1
   2
   3
   4
   5
   6
   7
   8
   9
  10
  11
  12
  13
  14
  15
  16
  17
  18
  19
  20
  21
  22
  23
  24
  25
  26
  27
  28
  29
  30
  31
  32
  33
  34
  35
  36
  37
  38
  39
  40
  41
  42
  43
  44
  45
  46
  47
  48
  49
  50
  51
  52
  53
  54
  55
  56
  57
  58
  59
  60
  61
  62
  63
  64
  65
  66
  67
  68
  69
  70
  71
  72
  73
  74
  75
  76
  77
  78
  79
  80
  81
  82
  83
  84
  85
  86
  87
  88
  89
  90
  91
  92
  93
  94
  95
  96
  97
  98
  99
 100
 101
 102
 103
 104
 105
 106
 107
 108
 109
 110
 111
 112
 113
 114
 115
 116
 117
 118
 119
 120
 121
 122
 123
 124
 125
 126
 127
 128
 129
 130
 131
 132
 133
 134
 135
 136
 137
 138
 139
 140
 141
 142
 143
 144
 145
 146
 147
 148
 149
 150
 151
 152
 153
 154
 155
 156
 157
 158
 159
 160
 161
 162
 163
 164
 165
 166
 167
 168
 169
 170
 171
 172
 173
 174
 175
 176
 177
 178
 179
 180
 181
 182
 183
 184
 185
 186
 187
 188
 189
 190
 191
 192
 193
 194
 195
 196
 197
 198
 199
 200
 201
 202
 203
 204
 205
 206
 207
 208
 209
 210
 211
 212
 213
 214
 215
 216
 217
 218
 219
 220
 221
 222
 223
 224
 225
 226
 227
 228
 229
 230
 231
 232
 233
 234
 235
 236
 237
 238
 239
 240
 241
 242
 243
 244
 245
 246
 247
 248
 249
 250
 251
 252
 253
 254
 255
 256
 257
 258
 259
 260
 261
 262
 263
 264
 265
 266
 267
 268
 269
 270
 271
 272
 273
 274
 275
 276
 277
 278
 279
 280
 281
 282
 283
 284
 285
 286
 287
 288
 289
 290
 291
 292
 293
 294
 295
 296
 297
 298
 299
 300
 301
 302
 303
 304
 305
 306
 307
 308
 309
 310
 311
 312
 313
 314
 315
 316
 317
 318
 319
 320
 321
 322
 323
 324
 325
 326
 327
 328
 329
 330
 331
 332
 333
 334
 335
 336
 337
 338
 339
 340
 341
 342
 343
 344
 345
 346
 347
 348
 349
 350
 351
 352
 353
 354
 355
 356
 357
 358
 359
 360
 361
 362
 363
 364
 365
 366
 367
 368
 369
 370
 371
 372
 373
 374
 375
 376
 377
 378
 379
 380
 381
 382
 383
 384
 385
 386
 387
 388
 389
 390
 391
 392
 393
 394
 395
 396
 397
 398
 399
 400
 401
 402
 403
 404
 405
 406
 407
 408
 409
 410
 411
 412
 413
 414
 415
 416
 417
 418
 419
 420
 421
 422
 423
 424
 425
 426
 427
 428
 429
 430
 431
 432
 433
 434
 435
 436
 437
 438
 439
 440
 441
 442
 443
 444
 445
 446
 447
 448
 449
 450
 451
 452
 453
 454
 455
 456
 457
 458
 459
 460
 461
 462
 463
 464
 465
 466
 467
 468
 469
 470
 471
 472
 473
 474
 475
 476
 477
 478
 479
 480
 481
 482
 483
 484
 485
 486
 487
 488
 489
 490
 491
 492
 493
 494
 495
 496
 497
 498
 499
 500
 501
 502
 503
 504
 505
 506
 507
 508
 509
 510
 511
 512
 513
 514
 515
 516
 517
 518
 519
 520
 521
 522
 523
 524
 525
 526
 527
 528
 529
 530
 531
 532
 533
 534
 535
 536
 537
 538
 539
 540
 541
 542
 543
 544
 545
 546
 547
 548
 549
 550
 551
 552
 553
 554
 555
 556
 557
 558
 559
 560
 561
 562
 563
 564
 565
 566
 567
 568
 569
 570
 571
 572
 573
 574
 575
 576
 577
 578
 579
 580
 581
 582
 583
 584
 585
 586
 587
 588
 589
 590
 591
 592
 593
 594
 595
 596
 597
 598
 599
 600
 601
 602
 603
 604
 605
 606
 607
 608
 609
 610
 611
 612
 613
 614
 615
 616
 617
 618
 619
 620
 621
 622
 623
 624
 625
 626
 627
 628
 629
 630
 631
 632
 633
 634
 635
 636
 637
 638
 639
 640
 641
 642
 643
 644
 645
 646
 647
 648
 649
 650
 651
 652
 653
 654
 655
 656
 657
 658
 659
 660
 661
 662
 663
 664
 665
 666
 667
 668
 669
 670
 671
 672
 673
 674
 675
 676
 677
 678
 679
 680
 681
 682
 683
 684
 685
 686
 687
 688
 689
 690
 691
 692
 693
 694
 695
 696
 697
 698
 699
 700
 701
 702
 703
 704
 705
 706
 707
 708
 709
 710
 711
 712
 713
 714
 715
 716
 717
 718
 719
 720
 721
 722
 723
 724
 725
 726
 727
 728
 729
 730
 731
 732
 733
 734
 735
 736
 737
 738
 739
 740
 741
 742
 743
 744
 745
 746
 747
 748
 749
 750
 751
 752
 753
 754
 755
 756
 757
 758
 759
 760
 761
 762
 763
 764
 765
 766
 767
 768
 769
 770
 771
 772
 773
 774
 775
 776
 777
 778
 779
 780
 781
 782
 783
 784
 785
 786
 787
 788
 789
 790
 791
 792
 793
 794
 795
 796
 797
 798
 799
 800
 801
 802
 803
 804
 805
 806
 807
 808
 809
 810
 811
 812
 813
 814
 815
 816
 817
 818
 819
 820
 821
 822
 823
 824
 825
 826
 827
 828
 829
 830
 831
 832
 833
 834
 835
 836
 837
 838
 839
 840
 841
 842
 843
 844
 845
 846
 847
 848
 849
 850
 851
 852
 853
 854
 855
 856
 857
 858
 859
 860
 861
 862
 863
 864
 865
 866
 867
 868
 869
 870
 871
 872
 873
 874
 875
 876
 877
 878
 879
 880
 881
 882
 883
 884
 885
 886
 887
 888
 889
 890
 891
 892
 893
 894
 895
 896
 897
 898
 899
 900
 901
 902
 903
 904
 905
 906
 907
 908
 909
 910
 911
 912
 913
 914
 915
 916
 917
 918
 919
 920
 921
 922
 923
 924
 925
 926
 927
 928
 929
 930
 931
 932
 933
 934
 935
 936
 937
 938
 939
 940
 941
 942
 943
 944
 945
 946
 947
 948
 949
 950
 951
 952
 953
 954
 955
 956
 957
 958
 959
 960
 961
 962
 963
 964
 965
 966
 967
 968
 969
 970
 971
 972
 973
 974
 975
 976
 977
 978
 979
 980
 981
 982
 983
 984
 985
 986
 987
 988
 989
 990
 991
 992
 993
 994
 995
 996
 997
 998
 999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
// *****************************************************************************
/*!
  \file      src/Inciter/ALECG.cpp
  \copyright 2012-2015 J. Bakosi,
             2016-2018 Los Alamos National Security, LLC.,
             2019-2021 Triad National Security, LLC.
             All rights reserved. See the LICENSE file for details.
  \brief     ALECG for a PDE system with continuous Galerkin + ALE + RK
  \details   ALECG advances a system of partial differential equations (PDEs)
    using a continuous Galerkin (CG) finite element (FE) spatial discretization
    (using linear shapefunctions on tetrahedron elements) combined with a
    Runge-Kutta (RK) time stepping scheme in the arbitrary Eulerian-Lagrangian
    reference frame.
  \see The documentation in ALECG.hpp.
*/
// *****************************************************************************

#include "QuinoaBuildConfig.hpp"
#include "ALECG.hpp"
#include "Vector.hpp"
#include "Reader.hpp"
#include "ContainerUtil.hpp"
#include "UnsMesh.hpp"
#include "ExodusIIMeshWriter.hpp"
#include "Inciter/InputDeck/InputDeck.hpp"
#include "DerivedData.hpp"
#include "CGPDE.hpp"
#include "Discretization.hpp"
#include "DiagReducer.hpp"
#include "NodeBC.hpp"
#include "Refiner.hpp"
#include "Reorder.hpp"
#include "Around.hpp"
#include "CGPDE.hpp"
#include "Integrate/Mass.hpp"
#include "FieldOutput.hpp"

#ifdef HAS_ROOT
  #include "RootMeshWriter.hpp"
#endif

namespace inciter {

extern ctr::InputDeck g_inputdeck;
extern ctr::InputDeck g_inputdeck_defaults;
extern std::vector< CGPDE > g_cgpde;

//! Runge-Kutta coefficients
static const std::array< tk::real, 3 > rkcoef{{ 1.0/3.0, 1.0/2.0, 1.0 }};

} // inciter::

using inciter::ALECG;

ALECG::ALECG( const CProxy_Discretization& disc,
              const CProxy_Ghosts&,
              const std::map< int, std::vector< std::size_t > >& bface,
              const std::map< int, std::vector< std::size_t > >& bnode,
              const std::vector< std::size_t >& triinpoel ) :
  m_disc( disc ),
  m_nsol( 0 ),
  m_ngrad( 0 ),
  m_nrhs( 0 ),
  m_nbnorm( 0 ),
  m_ndfnorm( 0 ),
  m_bnode( bnode ),
  m_bface( bface ),
  m_triinpoel( tk::remap( triinpoel, Disc()->Lid() ) ),
  m_bndel( Disc()->bndel() ),
  m_dfnorm(),
  m_dfnormc(),
  m_dfn(),
  m_esup( tk::genEsup( Disc()->Inpoel(), 4 ) ),
  m_psup( tk::genPsup( Disc()->Inpoel(), 4, m_esup ) ),
  m_u( Disc()->Gid().size(), g_inputdeck.get< tag::component >().nprop() ),
  m_un( m_u.nunk(), m_u.nprop() ),
  m_rhs( m_u.nunk(), m_u.nprop() ),
  m_rhsc(),
  m_chBndGrad( Disc()->Bid().size(), m_u.nprop()*3 ),
  m_dirbc(),
  m_chBndGradc(),
  m_diag(),
  m_bnorm(),
  m_bnormc(),
  m_symbcnodes(),
  m_farfieldbcnodes(),
  m_symbctri(),
  m_spongenodes(),
  m_timedepbcnodes(),
  m_timedepbcFn(),
  m_stage( 0 ),
  m_boxnodes(),
  m_edgenode(),
  m_edgeid(),
  m_dtp( m_u.nunk(), 0.0 ),
  m_tp( m_u.nunk(), g_inputdeck.get< tag::discr, tag::t0 >() ),
  m_finished( 0 ),
  m_newmesh( 0 )
// *****************************************************************************
//  Constructor
//! \param[in] disc Discretization proxy
//! \param[in] bface Boundary-faces mapped to side sets used in the input file
//! \param[in] bnode Boundary-node lists mapped to side sets used in input file
//! \param[in] triinpoel Boundary-face connectivity where BCs set (global ids)
// *****************************************************************************
//! [Constructor]
{
  usesAtSync = true;    // enable migration at AtSync

  auto d = Disc();

