Applied High Performance Computing (MCEN90031)

The use of physics-based computer simulation is a powerful tool in the scientific and engineering fields that allows for the investigation of phenomena that are often inaccessible by other means. As modern compute architectures continue to increase in terms of parallelism and power, so too can these simulations increase in scale and fidelity, but only when equipped with an understanding of the mathematics and underlying hardware, necessary to leverage this power. This subject will aim to develop such an understanding by tying together key tools and techniques used in the design of scientific software targeted at High Performance Computing (HPC) resources.

This subject will introduce several numerical methods that are ubiquitous in the solution of ordinary differential equations (e.g. Euler and Runge-Kutta methods), partial differential equations (e.g. finite difference and finite element methods), linear systems (e.g. conjugate gradient method), and apply these tools to solve governing equations commonly found in areas such as fluid dynamics and thermodynamics. This subject will investigate the development of software targeting shared memory multicore architectures with OpenMP, distributed memory architectures with MPI, and GPU accelerators with CUDA.