~grm08/fluidity/hilbert-for-pyop2 : revision 2249

1

<?xml version="1.0" encoding="UTF-8"?>

2

3

4

5

6

<a:documentation>Occasionally, it is desirable to apply operations or filters

7

to fields before using them for the purposes of adaptivity.</a:documentation>

8

9

<a:documentation>Invert a helmholtz operator to smooth out the field

10

before using it to adapt. This can help with noisy

11

fields.</a:documentation>

12

13

14

</element>

15

16

17

</element>

18

</element>

19

</element>

20

</optional>

21

</define>

22

23

24

25

26

27

<a:documentation>When specifying absolute measure

28

one specifies the absolute interpolation

29

error in the units of the field that is

30

being adapted, e.g. you can specify

31

the error to be 1.3 units </a:documentation>

32

33

<value>ADOPTT = 0</value>

34

</attribute>

35

36

37

38

</attribute>

39

40

<value>InterpolationErrorBound</value>

41

</attribute>

42

43

<value>ADWEIT</value>

44

</attribute>

45

46

47

</element>

48

</element>

49

50

51

<a:documentation>By default the interpolation error controlled is the L_inf

52

norm. Use this option to specify an alternative L_p norm. See

53

Chen Sun and Zu, Mathematics of Computation, Volume 76,

54

Number 257, January 2007, pp. 179-204.</a:documentation>

55

56

</element>

57

</optional>

58

</element>

59

60

<a:documentation>When specifying relative measure

61

one specifies the interpolation error

62

relative to the field that is

63

being adapted, e.g. you can specify

64

the error to be 5% (i.e. 0.05)</a:documentation>

65

66

<value>ADOPTT = 1</value>

67

</attribute>

68

69

70

71

</attribute>

72

73

<value>InterpolationErrorBound</value>

74

</attribute>

75

76

<value>ADADOT</value>

77

</attribute>

78

79

80

</element>

81

</element>

82

83

<a:documentation>The relative Hessian is calculated according to:

84

85

Q = H / max{ |psi|, psi_min}

86

87

where H is the Hessian, psi is the field value and

88

psi_min is the tolerance. The tolerance prevents

89

division by zero errors.

90

91

Source: Fluidity/ICOM manual draft version 1.2</a:documentation>

92

93

<value>ADATOT</value>

94

</attribute>

95

96

</element>

97

</element>

98

99

<a:documentation>Adapt using the anisotropic strategy of

100

Formaggia, Perotto, Micheletti.

101

Rather than taking two derivatives

102

and deriving the anisotropic information,

103

this approach computes an anisotropic Zienkiewicz-Zhu

104

error estimator for each element. The approach then

105

optimises the element orientation and length scales

106

to equidistribute the estimated error.</a:documentation>

107

108

<a:documentation>Tau is an anisotropic estimate for the H1 seminorm of the

109

error. This estimator is efficient and reliable, under the

110

caveat that the initial mesh is sufficiently fine so as to

111

prevent data oscillation. (Micheletti & Perotto, 2006)

112

Typically, tau will be ~= 6-8 * |e|_H1.</a:documentation>

113

114

</element>

115

</element>

116

</choice>

117

118

</element>

119

</optional>

120

</define>

121

122

123

124

<a:documentation>When specifying absolute measure

125

one specifies the absolute interpolation

126

error in the units of the field that is

127

being adapted, e.g. you can specify

128

the error to be 1.3 units </a:documentation>

129

130

<value>ADOPTT = 0</value>

131

</attribute>

132

133

134

135

</attribute>

136

137

<value>InterpolationErrorBound</value>

138

</attribute>

139

140

<value>ADWEIT</value>

141

</attribute>

142

143

144

</element>

145

</element>

146

147

148

<a:documentation>By default the interpolation error controlled is the L_inf

149

norm. Use this option to specify an alternative L_p norm. See

150

Chen Sun and Zu, Mathematics of Computation, Volume 76,

151

Number 257, January 2007, pp. 179-204.</a:documentation>

152

153

</element>

154

</optional>

155

</element>

156

157

<a:documentation>When specifying relative measure

158

one specifies the interpolation error

159

relative to the field that is

160

being adapted, e.g. you can specify

161

the error to be 5% (i.e. 0.05)</a:documentation>

162

163

<value>ADOPTT = 1</value>

164

</attribute>

165

166

167

168

</attribute>

169

170

<value>InterpolationErrorBound</value>

171

</attribute>

172

173

<value>ADADOT</value>

174

</attribute>

175

176

177

</element>

178

</element>

179

180

<a:documentation>The relative Hessian is calculated according to:

181

182

Q = H / max{ |psi|, psi_min}

183

184

where H is the Hessian, psi is the field value and

185

psi_min is the tolerance. The tolerance prevents

186

division by zero errors.

187

188

Source: Fluidity/ICOM manual draft version 1.2</a:documentation>

189

190

<value>ADATOT</value>

191

</attribute>

192

193

</element>

194

</element>

195

196

<a:documentation>Adapt using the anisotropic strategy of

197

Formaggia, Perotto, Micheletti.

198

Rather than taking two derivatives

199

and deriving the anisotropic information,

200

this approach computes an anisotropic Zienkiewicz-Zhu

201

error estimator for each element. The approach then

202

optimises the element orientation and length scales

203

to equidistribute the estimated error.</a:documentation>

204

205

<a:documentation>Tau is an anisotropic estimate for the H1 seminorm of the

206

error. This estimator is efficient and reliable, under the

207

caveat that the initial mesh is sufficiently fine so as to

208

prevent data oscillation. (Micheletti & Perotto, 2006)

209

Typically, tau will be ~= 6-8 * |e|_H1.</a:documentation>

210

211

</element>

212

</element>

213

</choice>

214

</define>

215

216

217

218

219

220

</element>

221

</optional>

222

</define>

223

224

225

226

<a:documentation>Adaptivity weights</a:documentation>

227

228

229

<a:documentation>When specifying absolute measure

230

one specifies the absolute interpolation

231

error in the units of the field that is

232

being adapted, e.g. you can specify

233

the error to be 1.3 units </a:documentation>

234

235

236

237

</attribute>

238

239

<value>InterpolationErrorBound</value>

240

</attribute>

241

242

<value>ADWEIU ADWEIV ADWEIW</value>

243

</attribute>

244

245

246

</element>

247

</element>

248

249

250

<a:documentation>By default the interpolation error controlled is the L_inf

251

norm. Use this option to specify an alternative L_p norm. See

252

Chen Sun and Zu, Mathematics of Computation, Volume 76,

253

Number 257, January 2007, pp. 179-204.</a:documentation>

254

255

</element>

256

</optional>

257

</element>

258

259

<a:documentation>When specifying relative measure

260

one specifies the interpolation error

261

relative to the field that is

262

being adapted, e.g. you can specify

263

the error to be 5% (i.e. 0.05)</a:documentation>

264

265

266

267

</attribute>

268

269

<value>InterpolationErrorBound</value>

270

</attribute>

271

272

<value>ADATOU ADATOV ADATOW</value>

273

</attribute>

274

275

276

</element>

277

</element>

278

279

<a:documentation>The relative Hessian is calculated according to:

280

281

Q = H / max{ |psi|, psi_min}

282

283

where H is the Hessian, psi is the field value and

284

psi_min is the tolerance. The tolerance prevents

285

division by zero errors.

286

287

Source: Fluidity/ICOM manual draft version 1.2</a:documentation>

288

289

</element>

290

</element>

291

</choice>

292

293

</element>

294

</optional>

295

</define>

296

297

298

299

<a:documentation>Adaptivity weights</a:documentation>

300

301

302

<a:documentation>When specifying absolute measure

303

one specifies the absolute interpolation

304

error in the units of the field that is

305

being adapted, e.g. you can specify

306

the error to be 1.3 units </a:documentation>

307

308

309

310

</attribute>

311

312

<value>InterpolationErrorBound</value>

313

</attribute>

314

315

<value>ADWEIU ADWEIV ADWEIW</value>

316

</attribute>

317

318

319

</element>

320

</element>

321

322

323

<a:documentation>By default the interpolation error controlled is the L_inf

324

norm. Use this option to specify an alternative L_p norm. See

325

Chen Sun and Zu, Mathematics of Computation, Volume 76,

326

Number 257, January 2007, pp. 179-204.</a:documentation>

327

328

</element>

329

</optional>

330

</element>

331

332

<a:documentation>When specifying relative measure

333

one specifies the interpolation error

334

relative to the field that is

335

being adapted, e.g. you can specify

336

the error to be 5% (i.e. 0.05)</a:documentation>

337

338

339

340

</attribute>

341

342

<value>InterpolationErrorBound</value>

343

</attribute>

344

345

<value>ADATOU ADATOV ADATOW</value>

346

</attribute>

347

348

349

</element>

350

</element>

351

352

<a:documentation>The relative Hessian is calculated according to:

353

354

Q = H / max{ |psi|, psi_min}

355

356

where H is the Hessian, psi is the field value and

357

psi_min is the tolerance. The tolerance prevents

358

division by zero errors.