  // Perform optional operator-access-pattern mesh node reordering
  if (g_inputdeck.get< tag::discr, tag::operator_reorder >()) {

    // Create new local ids based on access pattern of PDE operators
    std::unordered_map< std::size_t, std::size_t > map;
    std::size_t n = 0;

    for (std::size_t p=0; p<m_u.nunk(); ++p) {  // for each point p
      if (map.find(p) == end(map)) map[p] = n++;<--- Searching before insertion is not necessary.
      for (auto q : tk::Around(m_psup,p)) {     // for each edge p-q
        if (map.find(q) == end(map)) map[q] = n++;<--- Searching before insertion is not necessary.
      }
    }

    Assert( map.size() == d->Gid().size(), "Map size mismatch" );

    // Remap data in bound Discretization object
    d->remap( map );
    // Recompute elements surrounding points
    m_esup = tk::genEsup( d->Inpoel(), 4 );
    // Recompute points surrounding points
    m_psup = tk::genPsup( d->Inpoel(), 4, m_esup );
    // Remap boundary triangle face connectivity
    tk::remap( m_triinpoel, map );
  }

  // Query/update boundary-conditions-related data structures from user input
  queryBnd();

  // Activate SDAG wait for initially computing normals
  thisProxy[ thisIndex ].wait4norm();

  // Generate callbacks for solution transfers we are involved in

  // Always add a callback to be used when we are not involved in any transfers
  std::vector< CkCallback > cb;
  auto c = CkCallback(CkIndex_ALECG::transfer_complete(), thisProxy[thisIndex]);
  cb.push_back( c );

  // Generate a callback for each transfer we are involved in (either as a
  // source or a destination)
  auto meshid = d->MeshId();
  for (const auto& t : d->Transfers())
    if (meshid == t.src || meshid == t.dst)
      cb.push_back( c );

  // Send callbacks to base
  d->transferCallback( cb );
}
//! [Constructor]

void
ALECG::queryBnd()
// *****************************************************************************
// Query/update boundary-conditions-related data structures from user input
// *****************************************************************************
{
  auto d = Disc();

  // Query and match user-specified Dirichlet boundary conditions to side sets
  const auto steady = g_inputdeck.get< tag::discr, tag::steady_state >();
  if (steady) for (auto& deltat : m_dtp) deltat *= rkcoef[m_stage];<--- Consider using std::transform algorithm instead of a raw loop.
  m_dirbc = match( m_u.nprop(), d->T(), rkcoef[m_stage] * d->Dt(),
                   m_tp, m_dtp, d->Coord(), d->Lid(), m_bnode,
                   /* increment = */ false );
  if (steady) for (auto& deltat : m_dtp) deltat /= rkcoef[m_stage];<--- Consider using std::transform algorithm instead of a raw loop.

  // Prepare unique set of symmetry BC nodes
  auto sym = d->bcnodes< tag::bc, tag::bcsym >( m_bface, m_triinpoel );
  for (const auto& [s,nodes] : sym)
    m_symbcnodes.insert( begin(nodes), end(nodes) );

  // Prepare unique set of farfield BC nodes
  auto far = d->bcnodes< tag::bc, tag::bcfarfield >( m_bface, m_triinpoel );
  for (const auto& [s,nodes] : far)
    m_farfieldbcnodes.insert( begin(nodes), end(nodes) );

  // If farfield BC is set on a node, will not also set symmetry BC
  for (auto fn : m_farfieldbcnodes) m_symbcnodes.erase(fn);

  // Prepare boundary nodes contiguously accessible from a triangle-face loop
  m_symbctri.resize( m_triinpoel.size()/3, 0 );
  for (std::size_t e=0; e<m_triinpoel.size()/3; ++e)
    if (m_symbcnodes.find(m_triinpoel[e*3+0]) != end(m_symbcnodes))
      m_symbctri[e] = 1;

  // Prepare unique set of sponge nodes
  auto sponge = d->bcnodes< tag::sponge, tag::sideset >( m_bface, m_triinpoel );
  for (const auto& [s,nodes] : sponge)
    m_spongenodes.insert( begin(nodes), end(nodes) );

  // Prepare unique set of time dependent BC nodes
  m_timedepbcnodes.clear();
  m_timedepbcFn.clear();
  const auto& timedep =
    g_inputdeck.template get< tag::param, tag::compflow, tag::bctimedep >();
  if (!timedep.empty()) {
    m_timedepbcnodes.resize(timedep[0].size());
    m_timedepbcFn.resize(timedep[0].size());
    std::size_t ib=0;
    for (const auto& bndry : timedep[0]) {
      std::unordered_set< std::size_t > nodes;
      for (const auto& s : bndry.template get< tag::sideset >()) {
        auto k = m_bnode.find( std::stoi(s) );
        if (k != end(m_bnode)) {
          for (auto g : k->second) {      // global node ids on side set
            nodes.insert( tk::cref_find(d->Lid(),g) );
          }
        }
      }
      m_timedepbcnodes[ib].insert( begin(nodes), end(nodes) );

      // Store user defined discrete function in time. This is done in the same
      // loop as the BC nodes, so that the indices for the two vectors
      // m_timedepbcnodes and m_timedepbcFn are consistent with each other
      auto fn = bndry.template get< tag::fn >();
      for (std::size_t ir=0; ir<fn.size()/6; ++ir) {
        m_timedepbcFn[ib].push_back({{ fn[ir*6+0], fn[ir*6+1], fn[ir*6+2],
          fn[ir*6+3], fn[ir*6+4], fn[ir*6+5] }});
      }
      ++ib;
    }
  }

  Assert(m_timedepbcFn.size() == m_timedepbcnodes.size(), "Incorrect number of "
    "time dependent functions.");

  // Query ALE mesh velocity boundary condition node lists and node lists at
  // which ALE moves boundaries
  d->meshvelBnd( m_bface, m_bnode, m_triinpoel );
}

void
ALECG::norm()
// *****************************************************************************
// Start (re-)computing boundary point-, and dual-face normals
// *****************************************************************************
{
  auto d = Disc();

