Balancing resolution and response in computational steering with simulation trails

Computational steering provides many opportunities to gain additional insight into a numerical simulation, for example by facilitating what-if experimentation, detection of unstable situations and termination of uninteresting runs. When performing steering, it is important that steering changes are quickly reflected in the state of the simulation, so that cause and effect are clearly linked. However, this places constraints on the simulation: it must produce data quickly. The resolution of the simulation is often reduced to allow this. These two competing requirements of response and resolution must be balanced in a usable steering system. This paper proposes a technique, simulation trails, that addresses this issue of balance for simulations where the transient solutions are as important as the final state, and applies it to a simulation using the Smoothed Particle Hydrodynamics method from the domain of astrophysics.