Webpage for K. J. Roche

Contact Me, My Job, QMB Exascale Application, Computer Science Research, Building a Case for Federal Funding, OMB Software Metric


Contact Info

Kenneth J. Roche
High Performance Computing Group, Computational Sciences and Mathematics Division, Pacific Northwest National Laboratory
Nuclear Theory Group, Department of Physics, University of Washington
Box 351560
3910 15th Ave. NE
Seattle, WA 98195-1560
(tel) 1 (206) 661-9906
(fax) 1 (206) 685-0635
kenneth.roche AT pnl DOT gov ; k8r AT uw DOT edu


Professional Positions Held

I am a computational and information scientist at the Pacific Northwest National Laboratory. The US Department of Energy funds my research through Battelle. I am also affiliated with the University of Washington as an Associate Professor in the Department of Physics. My office is in the Physics Building at UW in Seattle.

Prior to moving to the northwest US, I lived in beautiful East Tennessee. I was a staff scientist in the Future Technologies Group of the Computer Science and Mathematics Division of the Oak Ridge National Laboratory for five years.

And, before joining the ORNL, I was a graduate student of physics and computer science, and then faculty in the Innovative Computing Laboratory (ICL) in the Computer Science Department (now EECE) of the University of Tennessee in Knoxville. I can say that parallel and distributed computing, and computational science are very well grounded at UTK and ORNL. I sharpened my skills in ET.

Exascale Application Development

Programming is simply part of my daily life. The problems that are brought to my attention tend to be unsolvable analytically, and hard to compute due to algorithms that require the biggest supercomputers for evaluation. One such problem space is many-body quantum mechanics -in particular studies of the dynamics of superfluid systems (ie, includes atomic nuclei). Interested readers can find my talks (and those of Professors A. Bulgac, P. Magierski, and I. Stetcu) at a national project designed to study low energy nuclear structure and dynamics, UNEDF. Here are some related posters and papers that describe my computational research into the dynamics of low-energy nuclear phenomena. To calibrate the scale of the software effort, we utilized over 70M cpu-hours in FY2010 executing problems. The entire (224,256 PEs) Cray XT5 at ORNL's National Center for Computational Sciences can be effectively utilized in a single run by the slda application software.

  • Extending the Kawai-Kerman-McVoy Statistical Theory of Nuclear Reactions to Intermediate Structure via Doorways, G. Arbanas, C.A. Bertulani, D.J. Dean, A.K. Kerman, and K.J. Roche, Third International Workshop on Compound Nuclear Reactions and Related Topics, Prague, Czech Republic, September 19-23, 2011
  • Phys. Rev. C 84, 051309(R) (2011) (arxiv.org/abs/1108.3064), Isovector Giant Dipole Resonance from 3D Time-Dependent Density Functional Theory for Superfluid Nuclei, I. Stetcu, A. Bulgac, P. Magierski, and K.J. Roche
  • arxiv.org/abs/1108.1779, Quantum Shock Waves and Domain Walls in Real-Time Dynamics of a Superfluid Unitary Fermi Gas, A. Bulgac, Y.-L. Luo, and K.J. Roche
  • arxiv.org/abs/1107.4925, UNDEF: Advanced Scientific Computing Transforms the Low-Energy Nuclear Many-Body Problem, M. Stoitsov, H. Nam, W. Nazarewicz, A. Bulgac, G. Hagen, M. Kortelainen, J.C. Pei, K.J. Roche, N. Schunck, I. Thompson, J.P. Vary, and S.M. Wild
  • Science, 10 June 2011: Vol. 332 no. 6035 pp. 1288-1291 DOI: 10.1126/science.1201968, Real-Time Dynamics of Quantized Vortices in a Unitary Fermi Superfluid, Aurel Bulgac, Yuan-Lung Luo, Piotr Magierski, Kenneth J. Roche, Yongle Yu ; Our public release position on this publication is here.
  • , 2011, site visit of US DOE ASCR
  • , 2008, invited US DOE SciDAC Conference
  • J.Phys. Conf. Ser. 012064 (2008), Time-Dependent Density Functional Theory Applied to Superfluid Nuclei, A. Bulgac, K.J. Roche
  • arxiv.org/abs/1011.5999,Generation and Dynamics of Quantized Vortices in a Unitary Fermi Superfluid, A. Bulgac, Y. Luo, P. Magierski, K.J. Roche, Y. Yu
  • AIP Conf. Proc. 597, pp218-226 (2001), New Results from Three-Dimensional Time Dependent Hartree-Fock, J. Mahrun, D.J. Dean, M.R. Strayer, K.J. Roche