  // Query nodes at which symmetry BCs are specified
  auto bn = d->bcnodes< tag::bc, tag::bcsym >( m_bface, m_triinpoel );

  // Query nodes at which farfield BCs are specified
  auto far = d->bcnodes< tag::bc, tag::bcfarfield >( m_bface, m_triinpoel );
  // Merge BC data where boundary-point normals are required
  for (const auto& [s,n] : far) bn[s].insert( begin(n), end(n) );

  // Query nodes at which mesh velocity symmetry BCs are specified
  std::unordered_map<int, std::unordered_set< std::size_t >> ms;
  for (const auto& s : g_inputdeck.template get< tag::ale, tag::bcsym >()) {
    auto k = m_bface.find( std::stoi(s) );
    if (k != end(m_bface)) {
      auto& n = ms[ k->first ];
      for (auto f : k->second) {
        n.insert( m_triinpoel[f*3+0] );
        n.insert( m_triinpoel[f*3+1] );
        n.insert( m_triinpoel[f*3+2] );
      }
    }
  }
  // Merge BC data where boundary-point normals are required
  for (const auto& [s,n] : ms) bn[s].insert( begin(n), end(n) );

  // Compute boundary point normals
  bnorm( bn );

  // Compute dual-face normals associated to edges
  dfnorm();
}

std::array< tk::real, 3 >
ALECG::edfnorm( const tk::UnsMesh::Edge& edge,
                const std::unordered_map< tk::UnsMesh::Edge,
                        std::vector< std::size_t >,
                        tk::UnsMesh::Hash<2>, tk::UnsMesh::Eq<2> >& esued )
const
// *****************************************************************************
//  Compute normal of dual-mesh associated to edge
//! \param[in] edge Edge whose dual-face normal to compute given by local ids
//! \param[in] esued Elements surrounding edges
//! \return Dual-face normal for edge
// *****************************************************************************
{
  auto d = Disc();
  const auto& inpoel = d->Inpoel();
  const auto& coord = d->Coord();
  const auto& x = coord[0];
  const auto& y = coord[1];
  const auto& z = coord[2];

  std::array< tk::real, 3 > n{ 0.0, 0.0, 0.0 };

  for (auto e : tk::cref_find(esued,edge)) {
    // access node IDs
    const std::array< std::size_t, 4 >
      N{ inpoel[e*4+0], inpoel[e*4+1], inpoel[e*4+2], inpoel[e*4+3] };
    // compute element Jacobi determinant
    const std::array< tk::real, 3 >
      ba{{ x[N[1]]-x[N[0]], y[N[1]]-y[N[0]], z[N[1]]-z[N[0]] }},
      ca{{ x[N[2]]-x[N[0]], y[N[2]]-y[N[0]], z[N[2]]-z[N[0]] }},
      da{{ x[N[3]]-x[N[0]], y[N[3]]-y[N[0]], z[N[3]]-z[N[0]] }};
    const auto J = tk::triple( ba, ca, da );        // J = 6V
    Assert( J > 0, "Element Jacobian non-positive" );
    // shape function derivatives, nnode*ndim [4][3]
    std::array< std::array< tk::real, 3 >, 4 > grad;
    grad[1] = tk::crossdiv( ca, da, J );
    grad[2] = tk::crossdiv( da, ba, J );
    grad[3] = tk::crossdiv( ba, ca, J );
    for (std::size_t i=0; i<3; ++i)
      grad[0][i] = -grad[1][i]-grad[2][i]-grad[3][i];
    // sum normal contributions
    // The constant 1/48: Eq (12) from Waltz et al. Computers & fluids (92) 2014
    // The result of the integral of shape function N on a tet is V/4.
    // This can be written as J/(6*4). Eq (12) has a 1/2 multiplier.
    // This leads to J/48.
    auto J48 = J/48.0;
    for (const auto& [a,b] : tk::lpoed) {
      auto s = tk::orient( {N[a],N[b]}, edge );
      for (std::size_t j=0; j<3; ++j)
        n[j] += J48 * s * (grad[a][j] - grad[b][j]);
    }
  }

  return n;
}

void
ALECG::dfnorm()
// *****************************************************************************
// Compute dual-face normals associated to edges
// *****************************************************************************
{
  auto d = Disc();
  const auto& inpoel = d->Inpoel();
  const auto& gid = d->Gid();

  // compute derived data structures
  auto esued = tk::genEsued( inpoel, 4, tk::genEsup( inpoel, 4 ) );

  // Compute dual-face normals for domain edges
  for (std::size_t p=0; p<gid.size(); ++p)    // for each point p
    for (auto q : tk::Around(m_psup,p))       // for each edge p-q
      if (gid[p] < gid[q])
        m_dfnorm[{gid[p],gid[q]}] = edfnorm( {p,q}, esued );

  // Send our dual-face normal contributions to neighbor chares
  if (d->EdgeCommMap().empty())
    comdfnorm_complete();
  else {
    for (const auto& [c,edges] : d->EdgeCommMap()) {
      decltype(m_dfnorm) exp;
      for (const auto& e : edges) exp[e] = tk::cref_find(m_dfnorm,e);
      thisProxy[c].comdfnorm( exp );
    }
  }

  owndfnorm_complete();
}

void
ALECG::comdfnorm( const std::unordered_map< tk::UnsMesh::Edge,
                    std::array< tk::real, 3 >,
                    tk::UnsMesh::Hash<2>, tk::UnsMesh::Eq<2> >& dfnorm )
// *****************************************************************************
// Receive contributions to dual-face normals on chare-boundaries
//! \param[in] dfnorm Incoming partial sums of dual-face normals associated to
//!   chare-boundary edges
// *****************************************************************************
{
  // Buffer up inccoming contributions to dual-face normals
  for (const auto& [e,n] : dfnorm) {
    auto& dfn = m_dfnormc[e];
    dfn[0] += n[0];
    dfn[1] += n[1];
    dfn[2] += n[2];
  }

  if (++m_ndfnorm == Disc()->EdgeCommMap().size()) {
    m_ndfnorm = 0;
    comdfnorm_complete();
  }
}

void
ALECG::bnorm( const std::unordered_map< int,
                std::unordered_set< std::size_t > >& bcnodes )
// *****************************************************************************
//  Compute boundary point normals
//! \param[in] bcnodes Local node ids associated to side set ids at which BCs
//!    are set that require normals
//*****************************************************************************
{
  auto d = Disc();

  m_bnorm = cg::bnorm( m_bface, m_triinpoel, d->Coord(), d->Gid(), bcnodes );

