Plastic Domain Decomposition Parallel Program

The Plastic Domain Decomposition (PDD) method implements parallel computational methodology designed for efficient and scalable computations with inelastic (elastic-plastic) finite elements on distributed memory parallel computers.

It is implemented using (UCD) modified part of OpenSees framework, ParMETIS graph library, PETSc numerical library, and other libraries from the UCDavis CompGeoMech toolset. This program is under active development, and we make this stable beta version available for trial, use and comments by the research community.

• Features:
  
  
  • Elements (currently working with)
  • Solids (8 and 20 node brick)
  • Beams (linear, force and/or displacement based nonlinear)
  • Zero length elements
  • any other that follows standard interface (needs addition of couple simple functions)
  
  
  • Material models (currently working with)
  • any elastic or elastic--plastic materials available for solids
  • linear or nonlinear, displacement and force based beam element materials
  
  
  • Multi-stage analysis
  • Self weight
  • Excavation
  • Construction sequencing
  • static loading
  • Dynamic (seismic) loading (using Domain Reduction Method)
  
  
  • Adaptive (multiobjective) computational load balancing by Plastic Domain Decomposition (PDD) method
  • Initial partitioning
  • Dynamic (adaptive) automatic repartitioning
  • No limit on a number of CPUs one can use on a given parallel machine
  • Transparent use of LAN (cluster) or WAN (grid) resources

• Availability:
• Please email Boris Jeremic (jeremic @ ucdavis . edu) if you wish to obtain up to date source files. New version is available as of September 2006.
  (The tar ball contains all the necessary files for running examples below. These files should not be mixed with the version of OpenSees currently available at the main OpenSees repository as there were some extensive additions and changes in parts of the source code).


• Makefiles for building on different architectures
• GeoWulf (our local Beowulf cluster, using GNU compilers for Linux) (Makefile.def)
• IBM Power4 DataStar at SDSC (using IBM C/C++ compilers for AIX) (Makefile.def)
• Teragrid parallel machines (x86 architecture, using Intel compilers for Linux) (Makefile.def)


• Executables are available (created from the above source files) and installed on machines listed below. Execution commands are given as well.
• GeoWulf (@ucdavis.edu)(mpirun, using MPICH-1.2.7)
• IBM Power4 DataStar at SDSC (llsubmit to loadleveler, /users09/ucd/gjie/bin/OpenSeesMPI.run)
• Teragrid parallel machines, TACC (@utexas.edu) and IA64 (@sdsc.edu) (qsub to queue, /users/gjie/bin/OpenSeesMPI.run)


• Writeups (draft, frequently changed) ( pdf , 2up.pdf )


• Development version repository (using SubVersion version control system, comming soon)


• Examples (with tcl scripts for OpenSees)
• Soil-Foundation-Structure Interaction example,
• Structural example script,
• more to come

This work was supported in part by a number of Agencies listed below:
Civil and Mechanical System program, Directorate of Engineering of the National Science Foundation, under Award NSF--CMS--0324661 (cognizant program director Dr. Steve McCabe) and under Award NSF-CMS-0337811 (cognizant program director Dr. Richarad Fragaszy);
Pacific Earthquake Engineering Research Center through Earthquake Engineering Research Centers Program of the National Science Foundation under Award Number NSF--EEC--9701568 (cognizant program director Dr. Joy Pauschke);
California Department of Transportation (Caltrans) under Award # 59A0433, (cognizant program director Dr. Saad El-Azazy).