Practical introduction to High Performance Computing
High Performance Computing
High Performance Computing (HPC) is a fundamental technology used in solving scientific and commercial problems. Many of the grand challenges of science depend on simulations and models run on HPC facilities to make progress, for example: protein folding, the search for the Higgs boson and developing nuclear fusion. In industry, sophisticated computer models are integral to the development of products such as jet engines, wind turbines and new drugs.
A selection of Case Studies illustrating the importance of HPC are available on the ARCHER website.
About the course
Modern supercomputers are parallel computers, gaining their power from many thousands of individual processors. Developing software to run on these systems requires using new parallel programming technologies. The course will cover all the fundamental concepts that underpin modern HPC. The course is practical in the sense that you will explore these topic by running parallel programs on real HPC systems such as the UK national supercomputer ARCHER. The same techniques can also be applied to smaller systems such as multi-core desktops, graphics processors and computing clusters.
On completion of this course we would expect students to:
- Understand the key components of HPC architectures and how they are put together to form complete systems.
- Be able to build and test code in a Linux/Unix environment, execute in parallel and understand the stages involved.
- Compare and contrast the message-passing and multithreaded parallel programming models, and describe four key parallelisation strategies used in modern HPC codes.
- Understand the importance of performance and Amdahl's law, and measure and comment on the performance of parallel codes.
- Explain the current uses of HPC and the key issues for HPC in the future.
This is a high-level decription of the curriculum for this course.
Hardware, Architectures and System Software
- overview of computational science and HPC
- HPC components and performance measurement
- multicore CPUs and HPC architectures
- system software and compilers
- batch systems and memory architecture
- HPC networks and data storage
- novel architectures and the future of HPC
- parallel programming concepts
- measuring and quantifying parallel performance
- message-passing and shared-variables approaches
- processes and threads: MPI and OpenMP
- new parallel languages
- using GPUs for parallel programming
- testing and verification
- scientific and commercial uses of HPC
- distributed and cloud computing
Each topic in the first two of the three themes will have an associated practical exercise, usually involving running programs on ARCHER and analysing their behaviour.