21 Sep 2017 Budapest - European projects PoP and EXA2CT results have been nominated for the Innovation Radar Prize 2017. You can cast your vote until October 15. The IT4Innovations National Supercomputing Center has been nominated with the ESPRESO, an open-source ExaScale PaRallel FETI (a numerical method - finite element tearing and interconnect) Solver developed in the FP7 EXA2CT project. The PoP project has been nominated with "Customer-Specific Performance Analysis for Parallel Code". The Innovation Radar (IR) is a European Commission initiative to identify high potential innovations and innovators in EU-funded research and innovation ICT projects. Using the radar, 47 of the best EU-funded innovators have been identified to compete with their EU-funded innovation for the Innovation Radar Prize 2017.
The public is invited to vote online for their favourite innovators from among these 47 in order to find the top 20, who will then get to compete in the Innovation Radar final, at the ICT Proposers' Day event in Budapest (taking place on 9 November 2017).
The EU-funded POP project is helping developers and users of computationally intensive programmes to accelerate their execution and thus enable them to address bigger problems in less time and at minimum cost. They do that by first providing deep insight on the actual behaviour of current programmes, identification of the fundamental causes of inefficiencies on today's machines and suggestions and support on how to improve the programmes to maximize the use of available resources. They target current parallel computers and their foreseen evolution.
PoP relies on performance analysis tools (fairly mature) developed by the partners over the years (BSC tools, Scalasca). The project focuses on the methodological and best practice aspects of applying these tools and background to real production codes, both from academic and industrial customers. The project has developed a unified set of metrics and analysis practices to maximize the insight provided to customers.
IT4Innovations National Supercomputing Center's ESPRESO solver is a massively parallel sparse linear solver developed for solving engineering problems in structural mechanics. It is designed to be highly scalable so that it can fully utilize today's most powerful petascale supercomputers and solve up to 225 billion unknown problems. By dynamically tuning several key hardware parameters ESPRESO can save over 20% of its energy consumption. It can be also accelerated by both GPU or Intel Xeon Phi accelerators to achieve up-to 5x performance improvements over general purpose CPUs. With all these properties, it has real potential to efficiently utilize future exascale systems where both performance and energy efficiency matter.