  // Send our nodal normal contributions to neighbor chares
  if (d->NodeCommMap().empty())
    comnorm_complete();
  else
    for (const auto& [ neighborchare, sharednodes ] : d->NodeCommMap()) {
      std::unordered_map< int,
        std::unordered_map< std::size_t, std::array< tk::real, 4 > > > exp;
      for (auto i : sharednodes) {
        for (const auto& [s,norms] : m_bnorm) {
          auto j = norms.find(i);
          if (j != end(norms)) exp[s][i] = j->second;
        }
      }
      thisProxy[ neighborchare ].comnorm( exp );
    }

  ownnorm_complete();
}

void
ALECG::comnorm( const std::unordered_map< int,
  std::unordered_map< std::size_t, std::array< tk::real, 4 > > >& innorm )
// *****************************************************************************
// Receive boundary point normals on chare-boundaries
//! \param[in] innorm Incoming partial sums of boundary point normal
//!   contributions to normals (first 3 components), inverse distance squared
//!   (4th component), associated to side set ids
// *****************************************************************************
{
  // Buffer up incoming boundary-point normal vector contributions
  for (const auto& [s,norms] : innorm) {
    auto& bnorms = m_bnormc[s];
    for (const auto& [p,n] : norms) {
      auto& bnorm = bnorms[p];
      bnorm[0] += n[0];
      bnorm[1] += n[1];
      bnorm[2] += n[2];
      bnorm[3] += n[3];
    }
  }

  if (++m_nbnorm == Disc()->NodeCommMap().size()) {
    m_nbnorm = 0;
    comnorm_complete();
  }
}

void
ALECG::registerReducers()
// *****************************************************************************
//  Configure Charm++ reduction types initiated from this chare array
//! \details Since this is a [initnode] routine, the runtime system executes the
//!   routine exactly once on every logical node early on in the Charm++ init
//!   sequence. Must be static as it is called without an object. See also:
//!   Section "Initializations at Program Startup" at in the Charm++ manual
//!   http://charm.cs.illinois.edu/manuals/html/charm++/manual.html.
// *****************************************************************************
{
  NodeDiagnostics::registerReducers();
}

void
ALECG::ResumeFromSync()
// *****************************************************************************
//  Return from migration
//! \details This is called when load balancing (LB) completes. The presence of
//!   this function does not affect whether or not we block on LB.
// *****************************************************************************
{
  if (Disc()->It() == 0) Throw( "it = 0 in ResumeFromSync()" );

  if (!g_inputdeck.get< tag::cmd, tag::nonblocking >()) next();
}

//! [setup]
void
ALECG::setup()
// *****************************************************************************
// Start setup for solution
// *****************************************************************************
{
  auto d = Disc();

  // Determine nodes inside user-defined IC box
  for (auto& eq : g_cgpde) eq.IcBoxNodes( d->Coord(), m_boxnodes );

  // Compute volume of user-defined box IC
  d->boxvol( m_boxnodes );

  // Query time history field output labels from all PDEs integrated
  const auto& hist_points = g_inputdeck.get< tag::history, tag::point >();
  if (!hist_points.empty()) {
    std::vector< std::string > histnames;
    for (const auto& eq : g_cgpde) {
      auto n = eq.histNames();
      histnames.insert( end(histnames), begin(n), end(n) );
    }
    d->histheader( std::move(histnames) );
  }
}
//! [setup]

void
ALECG::volumetric( tk::Fields& u, const std::vector< tk::real >& v )
// *****************************************************************************
//  Multiply solution with mesh volume
//! \param[in,out] u Solution vector
//! \param[in] v Volume to multiply with
// *****************************************************************************
{
  Assert( v.size() == u.nunk(), "Size mismatch" );

  for (std::size_t i=0; i<u.nunk(); ++i)
    for (ncomp_t c=0; c<u.nprop(); ++c)
      u(i,c,0) *= v[i];
}

void
ALECG::conserved( tk::Fields& u, const std::vector< tk::real >& v )
// *****************************************************************************
//  Divide solution with mesh volume
//! \param[in,out] u Solution vector
//! \param[in] v Volume to divide with
// *****************************************************************************
{
  Assert( v.size() == u.nunk(), "Size mismatch" );

  for (std::size_t i=0; i<u.nunk(); ++i)
    for (ncomp_t c=0; c<u.nprop(); ++c) {
      u(i,c,0) /= v[i];
    }
}

void
ALECG::box( tk::real v )
// *****************************************************************************
// Receive total box IC volume and set conditions in box
//! \param[in] v Total volume within user-specified box
// *****************************************************************************
{
  auto d = Disc();

  // Store user-defined box IC volume
  d->Boxvol() = v;

  // Set initial conditions for all PDEs
  for (auto& eq : g_cgpde)
    eq.initialize( d->Coord(), m_u, d->T(), d->Boxvol(), m_boxnodes );

  // Multiply conserved variables with mesh volume
  volumetric( m_u, Disc()->Vol() );

  // Initiate IC transfer (if coupled)
  Disc()->transfer( m_u );

  // Initialize nodal mesh volumes at previous time step stage
  d->Voln() = d->Vol();

  // Start computing the mesh mesh velocity for ALE
  meshvelstart();
}

void
ALECG::meshvelstart()
// *****************************************************************************
// Start computing the mesh mesh velocity for ALE
// *****************************************************************************
{
  auto d = Disc();

  // Apply boundary conditions on numerical solution
  BC();

  conserved( m_u, d->Vol() );

  // query fluid velocity across all systems integrated
  tk::UnsMesh::Coords vel;
  for (const auto& eq : g_cgpde) eq.velocity( m_u, vel );
  // query speed of sound in mesh nodes across all systems integrated
  std::vector< tk::real > soundspeed;
  for (const auto& eq : g_cgpde) eq.soundspeed( m_u, soundspeed );

  volumetric( m_u, d->Vol() );

  // Start computing the mesh mesh velocity for ALE
  d->meshvelStart( vel, soundspeed, m_bnorm, rkcoef[m_stage] * d->Dt(),
    CkCallback(CkIndex_ALECG::meshveldone(), thisProxy[thisIndex]) );
}

void
ALECG::meshveldone()
// *****************************************************************************
// Done with computing the mesh velocity for ALE
// *****************************************************************************
{
  // Assess and record mesh velocity linear solver conergence
  Disc()->meshvelConv();

  // Continue
  if (Disc()->Initial()) lhs(); else ale();
}

//! [start]
void
ALECG::start()
// *****************************************************************************
// Start time stepping
// *****************************************************************************
{
  // Set flag that indicates that we are now during time stepping
  Disc()->Initial( 0 );
  // Start timer measuring time stepping wall clock time
  Disc()->Timer().zero();
  // Zero grind-timer
  Disc()->grindZero();
  // Continue to first time step
  next();
}
//! [start]

//! [Compute lhs]
void
ALECG::lhs()
// *****************************************************************************
// Compute the left-hand side of transport equations
//! \details Also (re-)compute all data structures if the mesh changed.
// *****************************************************************************
{
  // No need for LHS in ALECG

  // (Re-)compute boundary point-, and dual-face normals
  norm();
}
//! [Compute lhs]

//! [Merge normals and continue]
void
ALECG::mergelhs()
// *****************************************************************************
// The own and communication portion of the left-hand side is complete
// *****************************************************************************
{
  // Combine own and communicated contributions of normals
  normfinal();

  if (Disc()->Initial()) {
    // Output initial conditions to file
    writeFields( CkCallback(CkIndex_ALECG::start(), thisProxy[thisIndex]) );
  } else {
    norm_complete();
  }
}
//! [Merge normals and continue]

void
ALECG::normfinal()
// *****************************************************************************
//  Finish computing dual-face and boundary point normals
// *****************************************************************************
{
  auto d = Disc();
  const auto& lid = d->Lid();

  // Combine own and communicated contributions to boundary point normals
  for (const auto& [s,norms] : m_bnormc) {
    auto& bnorms = m_bnorm[s];
    for (const auto& [p,n] : norms) {
      auto& norm = bnorms[p];
      norm[0] += n[0];
      norm[1] += n[1];
      norm[2] += n[2];
      norm[3] += n[3];
    }
  }
  tk::destroy( m_bnormc );

  // Divide summed point normals by the sum of inverse distance squared
  for (auto& [s,norms] : m_bnorm)
    for (auto& [p,n] : norms) {
      n[0] /= n[3];
      n[1] /= n[3];
      n[2] /= n[3];
      Assert( (n[0]*n[0] + n[1]*n[1] + n[2]*n[2] - 1.0) <
              1.0e+3*std::numeric_limits< tk::real >::epsilon(),
              "Non-unit normal" );
    }

  // Replace global->local ids associated to boundary point normals
  decltype(m_bnorm) bnorm;
  for (auto& [s,norms] : m_bnorm) {
    auto& bnorms = bnorm[s];
    for (auto&& [g,n] : norms)
      bnorms[ tk::cref_find(lid,g) ] = std::move(n);
  }
  m_bnorm = std::move(bnorm);

  // Count contributions to chare-boundary edges
  std::unordered_map< tk::UnsMesh::Edge, std::size_t,
    tk::UnsMesh::Hash<2>, tk::UnsMesh::Eq<2> > edge_node_count;
  for (const auto& [c,edges] : d->EdgeCommMap())
    for (const auto& e : edges)
      ++edge_node_count[e];

  // Combine and weigh communicated contributions to dual-face normals
  for (auto& [e,n] : m_dfnormc) {
    const auto& dfn = tk::cref_find( m_dfnorm, e );
    n[0] += dfn[0];
    n[1] += dfn[1];
    n[2] += dfn[2];
    auto count = static_cast< tk::real >( tk::cref_find( edge_node_count, e ) );
    auto factor = 1.0/(count + 1.0);
    for (auto & x : n) x *= factor;<--- Consider using std::transform algorithm instead of a raw loop.
  }

  // Generate list of unique edges
  tk::UnsMesh::EdgeSet uedge;
  for (std::size_t p=0; p<m_u.nunk(); ++p)
    for (auto q : tk::Around(m_psup,p))
      uedge.insert( {p,q} );

  // Flatten edge list
  m_edgenode.resize( uedge.size() * 2 );
  std::size_t f = 0;
  const auto& gid = d->Gid();
  for (auto&& [p,q] : uedge) {
    if (gid[p] > gid[q]) {
      m_edgenode[f+0] = std::move(q);
      m_edgenode[f+1] = std::move(p);
    } else {
      m_edgenode[f+0] = std::move(p);
      m_edgenode[f+1] = std::move(q);
    }
    f += 2;
  }
  tk::destroy(uedge);

