Scientific Computing
Dates: 14, 21, 28 Oct & 4 Nov 2015
Location:University of Edinburgh and Online
Lecture Live Stream Feed
http://www.media.is.ed.ac.uk/live/capturED/EPCC/EPCC.htmlCourse Hackpad
https://hackpad.com/Scientific-Computing-CY3nTvXX7C2The Hackpad is a live collaborative online document which we will use to share links, information and comments. All course participants are encouraged to contribute.
Lecture Recordings
Video recordings of the whole lecture series will be available on the ARCHER YouTube channel.
Wednesday 14th October 2015
Lecture Slides
Unless otherwise indicated all material is Copyright © EPCC, The University of Edinburgh, and is only made available for private study.
- Introduction to ARCHER training
- Overview of HPC
- Motivation
- HPC Architectures
- Simulation exercise
- Sharpen exercise
The PDF version of the "Overview of HPC" talk is not very useful as it does not reproduce animations or videos. A separate video of this talk is available as a youtube playlist.
Video of Lecture 1
Exercises
Unless otherwise indicated all material is Copyright © EPCC, The University of Edinburgh, and is only made available for private study.
- Exercise sheet for traffic modelling exercise
- Exercise sheet for sharpen exercise
- Quick-start guide to ARCHER (relevant for students attending remotely)
- Notes for running on Eddie rather than ARCHER (relevant for Edinburgh University Informatics attendees)
- Source code for sharpen exercise
Solutions
Unless otherwise indicated all material is Copyright © EPCC, The University of Edinburgh, and is only made available for private study.
- Serial solutions (C, Fortran and Python) to the traffic modelling exercise
- Example results from the Sharpen exercise.
Wednesday 21st October 2015
Lecture Slides
Unless otherwise indicated all material is Copyright © EPCC, The University of Edinburgh, and is only made available for private study.
Video of Lecture 2
Exercises
Unless otherwise indicated all material is Copyright © EPCC, The University of Edinburgh, and is only made available for private study.
- Attempt the parallel parts of the traffic modelling exercise given out in week 1.
- Quantify the performance measurements from the image sharpening exercise in terms of parallel speedup and efficiency, and compare with what you would expect from Amdahls's law.
- Exercise sheet for fractal exercise
- Source code for fractal exercise
Solutions
Unless otherwise indicated all material is Copyright © EPCC, The University of Edinburgh, and is only made available for private study.
- Discussion of parallelisation of traffic model.
- Parallel implementations of the traffic model in OpenMP and MPI (for both C and Fortran).
- Sample results from fractal exercise
Wednesday 28th October 2015
Lecture Slides
Unless otherwise indicated all material is Copyright © EPCC, The University of Edinburgh, and is only made available for private study.
Video of Lecture 3
Exercises
Unless otherwise indicated all material is Copyright © EPCC, The University of Edinburgh, and is only made available for private study.
- Investigate parallel scaling of traffic model solutions (see Week 2) for various problem sizes; compare to Amdahl's Law and Gustafson's Law.
- Simple C code illustrating precision issues
- Exercise sheet for orbits exercise
- Source code for orbits exercise
Solutions
Unless otherwise indicated all material is Copyright © EPCC, The University of Edinburgh, and is only made available for private study.
- Sample results from Orbits exercise.
- Sample performance scaling results from parallel traffic model.
Wednesday 4th November 2015
Lecture Slides
Unless otherwise indicated all material is Copyright © EPCC, The University of Edinburgh, and is only made available for private study.
Exercises
Unless otherwise indicated all material is Copyright © EPCC, The University of Edinburgh, and is only made available for private study.
Additional Slides
These slides are a brief introduction to how to discretise simple PDEs onto a regular grid, and how to solve the resulting linear equations using simple methods. They were not included in the lecture series due to lack of time, so do not appear in any of the videos, but I hope they are useful.
Unless otherwise indicated all material is Copyright © EPCC, The University of Edinburgh, and is only made available for private study.