|2015-02 Square Thinking, OpenCL, Proof, FFT|
Compucon CPD Seminar
4 February 2015 Wednesday
Compucon House Albany
4:00 - 4:30pm Thinking inside & outside the Square (TN)
4:30 - 5:00pm Why is Compucon placing the bet on OpenCL (TN)
5:00 - 5:30pm Proof of Parallel Computing Concept (Xu He)
5:30 - 6:00pm Fast Fourier Transform (Stefan Wong)
6:00 – 7:30pm Wine and Cheese
Thinking inside & outside the square
Thinking is one property that separates human from other living members of the world and achievers from non-achievers. We measure before we cut. We think before we act. There are 6 different types of serious thinking that the speaker can think of: expressive, perspective, critical, logical, creative, and inquisitive. This square session is meant for sharing the experience of the art of thinking. It will be given in an easy-to-relate manner, and illustrated with well accepted examples and some shallow humour. Peers are encouraged to suggest more types of serious thinking for discussion. Mind mapping and Bono 6-hats have been included already. Neural computing is outside the scope.
Betting on OpenCL
The high performance computing citizen scientist web forum www.compucon.co.nz/forum/ has a topic entitled “What is OpenCL”? It is an attempt by Compucon to raise the awareness of citizen scientists that OpenCL is a good candidate for implementing parallel computing. At the same time, Compucon has invested in new technology samples that claim to do parallel computing with OpenCL. The investments are not going to be light. Why Compucon bets on OpenCL? Is OpenCL real or a fantasy? Is it mature or in the work? Are there other options? Is it the best tool? Who else are betting on it? What is the chance of success of this bet? How to enhance or realise the chance of success? This session will employ the 6 thinking tools to find the answers and give us a clear and complete picture. Peers will get a very encouraging piece of info that our Compucon systems will benefit from this research very soon!
Proof of Parallel Computing Concept
This project is funded by MBIE of the Government for a university graduate student. Xu He is the student. He has completed a BTECH project in the University of Auckland in 2014 which was co-supervised by Professor Reinhard Klette and Compucon. The BTECH project was for developing the algorithms needed for detecting a person with a hands-up pose recorded by a surveillance camera. Xu succeeded and his project paper was subsequently accepted by the Image & Vision Conference held in November 2014 in the University of Waikato (IVCNZ). This MBIE-funded project is an extension of the BTECH project. The original application was written in C and C++ for running in Windows and it took 850ms of runtime. Xu has been able to refine the programming code and reduce the runtime to 157ms by 2015-01-16 in the same PC but with a CUDA card. Xu is now working on the same problem on an ARM based system-on-chip instead of a Compucon Diamond+ PC. He will explain how he achieved the speed up from 1 to 6fps in a PC and his progress on ARM SOC.
Fast Fourier Transform
Stefan is a core member of Compucon II development team as well as a Masters of Engineering student from AUT. Fourier Transform is a popular and essential algorithm in the computing world. In brief, it transforms a time series of signals into a frequency series, and vice versa. One of our hardware being explored is capable of cleaning a small Jpeg picture (128 x 128 pixels) in less than 1ms whilst an ARM CPU took nearly 10ms to do the same job. Both cleaning jobs employed Fourier Transform and Stefan will explain how Fourier Transform helps. The most popular source code is called FFTW where the first F stands for Fast and the last W stands for West being USA. FFTW is a clever chap. It can think- it measures before it cuts. Come and enjoy this high tech story written for Compucon peers.