Lorena Barba, an Assistant Professor of Mechanical Engineering at Boston University, has received an NVIDIA Academic Partnership award for her research using GPUs to speed-up simulations of complex fluid flow problems.
The Academic Partnership Program (formerly known as Professor Partnerships) was established by NVIDIA in 2006 and has since then awarded a total of 70 scientists with financial support for their research and/or equipment donations. These awards are very competitive; NVIDIA receives hundreds of proposals and awards only a few grants each year.
The use of GPU computing in scientific applications has exploded in the last two or three years, as researchers find that they can speed-up their simulations by at least 10 times, often more. GPUs are a modern hardware for which computer programs are written in an extended version of the C-language. They evolved from video cards developed over the years to satisfy the billion-dollar gaming market. But today, they are being used to crunch numbers for finance, to simulate proteins using molecular dynamics, to compute magnetic fields, fluid flow, seismic waves and much more.
Only last year, a GPU-based supercomputer allowed China to take the top place in the international supercomputing rankings, claiming #1 in the Top500 list. In the latest ranking of the world’s computers, three of the top-five systems are using GPUs, and thus no one doubts any more that this technology is a main player in the future of computing.
Prof. Barba started working with GPU computing in 2007, when it was not yet popular nor even known if this technology would be significant for science. She has since then developed a strong research program which involves investigating new numerical algorithms, and adapting old ones, for GPUs. Applications of interest in her group are mostly focused on unsteady fluid flow, but also consider problems in diverse areas, such as molecular physics.
This NVIDIA award is aimed at supporting Prof. Barba’s research in the use of GPUs for computational fluid dynamics, especially with immersed-boundary methods for flow around moving objects. A particular objective in Barba’s group is to develop the capability to compute the aerodynamics of small-scale flyers—such as bats, birds and micro-air vehicles. The development of immersed-boundary methods in GPU hardware would offer the capacity to simulate these flows in small computer clusters, which would enable many more researchers to address the science of unsteady flapping flight.