359

360

Source: Fluidity/ICOM manual draft version 1.2</a:documentation>

361

362

</element>

363

</element>

364

</choice>

365

366

</element>

367

</optional>

368

</define>

369

370

371

372

<a:documentation>Adaptivity weights</a:documentation>

373

374

375

<a:documentation>When specifying absolute measure

376

one specifies the absolute interpolation

377

error in the units of the field that is

378

being adapted, e.g. you can specify

379

the error to be 1.3 units </a:documentation>

380

381

382

383

</attribute>

384

385

<value>InterpolationErrorBound</value>

386

</attribute>

387

388

389

</element>

390

</element>

391

392

393

<a:documentation>By default the interpolation error controlled is the L_inf

394

norm. Use this option to specify an alternative L_p norm. See

395

Chen Sun and Zu, Mathematics of Computation, Volume 76,

396

Number 257, January 2007, pp. 179-204.</a:documentation>

397

398

</element>

399

</optional>

400

</element>

401

402

<a:documentation>When specifying relative measure

403

one specifies the interpolation error

404

relative to the field that is

405

being adapted, e.g. you can specify

406

the error to be 5% (i.e. 0.05)</a:documentation>

407

408

409

410

</attribute>

411

412

<value>InterpolationErrorBound</value>

413

</attribute>

414

415

416

</element>

417

</element>

418

419

<a:documentation>The relative Hessian is calculated according to:

420

421

Q = H / max{ |psi|, psi_min}

422

423

where H is the Hessian, psi is the field value and

424

psi_min is the tolerance. The tolerance prevents

425

division by zero errors.

426

427

Source: Fluidity/ICOM manual draft version 1.2</a:documentation>

428

429

</element>

430

</element>

431

</choice>

432

433

</element>

434

</optional>

435

</define>

436

437

438

439

<a:documentation>Adaptivity weights</a:documentation>

440

441

442

<a:documentation>When specifying absolute measure

443

one specifies the absolute interpolation

444

error in the units of the field that is

445

being adapted, e.g. you can specify

446

the error to be 1.3 units </a:documentation>

447

448

449

450

</attribute>

451

452

<value>InterpolationErrorBound</value>

453

</attribute>

454

455

456

</element>

457

</element>

458

459

460

<a:documentation>By default the interpolation error controlled is the L_inf

461

norm. Use this option to specify an alternative L_p norm. See

462

Chen Sun and Zu, Mathematics of Computation, Volume 76,

463

Number 257, January 2007, pp. 179-204.</a:documentation>

464

465

</element>

466

</optional>

467

</element>

468

469

<a:documentation>When specifying relative measure

470

one specifies the interpolation error

471

relative to the field that is

472

being adapted, e.g. you can specify

473

the error to be 5% (i.e. 0.05)</a:documentation>

474

475

476

477

</attribute>

478

479

<value>InterpolationErrorBound</value>

480

</attribute>

481

482

483

</element>

484

</element>

485

486

<a:documentation>The relative Hessian is calculated according to:

487

488

Q = H / max{ |psi|, psi_min}

489

490

where H is the Hessian, psi is the field value and

491

psi_min is the tolerance. The tolerance prevents

492

division by zero errors.

493

494

Source: Fluidity/ICOM manual draft version 1.2</a:documentation>

495

496

</element>

497

</element>

498

</choice>

499

500

</element>

501

</optional>

502

</define>

503

504

505

<a:documentation>Anisotropic mesh hr-adaptivity</a:documentation>

506

507

<value>ADMESH</value>

508

</attribute>

509

510

511

<a:documentation>Time interval (in simulation time) when mesh adaptivity performed.

512

513

DO NOT SET THIS EQUAL TO OR LESS THAN ONE TIME-STEP SMALLER

514

THAN YOUR DUMP PERIOD BECAUSE ALL YOUR DIAGNOSTIC FIELDS WILL

515

APPEAR AS ZERO OR JUNK

516

DIAGNOSTIC FIELDS GET ZEROED AFTER AN ADAPT

517

518

Usually set to 10-20 times the timestep.</a:documentation>

519

520

<value>TIMMES</value>

521

</attribute>

522

523

</element>

524

525

<a:documentation>Adapt period in timesteps.</a:documentation>

526

527

</element>

528

</choice>

529

530

531

<a:documentation>Time interval (in cpu time) when mesh adaptivity performed

532

Manual suggests disabling this option.</a:documentation>

533

534

<value>CPUMES</value>

535

</attribute>

536

537

</element>

538

</optional>

539

540

541

<a:documentation>The minimum number of nodes this simulation may use.

542

In parallel, by default, this is the global minimum number of nodes.

543

544

If the mesh adaptivity algorithm wants to place fewer nodes

545

than this, the desired mesh is refined everywhere in space

546

until it will exceed this limit. This option should generally

547

only be used if a specified node count is being targetted.

548

Default value: 0</a:documentation>

549

550

551

<a:documentation>Define minimum_number_of_nodes to be the minimum number of

552

nodes per process (rather than the global minimum number of

553

nodes).</a:documentation>

554

555

</element>

556

</optional>

557

558

</element>

559

</optional>

560

561

<a:documentation>The maximum number of nodes this simulation may use.

562

In parallel, by default, this is the global maximum number of nodes.

563

564

If the mesh adaptivity algorithm wants to place more

565

nodes than this, the desired mesh is coarsened

566

everywhere in space until it will fit within this limit.

567

In general, the error tolerances should be set so that

568

this is never reached; it should only be a safety catch.

569

A typical value is 100000.

570

571

When using vertically_structured_adaptivity this indicates the

572

maximum number of nodes in the horizontal mesh, i.e. the number

573

of nodes in the full mesh will be much bigger and depends on the

574

number of layers specified, or if using inhomogenous_vertical_resolution

575

on the resolution produced by the vertical adaptivity step.</a:documentation>

576

577

<value>MXNODS</value>

578

</attribute>

579

580

581

<a:documentation>Define maximum_number_of_nodes to be the maximum number of

582

nodes per process (rather than the global maximum number of

583

nodes).</a:documentation>

584

585

</element>

586

</optional>

587

588

</element>

589

590

591

<a:documentation>The maximum ratio by which the number of nodes is allowed to

592

increase in an adapt. e.g., a value of 1.1 indicates that the

593

number of nodes may be increased by at most 10%.</a:documentation>

594

595

</element>

596

</optional>

597

598

599

<a:documentation>Enable to lock nodes in the mesh.</a:documentation>

600

601

<a:documentation>Python function defining nodes to lock. Return 0 for free

602

nodes, and non-zero for locked nodes. Functions should be

603

of the form:

604

605

def val(x, t):

606

# Function code

607

return # Return value

608

609

The return value must be an integer.</a:documentation>

610

611

</element>

612

613

</element>

614

</optional>

615

616

617

<a:documentation>Specifies the minimum element functional value for which elements

618

are considered for adaptivity. For the Pain et al 2001 functional,

619

ideal tetrahdera have a functional value of 0.0. A functional value

620

of 0.5 corresponds to a tetrehedron with: unit edge lengths and

621

and in-sphere radius of 0.3, or alternatively unit in-sphere radius,

622

five edges of unit length and a single edge of length 2

623

(all measured in metric space).

624

625

The minimum permitted value is 0.15 - the value supplied to

626

libadaptivity is max(abs(user value), 0.15).

627

628

Default value if not specified: 0.15</a:documentation>

629

630

<value>MESTP1</value>

631

</attribute>

632

633

</element>

634

</optional>

635

636

637

638

<a:documentation>Metric advection algorithm.

