Gordon Bell Computational Methods
           
Event Type Start Time End Time Rm # Chair  

 

Paper 1:30PM 2:00PM 36-37 William Gropp (Argonne National Laboratory)
 
Title:
High Resolution Forward and Inverse Earthquake Modeling on Terascale Computers
  Speakers/Presenter:
Volkan Akcelik (Carnegie Mellon University), Jacobo Bielak (Carnegie Mellon University), George Biros (Courant Institute, New York University), Ioannis Epanomeritakis (Carnegie Mellon University), Antonio Fernandez (Carnegie Mellon University), Omar Ghattas (Carnegie Mellon University), Eui Joong Kim (Carnegie Mellon University), Julio Lopez (Carnegie Mellon University), David O'Hallaron (Carnegie Mellon University), Tiankai Tu (Carnegie Mellon University), John Urbanic (Pittsburgh Supercomputing Center)

 

Paper 2:00PM 2:30PM 36-37 William Gropp (Argonne National Laboratory)
 
Title:
IPSAP : A High-performance Parallel Finite Element Code for Large-scale Structural Analysis Based on Domain-wise Multifrontal Technique
  Speakers/Presenter:
Seung Jo Kim (Department of Aerospace Engineering, Seoul National University, KOREA), Chang Sung Lee (Department of Aerospace Engineering, Seoul National Universtiy, KOREA), Jeong Ho Kim (High Performance Computing and Networking Supercomputing Center, Korea), Minsu Joh (High Performance Computing and Networking Supercomputing Center, Korea), Sangsan Lee (High Performance Computing and Networking Supercomputing Center, Korea)

 

Paper 2:30PM 3:00PM 36-37 William Gropp (Argonne National Laboratory)
 
Title:
A new parallel kernel-independent fast multipole method
  Speakers/Presenter:
Lexing Ying (New York University), George Biros (New York University), Denis Zorin (New York University), Harper Langston (New York University)
             

 

     
  Session: Gordon Bell Computational Methods
  Title: High Resolution Forward and Inverse Earthquake Modeling on Terascale Computers
  Chair: William Gropp (Argonne National Laboratory)
  Time: Wednesday, November 19, 1:30PM - 2:00PM
  Rm #: 36-37
  Speaker(s)/Author(s):  
  Volkan Akcelik (Carnegie Mellon University), Jacobo Bielak (Carnegie Mellon University), George Biros (Courant Institute, New York University), Ioannis Epanomeritakis (Carnegie Mellon University), Antonio Fernandez (Carnegie Mellon University), Omar Ghattas (Carnegie Mellon University), Eui Joong Kim (Carnegie Mellon University), Julio Lopez (Carnegie Mellon University), David O'Hallaron (Carnegie Mellon University), Tiankai Tu (Carnegie Mellon University), John Urbanic (Pittsburgh Supercomputing Center)
   
  Description:
  For earthquake simulations to play an important role in the reduction of seismic risk, they must be capable of high resolution and high fidelity. We have developed algorithms and tools for earthquake simulation based on multiresolution hexahedral meshes. We have used this capability to carry out 1 Hz simulations of the 1994 Northridge earthquake in the LA Basin using 100 million grid points. Our code sustains 0.9 teraflop/s for 12 hours on 2048 AlphaServer processors at 87% parallel efficiency. Because of uncertainties in characterizing earthquake source and basin material properties, a critical remaining challenge is to invert for source and material parameter fields for complex 3D basins from records of past earthquakes. Towards this end, we present results for material and source inversion of high-resolution models of basins undergoing antiplane motion using parallel scalable inversion algorithms that overcome many of the difficulties particular to inverse heterogeneous wave propagation problems.
  Link: Download PDF
   

 

     
  Session: Gordon Bell Computational Methods
  Title: IPSAP : A High-performance Parallel Finite Element Code for Large-scale Structural Analysis Based on Domain-wise Multifrontal Technique
  Chair: William Gropp (Argonne National Laboratory)
  Time: Wednesday, November 19, 2:00PM - 2:30PM
  Rm #: 36-37
  Speaker(s)/Author(s):  
  Seung Jo Kim (Department of Aerospace Engineering, Seoul National University, KOREA), Chang Sung Lee (Department of Aerospace Engineering, Seoul National Universtiy, KOREA), Jeong Ho Kim (High Performance Computing and Networking Supercomputing Center, Korea), Minsu Joh (High Performance Computing and Networking Supercomputing Center, Korea), Sangsan Lee (High Performance Computing and Networking Supercomputing Center, Korea)
   
  Description:
  Most of researches for large-scale parallel structural analysis have focused on iterative solution methods since direct solution methods generally have many difficulties and disadvantages for large-scale problems. However, due to the numerical robustness of direct methods that guarantees the solution to be obtained within estimated time, direct methods are much more desirable for general application of large-scale structural analysis, if the difficulties and disadvantages can be overcome. In this research, we propose the domain-wise multifrontal solver as an efficient direct solver that can overcome most of these difficulties and disadvantages. By using our own structural analysis code IPSAP which uses the proposed solver, we can solve the largest problem ever solved by direct solvers and can sustain 191 Gflop/s with 256 CPUs on our self-made cluster system, Pegasus. By implementing the block Lanczos algorithm using our solver, IPSAP can solve eigen problems with 7 millions of DOFs within one hour.
  Link: Download PDF
   

 

     
  Session: Gordon Bell Computational Methods
  Title: A new parallel kernel-independent fast multipole method
  Chair: William Gropp (Argonne National Laboratory)
  Time: Wednesday, November 19, 2:30PM - 3:00PM
  Rm #: 36-37
  Speaker(s)/Author(s):  
  Lexing Ying (New York University), George Biros (New York University), Denis Zorin (New York University), Harper Langston (New York University)
   
  Description:
  We present a new adaptive fast multipole algorithm and its parallel implementation. The algorithm is kernel-independent in the sense that the acceleration of the computation of the far field does not rely on any analytic expansions, but only uses kernel evaluations. The new method enables scalable simulations for many important problems in science and engineering. Examples include viscous flows, fracture mechanics and screened Coulombic interactions. Our MPI based parallel implementation logically separates the computation and communication phases to avoid synchronization in the upward and downward computation passes, and it enables us to fully exploit computation and communication overlapping. We measure isogranular and fixed-size scalability for a variety of kernels on the Pittsburgh Supercomputing Center's TCS-1 Alphaserver on up to 2048 processors. Our largest experiments reached 1.2 billion unknowns, for which we have achieved 1 Tflops/s peak performance and 0.7 Tflops/s sustained performance. Overall, our implementation achieves excellent parallel efficiency.

This paper has been nominated for the Best Paper of SC2003 award.

This paper has been nominated for the Best Student Paper of SC2003 award.
  Link: Download PDF