Traditionally, astronomy has been almost purely an observational science. As opposed to most other physical disciplines, it's impossible to go into the laboratory and perform experiments.
Over the last 40 years, however, the increasing power of computers and the development of sophisticated numerical methods have allowed numerical simulations to emerge as one of the key components of modern astrophysics. It is now possible to run an ``experiment" on a computer and ``see" what happens when galaxies collide, or watch the development of a proto-planetary system around a young star, or follow what happens to gas as it falls into a black hole. Powerful computers are also essential for analyzing and interpreting modern observational data
There are two broad classes of numerical simulations which are being pursued at CITA. Hydrodynamic codes are used to simulate astrophysical gas flows while N-body codes are used to follow the evolution of systems with millions (even billions) of gravitationally-interacting particles. Thus, a hydro code might be used to study gas as it flows into a supermassive black hole (a process ocurring at the center of our Galaxy) while an N-body code could be used to study the dynamics of a galaxy (which can contain many billions of stars). It is also possible to combine these two types of codes and simultaneously follow the evolution of gas and particles in a system.
The ever-increasing desire for more CPU power and memory means that astrophysicists can benefit tremendously from developing efficient numerical codes which make the best use of computational resources. There's no point in having the fastest computer if your numerical codes run more slowly than those on a competitor's machine! Thus, much effort has gone into developing efficient and accurate codes that have low memory overhead and are optimized for performance.