639

By advecting the metric with the flow velocity,

640

we can push mesh resolution ahead of the flow dynamics,

641

rather than always lagging behind.</a:documentation>

642

643

<a:documentation>Spatial discretisation options</a:documentation>

644

645

646

</element>

647

648

<a:documentation>Conservative discretisation of field advection equation

649

TBETA=1. -- conservative (divergence form)

650

TBETA=0. -- non-conservative

651

0. < TBETA < 1.</a:documentation>

652

653

<value>TBETA, DEFALT (TBETA = 0.0)</value>

654

</attribute>

655

656

</element>

657

</element>

658

659

660

<a:documentation>Implicit/explicit control (TTHETA)

661

=0. -- explicit

662

=0.5 -- Crank-Nicolson

663

=1. -- implicit</a:documentation>

664

665

<value>TTHETA, DEFALT (TTHETA = 0.5)</value>

666

</attribute>

667

668

</element>

669

670

671

<a:documentation>Use subcycling to advect the metric.

672

673

Specify the maximum courant number per subcycle.</a:documentation>

674

675

676

</element>

677

678

<a:documentation>Use subcycling to advect the metric.

679

680

Specify the number of subcycles.</a:documentation>

681

682

</element>

683

</choice>

684

685

686

<a:documentation>Scale the time period over which the metric is

687

advected by this factor.

688

689

Default is 1.1</a:documentation>

690

691

</element>

692

</optional>

693

694

<a:documentation>Temporal discretisation options for the control volume discretisation</a:documentation>

695

696

697

<a:documentation>Number of iterations within an advection solve.

698

This increases the accuracy of the face values and ensures that

699

the pivoted solution is cancelled out.

700

Defaults to 1 if unselected.</a:documentation>

701

702

<value>INT(ABS(NDISOT)/10)</value>

703

</attribute>

704

705

</element>

706

</optional>

707

708

709

<a:documentation>If not active then the theta specified above will be used.

710

Otherwise use variable limited theta on individual faces.</a:documentation>

711

712

<value>MOD(INT(ABS(NDISOT)/1),10) = 1,3,5,7,9 (odd)</value>

713

</attribute>

714

715

</element>

716

</optional>

717

718

719

<a:documentation>Time discretisation of upwind discretisation off which the

720

higher order solution is pivotted.

721

- pivot_theta = 1 - implicit pivot (default if not set and

722

best choice if not intentionally modifying

723

scheme to be explicit)

724

- pivot_theta = 0 - explicit pivot</a:documentation>

725

726

</element>

727

</optional>

728

</element>

729

</element>

730

731

732

<a:documentation>Solver</a:documentation>

733

734

</element>

735

736

<a:documentation>Assume this field is being solved explicitly and skip the solver.

737

738

Assumes lhs matrix only has diagonal lumped mass

739

and divides the rhs by this.</a:documentation>

740

741

</element>

742

</choice>

743

744

<a:documentation>Debugging output options</a:documentation>

745

746

747

<a:documentation>Output vtus of the advected metric and edge lengths at every subcycle</a:documentation>

748

749

</element>

750

</optional>

751

752

753

<a:documentation>Output vtus of the final metric merged over all the subcycles</a:documentation>

754

755

</element>

756

</optional>

757

</element>

758

</element>

759

</optional>

760

761

762

<a:documentation>Apply geometric constraints to the metric formation.

763

764

As specified in (Pain, 2001), the mesh adaptivity

765

scheme attempts to formulate an appropriate edge length

766

for each direction at each point in space, independent

767

of problem, PDE or domain.

768

769

This option instructs the error metric formation

770

code to inspect the boundaries of the domain

771

and to bound the edge lengths requested appropriately.

772

This procedure stops the metric from asking for edge lengths

773

that are inappropriately large in comparison to the

774

resolution required to preserve the geometric accuracy

775

of the boundaries.

776

777

If you get 'knife elements' near domain boundaries,

778

turn this on.</a:documentation>

779

780

</element>

781

</optional>

782

783

784

<a:documentation>Bounding box factor.

785

786

If the length scales specified by the metric are

787

unrealistically large, the mesh optimisation

788

algorithm can get confused. An example

789

would be specifying a length scale in a direction to be an

790

order of magnitude greater than the width

791

of the domain.

792

793

In order to fix this, the edge lengths requested

794

are bounded by the bounding box of the domain

795

(the smallest cuboid that contains the domain).

796

However, it was found that bounding by the bounding

797

box impairs the generation of anisotropic elements

798

in the mesh optimisation algorithm.

799

800

This option is multiplied by the bounding box of the domain

801

before it bounds the metric formed from other

802

considerations. By default, it is set to 2.0.</a:documentation>

803

804

</element>

805

</optional>

806

807

808

<a:documentation>Goal-based adaptivity.

809

810

With this option,

811

rather than taking the user-specified interpolation

812

error bounds as the weights to form the error metric

813

from the Hessians of the solution fields,

814

the interpolation weight is computed to optimally

815

represent the value of some specified functional

816

of state. In other words, the mesh is optimised

817

for the representation of a particular goal.

818

819

This is currently experimental. Activating

820

this option induces the code to ignore

821

any error bounds associated with the fields

822

specified under a material_phase.

823

824

For more information on this scheme,

825

see (Venditti & Darmofal, 2003), or

826

(Power et. al, 2006).

827

828

Coding your own goal is also possible

829

but currently undocumented. See

830

error_measures/Goals.F90

831

for examples.</a:documentation>

832

833

834

<a:documentation>Optimise for the representation of

835

enstrophy,

836

0.5 * int( |curl(velocity)|**2 ) dV.</a:documentation>

837

838

<value>goal_enstrophy</value>

839

</attribute>

840

841

<value>Velocity%1 Velocity%2 Velocity%3</value>

842

</attribute>

843

</element>

844

845

<a:documentation>Optimise for the representation of

846

gradients of temperature,

847

int(|grad(temperature)|**2) dV.</a:documentation>

848

849

850

</attribute>

851

852

<value>Temperature</value>

853

</attribute>

854

</element>

855

856

<a:documentation>Optimise for the contribution of the standard

857

Smagorinsky LES tensor.

858

859

In effect, this goal minimises

860

the contribution of the sub-filter scale

861

model -- it applies mesh resolution where

862

the sub-grid scale model has an effect.

863

864

int( transpose(grad(u)) . kappa . grad(u) ) dV,

865

with u ranging over the components of (nonlinear)

866

velocity and kappa the LES tensor.</a:documentation>

867

868

<value>goal_les_velocity</value>

869

</attribute>

870

871

<value>NonlinearVelocity%1 NonlinearVelocity%2 NonlinearVelocity%3</value>

872

</attribute>

873

874

875

<a:documentation>The number of nonlinear iterations to perform when

876

forming the metric with this goal.

877

878

Because the LES tensor explicitly depends on mesh sizing,

879

we can form a metric and feed this back into the goal,

880

using the proposed mesh sizing instead of the current mesh.

881

This allows us to simulate adapts without actually incurring

882

the cost of adaptation, for the purposes of converging the metric.

883

The default value is 3.</a:documentation>

884

885

</element>

886

</optional>

887

</element>

888

889

<a:documentation>Optimise for the contribution of the new 4th-order

890

LES tensor.

891

892

In effect, this goal minimises

893

the contribution of the sub-filter scale

894

model -- it applies mesh resolution where

895

the sub-grid scale model has an effect.

896

897

int( transpose(grad(u)) . kappa . grad(u) ) dV -

898

int( transpose(grad_h(u)) . kappa . grad_h(u) ) dV

899

900

with u ranging over the components of (nonlinear)

901

velocity, kappa the LES tensor,

902

grad(.) differentiation of basis functions and

903

grad_h(.) the Galerkin projection of the first derivative.</a:documentation>

904

905

<value>goal_les_velocity_4th</value>

906

</attribute>

907

908

<value>NonlinearVelocity%1 NonlinearVelocity%2 NonlinearVelocity%3</value>

909

</attribute>

910

911

912

<a:documentation>The number of nonlinear iterations to perform when

913

forming the metric with this goal.

914

915

Because the LES tensor explicitly depends on mesh sizing,

916

we can form a metric and feed this back into the goal,

917

using the proposed mesh sizing instead of the current mesh.

918

This allows us to simulate adapts without actually incurring

919

the cost of adaptation, for the purposes of converging the metric.

920

The default value is 3.</a:documentation>

921

922

</element>

923

</optional>

924

</element>

925

</choice>

926

927

928

<a:documentation>The tolerance of the goal specifies the acceptable

929

error in the quantity computed. The adaptation scheme

930

attempts to adapt the mesh to ensure that the

931

goal computed from the primitive solution is

932

within the tolerance specified here.

