Trilinos: From a User\'s Perspective Russell Hooper Nov. 7, 2007 SAND 2007-7285P Sandia is a multiprogram research center worked by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy under Contract DE-AC04-94AL85000
Slide 2Aleph Overview Program Driver keeping in mind the end goal to meet significantly lessened item improvement objectives (cost and timetable), a prescient ability for supporting configuration and surveying execution is required. Specialized Drivers Proper operation is emphatically coupled to materials, geometry, and connected power: What are the prevailing factors? What is the outline edge? Is a given plan advanced? Is the plan unsteady? No codes right now exist which can deal with the extraordinary range in plasma thickness and the coupling to materials required for prescient execution. Best gauge is that 1-3 petaflops will be required for a forceful calculation joining immeasurably critical material science (2D counts with fractional physical science utilized 10 7 CPU-hours as a part of 2004!). The instrument will have the capacity to answer numerous major plan and execution questions.
Slide 3The Team The demonstrating task is multidisciplinary : 1100 – Paul Miller, Ed Barnat 1400 – Paul Crozier, David Day, Russell Hooper, Steve Plimpton 1500 – Matt Hopkins, Rick Buss, Alan Williams, Polly Hopkins 2700 – Juan Elizondo, Ben Cole Aleph Team
Slide 4Aleph - Etymology First letter of Hebrew letters in order, Represents cardinality of unbounded sets, Integers Reals Arc displaying code quantz.sandia.gov/arcmodel/wiki
Slide 5Aleph Physics Field understand (continuum) : Particle moves :
Slide 7Particles traveling through a 3D unstructured tetrahedral work in parallel with load adjusting. Translucent shading demonstrates processor number. (Genuine reproductions will have numerous, numerous, more particles.)
Slide 8Aleph Dependencies Justifiable Complexity !! Zoltan Exodus Aleph IO_Standalone Trilinos AztecOO Epetra Teuchos Triutils
Slide 9Current Trilinos Usage - Overview Construction of Solver LHS - Epetra_FECrsMatrix Construction of Solver RHS - Epetra_FEVector Linear Solve - AztecOO (ML, Belos) Particle spatial averaging - Epetra_FEVector Particle worldly averaging - Epetra_FEVector Load adjusting - Zoltan, Epetra_FEVector
Slide 10Dynamic Balancing Using Zoltan
Slide 11Current Trilinos Usage - Field Solve Epetra_Import
Slide 12Current Trilinos Usage - Spatial Particle Averaging (pre-PIC) Epetra_FEVector NodResult1; NodResult1.SumIntoGlobalValues(); NodResult1.GlobalAssemble(Add);
Slide 13Current Trilinos Usage - Temporal Particle Averaging NodResult1.SumIntoGlobal(); NodResult2.SumIntoGlobal(); ElemResult1.SumIntoGlobal(); Exodus Time Steps NodResult1.GlobalAssemble(Add); SharedNodResult1.Import(NodResult1);
Slide 14Current Trilinos Usage - Temporal Particle Averaging with Dynamic Balancing NodResult1.SumIntoGlobal(); NodResult1.GlobalAssemble(Add); Epetra_Import importer(New,Old); NodResult1.Import(NodResult1,importer); Exodus t Balance Time Steps
Slide 15Future Trilinos Usage - Performance Tuned Poisson Solves - ML, Belos, Epetra(Ext ?)_CrsSingletonFilter Smoothed Gradient Projections : Improved starting conjecture - Epetra_Parks_Day? FEM variations for EMS, EMD - ML, new package(s)
Slide 16Future Trilinos Usage - Performance Tuned Poisson Solves - ML, Belos, Epetra(Ext ?)_CrsSingletonFilter Smoothed Gradient Projections : Improved starting supposition - Epetra_Parks_Day? FEM variations for EMS, EMD - ML, new package(s)
Slide 17Feedback Working tests/cases are precious !! Documentation, esp. in "Develop bundles" (I can\'t toss stones!) Apps + Trilinos show, esp. wrt Meta bundles
Slide 18Post-TUG Aleph Dependencies Zoltan Exodus Aleph Trilinos IO_Standalone AztecOO Epetra Teuchos Triutils Isorropia EpetraExt ML Belos