Parallel Computational Fluid Dynamics 2005 (eBook)

Front Cover 1
Parallel Computational Fluid Dynamics 4
Copyright Page 5
Preface 6
Acknowledgements 7
Table of Contents 8
1. Invited Speakers and Special Sessions 16
Parallel Simulation of Lifting Rotor Wakes 16
1. Introduction 16
2. Flow solver 17
3. Grid generation 18
4. 7A rotor lifting cases 20
5. Computational performance 24
6. Conclusions 24
Acknowledgements 25
References 25
A Distributed Lagrange Multipliers Based Fictitious Domain Method for the Numerical Simulation of Particulate Flow and its Parallel Implementation 26
1. A model problem 26
2. A fictitious domain formulation 28
3. Solving problem (18)-(25) by operator-splitting 30
4. Numerical Implementation and its Parallelization 31
5. Numerical results 31
6. Conclusion 34
Acknowledgments 34
References 35
Solving Fluid Flow Problems on Moving and Adaptive Overlapping Grids 36
1. Introduction 36
2. Multigrid 38
3. Solution of the reactive Euler Equations 39
4. Incompressible Flow 42
5. Parallel computations 43
6. Acknowledgments 44
References 44
Towards Optimal Performance for Lattice Boltzmann Applications on Terascale Computers 46
1. Introduction 46
2. Basics of the Lattice Boltzmann Method 47
3. Implementation and Optimization Strategies 48
4. Single Node Specification and Performance 48
5. Scalability and Parallel Performance 52
6. Conclusion 55
Acknowledgments 55
References 55
Moore’s Law, the Life Cycle of Scientific Computing Codes and the Diminishing Importance of Parallel Computing 56
1. Moore's law 56
2. The life cycle of scientific computing codes 58
3. Examples 60
4. Discussion 62
References 63
“Letting physicists be physicists,” and Other Goals of Scalable Solver Research 66
Introduction and motivation 66
Demand for scalable solvers 69
Detailed requirements specification 73
Solver toolchain 76
Illustrations from the TOPS solver project 77
Future of solver software 81
Conclusions 83
Acknowledgments 83
Bibliography 84
2. Turbulence 92
Investigation of Different Parallel Models for DNS of Turbulent Channel Flow 92
Abstract 92
1. Parallel Investigation of Channel DNS Code 92
2. Statistics and visualization of High Reynolds number simulation 95
3. Summary and Conclusion 96
Acknowledgments 96
References 96
Flow Simulation Around a Finite Cylinder on Massively Parallel Computer Architecture 100
1. Introduction 100
2. Configuration and Physics 101
3. Numerical Method 101
4. Suitability of Parallel Computers 102
5. Discretisation and Simulation Parameters 102
6. Results 103
7. Conclusion and Outlook 107
Acknowlegdements 107
References 107
Parallel Computation of Turbulent Fluid Flows with the Piecewise-Parabolic Method 108
1. Introduction 108
2. New features of PPM used in this work 109
3. SHMOD parallel code framework 109
4. Homogeneous, Mach 1 turbulence simulation on a 20483 grid 110
References 115
Parallel Large-Eddy Simulations of Turbulent Flows with Complex Moving Boundaries on Fixed Cartesian Grids 116
1. Introduction 116
2. Methodology 117
3. Parallelization 118
4. Results 118
5. Conclusions 122
References 123
Direct Numerical Simulation of Turbulent Flows on a Low Cost PC Cluster 124
1. Introduction 124
2. Overview of the Navier-Stokes solver 125
3. Direct Schur-Fourier decomposition 125
4. Parallel performance and illustrative results 129
5. Conclusions 130
References 131
Large Eddy Simulation of Turbulent Couette-Poiseuille Flows in a Square Duct 132
1. Introduction 132
2. Governing Equations and Modeling 133
3. Numerical Algorithms 133
4. Results 134
5. Conclusions 136
6. Acknowledgments 136
References 137
Development of a Framework for a Parallel Incompressible LES Solver Based on Free and Open Source Software 140
1. Solver characteristics 140
2. Framework characteristics 141
3. Solver status and Perspectives 146
4. Conclusions 147
References 147
3. Grid Computing 148
On a Fault Tolerant Algorithm for a Parallel CFD Application 148
1. Introduction 148
2. Fault Tolerant Algorithms 148
3. Reconstruction of the solution 149
4. Conclusion 155
References 155
Computational Fluid Dynamics Applications on TeraGrid 156
1. Introduction 156
2. Parallel computing on TeraGrid 157
3. Visualization on TeraGrid 158
4. Results 158
5. Conclusions 162