    • Computer Science Research

    I am interested in software libraries, algorithm research, and complexity theory. Self-adaptive software methods attempt to exploit the notion that computers can be programmed to make decisions that normally involve human intervention. Sample reasons for pursuing such techniques are to search for correct algorithms, performance optimization, adjust to changes in hardware or software environments, scheduling, etc. I have employed such techniques for libraries in tightly coupled distributed environments, shared memory environments, clusters of clusters in wide area environments (such as those that support the back-end of cloud or grid computing platforms), the internet, and heterogenous environments. Here are some examples of this work:

  • US DOE Journal of Undergraduate Research v7 pp48-54 (2007), Adaptively Improving Long Distance Network Transfers with Logistics, D. LaBissoniere, K.J. Roche
  • (Sp. Ed.) Parallel Computing 29(11-12) pp1723-1743 (2003), Self-Adapting Software for Numerical Linear Algebra and LAPACK for Clusters, (also LAPACK Working Note (LAWN) 160), Z. Chen, J.J. Dongarra, P. Luszczek, K.J. Roche
  • IEEE Computer Society, Proc. Euromicro Parallel, Distributed and Network-Based Processing (PDP), IEEE Computer Society, Genova, Italy (2003), Automatic Optimizations of Parallel Linear Algebra Routines in Systems with Variable Load, J. Cuenca, D. Giminez, J. Gonzalez, J. Dongarra, K. Roche
  • Parallel Computing -Advances and Current Issues, Imperial College, London ISBN 1-86094-315-2 pp3-30 (invited) (2002), Deploying Parallel Numerical Library Routines to Cluster Computing in a Self Adapting Fashion, K.J. Roche, J.J. Dongarra
  • International Journal of High Performance Computing Applications 15(4) pp359-374 (2001), Numerical Libraries and the Grid , (also Proc.of Supercomputing (SC01), IEEE Computer Society, ACM SIGARCH, Denver, CO), A. Petitet, S. Blackford, J. Dongarra, B. Ellis, G. Fagg, K. Roche, S. Vadhiyar

    • GPUs and m-core chipsets together define programming research targets that I am still learning how to program in hybrid manner.

    Supercomputing: Making A Case for Federal Support

    I am interested to promote the continued federal support of the design and deployment of massively parallel computing systems for computational research, engineering, and security. In the last decade, there have been dozens of strategically formed, federally sponsored and academic working groups with the goal to solve problems related to this agenda. Here is a short list of efforts to which I contributed:

  • From Petascale to Exascale:R&D Challenges for HPC Simulation Environments, Advanced Simulation Computing, National Nuclear Security, US Department of Energy, San Francisco, CA, March 22-24, 2011.
  • Exascale Research Kickoff Meeting, Office of Advanced Scientific Computing Research, US Department of Energy, San Diego, CA, March 7-11, 2011.
  • Cross-Cutting Technologies for Computing at the Exascale, (report), US Department of Energy, Office of Science, Advanced Scientific Computing Research, Washington, D.C., February 2-4, 2010.
  • Scientific Grand Challenges for National Security: the Role of Computing at the Extreme Scale, (report), US Department of Energy, National Nuclear Security Administration, Washington, D.C., October 6-8, 2009
  • Nuclear Physics Workshop: Forefront Questions in Nuclear Science and the Role of High Performance Computing (report), US Department of Energy, Nuclear Physics and Advanced Scientific Computing Research Programs, Washington, D.C. , January 26-28, 2009
  • E3SGS Simulation and Modeling at the Exascale for Energy, Ecological Sustainability and Global Security, US Department of Energy, Oak Ridge National Laboratory, Oak Ridge, TN, May 17-18, 2007
  • Computational Subsurface Sciences Workshop, (report) , US Department of Energy, Biological and Environmental Sciences Programs, Bethesda, MD, January 9-12, 2007
  • Report on Simulation and Modeling for Advanced Nuclear Energy Systems, (report), US DOE Office of Nuclear Energy and Office of Science - Advanced Scientific Computing Research, P. Finck, D. Keyes, and R. Stevens (Eds), Washington, DC, August 15-17, 2006


