Large offshore engineering projects – such as designing and manufacturing coastal flood defences, ships or oil-rigs that can withstand the pressure from vast bodies of rapidly-moving water and the impact of crashing waves – are some of the more unusual items on the agenda at a computing conference being held this month in Paris.
The work being done by the STFC Hartree Centre and Scientific Computing Department will be showcased by STFC’s Dr Xiaohu Guo and Sergi Siso, along with collaborators from China, Italy, Spain, the UK and the USA, at a mini-symposium at the Society for Industrial and Applied Mathematics (SIAM) Conference on Parallel Processing for Scientific Computing 2016.
Together Xiaohu and Sergi will showcase the departments’ expertise of using computer simulations to efficiently solve these major offshore engineering problems.
Dr Xiaohu Guo
Dr Guo says, "The increasing scale and complexity of current and future computing systems is putting pressure on numerical algorithms and the multiple software systems we use. In other words, it can cause problems in running the software efficiently, or completely fail to run them, so we need to find ways to balance this out.
"We’re holding this symposium to discuss the challenges posed by different application software and computing architectures, and find some possible solutions. Hopefully we will also forge new international collaborations at the end of the workshop."
"For these engineering projects we work with colleagues at the University of Manchester and the University of Vigo in Spain to simulate the impact large numbers of water particles in waves have on solid structures," says Sergi Siso. "They focus more on the physics in the simulations, while we focus on improving the performance. We use DualSPHysics, a code which the Manchester team developed with other partners. We run this code on the Intel Xeon Phi high performance processor, to simulate wave interaction with solid structures."
Programming for the Intel Xeon Phi is carried out at the Hartree Centre and is funded by Intel’s Parallel Computing Centre programme. The processor has 60 cores (processing units) which can run hundreds of threads (small sequences of programmed instructions) all at the same time. The performance challenges of other computing platforms, including nVidia GPU, Intel CPU, IBM Power8 and possibly even a Chinese self-made processor, will also be discussed.
Xiaohu and Sergi will be presenting at the SIAM Conference (link opens in a new window) on 14 April. See the conference website for details of their morning (link opens in a new window) and afternoon (link opens in a new window) symposium sessions. For an idea of how the DualSPHysics simulations work, check out these animations (link opens in a new window).