| 2014-11 Embrace ARM, Parallel Computing, Modelling |
|
| October 2014 | |
|
Compucon CPD Seminar 12 November 2014 Wednesday Compucon House Albany 4:00 – 4:30pm Embrace ARM (TN) 4:30 – 5:15pm 4D Parallel Computing (TN) 5:15 – 6:00pm Managing Parallel Computing (Stefan Wong) 6:00 – 7:30pm Wine and Cheese Embrace ARM We mentioned ARM a few times in passing in previous seminars. This time ARM is the theme. We will explain the way ARM makes its way up to the horizon from the base of the iceberg, what it is, how it differs to Intel and AMD CPU's, and how it may become a standard for Compucon in the near future. 4D Parallel Computing Sequential computing is one dimensional (1D) as we know of. Computing instructions are executed one by one in a prescribed sequence in hardware real estate. We may say parallel computing is 2D, with portions of the computing taken out to execute in the GPU. GPU also executes instructions sequentially but one single instruction would be executed in multiple threads simultaneously. GPU is 2D in its own right. We can install multiple copies of GPU in a single PC, or execute the same application software on multiple PC's each installed with a GPU. This scenario qualifies for 3D. Can we have 4D? Yes we can and the 4th dimension is time as Albert Einstein led us to understand. This session attempts to examine the current developments in 3D and 4D parallel computing. We may introduce FPGA conceptually for 4D and 5D computing. Managing Parallel Computing The previous presentation by Stefan in June was about his Masters of Engineering project on computer modelling for object tracking and software verification of the model for suitability to run in a hardware device. The first part is called data modelling and the second part is called device modelling. We gave Stefan 30 minutes to rush through 2 years of research work and did not give Stefan any justice. We will give Stefan another 45 minutes this time to explain the two important activities to prepare for parallel computing. Stefan will explain why we need modelling for FPGA at all. An important theme is the definition of "correct" for a certain problem solution and questioning what verification can and cannot answer (20 min) and how to verify what we do is correct (25 min). He will re-use the object tracking problem in June as the example for the explanation, and speak in a layman language this time to assure that we will all understand and enjoy. |