IT4Innovations National Supercomputing Center

  • Ivica Cubic profile
    Ivica Cubic
    18 September 2017 - updated 3 years ago
Voting is closed. 213 users have voted.
Country: 
Czech Republic

The 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.

About the Innovator

/futurium/en/file/espresoteamresizedjpgespreso_team_resized.jpg

ESPRESO team

Our research team was established on the 1st of September, 2011 along with other teams and research programs at IT4Innovations National Supercomputing Center in the Czech Republic. Our work is focused on development of scalable numerical algorithms. These novel algorithms have been successfully used to solve challenging engineering problems (e.g. heat transfer, optimization in structural mechanic, air and water pollution transport etc.). We have also the advantage of long-term collaboration with renowned domestic and international research centers and industrial partners.

Twitter: https://twitter.com/it4innovations

What is the innovation

ESPRESO, an open-source ExaScale PaRallel FETI (a numerical method - finite element tearing and interconnect) Solver for problems of engineering mechanics developed at IT4Innovations center within the FP7 EXA2CT project. ESPRESO contains several FETI based domain decomposition algorithms including Hybrid Total FETI method suitable for parallel machines with tens or hundreds of thousands of cores. The solver is based on highly efficient communication layer on top of pure MPI. Its scalability has been tested up to 18,000 compute nodes of the ORNL Titan supercomputer. ESPRESO supports GPUs and MICs acceleration.

Out of the lab – Into the Market

The important mission of our team is collaboration with external research institutes and industrial partners coming from automotive, aerospace, and energy area. We capitalize our technology by contractual and applied research. As a supercomputing centre, IT4Innovations frequently organizes workshops and tutorials for both academic and industry users of our supercomputers. The ESPRESO toolbox, its potential and results are frequently presented on these workshops. Tutorial sessions are used to teach users how to use project framework for increasing the partners’ knowledge.

Benefits of participation in the Framework Programme

  • Deeper understanding of numerical simulations of complex extreme scale engineering problems.
  • Expand our expertise in the development of massively parallel algorithms based on domain decomposition techniques.
  • Using the EXA2CT framework to produce new tools for industry partners to increase the simulation capabilities with extreme scale computation and in-situ data processing.
  • By raising the awareness of EXA2CT potential among researchers, students and industrial partners of IT4I, we aim to increase impact of newly developed optimization methods.