Lorena Barba, Ph.D.

Photo of Lorena Barba

Adjunct Research Professor

Ph.D., California Institute of Technology

phone: (202) 994-3715
email: labarba@bu.edu
website: http://lorenabarba.com
office: Staughton Hall, George Washington University

Research Interests

Fluid dynamics * Novel computer architectures * Development of fast and efficient algorithms * Particle methods used for fluid simulation

Professor Barba is a computational scientist and a fluid dynamicist. That is, she does research by means of computer simulations, studies and develops numerical methods, and the types of applications she is most interested in have to do with problems of fluid flow.

One of the challenges in fluid simulation continues to be the need to straddle many scales. New research in computational methods is still sorely needed to build schemes that are intelligent, detect the scales of the problem, and adapt. Moreover, we need programs that are hardware-aware.

A new-wave in computational fluid and solid mechanics is the development of meshfree methods. They promise to offer solutions to many enduring problems, such as dealing with discontinuities such as cracks and simulating flows with low numerical diffusion. Prof Barba believes that, in addition, meshfree methods should be especially well-suited to exploit the new hardware technologies that are revolutionizing the scene. They should also more naturally offer ways to deal with multiple scales in a problem, and to construct adaptive schemes.

Meshfree methods also offer enormous potential for the simulation of problems involving moving and/or deformable boundaries, such as happens in most biological applications. In these problems, the need to adapt or regenerate a mesh is a huge burden for the computation and removing this burden will allow the solution of larger problems, in less time, and in smaller computers. Here, as in multi-scale problems, Prof Barba believes we should also discover ways to combine continuum and particle approaches, hence build and test hybrid methods.

Selected Publications
  • Rio Yokota and L A Barba. Hierarchical N -body simulations with autotuning for heterogeneous systems. Computing in Science and Engineering , 2012 . Accepted. Preprint at arXiv:1108.5815 .
  • Rio Yokota and L A Barba. A tuned and scalable fast multipole method as a preeminent algorithm for exascale systems. Int. J. High-perf. Comput. Appl. , 2011 . In press. Preprint on arXiv:1106.2176 .
  • Felipe A. Cruz , Simon K. Layton , and L A Barba. How to obtain effi cient GPU kernels: an illustration using FMM & FGT. Comp. Phys. Comm. , 182 (10 ):2084 –2098 , 2011 . doi:10.1016/j.cpc.2011.05.002 .2
  • Rio Yokota, Jaydeep P. Bardhan, Matthew G. Knepley, L A Barba, and Tsuyoshi Hamada. Biomolecular electrostatics using a fast multipole BEM on up to 512  GPUs and a billion unknowns. Comp. Phys. Comm. , 182 (6 ):1271 –1283 , 2011 . doi:10.1016/j.cpc.2011.02.013 .
  • Rio Yokota and L A Barba. Comparing the treecode with FMM on GPUs for vortex particle simulations of leapfrogging vortex rings. Computers & Fluids , 45 (1 ):155 –161 , June 2011 . doi:10.1016/j.compfluid.2010.11.029 .
  • Felipe A Cruz, Matthew G Knepley, L A Barba, “PetFMM–A dynamically load-balancing parallel fast multipole library,”  Int. J. Num. Meth. Eng., 2010, doi:10.1002/nme.2972.
  • Rio Yokota, L A Barba, Matthew G Knepley, “PetRBF—A parallel O(N) algorithm for radial basis function interpolation,” Comput. Meth. Appl. Mech. Eng., Vol. 199(25–28): 17931804, 2010. doi:10.1016/j.cma.2010.02.008