Europe and Russia are to work together towards exascale supercomputing for science and industry. A two-year project, called APOS, will address the programmability challenges of HPC applications in strategically important areas including energy prospecting, seismic modelling, CFD and nanotechnology.
Faster computers do not immediately lead to better results. Incompatibility between existing software and the capabilities of new supercomputers is a growing problem that the new Russian-European collaboration will address.
'A key consideration for exascale computers is programmability. The APOS team is working with some of today's most important HPC applications, developing the tools and techniques required to help them realise the benefits of future-generation supercomputers,' said George Beckett, APOS project manager at EPCC, University of Edinburgh.
Processor manufacturers are squeezing ever more computational cores onto their CPUs, and HPC vendors are augmenting these many-core chips with GPU-like accelerators. While such developments give scientists the potential to run ever bigger, faster, or more detailed simulations, the combination of massive parallelism and heterogeneity presents significant challenges.
Funded by the European Union's 7th Framework Programme and the Ministry of Education and Science of the Russian Federation, APOS will target popular codes from the strategically important application areas of seismic modelling, oil- and gas-reservoir simulation, computational fluid dynamics, fusion energy, and nanotechnology. It will pool expertise from computational science, high-performance computing, and software engineering to promote next-generation science on next-generation supercomputers.
APOS will couple research centres in Europe and Russia, along with the petrochemical company Total, and CAPS, an innovative European HPC tools producer.
'Improving seismic image quality to find oil and gas deposits thousands of metres below the seafloor requires ever increasing computing power. Simulating oil and gas reservoir behaviour with various physics and detailed meshes also poses serious challenges in terms of programming methods that should match future architectures. We expect significant benefit from joint research with academia,' according to Pascal Dauboin of Total.
'Collaboration in the framework of APOS allows Russian scientists to adapt their numerical methods to modern supercomputer systems using the experience and innovative tools of European partners, and to carry out joint theoretical investigations on the basis of developed scientific software tools, to be used in practically important application areas,' said Boris Chetverushkin, Keldysh Institute of Applied Mathematics, Russian Academy of Sciences.
