"The grant from Intel opens a fantastic opportunity for us to focus effort on optimization and parallelization of codes that are key to the users of CSC's present and future supercomputer systems", stated Per Öster, Director of Research Infrastructures and leader of the project, as he also stresses the importance to prepare for the shift in processor architectures that will take place in the next few years.
Professor Markku Kulmala, Helsinki University, research leader of the Finnish Centre of Excellence in Atmospheric Science is also very enthusiastic. "We are now coupling models of processes in the atmosphere and biosphere at all scales in time and space and with different physics, chemistry, meteorology, and biology to better understand the Earth's climate. This results in extreme demands on computing capacity and optimal use of the latest computing technologies is critical for our success."
"CSC and the Intel Parallel Computing Center program combine to create an outstanding resource for the high-performance community", stated Joe Curley, Director of Marketing in Intel's Technical Computing Group. "By the end of this decade, powerful and highly parallel machines will evolve to help us find solutions to the most important problems in science, industry, and government. CSC's expertise in application, computing, and data services uniquely positions this collaboration to enable modern, parallel applications that convert problems to insight."
In addition to atmospheric research, CSC will work with a set of open source applications codes from environmental and materials sciences. The selected applications are widely used both in national and international level, and the optimization work will benefit large group of researchers. The initial list of applications to be optimized for large-scale simulations is:
Parallelization and optimization of the codes will be carried out by the CSC HPC team using an extremely energy-efficient liquid-cooled Bull B715 supercomputer prototype and CSC's other supercomputer resources. The optimizations will target both the latest generations of Intel Xeon processors as well as Intel Xeon Phi coprocessors.
This modernizing of key applications is expected to enable large performance increases, and will help to make the next leap in discovery. "Validating current approximations and obtaining new and more detailed insight into the mechanisms of atmospheric new-particle formation is computationally very demanding - this can be only achieved by very efficient parallelization", stated researcher Ville Loukonen from Professor Hanna Vehkamäki's Computational aerosol physics group from the University of Helsinki.
Training researchers and engineers on modernizing their code is also necessary. CSC will carry out comprehensive training to improve the HPC programming and optimization skills of the research groups.