Research Unit Leader:
Objective
The objective of this research unit is (I) the optimization and the parallelization of the numerical models for climate change simulations (both climate and impacts models); (II) the evaluation of how new emergent technologies impact on the implementation and the design of the current climate models; (III) the innovation of the main computational kernels by means of a re-design at numerical, algorithmic and software level.
Activities
- Models optimization and parallelization
This activity supports all of the CMCC divisions and aims at the use of “co-design” techniques for the optimization of the climate models used at CMCC. Moreover the activity focuses on the analysis of the main computational kernels, which characterize the models. The aim is the kernels optimization on “multi-cores” hybrid architectures taking into consideration the performance impacts of optimized compilers, numerical libraries and new parallel paradigms.
- Re-design of computational kernels towards Exascale
This activity focuses on the study of the impact of exascale computing architectures on the numerical algorithms used in the main climate models studied at CMCC. In particular, the following aspects are analyzed from exascale point of view: (I) the optimization of Earth System coupled Models; (II) the use of advanced parallel algorithmic structures to reduce the current “dynamical cores” communication overhead; (III) the optimal management of “multi-model” and “multi-emission” ensemble experiments; (IV) the analysis of the “super-parameterization” approaches used to solve the CRM (Cloud Resolving Models) at highest resolution; (V) the optimal management of the memory system and its hierarchies; (VI) the rationalization of I/O operations; (VII) the fault-tolerance management; (VIII) the adoption of new communication parallel paradigms and parallel tasks synchronization mechanisms.
This activity is carried out under the international initiatives IESP (International Exascale Software Programme) and EESI (European Exascale Software Initiative).