| layout | title | permalink |
|---|---|---|
page |
About |
/about/ |
Escalating demand for High Performance Computing (HPC) is driven by the need to solve complex simulations on increasingly large scales that cannot be approached using other methods. Modelling and simulation of complex problems has become an established ‘third pillar’ of science, complementary to theory and experimentation. In order to deliver the required computational research support and advancement on an international level and scale, an EPSRC appointed panel of experts (on HPC) suggests
“Many areas of research in the future will be only accessible to those with advanced computational technology and platforms” [1].
In the last ten years, there has been an increased drive toward the use of many-core architectures which exploit larger amounts of data parallelism. The term "accelerated computing" is a model used in scientific and engineering applications, where specialised many core architectures are applied to "accelerate" computational performance. Traditionally, accelerated computing is has been carried out by co-processors such as GPUs however modern Xeon Phi processors integrate data parallelism and highly vectorised compute units within more traditional looking CPUs to accelerate computational performance.
Currently 50% of the top ten supercomputers are reliant on accelerated computing. In 2017, the Titan supercomputer (currently ranked 2nd in the world) will be superseded by Summit due, in part, to its higher GPU capacity [2]. When considering energy performance, accelerators demonstrate substantially higher efficiency than traditional CPU processing cores due to their design, which commits higher die space to raw computing. The top 22 of the Green 500 (highest ranked energy efficient supercomputers) machines are reliant on accelerators. Given the rising energy costs associated with HPC, it is difficult to imagine a future where accelerators will not play a critical role.
The acceleratedcomputing.ac.uk website (and gpucomputing.ac.uk) is a community driven website and blog for the UK accelerated computing community. The site provides links to research groups, teaching and training opportunities and software as well as a community contributed blog. The website is hosted on Github and blog contributions and updates are welcome via pull requests. Blog posts should feature interesting technical information, benchmarking, software and workshop announcements and training opportunities. Insights from Research Software Engineers (RSEs) working on accelerated computing are particularly welcome.
[1] D. Willet, “Eight Great Technologies”, Technical Report, Policy Exchange, 2013.
[2] NVIDIA, “Summit and Sierra Supercomputers: An Inside Look at the U.S. Department of Energy's New Pre-Exascale Systems”, Whitepaper, 2014.