 At
the start of 2007, BiLCEM (Computational Electromagnetics Research Center of Bilkent
University) Director Prof. Levent Gürel announced that they had solved the world's
largest problem in computational electromagnetics involving 42 million unknowns. In less
than one year, BiLCEM researchers doubled the world record by solving huge mathematical
problems involving as many as 85 million unknowns. This achievement paves the way for
solutions to a vast variety of physical problems that were previously unsolvable. Many
disciplines, such as medical imaging, nanotechnology, remote sensing, and optics, can
benefit. For example, BiLCEM researchers have been able to study the electromagnetic
properties of the world's largest metamaterial structures. Metamaterials are artificial
materials that exhibit very interesting electromagnetic properties. They have potential
applications for invisibility shields and sub-wavelength focusing. Prominent institutions,
such as SSM, ASELSAN, TÜBİTAK, and TÜBA, have been supporting BiLCEM's research.
"Our next goal is to solve 100 million by 100 million (100,000,000 x 100,000,000)
matrix equations." declared Özgür Ergül, a senior PhD student at BiLCEM. Computing
facilities of BiLCEM include a number of parallel computers constructed at Bilkent
University. The 64-core cluster is the main workhorse, however, Tahir Malas, another PhD
student at BiLCEM, was happy to announce that "We are currently in the process of
constructing a new parallel computer with 128 cores and 512 GB of memory." Although
BiLCEM researchers work with such large parallel supercomputers, their hardware is not
even close to the top 500 most powerful computers (www.top500.org).
The secret of solving the world's largest integral-equation problems on relatively modest
computing resources is explained by Prof. Gürel: "What enables us to break the world
records are the mathematical and electromagnetic methods we have been developing. For
about 20 years, we have been focusing our intense studies on the advancement of fast
algorithms and parallel computing for the solution of large problems."
Prof. Gürel invites graduate and undergraduate students to get involved in BiLCEM
to help contribute to projects, and gain valuable experience. For more information, visit www.cem.bilkent.edu.tr.
|
