Sunday, June 9, 2013

CERN Computing School at Split, Croatia

About three years ago, I went to the Computing school co-organized by CERN and Brunel University at Uxbridge, London and enjoyed it a lot. This year I had the chance again to participate in the smaller version of it - thematic school, which means instead of focusing on network, computing and data analysis, it is now focusing on one single topic, which is computing technologies for this year. The school director is a very wonderful guy and his passion has bring the summer school to a higher dimension - study hard, play hard and social hard! I have been to many summer schools and have never experienced something like this. And this is going to be his last year organising such schools and he is retiring by the end of this year. :(

Port of Split, Croatia
The venue of the school was at Split in Croatia, and is a tremendously scenic city alongside with Dubrovnik. The city lies along the coast of Adriatic sea and within the region of Dalmatian. It has a very rich history, from Roman period until Communist period just 20 years ago. As they have became part of European Union (EU) since 1st of July, the country is developing fast and the standard of living will be improved over the coming years. The living cost now is just slightly higher than in Malaysia so act faster if you wanna enjoy this country with less $$.

Back to the school, we spent the whole week learning about the front end of computing technologies. Basically our situation now is different from like 15 years ago where every few months you will see a new processor with higher and higher frequency. The era of parallel computing is evolving very fast, not only CPU but GPU and accelerator are heavily involved in delivering high performance computing. Of course, without the software part, hardware is nothing. Currently there are several ways of parallelise your program such as OpenMP and Intel TBB. The former parallelise your program at compiler level whereas the latter is library-based. OpenMP is currently the standard and I have started to implement it in my simulation and analysis programs. It's pretty simple.

We had also other lectures about comparison of different compilers, awareness about memory when programming something, and data oriented design. Overall I have benefited a lot from it and got to know more people in the physics computing field. I am also very fascinated by the development of the Intel Xeon Phi coprocessor, which has 61 cores and 244 threads. This was unimaginable for me a few years back.

Closing dinner of the school (my table)

No comments: