The Art of Scientific Computation (COMP90072)

The physical, social and engineering sciences make widespread use of numerical simulations and graphical representations that link underlying their theoretical foundations with experimental or empirical data. These approaches are routinely designed and conducted by researchers with little or no formal training in computation, assembling instead the necessary skills from a variety of sources. There is an art to assembling computational tools that both achieve their goals and make good effective use of the available computational resources.

This subject introduces students to a wide range of skills that are commonly encountered in the design and construction of computational tools in research applications:

• Formulation of the task as a sequence of operations or procedures that express the context of the assigned problem in a form accessible to digital computing (Mathematica).
• Implementation of this formulation using computer languages appropriate for numerically intensive computation (C, C++, Fortran)
• Modularization of computationally intensive tasks, either as user-written procedures or existing libraries (for example BLAS, lapack)
• Documentation of the code to explain both its design, operation and limitations (LaTeX)
• Instrumentation of the code to verify its correct operation and monitor its performance (gprof)
• Optimization of the code, including the use of parallelization (OpenMPI)
• Visualization of data using graphical packages or rendering engines (Geomview, OpenGL)
• Interaction with the code through a graphical user interface (Python, Matlab)

These skills are introduced to the student by undertaking a short project that is selected in consultation with the Subject Coordinator.