SCIENCE
NYU Courant researchers receive supercomputing award for creating software that simulates blood flow
Researchers at New York University's Courant Institute of Mathematical Sciences have received the George Bell Prize, given to the world's fastest supercomputing application, for creating software that simulates blood flow. The award, which includes a cash prize of $10,000 and is given by the Association for Computing Machinery (ACM), is shared with researchers at Georgia Tech and Oak Ridge National Laboratory (ORNL).
Ultimately, researchers aim to create a virtual circulatory system that can study blood clots and help scientists develop safer, more effective devices such as stints and heart pumps.
The software, which creates a blood-flow simulation of 260 million deformable red blood cells flowing in plasma, ran on ORNL's Jaguar supercomputer at 700 trillion calculations per second.
The research team was headed by George Biros, an associate professor in Georgia Tech's School of Computational Science & Engineering, who began the software project as a post-doctoral fellow at the Courant Institute, working with Courant Professor Denis Zorin and doctoral student Lexing Ying.
The application was notable for its realistic modeling—it simulated cells that deform as they move through plasma, not artificially spherical blood cells that retain their shape. Shravan Veerapaneni, a post-doctoral researcher at the Courant Institute, along with Abtin Rahimian, a doctoral student at Georgia Tech, Biros, and Zorin, developed the algorithm that produced this realistic simulation of how individual, three-dimensional cells deform.
Ilya Lashuk, a former postdoctoral fellow at Georgia Tech and currently at Lawrence Livermore National Laboratory, and a team led by Georgia Tech Assistant Professor Richard Vuduc contributed key algorithms making it possible to run the simulations on Jaguar.
The process relies on "fast multipole method," which was developed by Courant Director Leslie Greengard and Vladimir Rokhlin, a professor a Yale University, in 1987. The method can be used to solve a range of equations, in particular ones describing fluid motion on the small scale that characterize the blood flow software.