  // Convert dual-face normals to streamable (and vectorizable) data structure
  m_dfn.resize( m_edgenode.size() * 3 );      // 2 vectors per access
  std::unordered_map< tk::UnsMesh::Edge, std::size_t,
                      tk::UnsMesh::Hash<2>, tk::UnsMesh::Eq<2> > eid;
  for (std::size_t e=0; e<m_edgenode.size()/2; ++e) {
    auto p = m_edgenode[e*2+0];
    auto q = m_edgenode[e*2+1];
    eid[{p,q}] = e;
    std::array< std::size_t, 2 > g{ gid[p], gid[q] };
    auto n = tk::cref_find( m_dfnorm, g );
    // figure out if this is an edge on the parallel boundary
    auto nit = m_dfnormc.find( g );
    auto m = ( nit != m_dfnormc.end() ) ? nit->second : n;
    m_dfn[e*6+0] = n[0];
    m_dfn[e*6+1] = n[1];
    m_dfn[e*6+2] = n[2];
    m_dfn[e*6+3] = m[0];
    m_dfn[e*6+4] = m[1];
    m_dfn[e*6+5] = m[2];
  }

  tk::destroy( m_dfnorm );
  tk::destroy( m_dfnormc );

  // Flatten edge id data structure
  m_edgeid.resize( m_psup.first.size() );
  for (std::size_t p=0,k=0; p<m_u.nunk(); ++p)
    for (auto q : tk::Around(m_psup,p))
      m_edgeid[k++] = tk::cref_find( eid, {p,q} );
}

void
ALECG::BC()
// *****************************************************************************
// Apply boundary conditions
// \details The following BC enforcement changes the initial condition or
//!   updated solution (dependending on when it is called) to ensure strong
//!   imposition of the BCs. This is a matter of choice. Another alternative is
//!   to only apply BCs when computing fluxes at boundary faces, thereby only
//!   weakly enforcing the BCs. The former is conventionally used in continunous
//!   Galerkin finite element methods (such as ALECG implements), whereas the
//!   latter, in finite volume methods.
// *****************************************************************************
{
  const auto& coord = Disc()->Coord();

  conserved( m_u, Disc()->Vol() );

  // Apply Dirichlet BCs
  for (const auto& [b,bc] : m_dirbc)
    for (ncomp_t c=0; c<m_u.nprop(); ++c)
      if (bc[c].first) m_u(b,c,0) = bc[c].second;

  // Apply symmetry BCs
  for (const auto& eq : g_cgpde)
    eq.symbc( m_u, coord, m_bnorm, m_symbcnodes );

  // Apply farfield BCs
  for (const auto& eq : g_cgpde)
    eq.farfieldbc( m_u, coord, m_bnorm, m_farfieldbcnodes );

  // Apply sponge conditions
  for (const auto& eq : g_cgpde)
    eq.sponge( m_u, coord, m_spongenodes );

  // Apply user defined time dependent BCs
  for (const auto& eq : g_cgpde)
    eq.timedepbc( Disc()->T(), m_u, m_timedepbcnodes, m_timedepbcFn );

  volumetric( m_u, Disc()->Vol() );
}

void
ALECG::next()
// *****************************************************************************
// Continue to next time step
// *****************************************************************************
{
  dt();
}

void
ALECG::dt()
// *****************************************************************************
// Compute time step size
// *****************************************************************************
{
  tk::real mindt = std::numeric_limits< tk::real >::max();

  auto const_dt = g_inputdeck.get< tag::discr, tag::dt >();
  auto def_const_dt = g_inputdeck_defaults.get< tag::discr, tag::dt >();
  auto eps = std::numeric_limits< tk::real >::epsilon();

  auto d = Disc();

  // use constant dt if configured
  if (std::abs(const_dt - def_const_dt) > eps) {

    mindt = const_dt;

  } else {      // compute dt based on CFL

    //! [Find the minimum dt across all PDEs integrated]
    conserved( m_u, Disc()->Vol() );
    if (g_inputdeck.get< tag::discr, tag::steady_state >()) {

      // compute new dt for each mesh point
      for (const auto& eq : g_cgpde)
        eq.dt( d->It(), d->Vol(), m_u, m_dtp );

      // find the smallest dt of all nodes on this chare
      mindt = *std::min_element( begin(m_dtp), end(m_dtp) );

    } else {    // compute new dt for this chare

      // find the smallest dt of all equations on this chare
      for (const auto& eq : g_cgpde) {
        auto eqdt = eq.dt( d->Coord(), d->Inpoel(), d->T(), d->Dtn(), m_u,
                           d->Vol(), d->Voln() );
        if (eqdt < mindt) mindt = eqdt;
      }

    }
    volumetric( m_u, Disc()->Vol() );
    //! [Find the minimum dt across all PDEs integrated]

  }

  //! [Advance]
  // Actiavate SDAG waits for next time step stage
  thisProxy[ thisIndex ].wait4grad();
  thisProxy[ thisIndex ].wait4rhs();

  // Contribute to minimum dt across all chares and advance to next step
  contribute( sizeof(tk::real), &mindt, CkReduction::min_double,
              CkCallback(CkReductionTarget(ALECG,advance), thisProxy) );
  //! [Advance]
}

void
ALECG::advance( tk::real newdt )
// *****************************************************************************
// Advance equations to next time step
//! \param[in] newdt The smallest dt across the whole problem
// *****************************************************************************
{
  auto d = Disc();

  // Set new time step size
  if (m_stage == 0) d->setdt( newdt );

  // Compute gradients for next time step
  chBndGrad();
}

void
ALECG::chBndGrad()
// *****************************************************************************
// Compute nodal gradients at chare-boundary nodes. Gradients at internal nodes
// are calculated locally as needed and are not stored.
// *****************************************************************************
{
  auto d = Disc();

  // Divide solution with mesh volume
  conserved( m_u, Disc()->Vol() );
  // Compute own portion of gradients for all equations
  for (const auto& eq : g_cgpde)
    eq.chBndGrad( d->Coord(), d->Inpoel(), m_bndel, d->Gid(), d->Bid(), m_u,
                  m_chBndGrad );
  // Multiply solution with mesh volume
  volumetric( m_u, Disc()->Vol() );

