Speaker: Prof Jonathan Hirst, University of Nottingham

Title: Molecular Simulation: Exploiting New Architectures and New Ideas

Abstract:The molecular simulation community is embracing GPGPUs. In the context of biomolecular simulations, we review some of the algorithms that have been implemented and some of the aspects that distinguish the GPU from previous parallel environments. However, new ideas as well as new hardware are key to faster simulations. In the second part of the talk, we present a new approach to computing fundamental quantities in statistical thermodynamics, which makes much more efficient use of sampled configurations.

CSC Coffee and Cookies are held at 14:00 hours in the Computer Science Common Room, 1st Floor, Computer Science Building.

Please come along, All are welcome

Speaker: Dr Eike Mueller, Mathematics, Bath

Title: Scalability of Elliptic Solvers in Numberical Weather-and Climate-Prediction

Abstract:

Numerical weather- and climate- prediction requires the solution of elliptic partial differential equations with a large number of unknows on a spherical grid. If implicit time stepping is used in the dynamical core of the forecast model, an elliptic PDE for the pressure correction has to be solved at each time step. This often amounts to a significant proportion of the model runtime. The goal of the Next Generation Weather and Climate Prediction (NGWCP) project is the development of a new dynamical core for the UK Met Office Unified Model with a significantly increased global model resolution, resulting in more than 10^{10} degrees of freedom for each atmospheric variable. To run the model operationally, the solver has to scale to hundreds of thousands of processor cores on modern computer architectures.

We carried out first tests on a model equation and demonstrate that multigrid solvers allow us to keep the total solution time fixed as the horizontal resolution increases. The large ratio between the horizontal and vertical extent of the atmosphere leads to a strong vertical anisotropy which has to be dealt with by using a suitable smoother, such as vertical line relaxation.

For this we tested existing solvers in the Distributed and Unified Numerics Environment (DUNE) and have demonstrated the scalability of the existing algebraic multigrid solver on up to 65536 cores of the Hector supercomputer for a system with 10^{10} degrees of freedom for the elliptic PDE arising from semi-implicit semi-Lagrangian time stepping.

As algebraic multigrid suffers from large setup costs, we also implemented a matrix-free parallel geometric multigrid code with a vertical line smoother and demonstrate its performance and scalability.