933

934

A relative tolerance specifies that the acceptable error

935

in the goal is some fraction of the value as computed

936

from the primitive solution. It is generally

937

the easiest to use. This is a unitless percentage.</a:documentation>

938

939

</element>

940

941

<a:documentation>The tolerance of the goal specifies the acceptable

942

error in the quantity computed. The adaptation scheme

943

attempts to adapt the mesh to ensure that the

944

goal computed from the primitive solution is

945

within the tolerance specified here.

946

947

An absolute tolerance specifies the acceptable error

948

in the goal, in the units of the goal itself.</a:documentation>

949

950

</element>

951

</choice>

952

</element>

953

</optional>

954

955

<a:documentation>Mesh size constraints: the minimum edge length bound.</a:documentation>

956

957

<value>MinimumEdgeLengths</value>

958

</attribute>

959

960

961

</element>

962

</element>

963

964

<a:documentation>Mesh size constraints: the maximum edge length bound.</a:documentation>

965

966

<value>MaximumEdgeLengths</value>

967

</attribute>

968

969

970

</element>

971

</element>

972

973

974

<a:documentation>Supply a reference mesh with which to bound the metric</a:documentation>

975

976

977

</attribute>

978

979

980

</attribute>

981

982

983

<a:documentation>Use this reference mesh as a bound on the minimum edge

984

length of the metric</a:documentation>

985

986

</element>

987

988

<a:documentation>Use this reference mesh as a bound on the maximum edge

989

length of the metric</a:documentation>

990

991

</element>

992

</choice>

993

994

</element>

995

</zeroOrMore>

996

997

998

<a:documentation>Maximum aspect ratio in the adapted mesh.</a:documentation>

999

1000

</element>

1001

</optional>

1002

1003

1004

<a:documentation>Adapt at first timestep</a:documentation>

1005

1006

<a:documentation>Number of adapts done after initialisation but

1007

before the actual simulation starts</a:documentation>

1008

1009

</element>

1010

1011

1012

<a:documentation>Write out the first timestep adapted mesh.

1013

This is useful when needing to re-run simulations

1014

without waiting for the first timestep adapt</a:documentation>

1015

1016

</element>

1017

</optional>

1018

</element>

1019

</optional>

1020

1021

1022

<a:documentation>Enable this option to preserve any regions in your

1023

mesh (i.e. those specified by region_ids).

1024

Also, any prescribed fields using region_ids will be

1025

reinitialised using them on the new mesh.

1026

1027

Therefore this is a required option if you want your

1028

prescribed region_id fields to survive adapts!

1029

Obviously this does not apply to initial conditions

1030

set using region_ids.</a:documentation>

1031

1032

</element>

1033

</optional>

1034

1035

1036

<a:documentation>Vertically structured adaptivity.

1037

1038

The mesh will be unstructured in the horizontal, but columnar

1039

in the vertical.

1040

To enable this, your meshes must be derived by extrusion

1041

from a lower-dimensional horizontal mesh.

1042

This will give a columnar, layered mesh.</a:documentation>

1043

1044

1045

<a:documentation>If this is enabled, the resolution along each column

1046

will be computed from the error metric.

1047

This will give a columnar mesh, but not a layered one.</a:documentation>

1048

1049

1050

<a:documentation>Ignore the horizontal adapt, and *only* adapt in the vertical</a:documentation>

1051

1052

</element>

1053

</optional>

1054

</element>

1055

</optional>

1056

1057

1058

<a:documentation>Apply a separate step of horizontal (and vertical, if using inhomogeneous_vertical_resolution)

1059

gradation to the horizontal (and vertical) metric.

1060

This replaces is the gradation applied to the full metric and uses the gradation

1061

algorithm specified above.</a:documentation>

1062

1063

</element>

1064

</optional>

1065

1066

1067

<a:documentation>Remove components from each field's metric that aren't aligned either horizontally

1068

or vertically. This can help prevent the horizontal metric from becoming

1069

contaminated by parts of the vertical metric.

1070

1071

NOTE: Vertical is defined using the GravityDirection field. This is unlike some other

1072

parts of the vertically_structured_adaptivity and extrusion routines where vertical is defined as the

1073

last dimension.</a:documentation>

1074

1075

</element>

1076

1077

<a:documentation>When separating the horizontal metric from the vertical use the full metric. This

1078

can lead to more constrained horizontal resolution due to leakage between the vertical

1079

and horizontal components.</a:documentation>

1080

1081

</element>

1082

</choice>

1083

1084

1085

<a:documentation>When constructing the horizontal metric incorporate the components of the

1086

full metric tangential to the bottom boundary. For example, this is useful

1087

when horizontal contours of a field intersect the bathymetry and this information

1088

is not automatically incorporated into the horizontal metric leading to

1089

the contact point being underresolved.

1090

1091

Depends on /geometry/ocean_boundaries.</a:documentation>

1092

1093

</element>

1094

</optional>

1095

</element>

1096

</optional>

1097

1098

1099

<a:documentation>Zoltan Options

1100

1101

Set up some Zoltan options, including which partitioner to use.

1102

Only works if you have the --with-zoltan option in your configure.

1103

1104

Note that Zoltan will become the default (replacing Sam at some point)

1105

and these options are currently here only for development and testing

1106

purposes. They are subject to change and may not be in the final "release".</a:documentation>

1107

1108

1109

<a:documentation>Select which partitioner to use. Graph partitioners available are ParMETIS

1110

and Zoltan PHG. The Zoltan PHG hypergraph partitoner is also available.</a:documentation>

1111

1112

1113

<a:documentation>Use the ParMETIS graph partitioner. ParMETIS setup to match as

1114

closely as possible the setup used previously by Sam.</a:documentation>

1115

1116

</element>

1117

1118

<a:documentation>Use the PT-Scotch graph partitioner. </a:documentation>

1119

1120

</element>

1121

1122

<a:documentation>Use the Zoltan PHG partitioner.</a:documentation>

1123

1124

<a:documentation>Select the partitioning method you would like used by Zoltan PHG.

1125

Currently hypergraph partitioning is the simplest implementation

1126

and can produce non-contiguous partitions for certain problems.</a:documentation>

1127

1128

<value>graph</value>

1129

<value>hypergraph</value>

1130

</choice>

1131

</element>

1132

</element>

1133

</choice>

1134

</element>

1135

</optional>

1136

1137

1138

<a:documentation>Element quality cut off.

1139

Elements with a quality greater than this will be liable to be

1140

on the parition boundary. Below this, they will be free to be adapted

1141

and improved. Range is 0 to 1, with default 0.6</a:documentation>

1142

1143

</element>

1144

</optional>

1145

1146

1147

<a:documentation>Zoltan Debugging

1148

1149

Turn on more verbose output for use when debugging Zoltan.</a:documentation>

1150

1151

1152

<a:documentation>Turn on graph checking.

1153

When using ParMETIS or PT-Scotch options for turning on

1154

graph checking are provided by Zoltan.

1155

1 - on process checking,

1156

2 - full checking (very slow)</a:documentation>

1157

1158

1159

1160

</choice>

1161

</element>

1162

</optional>

1163

1164

1165

<a:documentation>Print out a dump file of the edge counts.

1166

Edge counts for each owned node are calculated in zoltan_cb_get_num_edges.

1167

This option dumps the edge count for each owned node.

1168

Dump is to the current directory, in a file called edge_counts_*.dat

1169

One dump file is created for each rank.

1170

Will get overwritten at each adapt.</a:documentation>

1171

1172

</element>

1173

</optional>

1174

1175

1176

<a:documentation>Print out a dump file of the edge weights.

1177

Edge weights are calculated for each edge in zoltan_cb_get_edge_list.

1178

Permits a visualisation of how the edge weights are distributed.

1179

Dump is to the current directory, in a file called edge_weights_*.dat

1180

One dump file is created for each rank.

1181

Will get overwritten at each adapt.</a:documentation>

1182

1183

</element>

1184

</optional>

1185

1186

1187

<a:documentation>Print out a dump file of node sizes.

1188

Zoltan needs to be told how much data is associated with each node when

1189

doing phase one migration.

1190

Here we dump the size calculated by zoltan_cb_pack_node_sizes for each

1191

owned node.

1192

Dump is to the current directory, in a file called node_sizes_*.dat

1193

One dump file is created for each rank.</a:documentation>

1194

1195

</element>

1196

</optional>

1197

1198

1199

<a:documentation>Print out a dump file of halo node sizes.

1200

Zoltan needs to be told how much data is associated with each halo node when

1201

doing phase two migration.

1202

Here we dump the size calculated by zoltan_cb_pack_halo_node_sizes for each

1203

owned node.

