We use cookies to ensure that we give you the best experience on our website. You can change your cookie settings at any time. Otherwise, we'll assume you're OK to continue.

Durham University

Advanced Research Computing

Intel® Parallel Computing Center

Durham University’s Institute for Computational Cosmology, together with the Institute of Advanced Research Computing, becomes an Intel® Parallel Computing Center.

The Institute for Computational Cosmology (ICC), part of Durham University, is world renowned for its research on how galaxies form. Researchers perform large cosmological simulations that follow how stars and galaxies grow from tiny imperfections in the big bang. The goal of the institute is to generate a virtual universe, creating such realistic galaxies that it is almost impossible to distinguish them from systems in the real world.

This research has now been given a great boost with the establishment of the ICC and iARC as an Intel® Parallel Computing Center (Intel® PCC). The current computer calculations take months to run on the DiRAC data centric computer, hosted by the ICC. Such long run times, even on a computer as powerful as DiRAC, limit the ability of researchers to experiment with the physical laws that determine the appearance of galaxies. In order to advance the simulations further, generating galaxies in more detail and capturing even greater expanses of the observable Universe, a leap forward in the efficiency of the computer code is required. Researchers at the ICC will work in close collaboration with Intel to develop a new simulation code, called Swift, which will aid in revolutionizing simulations of the Universe.

Swift uses fine-grained task-based parallelism to exploit efficiently the thousands of processors on current computers. Through the Intel collaboration, this code will be optimized to work with the latest generation of Intel® processors, including the Intel® Xeon Phi™ coprocessor.

“We are very excited about this”, commented Durham University’s Prof. Richard Bower, “In the EAGLE project, we have been making great strides in understanding the physics that fills the Universe with stars and galaxies. At present the detail we can achieve is limited by the long-run time of the calculation. To do better, we need to more effectively harness the power of massively parallel computing systems.”

Robert Maskell, Intel’s Director of High Performance Computing, added: “To meet the processing demands of the next generation of simulations, the computer code will need to make use of millions, rather than thousands, of cores. This requires a new approach to the computer code. The collaboration between Intel and the ICC researchers, of Durham University, will foster an opportunity to demonstrate a new approach to task parallelism through optimizing the SWIFT code using Intel® Xeon Phi™ coprocessor. We anticipate this will allow for the work being done by each processor to be established dynamically. Through the building of a tree of task dependencies and conflicts, it is possible to avoid the traditional bottle-neck where one compute node must wait to exchange calculation results with another. This technique has applications that go well beyond computer simulation of the cosmos.”

Further details about the EAGLE simulation project can be found at: