Professor of Computer Science, Physics & Astronomy, Chemical Engineering & Materials Science, and Biological Sciences
Ph.D. University of Tokyo, Japan, 1989
Contact InformationCollaboratory for Advanced Computing and Simulations (CACS)
High-end scientific computing on geographically distributed parallel supercomputers and virtual environment:
Divide-conquer-recombine simulation algorithms based on spatial locality with low time/space/bandwidth complexity and tight error control.
A space-time-ensemble parallel approach based on temporal locality to predict long-time dynamics.
Metascalable ("design once, scale on new architectures") parallel-and-distributed supercomputing frameworks.
Immersive and interactive visualization and analytics of large scientific datasets (billion-atom chemical bond networks).
Hierarchical simulations and validation that automatically embed quantum-mechanical and atomistic calculations within continuum calculation on demand with guaranteed quality-of-solutions.
High-end computational materials science.
We have demonstrated:
Unprecedented scales of quantum-mechanically accurate and well validated, chemically reactive molecular dynamics (MD) simulations--8.5 billion-atom reactive molecular dynamics (RMD) and 39.8 trillion electronic degrees-of-freedom (50.3 million-atom) quantum molecular dynamics (QMD) in the framework of density functional theory--in addition to 5.0 trillion-atom space-time multiresolution MD, with parallel efficiency over 0.98 on 786,432 Blue Gene/Q processors.
An automated execution of hierarchical QMD/MD simulation on a Grid of 6 supercomputer centers in the US and Japan, in which the number of processors changed dynamically on demand and resources were allocated and migrated dynamically in response to unexpected faults.
Real-time visualization of a billion-atom chemical bond network, with an embedded graph-based topological analysis.
To follow advances in computing technologies (hardware, software, algorithms) from teraflops to petaflops and beyond, to establish a comprehensive collaborative environment for geographically distributed computational scientists and information technology (IT) experts to perform the largest bio-nano simulations.
To establish educational programs to propel students into careers in emerging areas of nano, bio, and information technologies both in academic and industrial settings. CACS has excellent computing and visualization facilities: a 4,096-processor Linux cluster, and a visualization laboratory with an 8' by 14' tiled display and an immersive and interactive 3D visualization environment.
Dual-degree Graduate Education in High Performance Computing and Simulations
At USC, we have introduced a dual-degree program that allows students to obtain a Ph.D. in the physical sciences/engineering and an MS in Computer Science (CS). I have developed an MSCS program with specialization in High Performance Computing and Simulations (MSCS-HPCS), for which I serve as the faculty coordinator. For the MSCS-HPCS program, I have developed HPCS courses: CSCI596 (Scientific Computing and Visualization), CSCI653 (High Performance Computing and Simulations), PHYS516 (Methods of Computational Physics), and CSCI699 (Extreme-Scale Quantum Simulations).
Minority Research and Education
CACS organizes annual Computational Science Workshops for Underrepresented Groups (CSWUG) to provide undergraduate students and mentors from underrepresented groups with hands-on experience in HPCS.