Meros: Software for Block Preconditioning the Navier-Stokes Equations

Slides:



Advertisements
Similar presentations
Stability Analysis Algorithms for Large-Scale Applications Andy Salinger, Roger Pawlowski, Ed Wilkes Louis Romero, Rich Lehoucq, John Shadid Sandia National.
Advertisements

Coupled Fluid-Structural Solver CFD incompressible flow solver has been coupled with a FEA code to analyze dynamic fluid-structure coupling phenomena CFD.
12/21/2001Numerical methods in continuum mechanics1 Continuum Mechanics On the scale of the object to be studied the density and other fluid properties.
Sparse Matrix Methods Day 1: Overview Day 2: Direct methods
Network and Grid Computing –Modeling, Algorithms, and Software Mo Mu Joint work with Xiao Hong Zhu, Falcon Siu.
The Landscape of Ax=b Solvers Direct A = LU Iterative y’ = Ay Non- symmetric Symmetric positive definite More RobustLess Storage (if sparse) More Robust.
ENGG 1801 Engineering Computing MATLAB Lecture 7: Tutorial Weeks Solution of nonlinear algebraic equations (II)
MECH300H Introduction to Finite Element Methods Lecture 10 Time-Dependent Problems.
Laminar Incompressible Flow over a Rotating Disk Numerical Analysis for Engineering John Virtue December 1999.
Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy’s National Nuclear.
ITERATIVE TECHNIQUES FOR SOLVING NON-LINEAR SYSTEMS (AND LINEAR SYSTEMS)
Tutorial 5: Numerical methods - buildings Q1. Identify three principal differences between a response function method and a numerical method when both.
Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy’s National Nuclear.
Lecture Objectives Review SIMPLE CFD Algorithm SIMPLE Semi-Implicit Method for Pressure-Linked Equations Define Residual and Relaxation.
Are their more appropriate domain-specific performance metrics for science and engineering HPC applications available then the canonical “percent of peak”
Continuation Methods for Performing Stability Analysis of Large-Scale Applications LOCA: Library Of Continuation Algorithms Andy Salinger Roger Pawlowski,
Scalable Multi-Stage Stochastic Programming
Next Class Final Project and Presentation – Prepare and me the ppt files 5-7 slides Introduce your problem (1-2 slides) – Problem – Why CFD? Methods.
1 Using the PETSc Parallel Software library in Developing MPP Software for Calculating Exact Cumulative Reaction Probabilities for Large Systems (M. Minkoff.
Large-Scale Stability Analysis Algorithms Andy Salinger, Roger Pawlowski, Ed Wilkes Louis Romero, Rich Lehoucq, John Shadid Sandia National Labs Albuquerque,
Short introduction for the quasi-equilibrium binary neutron star solutions. Introducing two patches of fluid coordinate grids, the initial data code can.
ANS 1998 Winter Meeting DOE 2000 Numerics Capabilities 1 Barry Smith Argonne National Laboratory DOE 2000 Numerics Capability
ParCFD Parallel computation of pollutant dispersion in industrial sites Julien Montagnier Marc Buffat David Guibert.
CFD Lab - Department of Engineering - University of Liverpool Ken Badcock & Mark Woodgate Department of Engineering University of Liverpool Liverpool L69.
ML: Multilevel Preconditioning Package Trilinos User’s Group Meeting Wednesday, October 15, 2003 Jonathan Hu Sandia is a multiprogram laboratory operated.
Efficient Integration of Large Stiff Systems of ODEs Using Exponential Integrators M. Tokman, M. Tokman, University of California, Merced 2 hrs 1.5 hrs.
Lecture Objectives: Explicit vs. Implicit Residual, Stability, Relaxation Simple algorithm.
MECH593 Introduction to Finite Element Methods Eigenvalue Problems and Time-dependent Problems.
Danny Dunlavy, Andy Salinger Sandia National Laboratories Albuquerque, New Mexico, USA SIAM Parallel Processing February 23, 2006 SAND C Sandia.
Lecture Objectives Review Define Residual and Relaxation SIMPLE CFD Algorithm SIMPLE Semi-Implicit Method for Pressure-Linked Equations.
Circuits Theory Examples Newton-Raphson Method. Formula for one-dimensional case: Series of successive solutions: If the iteration process is converged,
MECH4450 Introduction to Finite Element Methods Chapter 9 Advanced Topics II - Nonlinear Problems Error and Convergence.
Domain Decomposition in High-Level Parallelizaton of PDE codes Xing Cai University of Oslo.
Cracow Grid Workshop, November 5-6, 2001 Concepts for implementing adaptive finite element codes for grid computing Krzysztof Banaś, Joanna Płażek Cracow.
FALL 2015 Esra Sorgüven Öner
On the Use of Finite Difference Matrix-Vector Products in Newton-Krylov Solvers for Implicit Climate Dynamics with Spectral Elements ImpactObjectives 
MECH593 Introduction to Finite Element Methods
An Overview of Meros Trilinos User’s Group Wednesday, November 2, 2005 Victoria Howle Computational Sciences and Mathematics Research Department (8962)
Programming Massively Parallel Graphics Multiprocessors using CUDA Final Project Amirhassan Asgari Kamiabad
Scientific Computing Lab Outlook / State of Research Dr. Miriam Mehl Institut für Informatik Scientific Computing in Computer Science.
Adaptive grid refinement. Adaptivity in Diffpack Error estimatorError estimator Adaptive refinementAdaptive refinement A hierarchy of unstructured gridsA.
Quality of Service for Numerical Components Lori Freitag Diachin, Paul Hovland, Kate Keahey, Lois McInnes, Boyana Norris, Padma Raghavan.
Multipole-Based Preconditioners for Sparse Linear Systems. Ananth Grama Purdue University. Supported by the National Science Foundation.
Finding Rightmost Eigenvalues of Large, Sparse, Nonsymmetric Parameterized Eigenvalue Problems Minghao Wu AMSC Program Advisor: Dr. Howard.
Solving linear systems in fluid dynamics P. Aaron Lott Applied Mathematics and Scientific Computation Program University of Maryland.
Model Anything. Quantity Conserved c  advect  diffuse S ConservationConstitutiveGoverning Mass, M  q -- M Momentum fluid, Mv -- F Momentum fluid.
Conjugate gradient iteration One matrix-vector multiplication per iteration Two vector dot products per iteration Four n-vectors of working storage x 0.
Hui Liu University of Calgary
G. Cheng, R. Rimmer, H. Wang (Jefferson Lab, Newport News, VA, USA)
Xing Cai University of Oslo
Objective Numerical methods SIMPLE CFD Algorithm Define Relaxation
A computational loop k k Integration Newton Iteration
ECE 1270: Introduction to Electric Circuits
Lecture Objectives: Review Explicit vs. Implicit
© Fluent Inc. 1/10/2018L1 Fluids Review TRN Solution Methods.
Robert Shuttleworth Applied Math & Scientific Computation (AMSC)
Lecture Objectives: Advance discretization methods
Objective Unsteady state Numerical methods Discretization
ENGG 1801 Engineering Computing
A robust preconditioner for the conjugate gradient method
finite element method node point based strong form
GENERAL VIEW OF KRATOS MULTIPHYSICS
finite element method node point based strong form
Supported by the National Science Foundation.
Objective Numerical methods Finite volume.
Introduction to Fluid Dynamics & Applications
Mathematical Solution of Non-linear equations : Newton Raphson method
Parallelizing Unstructured FEM Computation
Ph.D. Thesis Numerical Solution of PDEs and Their Object-oriented Parallel Implementations Xing Cai October 26, 1998.
A computational loop k k Integration Newton Iteration
Presentation transcript:

