Performance optimisation and application scaling activities at EPCC
Posted: 21 Jun 2021 | 10:32
EPCC has a history of working with some of the most powerful supercomputers in the world. As such, we have decades of experience optimising and fine-tuning applications for the best performance. One of our key areas of research is HPC software optimisation to achieve maximum efficiency with the aim of delivering scientific results quickly and accurately. Here are some highlights from our current work in this area.
EPCC is a partner in the EXCELLERAT project, a single point of access for expertise on how data management, data analytics, visualisation, simulation-driven design and Co-design with high-performance computing (HPC) can benefit engineering, especially in the aeronautics, automotive, energy and manufacturing sectors. https://www.excellerat.eu
2:45-3:00pm, June 29
EXCELLERAT presentation: “Exascale for Industry” by Mark Sawyer, EPCC.
Abstract: The presentation will look at the ambitions for industrialists using Exascale computing, drawing on examples from Excellerat and other initiatives and outlining some of the technical challenges that we experience and foresee. In addition I will comment on whether Exsascale can be a realistic opportunity for SMEs.
The ASiMoV Strategic Prosperity Parntership is jointly led by EPCC, the supercomputing centre at the University of Edinburgh, and Rolls-Royce PLC, in collaboration with the Universities of Bristol, Cambridge, Oxford and Warwick. ASiMoV uniquely combines Computational and Physical Science research to address the fundamental scaling and modelling challenges faced by the growing need to run ever larger and more accurate simulations.
Our 5-year goal is the high-fidelity simulation of a complete gas-turbine engine during operation, simultaneously including thermo-mechanics, electromagnetics, and computational fluid dynamics. The partnership has the tagline: trillion cell simulations running on millions of computing cores. Crucially, ASiMoV integrates the multi-facetted challenges of large-scale industrial simulations with fundamental research into areas such as extreme scaling, physical modelling, cost of computing, security of data and trust in the simulation results.
The EuroCC network of national centres of excellence will increase the accessibility and availability of HPC, HPDA and AI, and provide a single “front door” into HPC research and development capabilities for industrial customers and academic researchers alike. The UK centre will be delivered jointly by EPCC and the Hartree Centre.
EPiGRAM-HS: Enabling extreme scale applications on heterogeneous hardware.
While supercomputers are moving towards Exascale and steadily becoming more powerful every year, they are also becoming more difficult to program. EPiGRAM-HS is a European Commission Funded project with the goal of designing and delivering a programming environment for Exascale heterogeneous systems in order to support the execution of large scale applications.
VESTEC is a European funded project that builds a flexible toolchain to combine multiple data sources, efficiently extract essential features, enable flexible scheduling and interactive supercomputing, and realise 3D visualization environments for interactive explorations by stakeholders and decision makers. VESTEC will develop and evaluate methods and interfaces to integrate high-performance data analytics processes into running simulations and real-time data environments. Interactive ensemble management will launch new simulations for new data, building up statistically more and more accurate pictures of emerging, time-critical phenomena. Innovative data compression approaches, based on topological feature extraction and data sampling, will result in considerable reductions in storage and processing demands by discarding domain-irrelevant data.
Extreme-scale precision imaging in radio astronomy
EPCC is collaborating with Prof. Yves Wiaux (Heriot-Watt University) to advance algorithms for high-precision and high-sensitivity computational imaging.
The EIRA (Extreme-Scale Precision Imaging in Radio Astronomy) collaboration will focus on radio astronomy, which uses radio telescopes to collect data. This allows observation of the sky with antennae arrays at otherwise inaccessible angular resolutions and sensitivities. Algorithms being developed at Heriot-Watt University will address the challenges of building images from these incomplete linear data sets.