½-1 day aimed at participants with little or no exposure to running simulations or analyses in parallel. Parallelism allows faster return or results, or to scale up to bigger problems and higher resolutions. The course comprises an online pre-course element covering basic terminology of programming and parallelism and a face to face session focussed on exploring (without any coding) what it means to “go parallel” and to consider how to apply to their own research. Exercises may use of jigsaws and coloured balls. Having grasped the advantages, and potential limitations, as to what can be accelerated by use of more resources, we cover at a high level some theory of parallel programming. Optionally, for those with some programming experience (or wishing to gain some!) we give examples for participants to run on real life parallel computers to illustrate the topics covered. Skills gained in this course will be transferable to analysis of model simulations in future careers.
Related Training
Also see our main training page
- introduction to thinking parallel
- getting started with FORTRAN
- designing and writing efficient programs
- the art of compilation (and how to use compilers to tell you what they're not doing but could be with your hints)
- introduction to using OpenMP to make the most of your multicore desktop
- introduction to MPI for multi-node processing (and for Xeon Phi)
- using profilers to improve the efficiency of your parallel code
- introduction to measuring the energy consumed by your model simulation for #greenerCompute
- batch schedules and running "high throughput" eg Monte Carlo
- Using MS Azure for high end compute
- use of directives for programming GPUs and Xeon Phi
- CUDA, OpenAcc and OpenMP 4.5