Meros: Software for Block Preconditioning the Navier-Stokes Equations Robert Shuttleworth (CSCAMM/AMSC) Howard Elman (CS/AMSC) Vicki Howle, John Shadid, and Ray Tuminaro (Sandia National Labs) Motivation: Develop fully implicit solution methods to the incompressible Navier-Stokes Efficient, robust solution of flow problems requires block preconditioning Linear Solvers: Operator Based Block Preconditioning Focus: Adapt block preconditioners to the linear subproblems that arise in realistic fluid flow problems MPSalsa: Realistic, massively parallel, chemically reactive fluid flow code 11/12/2018 SC Student Seminar

Introduction Given the Navier-Stokes Equations: Nonlinear Term: Oseen: Newton: Jacobian of Momentum Eq. Discretization and Linearization: Vectors, example sizes, sparsity, computation gets expensive because … 11/12/2018 SC Student Seminar

Where Meros fits Packages: Nonlinear Solver Linear block precondition Methods Component Nonlinear Solver Linear block precondition MPSalsa Meros Newton-Krylov Finite Element GMRESR Time Loop Nonlinear Loop Linear Solver Block Precond End NonLin Loop End Time Loop 11/12/2018 SC Student Seminar

Block Preconditioners Discretization Consider: Optimal preconditioner is when X is the Schur Complement, Question: How to approximate the Schur complement? 11/12/2018 SC Student Seminar

MPSalsa Steady Problem Results Mesh Size Incomplete LU Fp 10 64 x 64 88.0 25.4 128 x 128 194.2 23.2 100 95.7 40.8 335.3 40.7 200 95.9 56.6 364.7 53.7 Challenges in mpsalsa, salsa uses stabilization, less control over discretization, salsa supplied fp, ap from salsa, issues, fp mesh indy not ilut, issue with parallelize Mesh Refinement Results – 2D Lid Driven Cavity The values in each column represent the average number of Outer Saddle Point Solves per Newton Step. 11/12/2018 SC Student Seminar

MPSalsa Steady Problem Results The Fp preconditioner converges in less computational time than ILU on a lid driven cavity problem with 200,000 unknowns. This graph represents the average number of linear iterations per nonlinear step for the 2D lid driven cavity problem with 12,000 unknowns. 11/12/2018 SC Student Seminar

Future Work Parallel Trials More intricate flow problems 512 Dual Node Sandia Linux Cluster Sandia’s ASCI Red Supercomputer Sandia’s ASCI Red Storm Supercomputer More intricate flow problems Backward facing step Diamond obstruction Chemically reactive flow 11/12/2018 SC Student Seminar