6. Acknowledgement 163
7. References 163
Mapping LSE Method on a Grid: Software Architecture and Performance Gains 164
1. Introduction and Motivations 164
2. LSE Method 165
3. Mapping LSE method onto a grid of computer 165
4. Realizing LSE method 166
5. Conclusion 170
References 171
Acceleration of Fully Implicit Navier-Stokes Solvers with Proper Orthogonal Decomposition on GRID Architecture 172
1. Motivations 172
2. Initial guess for the Newton solution with POD 173
3. POD-reduced model acceleration and its GRID applications 175
4. Numerical experiments on homogeneous and grid architecture 176
Conclusions 178
References 179
4. Software Frameworks and Component Architectures 180
A Gentle Migration Path to Component-Based Programming 180
1. Introduction 180
2. CCAIN Component Model 181
3. Summary 184
4. Future Work 185
Acknowledgements 185
PyNSol: A Framework for Interactive Development of High Performance Continuum Models 186
1. The PyNSol working environment 186
2. Motivation and requirements 187
3. Architecture 189
4. Innovations 192
5. Acknowledgements 193
References 193
The Exchange Grid: A Mechanism for Data Exchange Between Earth System Components on Independent Grids 194
1. Introduction 194
2. Definition of an exchange grid 196
3. Implicit coupling 198
4. Parallelization 199
5. Conclusions 200
Acknowledgments 201
References 201
A Generic Coupler for Earth System Models 202
1. Introduction 202
2. Model Coupling issues 203
3. A generic coupler 204
4. Discussion 206
5. Summary 207
6. Acknowledgement 207
References 207
CFD Analyses on Cactus PSE 210
1. Introduction 210
2. Cactus-based CFD analysis 211
3. Computational Supports for CFD Analyses 214
4. Conclusion 217
References 217
The Scalability Impact of a Component-Based Software Engineering Framework on a Growing SAMR Toolkit: a Case Study 218
1. Introduction 218
2. Software development scalability 220
3. Performance 221
4. Conclusions 224
5. Acknowledgments 225
References 225
5. BioFluids 226
Dynamics of Biological and Synthetic Polymers Through Large-Scale Parallel Computations 226
1. Introduction 226
2. Mathematical formulation 227
3. Relevant time scales 228
4. Parallel algorithm for polymer dynamics 228
5. Conclusions 230
References 230
Parallel Simulation of High Reynolds Number Vascular Flows 234
1. Introduction 234
2. Navier-Stokes Discretization 234
3. Boundary Conditions 236
4. Parallel Performance 238
5. Transition in an Arteriovenous Graft 239
Acknowledgments 240
References 240
6. Multiphysics and MHD 242
Large-Scale Multidisciplinary Computational Physics Simulations Using Parallel Multi-Zone Methods 242
Abstract 242
1. Introduction 242
2. Computer Framework 243
3. Parallel Approach 247
4. Representative Applications 249
References 249
Parallelization of Phase-Field Model for Phase Transformation Problems in a Flow Field 250
1. Introduction 250
2. Governing Equations of Phase-Field Model with Convection 251
3. Numerical Methods and Results 253
4. Approach to Parallelization and Results 254
5. Summary and Conclusions 256
6. Acknowledgements 257
References 257
A Parallel Unsplit Staggered Mesh Algorithm for Magnetohydrodynamics 258
1. Introduction 258
2. Numerical MHD 259
3. Numerical Results 261
4. Conclusion 264
References 264
Coupled Magnetogasdynamics – Radiative Transfer Parallel Computing Using Unstructured Meshes 266
Application of new RMHD numerical technologies to plasma physics studies 266
Governing system: 2 temperature MHD 267
The splitting scheme 268
Grids and discretization 269
Radiative transfer 270
Parallel implementation 271
Numerical results 273
References 274
7. Aerodynamics 276
Parallel Performance of a UKAAC Helicopter Code on HPCx and Other Large-Scale Facilities 276
1. Introduction 276
2. Computing facilities 277
3. Helicopter simulations and the ROTORMBMGP code 278
4. Profiling on HPCX and code optimizations 278
5. Benchmark timings on high performance computing architectures 281
6. PMB flow solver 282
7. Conclusions 283
Acknowledgements 283
Parallel Computations of Unsteady Aerodynamics and Flight Dynamics of Projectiles 284
1. Introduction 284
2. Computational methodology 285
3. Parallel computational issues 286
4. Results 288
5. Concluding remarks 290
References 291