1204

Dump is to the current directory, in a file called halo_node_sizes_*.dat

1205

One dump file is created for each rank.</a:documentation>

1206

1207

</element>

1208

</optional>

1209

1210

1211

<a:documentation>Print out a dump file of field sizes.

1212

Zoltan needs to be told how much data is associated with the fields for each

1213

element when we're transfering fields.

1214

Here we dump the size calculated by zoltan_cb_pack_field_sizes for each

1215

owned node.

1216

Dump is to the current directory, in a file called field_sizes_*.dat

1217

One dump file is created for each rank.</a:documentation>

1218

1219

</element>

1220

</optional>

1221

</element>

1222

</optional>

1223

</element>

1224

</optional>

1225

1226

1227

<a:documentation>Select the adaptivity library used by hr-adaptivity. If disabled,

1228

the defaults are:

1229

In 3D: libadaptivity

1230

In 2D: libmba2d

1231

In 1D: adaptivity_1d</a:documentation>

1232

1233

1234

<a:documentation>libadaptivity. 3D, parallelised.</a:documentation>

1235

1236

1237

<a:documentation>The number of adaptivity sweeps. Default value: 10</a:documentation>

1238

1239

</element>

1240

</optional>

1241

1242

1243

<a:documentation>Enable this option to turn off edge splitting</a:documentation>

1244

1245

</element>

1246

</optional>

1247

1248

1249

<a:documentation>Enable this option to turn off edge collapsing</a:documentation>

1250

1251

</element>

1252

</optional>

1253

1254

1255

<a:documentation>Enable this option to turn off edge/face, face/edge and

1256

edge/edge swapping</a:documentation>

1257

1258

</element>

1259

</optional>

1260

1261

1262

<a:documentation>Enable this option to turn off node movement, and use

1263

h-adaptivity only.</a:documentation>

1264

1265

</element>

1266

</optional>

1267

1268

1269

<a:documentation>Writes vtus containing the Pain 2001 P0 element functionals

1270

of the adapted meshes.</a:documentation>

1271

1272

</element>

1273

</optional>

1274

1275

</element>

1276

1277

<a:documentation>libmba2d. 2D. A testing parallel implementation is available.</a:documentation>

1278

1279

1280

<a:documentation>Desired output mesh quality, 0 <= quality <= 1.

1281

Default value 0.6.</a:documentation>

1282

1283

</element>

1284

</optional>

1285

1286

</element>

1287

1288

<a:documentation>Re-uses 1D adaptivity code from 2+1D adaptivity. 1D, serial only.</a:documentation>

1289

1290

</element>

1291

1292

<a:documentation>libmba3d. 3D, serial only.</a:documentation>

1293

1294

1295

<a:documentation>Desired output mesh quality, 0 <= quality <= 1.

1296

Default value 0.6.</a:documentation>

1297

1298

</element>

1299

</optional>

1300

1301

1302

<a:documentation>Maximum number of mesh optimisations. Default value 10^5.</a:documentation>

1303

1304

</element>

1305

</optional>

1306

1307

</element>

1308

</choice>

1309

</element>

1310

</optional>

1311

1312

1313

<a:documentation>hr adptivity debugging options</a:documentation>

1314

1315

1316

<a:documentation>Write out error metric at each stage of the processing

1317

pipeline. This can be very useful in diagnosing why

1318

adaptivity is doing something you don't expect.</a:documentation>

1319

1320

</element>

1321

</optional>

1322

1323

1324

<a:documentation>Write out the Coordinate field to a triangle mesh for every

1325

state adapt. In parallel, a triangle mesh and .halo files

1326

will be written for every parallel adapt iteration.</a:documentation>

1327

1328

</element>

1329

</optional>

1330

1331

1332

<a:documentation>Write out the system state to a vtu adapt every state adapt.

1333

In parallel, a vtu will be written for every parallel

1334

adapt iteration.</a:documentation>

1335

1336

</element>

1337

</optional>

1338

1339

1340

<a:documentation>Checkpoint the simulation after every adapt. In parallel, a

1341

checkpoint will be written only after the final adapt

1342

iteration. Checkpoints are postfixed with "adapt_checkpoint".</a:documentation>

1343

1344

1345

<a:documentation>Number of checkpoints to write before overwriting existing

1346

checkpoints.</a:documentation>

1347

1348

</element>

1349

</optional>

1350

1351

</element>

1352

</optional>

1353

1354

</element>

1355

</optional>

1356

</element>

1357

</define>

1358

1359

1360

<a:documentation>Mesh adaptivity, with prescribed adapt interval and target meshes.

1361

Serial only.</a:documentation>

1362

1363

<a:documentation>Options relating the the frequency of mesh adaptivity</a:documentation>

1364

1365

<a:documentation>Python code defining whether to adapt the mesh, evaluated at the

1366

end of each timestep. Return non-zero to signal a mesh adapt, and

1367

zero otherwise. Functions should be of the form:

1368

1369

def val(t):

1370

# Function code

1371

return # Return value

1372

1373

The return value must be an integer.</a:documentation>

1374

1375

</element>

1376

1377

</element>

1378

1379

<a:documentation>Options relating to the target meshes</a:documentation>

1380

1381

<a:documentation>The target mesh. If not reading from file, this must be a mesh

1382

specified under /geometry.</a:documentation>

1383

1384

<a:documentation>Python code defining the target mesh. Functions should be of

1385

the form:

1386

1387

def val(t):

1388

# Function code

1389

return # Return value

1390

1391

The return value must be a string.</a:documentation>

1392

1393

</element>

1394

1395

</element>

1396

1397

1398

<a:documentation>Read the mesh from file, rather than extracting it from the

1399

system state.</a:documentation>

1400

1401

<a:documentation>Input mesh file format.</a:documentation>

1402

<value>triangle</value>

1403

1404

</element>

1405

1406

</element>

1407

</optional>

1408

1409

</element>

1410

1411

</element>

1412

</define>

1413

1414

1415

<a:documentation>Basis function interpolation.

1416

The standard algorithm. It is quick

1417

and bounded, but non-conservative and dissipative.

1418

All other algorithms require construction of a supermesh.</a:documentation>

1419

1420

</element>

1421

</define>

1422

1423

1424

<a:documentation>Basis function pseudo-interpolation, with averaging for nodes on

1425

element boundaries. This can accept a discontinuous input, but always

1426

gives a continuous output. It is strong recommended that you use

1427

interpolation_galerkin instead of this interpolation method.

1428

1429

The distance in ideal space from the boundary where averaging is

1430

applied is currently hard coded to 1.0e3 * epsilon(0.0).</a:documentation>

1431

1432

</element>

1433

</define>

1434

1435

1436

<a:documentation>Grandy interpolation. Conservative, but highly diffusive.

1437

See doi:10.1006/jcph.1998.6125 .</a:documentation>

1438

1439

</element>

1440

</define>

1441

1442

1443

<a:documentation>Disable interpolation</a:documentation>

1444

1445

</element>

1446

</define>

1447

1448

1449

1450

1451

1452

<a:documentation>Galerkin projection. By default, conservative, non-dissipative and

1453

non-bounded. The most accurate choice, in the sense of minimising

1454

the L2 norm of the residual</a:documentation>

1455

1456

</element>

1457

1458

</choice>

1459

</define>

1460

1461

1462

</define>

1463

1464

1465

</define>

1466

1467

1468

</define>

1469

1470

1471

</define>

1472

1473

1474

1475

1476

1477

<a:documentation>Galerkin projection. By default, conservative, non-dissipative and

1478

non-bounded. The most accurate choice, in the sense of minimising

1479

the L2 norm of the residual</a:documentation>

1480

1481

</element>

1482

1483

1484

<a:documentation>Helmholtz decomposed projection of the Coriolis acceleration. Suitable

1485

only for Velocity fields. This interpolation happens in three stages:

1486

1487

1. Computation of Coriolis and its Helmholtz decomposition

1488

2. Interpolation of the Helmholtz decomposition

1489

3. Formation of Coriolis from the decomposition and inversion for velocity

1490

1491

Notes for balance preserving interpolants:

1492

1493

The spatial discretisation options for the conservative potential

1494

must match those used for Pressure.

1495

1496

With weak boundary conditions for the conservative potential, if

1497

no-normal-flow is satisfied on the boundary this must be preserved by

1498

the interpolation. For 2D domains this can be achieved by using

1499

consistent interpolation for the conservative potential or, for more

1500

general interpolants, by performing a further decomposition of the

1501

conservative potential (see

1502

geostrophic_interpolation/conservative_potential/decompose).

