Towards petascaling of the NEMO ocean model

This poster describes different efforts to improve the scalability of the NEMO ocean model. Firstly, advanced parameters of the MPI-communication library were optimized. Secondly, NEMO uses a tri-polar and staggered grid, which involves a complicated fold across the northpole. Communication along this fold involves collective gather and scatter operations which create a bottleneck at a single core, so these were replaced by a small series of send-receive messages.