On Thursday 14th January I attended the inaugural talk by Professor Mark Parsons, ‘“A billion billion calculations in a second – a step too far?”. Prof. Parsons is Director of the Edinburgh Parallel Computing Centre which hosts the UK’s national supercomputer, ARCHER.
As an experimentalist, parallel computing has never really been my thing although I collaborate with computer simulators who run intensive simulations in this way. However, it was highly interesting to hear how the world’s superpowers are competing to create faster and faster computers, with the Chinese winning with Tianhe-2, currently. Ironically, Prof. Parsons then pointed out that current software techniques cannot cope and such machines cannot be efficiently programmed. This is because we are still using linear mathematical procedures like Fast Fourier Transforms and Finite Element Methods – to advance our mathematicians need to help find new mathematical techniques suitable for efficient parallel programming.
For those of you interested in the weather, apparently the ECMWF weather model is the world’s best for weather prediction and more reliable for predicting hurricane routes!