1503

1504

For shallow-water modelling the interpolants for layer thickness and

1505

the conservative potential must be identical and degree one

1506

homogenenous (see

1507

geostrophic_interpolation/conservative_potential/project_pressure/scale_factor).</a:documentation>

1508

1509

<a:documentation>Options relating to the Coriolis acceleration</a:documentation>

1510

1511

1512

1513

<a:documentation>The mesh used for the Coriolis acceleration. Defaults to the

1514

Velocity mesh if not supplied.</a:documentation>

1515

1516

<value>VelocityMesh</value>

1517

</attribute>

1518

</element>

1519

1520

<a:documentation>The mesh used for the Coriolis acceleration. Defaults to the

1521

Velocity mesh if not supplied.</a:documentation>

1522

1523

<value>PressureMesh</value>

1524

</attribute>

1525

</element>

1526

1527

<a:documentation>The mesh used for the Coriolis acceleration. Defaults to the

1528

Velocity mesh if not supplied.</a:documentation>

1529

1530

<value>CoordinateMesh</value>

1531

</attribute>

1532

</element>

1533

1534

<a:documentation>The mesh used for the Coriolis acceleration. Defaults to the

1535

Velocity mesh if not supplied.</a:documentation>

1536

1537

1538

</attribute>

1539

</element>

1540

</choice>

1541

</optional>

1542

1543

<a:documentation>Options relating to the diagnostic solve for the Coriolis

1544

acceleration from Velocity on the donor mesh.</a:documentation>

1545

1546

1547

</optional>

1548

1549

1550

<a:documentation>Lump the RHS term. Requires Velocity and the Coriolis

1551

acceleration to be on the same mesh.</a:documentation>

1552

1553

</element>

1554

</optional>

1555

1556

</element>

1557

1558

<a:documentation>Options relating to the diagnostic solve for Velocity from

1559

the Coriolis acceleration on the target mesh.</a:documentation>

1560

1561

1562

</optional>

1563

1564

1565

<a:documentation>Lump the RHS term. Requires Velocity and the Coriolis

1566

acceleration to be on the same mesh.</a:documentation>

1567

1568

</element>

1569

</optional>

1570

1571

</element>

1572

1573

</element>

1574

1575

<a:documentation>Options relating to the conservative potential component of the

1576

Helmholtz decomposition</a:documentation>

1577

1578

1579

<a:documentation>The mesh used for the conservative potential. Note that this is

1580

computed using the same method as the pressure projection, and

1581

hence LBB constraints apply.</a:documentation>

1582

1583

<value>PressureMesh</value>

1584

</attribute>

1585

</element>

1586

1587

<a:documentation>The mesh used for the conservative potential. Note that this is

1588

computed using the same method as the pressure projection, and

1589

hence LBB constraints apply.</a:documentation>

1590

1591

<value>VelocityMesh</value>

1592

</attribute>

1593

</element>

1594

1595

<a:documentation>The mesh used for the conservative potential. Note that this is

1596

computed using the same method as the pressure projection, and

1597

hence LBB constraints apply.</a:documentation>

1598

1599

<value>CoordinateMesh</value>

1600

</attribute>

1601

</element>

1602

1603

<a:documentation>The mesh used for the conservative potential. Note that this is

1604

computed using the same method as the pressure projection, and

1605

hence LBB constraints apply.</a:documentation>

1606

1607

1608

</attribute>

1609

</element>

1610

</choice>

1611

1612

<a:documentation>Spatial discretisation options</a:documentation>

1613

1614

<a:documentation>Options relating to the mass matrix</a:documentation>

1615

1616

1617

<a:documentation>Lump the mass matrix. Required for continuous fields.</a:documentation>

1618

1619

</element>

1620

</optional>

1621

</element>

1622

1623

<a:documentation>Use a continuous Galerkin discretisation</a:documentation>

1624

1625

1626

<a:documentation>Integrate the divergence operator by parts</a:documentation>

1627

1628

</element>

1629

</optional>

1630

1631

1632

<a:documentation>Remove the stabilisation term from the projection operator.

1633

1634

Automatic when not using P1P1.</a:documentation>

1635

1636

</element>

1637

</optional>

1638

1639

</element>

1640

1641

</element>

1642

1643

1644

<a:documentation>Reference node, at which the solution value is pinned to zero</a:documentation>

1645

1646

</element>

1647

</optional>

1648

1649

<a:documentation>Solver options for the conservative potential calculation</a:documentation>

1650

1651

</element>

1652

1653

1654

<a:documentation>Galerkin projection. By default, conservative, non-dissipative and

1655

non-bounded. The most accurate choice, in the sense of minimising

1656

the L2 norm of the residual</a:documentation>

1657

1658

1659

1660

1661

</optional>

1662

1663

</element>

1664

1665

1666

</choice>

1667

1668

1669

1670

<a:documentation>Supply a Pressure field. This enables better initial guesses for

1671

the decomposition solvers, and also allows decomposed interpolants

1672

for Pressure (see

1673

geostrophic_interpolation/conservative_potential/decompose and

1674

geostrophic_interpolation/conservative_potential/interpolate_boundary).</a:documentation>

1675

1676

<value>Pressure</value>

1677

</attribute>

1678

1679

1680

</optional>

1681

1682

</element>

1683

1684

<a:documentation>Supply a Pressure field. This enables better initial guesses for

1685

the decomposition solvers, and also allows decomposed interpolants

1686

for Pressure (see

1687

geostrophic_interpolation/conservative_potential/decompose and

1688

geostrophic_interpolation/conservative_potential/interpolate_boundary).</a:documentation>

1689

1690

<value>LayerThickness</value>

1691

</attribute>

1692

1693

1694

</element>

1695

1696

<a:documentation>Supply a Pressure field. This enables better initial guesses for

1697

the decomposition solvers, and also allows decomposed interpolants

1698

for Pressure (see

1699

geostrophic_interpolation/conservative_potential/decompose and

1700

geostrophic_interpolation/conservative_potential/interpolate_boundary).</a:documentation>

1701

1702

1703

</attribute>

1704

1705

1706

</optional>

1707

1708

</element>

1709

</choice>

1710

</optional>

1711

1712

1713

1714

<a:documentation>Decompose the conservative potential into a component constant

1715

on the boundary, and a residual.

1716

1717

If interpolating Pressure, a similar decomposition is applied

1718

to the Pressure projection.

1719

1720

This requires the domain to be 2D and simply connected.</a:documentation>

1721

1722

1723

<a:documentation>Choose a boundary value that minimises the l2 norm of the

1724

residual</a:documentation>

1725

1726

</element>

1727

1728

<a:documentation>Use the mean boundary value</a:documentation>

1729

1730

</element>

1731

</choice>

1732

1733

<a:documentation>Solver options for the decomposition</a:documentation>

1734

1735

</element>

1736

1737

</element>

1738

1739

<a:documentation>Interpolate the boundary values using consistent interpolation and

1740

use these as a strong Dirichlet boundary condition on the

1741

conservative potential.

1742

1743

If interpolating Pressure, a similar boundary condition is applied

1744

to the Pressure projection.</a:documentation>

1745

1746

</element>

1747

</choice>

1748

</optional>

1749

1750

</element>

1751

1752

<a:documentation>Options relating to the non-conservative residual component of the

1753

Helmholz decomposition</a:documentation>

1754

1755

1756

<a:documentation>Galerkin projection. By default, conservative, non-dissipative and

1757

non-bounded. The most accurate choice, in the sense of minimising

1758

the L2 norm of the residual</a:documentation>

1759

1760

1761

1762

</optional>

1763

1764

</element>

1765

1766

1767

1768

</choice>

1769

1770

1771

<a:documentation>Enforce divergence free after the projection</a:documentation>

1772

1773

</element>

1774

</optional>

1775

1776

</element>

1777

1778

1779

<a:documentation>If enabled, preconditions the Helmholtz decomposition by solving

1780

for the conservative potential using a geopressure solver. The

1781

projection equation then becomes:

1782

M f = M f_* + C \phi + C_{gp} \phi_{gp},

1783

where f_* is the coriolis acceleration, f is divergence free, phi is

1784

the conservative potential, phi_gp is the geopressure conservative

1785

potential and:

1786

C_{gp,ij}^q = \int_Omega N_j \partial_q M_i,

1787

is the geopressure gradient matrix, where N_i are the velocity shape

1788

functions and M_i the geopressure conservative potential shape

1789

functions.</a:documentation>

1790

1791

<a:documentation>The mesh used for the geopressure conservative potential</a:documentation>