    • DOE SC ASCR GPRA-PMM Software Metric SC GG 3.1/2.5.2: Improve Computational Science Capabilities
    Since FY2004, I have led a national benchmarking exercise for the US DOE SC     Office of Advanced Scientific Computing Research that annually demonstrates to the Office of Management and Budget (OMB) at the White House the effective use of federally funded supercomputers to simulate problems of interest to the Energy program, and thus the nation. The applications are recommended annually by laboratory leaders at all the major national laboratories and the final list is selected by committee, the Advanced Scientific Computing Research Committee. I may provide access to the FY04-FY10 reports in the future. For now, they will be available on request from me or DOE headquarters only -(unless you are a lucky participant!). To help calibrate the scale of these efforts, this project had a 150M cpu hour allocation on the largest open science computer in the nation in FY10. Get the facts. Below is some summary data for consideration, and here is the most recent talk on this research delivered on November 2, 2011 at the American Geophysical Union building in Washington, D.C. to the ASCAC committee.

    Application--Institute--Domain--Architecture/Platform
    FY11: (current codes)
    LAMMPS , SNL , molecular dynamics, Intel / NVIDIA Tesla hybrid cluster
    STOMP , PNNL , subsurface modeling , Cray XT5
    OMEN , Purdue, nano-devices / icc design , Cray XT5
    COMPASS , UCLA / FNAL , plasma physics , Cray XT5
    FY10
    TD_SLDA , UW / PNNL , condensed matter / nuclear theory , Cray xt5
    POP , LANL , CCSM ocean / climate , Cray XT5
    LS3DF , Berkeley , nano-structure , Cray XT5
    Denovo , ORNL , nuclear reactor modeling , Cray XT5
    FY09
    CAM , NOAA/ORNL , CCSM atmosphere / climate , Cray XT5
    RAPTOR , SNL , combustion chemistry , Cray XT5
    VisIt , LLNL , visualization / data analytics , Cray XT5
    XGC1 , NYU , plasmas / tokamaks , Cray XT5
    FY08
    DCA++ , ORNL , condensed matter , Cray XT5
    GYRO , GA / ORNL , plasmas / tokomak , Cray XT5
    PFLOTRAN , LANL , subsurface modeling , Cray XT5
    FY07
    CHIMERA , UT/ORNL , supernova modeling , Cray XT3
    GTC-S , Princeton, plasmas / tokamaks , Cray XT4
    S3D INCITE , SNL , combustion chemistry , Cray XT3 U XT4
    FY06
    DCA/QMC , ORNL , condensed matter , Cray X1e
    ENZO , UCSD , cosmology , IBM Power5
    MADNESS , ORNL , math framework , Cray XT3
    ScalaBLAST , PNNL , bioinformatics / genome sequencing , HP Itanium2 lp
    FY05
    AORSA , ORNL / MIT , plasmas / tokomaks , Cray X1e
    CCSM , NOAA / LANL / ORNL / CO , community climate system model , Cray X1e
    LAMMPS , SNL , molecular dynamics , IBM BG/L
    Omega3P , Stanford / SLAC , accelerator physics , IBM SP Power3
    S3D INCITE , SNL , combustion chemistry , Cray X1
    S3D SciDAC , Michigan , combustion chemistry , HP Itanium2 lp
    FY04
    CCSM NOAA / LANL / ORNL / CO , climate , IBM p690
    MILC , US QCD , high energy nuclear theory , QCDOC
    NSM MC , ORNL , nuclear theory , IBM SP Power3
    RMPS , Auburn , atomic theory , IBM SP Power3
    VH-1 , North Carolina State , hydrodynamics , Cray X1

    Last updated November 21, 2011