High-fidelity simulation of turbulent fluid flow is a natural driver for exascale computing, with a virtually unbounded need of computational resources for accurately model problems of industrial and academic relevance. Besides adequate computational resources, numerical methods offering both accuracy and computational efficiency is necessary to carry out these high-fidelity simulations. Adaptive spectral element methods offer a framework with sufficient accuracy, fast convergence as well as large scale parallelism and high computational intensity for efficient utilization of recent computer hardware. A key issue in efficiently exploiting future exascale systems is the amount of parallelism exposed, likely reaching billions of parallel activities. Current spectral element codes scale reasonably well as long as each core has sufficient amount of local work. However, achieving efficient performance at exascale will require either unreasonable large problem sizes, or algorithmic developments need to be achieved which retain parallel scaling even for very few elements per core. The purpose of this project is to address extreme-scale computing challenges, develop novel parallelisation schemes and numerical methods to enable the use of accurate, scalable and efficient computational fluid dynamics simulations for both academic and industrial applications on the next generation of supercomputers.