Parallel Adaptive Solvers in Compressible PETSc-FUN3D Simulations 292
1. Introduction 292
2. Parallel compressible flow example 293
3. Adaptive solvers 295
4. Experimental results 296
5. Conclusions and future work 298
References 299
Numerical Simulation of Transonic Flows by a Double Loop Flexible Evolution 300
1. Introduction 300
2. The problem to solve 301
3. Partial grid sampling 302
4. Evolutionary algorithms as a meshless method 302
5. Flexible evolution agent 303
6. Double loop strategy with evolutionary algorithms 305
7. Parallel implementation 305
8. Conclusions and future work 306
References 306
Numerical Simulation of 2D Radiation Heat Transfer for Reentry Vehicles 308
1. Introduction 308
2. Numerical procedure 309
3. Database 310
4. Parallel computing 312
5. Numerical results 312
6. Concluding remarks 314
References 314
8. Parallel Algorithms and Solvers 316
Parallelism Results on Time Domain Decomposition for Stiff ODEs Systems 316
1. Introduction 316
2. Parallelization across the method 317
3. The Parareal algorithm 318
4. Adaptive Parareal: numerical and parallelism results 318
5. Conclusions and future works 322
References 323
Application of an Implicit Dual-Time Stepping Multi-Block Solver to 3D Unsteady Flows 324
1. Introduction 324
2. CFD method and parallelisation 325
3. Examples of 3D unsteady applications 325
4. Conclusions 330
Acknowledgements 330
References 330
Improving the Resolution of an Elliptic Solver for CFD Problems on the Grid 332
1. Introduction and Motivation 332
2. The subdomain solver tuning 333
3. Domain Decomposition Technique 333
4. Results 334
5. Conclusion 338
References 339
9. Structured AMR 340
FLASH: Applications and Future 340
1. Introduction 340
2. Architecture 341
3. Applications 343
4. Software Process 344
5. Summary and Conclusions 345
References 345
An Adaptive Cartesian Detonation Solver for Fluid-Structure Interaction Simulation on Distributed Memory Computers 348
1. Introduction 348
2. Governing Equations 349
3. Cartesian Finite Volume Scheme with Embedded Boundaries 349
4. Structured Adaptive Mesh Refinement 350
5. Fluid-Structure Coupling with SAMR 352
6. HMX Detonation in a Tantalum Cylinder 353
7. Conclusions 355
References 355
PARAMESH: A Parallel, Adaptive Grid Tool 356
1. Introduction to PARAMESH 356
2. Applications 359
References 363
10. General Fluid Dynamics 364
Embarrassingly Parallel Computations of Bubbly Wakes 364
Abstract 364
1 Introduction 364
2 Method 365
3 Results 367
4 Conclusions 367
Acknowledgments 368
References 368
Parallelisation of Inundation Simulations 372
1. Introduction 372
2. Numerical modelling of inundations 373
3. Parallelisation of the solver 374
4. Parallelisation of the inundation model 376
5. Summary and conclusions 378
References 379
Towards Numerical Modelling of Surface Tension of Microdroplets 380
1. Introduction 380
2. Governing equations 381
3. Tangent transformation 382
4. Calculation of interfacial curvature 382
5. The surface tension force in a collocated grid 383
6. Parallelisation strategy 384
7. Test cases 384
8. Conclusions 387
Acknowledgements 387
References 387
DNS Simulation of Sound Suppression in a Resonator with Upstream Flows 388
1. Introduction 388
2. Problem formulation 389
3. Numerical method 390
4. Parallel implementation 391
5. Numerical results 392
References 393
Performance Characterization and Scalability Analysis of a Chimera Based Parallel Navier-Stokes Solver on Commodity Clusters 396
1. Abstract 396
2. Introduction and Methodology 397
3. High order compact Scheme and Filtering 398
4. Parallel Algorithm 398
5. Results and Discussions 399
6. Conclusions 400
7. Acknowledgements 400
8. References 401
A Parallel Unstructured Implicit Solver for Hypersonic Reacting Flow Simulation 404
1. Introduction 404
2. Solver Implementation 405
3. Flux Evaluation Methods 405
4. Implicit Method of Solution 406
5. Partitioning and Parallelization 409
6. Parallel Performance of the Solver 409
References 410
11. Boundary Methods 412
Toward A MatlabMPI Parallelized Immersed Boundary Method 412
1. Algorithm 412
2. Computational cost 414
3. The Parallel algorithm 415
4. The IBM case 417
5. Conclusion 418
References 419