1792

1793

1794

</attribute>

1795

</element>

1796

1797

1798

<a:documentation>Reference node, at which the solution value is pinned to zero</a:documentation>

1799

1800

</element>

1801

</optional>

1802

1803

<a:documentation>Solver options for the geopressure conservative potential

1804

calculation</a:documentation>

1805

1806

</element>

1807

1808

1809

<a:documentation>Galerkin projection. By default, conservative, non-dissipative and

1810

non-bounded. The most accurate choice, in the sense of minimising

1811

the L2 norm of the residual</a:documentation>

1812

1813

1814

1815

</optional>

1816

1817

</element>

1818

1819

1820

1821

</choice>

1822

</element>

1823

</optional>

1824

1825

1826

<a:documentation>Options relating to the vertical velocity. Required in 3D.</a:documentation>

1827

1828

1829

<a:documentation>Galerkin projection. By default, conservative, non-dissipative and

1830

non-bounded. The most accurate choice, in the sense of minimising

1831

the L2 norm of the residual</a:documentation>

1832

1833

1834

1835

</optional>

1836

1837

</element>

1838

1839

1840

1841

</choice>

1842

1843

</element>

1844

</optional>

1845

1846

1847

<a:documentation>Debug options</a:documentation>

1848

1849

1850

<a:documentation>If enabled, pre and post interpolation decomposition vtus are

1851

written</a:documentation>

1852

1853

1854

<a:documentation>Maximum number of debug vtus that will be written before

1855

over-writing existing vtus</a:documentation>

1856

1857

</element>

1858

</optional>

1859

1860

</element>

1861

</optional>

1862

1863

</element>

1864

</optional>

1865

1866

</element>

1867

</choice>

1868

</define>

1869

1870

1871

<a:documentation>Scale the pressure field by some factor before interpolation,

1872

and apply the inverse after interpolation. This should be set

1873

if the pressure field is divided by some reference value

1874

e.g., in a shallow water with gravity magnitude g, this should

1875

take the value g. This enables better initial guesses for the

1876

decomposition solvers, and means that non degree one

1877

homonogeneous interpolants can be used for the conservative

1878

potential and pressure, while still being balance preserving.</a:documentation>

1879

1880

</element>

1881

</define>

1882

1883

1884

<a:documentation>Honour strong Dirichlet boundary conditions in the Galerkin projection</a:documentation>

1885

1886

</element>

1887

</define>

1888

1889

1890

1891

1892

</optional>

1893

1894

1895

</optional>

1896

1897

1898

</optional>

1899

</define>

1900

1901

1902

1903

1904

</optional>

1905

1906

1907

</optional>

1908

1909

1910

</optional>

1911

</define>

1912

1913

1914

<a:documentation>Continuous field Galerkin projection.

1915

If the field you are interpolating is continuous, then

1916

a linear solver is required to invert the mass matrix.</a:documentation>

1917

1918

1919

1920

<a:documentation>Use a bounded Galerkin projection. Conservative, bounded in the

1921

limit, and minimally dissipative. This algorithm starts with the

1922

Galerkin projection and dissipates it until it achieves

1923

boundedness.

1924

If it does not converge, it may not be exactly bounded.

1925

Note well: this only works for linear fields.</a:documentation>

1926

1927

<value>Diffuse</value>

1928

</attribute>

1929

1930

<a:documentation>The number of dissipation iterations attempted to bound the

1931

Galerkin projection.</a:documentation>

1932

1933

1934

1935

<a:documentation>Specify the tolerance to which boundedness is to be tested during the iterations.

1936

Defaults to computer precision if unspecified.</a:documentation>

1937

1938

</element>

1939

</optional>

1940

</element>

1941

1942

1943

<a:documentation>If the bounds on this field are known then they can be set here.

1944

These can either further constrain the limits worked out by the

1945

lumped version of the projection (i.e. to make sure that errors

1946

don't accumulate with succesive interpolations) or if apply_globally

1947

is set they are just made to be bounded within the bounds globally

1948

(i.e. anything between those bounds is not smoothed).</a:documentation>

1949

1950

1951

1952

1953

1954

<a:documentation>If this is set the upper_bound is used everywhere.

1955

If left unset the upper_bound is only used to constrain

1956

the smoothed bounds calculated by the code</a:documentation>

1957

1958

</element>

1959

</optional>

1960

1961

1962

<a:documentation>This field is to be considered as being coupled to another field

1963

such that the sum of the two fields is constrained to be less than

1964

the upper_bound specified above.

1965

1966

The relationships between fields are worked out according to their

1967

priority ordering.

1968

1969

This method is akin to the coupled_cv advection method.</a:documentation>

1970

1971

</element>

1972

</optional>

1973

</element>

1974

</optional>

1975

1976

1977

1978

1979

1980

<a:documentation>If this is set the upper_bound is used everywhere.

1981

If left unset the upper_bound is only used to constrain

1982

the smoothed bounds calculated by the code</a:documentation>

1983

1984

</element>

1985

</optional>

1986

</element>

1987

</optional>

1988

</element>

1989

</optional>

1990

1991

1992

<a:documentation>If, after performing all the boundedness_iterations, the field

1993

is still not bounded then perform surgery to redistribute the

1994

deviations to nodes that have less than their bounds.</a:documentation>

1995

1996

1997

<a:documentation>Specify the tolerance to which boundedness is to be tested during the repair.

1998

Defaults to computer precision if unspecified.</a:documentation>

1999

2000

</element>

2001

</optional>

2002

</element>

2003

</optional>

2004

</element>

2005

2006

<a:documentation>Use a bounded Galerkin projection. Conservative, bounded in the

2007

limit, and hopefully minimally dissipative. This algorithm starts with the

2008

Galerkin projection and uses the optimisation library Algencan to bound

2009

it by minimising a functional within constraints on boundedness and conservation.

2010

Note well: this only works for linear fields.</a:documentation>

2011

2012

<value>Algencan</value>

2013

</attribute>

2014

2015

2016

2017

2018

</attribute>

2019

2020

2021

2022

</element>

2023

</optional>

2024

</element>

2025

2026

2027

<value>LumpedMassL2</value>

2028

</attribute>

2029

2030

2031

2032

</element>

2033

</optional>

2034

</element>

2035

2036

2037

<value>IntegralL2</value>

2038

</attribute>

2039

2040

2041

2042

</element>

2043

</optional>

2044

</element>

2045

2046

2047

2048

</attribute>

2049

2050

2051

2052

</element>

2053

</optional>

2054

</element>

2055

</choice>

2056

2057

2058

<a:documentation>If the bounds on this field are known then they can be set here.

2059

These can either further constrain the limits worked out by the

2060

lumped version of the projection (i.e. to make sure that errors

2061

don't accumulate with succesive interpolations) or if apply_globally

2062

is set they are just made to be bounded within the bounds globally

2063

(i.e. anything between those bounds is not smoothed).</a:documentation>

2064

2065

2066

2067

2068

2069

<a:documentation>If this is set the upper_bound is used everywhere.

2070

If left unset the upper_bound is only used to constrain

2071

the smoothed bounds calculated by the code</a:documentation>

2072

2073

</element>

2074

</optional>

2075

2076

2077

<a:documentation>This field is to be considered as being coupled to another field

2078

such that the sum of the two fields is constrained to be less than

2079

the upper_bound specified above.

2080

2081

The relationships between fields are worked out according to their

2082

priority ordering.

2083

2084

This method is akin to the coupled_cv advection method.</a:documentation>

2085

2086

</element>

2087

</optional>

2088

</element>

2089

</optional>

2090

2091

2092

2093

2094

2095

<a:documentation>If this is set the upper_bound is used everywhere.

2096

If left unset the upper_bound is only used to constrain

2097

the smoothed bounds calculated by the code</a:documentation>

2098

2099

</element>

2100

</optional>

2101

</element>

2102

</optional>

2103

</element>

2104

</optional>

2105

</element>

2106

</choice>

2107

</optional>

2108

2109

2110

<a:documentation>Solver options for the linear solve.

2111

This method requires the inversion of a mass matrix. Note that

2112

conservation properties are affected by the tolerance of the

2113

linear solve.</a:documentation>

2114

2115

</element>

2116

2117

<a:documentation>Lump the mass matrix on the left hand side of the galerkin projection.

2118

Hence solver options aren't necessary.

2119

2120

This is much more diffusive than a non-lumped Galerkin projection

2121

for only a minimal saving in computational cost.</a:documentation>

2122

2123

</element>

2124

</choice>

2125

</element>

2126

</define>

2127

2128

2129

2130

2131

<a:documentation>Discontinuous field Galerkin projection.

