Current and future supercomputers provide a huge computational power but their architectures are more and more complex to counteract the low power increasing of cores. From the application point of view, the main challenge is to use efficiently these supercomputers in order to provide accurate and robust large multiphysics and multiscale numerical simulations. A scientific challenge is to strongly reduce the gap between actual and future supercomputers potential and current HPC methods and applications. Among all the key points to address, the aim of this workshop is to focus on specific topics which occurred in large numerical simulation on new supercomputing architectures as robustness, efficiency and accuracy of algorithms, scalablity and performance optimization, management and analysis of huge datasets of numerical results.