In this talk, we present a new open-source cosmological code, called SWIFT, designed to solve the equations of gravity and hydrodynamics using the SPH approach on hybrid shared/distributed clusters. The code relies on three main aspects to make efficient use of current architectures:

- Task-based parallelism to exploit shared-memory parallelism. This provides fine-grained load balancing enabling strong scaling, combined with mixing communication and computation, both on each node with multiple cores as well as on external devices (accelerators, GPUs).

- Hybrid shared/distributed memory parallelism, using the task-based schemes. Parts of the computation are scheduled only once the asynchronous transfers of the required data have completed. Communication latencies are thus hidden by computation, providing for strong scaling across multi-core nodes.

- Graph-based domain decomposition, which uses information from the task graph to decompose the simulation domain such that the work, as opposed to just the data, as in other space-filling curve schemes, is equally distributed
amongst all nodes.

These three main aspects alongside improved cache-efficient algorithms for neighbour finding allow the code to be 40x faster than the standard code Gadget-2 widely used by researchers on the same architecture. Our code also displays a strong scaling parallel efficiency of more than 60% when
running on 1024 cores.

Multiple variants of SPH, SPH-ALE and Voronoi hydrodynamics are present in the
code as examples.

Details of the project: http://www.swiftsim.com
Some early results present in arXiv:1404.2303