2132

In this case, no linear solver is required to invert the mass matrix.</a:documentation>

2133

2134

</element>

2135

</choice>

2136

</define>

2137

2138

2139

<a:documentation>Enables a supermesh free Galerkin projection. Uses incomplete

2140

quadrature, and hence is not conservative.</a:documentation>

2141

2142

</element>

2143

</define>

2144

2145

2146

<a:documentation>Options for checking the supermesh conservation properties</a:documentation>

2147

2148

2149

<a:documentation>Specify the fraction of the original elemental area/volume

2150

to be used to check the conservation of the supermesh.

2151

2152

Since all fields are supermeshed together the minimum tolerance

2153

specified over all fields will be used.

2154

2155

Defaults to 0.001 if unspecified.

2156

i.e. 0.1% of the area/volume of an element in the new mesh may

2157

be lost without warning or attempts to fix (if compiled with cgal)

2158

during the construction of the supermesh between the old

2159

and new meshes.</a:documentation>

2160

2161

</element>

2162

</optional>

2163

2164

2165

<a:documentation>Compute the field integral after the interpolation and print the relative

2166

mass loss to the logfile (level 2 verbosity).

2167

2168

Note this is a post interpolation step and offers no chance of

2169

fixing the conservation error (unlike the tolerance above if compiled

2170

with cgal)</a:documentation>

2171

2172

<a:documentation>Relative tolerance with which to test the conservation of the field

2173

integral. If the conservation fails this tolerance a warning is issued

2174

(level 0 verbosity) and vtus containing the field are output.</a:documentation>

2175

2176

</element>

2177

</element>

2178

</optional>

2179

</element>

2180

</define>

2181

2182

2183

2184

2185

<a:documentation>Options involving mesh movement (Lagrangian, ALE methods)

2186

Allow a moving mesh.

2187

Assigns memory for grid velocities

2188

Amends previous timestep`s mass matrix</a:documentation>

2189

2190

<value>MVMESH = TRUE, ZERQG, CMCHAN = TRUE</value>

2191

</attribute>

2192

2193

2194

<a:documentation>enable movement of mesh with the free surface</a:documentation>

2195

2196

2197

<a:documentation>Only move the nodes on the free surface using

2198

the surface height calculated at it.</a:documentation>

2199

2200

</element>

2201

2202

<a:documentation>Move the whole mesh according to the free surface

2203

height, scaled linearly by the depth from the surface.

2204

2205

Requires the specification of ocean_boundaries under

2206

geometry.</a:documentation>

2207

2208

</element>

2209

</choice>

2210

2211

2212

<a:documentation>Activates wetting and drying. Note: Works only with theta=1.0 and relaxation=1.0 under temporal discretisation</a:documentation>

2213

2214

<a:documentation>Specifies the d0 value which is the waterlevel kept in dry areas.</a:documentation>

2215

2216

</element>

2217

2218

2219

<a:documentation>Specifies the amount of absorption in dry areas. If a pressure corrected absorption term is desired, please add a pressure corrected absorption under Velocity (of magnitude 0).

2220

Note: If you double d0, this value has to be doubled as well to achieve the same amount of absorption.</a:documentation>

2221

2222

</element>

2223

</optional>

2224

2225

2226

<a:documentation>Use a smooth maximum operator discretisation. This is deprecated and will be removed soon.</a:documentation>

2227

2228

</element>

2229

</optional>

2230

</element>

2231

</optional>

2232

</element>

2233

2234

<a:documentation>Enable movement of mesh by an imposed Grid Velocity.

2235

Requires a prescribed GridVelocity field (see below).</a:documentation>

2236

2237

</element>

2238

2239

<a:documentation>Enable movement of mesh by the Velocity.

2240

Note that this is only pseudo-lagrangian as the

2241

Velocity and GridVelocity only match at the beginning of

2242

the timestep. Therefore the advection terms are still

2243

present and evaluated.</a:documentation>

2244

2245

2246

<a:documentation>Specify the material phase from which to extract the Velocity

2247

field. Defaults to state(1) if unspecified. Note that this

2248

doesn't necessarily correspond to the first material_phase

2249

in a flml file!</a:documentation>

2250

2251

2252

</attribute>

2253

2254

</element>

2255

</optional>

2256

</element>

2257

2258

<a:documentation>enable full ale movement of mesh</a:documentation>

2259

2260

2261

</element>

2262

2263

2264

</element>

2265

2266

2267

</element>

2268

2269

2270

</element>

2271

2272

2273

</element>

2274

2275

2276

</element>

2277

2278

2279

</element>

2280

2281

2282

</element>

2283

2284

2285

</element>

2286

2287

2288

</element>

2289

2290

2291

</element>

2292

2293

2294

</element>

2295

2296

2297

</element>

2298

2299

2300

</element>

2301

2302

2303

</element>

2304

2305

2306

2307

</element>

2308

</optional>

2309

2310

2311

2312

</element>

2313

</optional>

2314

2315

2316

2317

</element>

2318

</optional>

2319

</element>

2320

2321

<a:documentation>enable vertical movement of mesh (TBD)</a:documentation>

2322

2323

<value>NCOLOP < 0 </value>

2324

</attribute>

2325

<group>

2326

2327

<a:documentation>Functionals</a:documentation>

2328

2329

2330

<a:documentation>move the nodes along an isosurface

2331

Please specify a scalar field</a:documentation>

2332

2333

<value>NCOLOP = -1</value>

2334

</attribute>

2335

2336

2337

</attribute>

2338

</element>

2339

2340

<a:documentation>Lock a node to a high curvature region,

2341

Please specify a scalar field</a:documentation>

2342

2343

<value>NCOLOP = -2</value>

2344

</attribute>

2345

2346

2347

</attribute>

2348

</element>

2349

</choice>

2350

</element>

2351

2352

2353

<a:documentation>Mesh Quality terms</a:documentation>

2354

2355

2356

<a:documentation>Spring term, default value is 1.0</a:documentation>

2357

2358

</element>

2359

</optional>

2360

2361

2362

<a:documentation>Exponential term, default value is 1.0</a:documentation>

2363

2364

</element>

2365

</optional>

2366

</element>

2367

</optional>

2368

2369

2370

2371

</element>

2372

</optional>

2373

2374

2375

2376

</element>

2377

</optional>

2378

2379

2380

2381

</element>

2382

</optional>

2383

</group>

2384

</element>

2385

</choice>

2386

2387

<a:documentation>The velocity of the mesh.</a:documentation>

2388

2389

<value>GridVelocity</value>

2390

</attribute>

2391

2392

2393

</attribute>

2394

2395

2396

2397

2398

2399

<value>CoordinateMesh</value>

2400

</attribute>

2401

</element>

2402

2403

</element>

2404

2405

2406

2407

<value>CoordinateMesh</value>

2408

</attribute>

2409

</element>

2410

2411

</element>

2412

2413

2414

</element>

2415

</choice>

2416

</element>

2417

</element>

2418

</optional>

2419

2420

2421

2422

2423

</choice>

2424

</optional>

2425

</element>

2426

</define>

2427

2428

2429

</define>

2430

2431

2432

</define>

2433

2434

2435

2436

<a:documentation>Gradation constrains the jump

2437

in desired edge lengths along an edge, i.e.

2438

it controls how fast the mesh size may change.</a:documentation>

2439

2440

2441

<a:documentation>The gradation parameter. Must be a real >= 1.0.

2442

2443

The gradation parameter constrains the jump

2444

in desired edge lengths along an edge, i.e.

2445

it controls how fast the mesh size may change.

2446

A constant of 1.0 enforces a mesh of constant

2447

edge length everywhere. A value of 2.0 would

2448

allow the element size to double from element

2449

to element. The default value is 1.5.</a:documentation>

2450

2451

</element>

2452

</optional>

2453

</element>

2454

2455

<a:documentation>Anisotropic gradation algorithm, allowing for

2456

anisotropic bounds on the gradient of the sizing

2457

function.</a:documentation>

2458

2459

<a:documentation>Gamma is the tensor field that contains

2460

the bounds on the edge length specified by the error metric.</a:documentation>

2461

2462

<value>Gamma</value>

2463

</attribute>

2464

2465

2466

</element>

2467

</element>

2468

</element>

2469

</choice>

2470

</define>

2471

2472

2473

<a:documentation>Gradation constrains the jump

2474

in desired edge lengths along an edge, i.e.

2475

it controls how fast the mesh size may change.</a:documentation>

2476

2477

</element>

2478

</define>

2479

</grammar>