Dynamics of Multiphase Flows via Spectral Boundary Elements and Parallel Computations 420
1. Introduction 420
2. Mathematical formulation 421
3. Parallel interfacial algorithm 422
4. Conclusions 425
References 427
12. Parallel Tools and Load Balancing 428
SDLB - Scheduler with Dynamic Load Balancing for Heterogeneous Computers 428
Abstract 428
1. Introduction 428
2. System organization 429
3. Scheduling 430
4. Fault tolerance 431
5. Event logging system 432
6. Conclusion 434
References 434
Performance Evaluation of Adaptive Scientific Applications using TAU 436
1. Introduction 436
2. TAU Performance System 436
3. CFD Application Performance Mapping 437
4. Case Study: Uintah 439
5. Other frameworks 441
6. Conclusions 442
References 442
Computational Quality of Service in Parallel CFD 444
Abstract 444
1. Introduction 444
2. Motivating applications and algorithms 445
3. Computational quality of service for parallel CFD 446
4. Application example 449
5. Conclusions and future work 450
References 451
13. Combustion 452
Parallel Numerical Simulation of Flame Extinction and Flame Lift-Off 452
1. Introduction 452
2. LEMLES Formulation and Implementation 453
3. Efficient Computation of the Chemical Rates 454
4. Computational Performances 455
5. Flame Extinction 456
6. Conclusion 457
References 458
Parallel Computation of the Flow- and Acoustic Fields in a Gas Turbine Combustion Chamber 460
1. Introduction 460
2. Numerical methods 461
3. Results 463
4. Conclusions 466
5. Acknowledgements 466
References 467
14. Discrete Methods and Particles 468
Performance of Lattice Boltzmann Codes for Navier-Stokes and MHD Turbulence on High-End Computer Architectures 468
1. Lattice Boltzmann algorithm 468
2. HEC Platforms, Computational Implementation and Performance of LB 470
3. ELB simulations for 2D Navier-Stokes Turbulence 474
Acknowledgments 475
References 475
An Incompressible Navier-Stokes with Particles Algorithm and Parallel Implementation 476
1. Introduction 476
2. Problem Description 477
3. Projecting the Particle Force 478
4. Discretization of Advance 479
5. Evaluating PI(f) 480
6. AMR implementation 481
7. Convergence – Single Particle Settling 481
8. Parallel Performance – particle cloud with a vortex ring 481
References 483
Determination of Lubrication Characteristics of Bearings Using the Lattice Boltzmann Method 484
1. Introduction 484
2. Problem description 485
3. Numerical method 486
4. Results 487
5. Summary and conclusion 488
References 489
MPI-OpenMP Hybrid Parallel Computation in Continuous-Velocity Lattice-Gas Model 492
1. Introduction 492
2. Continuous-velocity lattice gas model 493
3. Parallel method 494
4. Results 495
5. Concluding Remarks 499
References 499
15. High Order Methods and Domain Decomposition 500
Parallel Atmospheric Modeling with High-Order Continuous and Discontinuous Galerkin Methods 500
1. Introduction 500
2. Shallow Water Equations 501
3. Space Discretization 501
4. Time Discretization 502
5. Numerical Experiments 502
6. Parallel Performance Results 503
7. Conclusions 503
Acknowledgements 504
References 504
A New Formulation of NUDFT Applied to Aitken-Schwarz DDM on Nonuniform Meshes 508
1. Motivations 508
2. Non Uniform Discrete Fourier Transform 509
3. Results on NUDFT approximation 511
4. Application in the Aitken-Schwarz DDM to solve Poisson problem 512
5. Perspectives and conclusions 514
References 515
16. Unstructured Grid Methods 516
Unstructured Mesh Processing in Parallel CFD Project GIMM 516
1. Introduction 516
2. General description 517
3. Numerical methods 517
4. Geometric models and meshing 518
5. The data treatment for the high-dimension meshes 520
6. Numerical results 521
7. Conclusion 523
References 523
Investigation of the Parallel Performance of the Unstructured DLR-TAU-Code on Distributed Computing Systems 524
1. Introduction 524
2. Performance evaluation method 525
3. Results of the benchmarks 528
4. Conclusion 530
Acknowledgments 531
References 531
17. Visualization 532
Parallel Visualization of CFD Data on Distributed Systems 532
1. Interactive vizualization 532
2. Scalar fields visualization 534
3. Parallel algorithm for 3D scalar datasets simplification 537
4. Vector fields visualization 539
5. Input/output and storing of mesh data 539
References 539