amount of data movements and storage and improve the performance of simulations and data processing.
  • Domain Specific Languages
    Help improve the portability of climate and weather models using Domain Specific Language (DSL) approaches. This might involve the application of PSyclone and/or GT4Py, two tools that can help to improve performance portability by representing (parts of) the code in a DSL and ease the transition to accelerator hardware. A potential service activity picks up developments of ESiWACE3 WP2.
  • Benchmarking ESMs
    Support could be provided for analysing the performance of models by running benchmarks on current and upcoming HPC systems. This might include help with porting to the benchmark system, running a standardised set of tests, and assessing the results.
  • Kernel Tuner
    KernelTuner is a tool for automatic optimisation of GPU code. Support can be provided to apply this tool to weather and climate models. As a new development, KernelTuner adds support for mixed precision and tuning accuracy, which comes in addition to the existing functionality of tuning parameters and code optimisations for compute performance and/or energy efficiency.
  • Optimised I/O or other infrastructure components in ESMs
    Support for the efficient use of complex HPC architectures about infrastructure components such as I/O operations, MPI communication, and connection of data analysis processes. In addition to analysing current problems in existing ESMs, this includes consulting and prototypical implementation of generic approaches (e.g. libraries for asynchronous I/O) and corresponding interfaces in the codes.
  • Coupling ESMs
    Current ESMs often consist of different model components to simulate the different domains of the Earth system. An efficient data exchange between them must occur via couplers such as YAC or OASIS. The support offered here can, in particular, represent the implementation of coupled models on heterogeneous HPC systems, where, e.g. parts of the ESM run on CPUs. In contrast, others are already ported to GPUs.
  • EuroHPC application support
    Computational resources for model development or climate simulations are provided by, among others, EuroHPC through a competitive application process. Many applications have failed in the past due to shortcomings in the formal technical descriptions of the requested resources. Support could be provided to groups that go through the EuroHPC application process by offering a simple review of the technical aspects of the application before submission.