WP3. Modelling, computing and simulating
Mathematical modelling and numerical simulation are by now highly important tools currently used in physics,chemistry, biology, medicine and humanities. The forthcoming development of exaflopic computers should allow to simulate with High Performance Computing new multi-scale and multi-physics phenomena, generating massive data (High Performance Data Analytics). One of the major challenges of the years to come is thus both to build new robust numerical modelling tools, working on several millions processors computers and to be able to accommodate on a single platform both HPC and BigData applications.
The work in this work package is devoted into two tasks which are aimed to have strong and permanent interactions.
T3.1: Numerical Simulation
Mejdi AZAIEZ, I2M
T3.1 is focussed on developing new numerical models and methods motivated by new applications, in particular in the context of WP5 on digital ecological systems.
T3.2: Convergence of High Performance Computing and Big Data
Olivier BEAUMONT, Inria
T3.2 is aimed to develop efficient algorithms to tackle the huge amount of data generated in T3.1 and to design a common framework at the architecture, applications and language levels to efficiently execute on HPC platforms applications designed in T3.1.
T3.1 and T3.2 will mutually enrich all over the duration of the project, in particular by transposing some of the sophisticated tools used to optimize simulations towards data analysis,whereas, conversely, data analysis strategies should be transferred to improve the synthesis of data produced by the numerical simulations.
The action of this work package can be viewed as matrix framework. The lines of the matrix are the methodological axis while the rows are the challenging application fields (WP5 and WP6).
The methodological actions concern: New mathematical models for complex phenomena, Numerical schemes for complex phenomena, Algorithms and solvers for high performance computing and heterogeneous architectures, Data assimilation, Reduced Order Methods and propagation of uncertainty.