  // Communicate gradients to other chares on chare-boundary
  if (d->NodeCommMap().empty())        // in serial we are done
    comgrad_complete();
  else // send gradients contributions to chare-boundary nodes to fellow chares
    for (const auto& [c,n] : d->NodeCommMap()) {
      std::vector< std::vector< tk::real > > g( n.size() );
      std::size_t j = 0;
      for (auto i : n) g[ j++ ] = m_chBndGrad[ tk::cref_find(d->Bid(),i) ];
      thisProxy[c].comChBndGrad( std::vector<std::size_t>(begin(n),end(n)), g );
    }

  owngrad_complete();
}

void
ALECG::comChBndGrad( const std::vector< std::size_t >& gid,
                     const std::vector< std::vector< tk::real > >& G )
// *****************************************************************************
//  Receive contributions to nodal gradients on chare-boundaries
//! \param[in] gid Global mesh node IDs at which we receive grad contributions
//! \param[in] G Partial contributions of gradients to chare-boundary nodes
//! \details This function receives contributions to m_chBndGrad, which stores
//!   nodal gradients at mesh chare-boundary nodes. While m_chBndGrad stores
//!   own contributions, m_chBndGradc collects the neighbor chare
//!   contributions during communication. This way work on m_chBndGrad and
//!   m_chBndGradc is overlapped. The two are combined in rhs().
// *****************************************************************************
{
  Assert( G.size() == gid.size(), "Size mismatch" );

  using tk::operator+=;

  for (std::size_t i=0; i<gid.size(); ++i) m_chBndGradc[ gid[i] ] += G[i];

  if (++m_ngrad == Disc()->NodeCommMap().size()) {
    m_ngrad = 0;
    comgrad_complete();
  }
}

void
ALECG::rhs()
// *****************************************************************************
// Compute right-hand side of transport equations
// *****************************************************************************
{
  auto d = Disc();

  // Combine own and communicated contributions to nodal gradients
  for (const auto& [gid,g] : m_chBndGradc) {
    auto bid = tk::cref_find( d->Bid(), gid );
    for (ncomp_t c=0; c<m_chBndGrad.nprop(); ++c)
      m_chBndGrad(bid,c,0) += g[c];
  }

  // clear gradients receive buffer
  tk::destroy(m_chBndGradc);

  const auto steady = g_inputdeck.get< tag::discr, tag::steady_state >();

  // Compute own portion of right-hand side for all equations
  auto prev_rkcoef = m_stage == 0 ? 0.0 : rkcoef[m_stage-1];
  if (steady)
    for (std::size_t p=0; p<m_tp.size(); ++p) m_tp[p] += prev_rkcoef * m_dtp[p];
  conserved( m_u, Disc()->Vol() );
  for (const auto& eq : g_cgpde) {
    eq.rhs( d->T() + prev_rkcoef * d->Dt(), d->Coord(), d->Inpoel(),
            m_triinpoel, d->Gid(), d->Bid(), d->Lid(), m_dfn, m_psup, m_esup,
            m_symbctri, m_spongenodes, d->Vol(), m_edgenode, m_edgeid,
            m_boxnodes, m_chBndGrad, m_u, d->meshvel(), m_tp, d->Boxvol(),
            m_rhs );
  }
  volumetric( m_u, Disc()->Vol() );
  if (steady)
    for (std::size_t p=0; p<m_tp.size(); ++p) m_tp[p] -= prev_rkcoef * m_dtp[p];

  // Query/update boundary-conditions-related data structures from user input
  queryBnd();

  // Communicate rhs to other chares on chare-boundary
  if (d->NodeCommMap().empty())        // in serial we are done
    comrhs_complete();
  else // send contributions of rhs to chare-boundary nodes to fellow chares
    for (const auto& [c,n] : d->NodeCommMap()) {
      std::vector< std::vector< tk::real > > r( n.size() );
      std::size_t j = 0;
      for (auto i : n) r[ j++ ] = m_rhs[ tk::cref_find(d->Lid(),i) ];
      thisProxy[c].comrhs( std::vector<std::size_t>(begin(n),end(n)), r );
    }

  ownrhs_complete();
}

void
ALECG::comrhs( const std::vector< std::size_t >& gid,
               const std::vector< std::vector< tk::real > >& R )
// *****************************************************************************
//  Receive contributions to right-hand side vector on chare-boundaries
//! \param[in] gid Global mesh node IDs at which we receive RHS contributions
//! \param[in] R Partial contributions of RHS to chare-boundary nodes
//! \details This function receives contributions to m_rhs, which stores the
//!   right hand side vector at mesh nodes. While m_rhs stores own
//!   contributions, m_rhsc collects the neighbor chare contributions during
//!   communication. This way work on m_rhs and m_rhsc is overlapped. The two
//!   are combined in solve().
// *****************************************************************************
{
  Assert( R.size() == gid.size(), "Size mismatch" );

  using tk::operator+=;

  for (std::size_t i=0; i<gid.size(); ++i) m_rhsc[ gid[i] ] += R[i];

  // When we have heard from all chares we communicate with, this chare is done
  if (++m_nrhs == Disc()->NodeCommMap().size()) {
    m_nrhs = 0;
    comrhs_complete();
  }
}

void
ALECG::solve()
// *****************************************************************************
//  Advance systems of equations
// *****************************************************************************
{
  auto d = Disc();

  // Combine own and communicated contributions to rhs
  for (const auto& b : m_rhsc) {
    auto lid = tk::cref_find( d->Lid(), b.first );
    for (ncomp_t c=0; c<m_rhs.nprop(); ++c) m_rhs(lid,c,0) += b.second[c];
  }

  // clear receive buffer
  tk::destroy(m_rhsc);

  // Update state at time n
  if (m_stage == 0) {
    m_un = m_u;
    if (g_inputdeck.get< tag::ale, tag::ale >()) d->UpdateCoordn();
  }

  // Solve the sytem
  if (g_inputdeck.get< tag::discr, tag::steady_state >()) {

    // Advance solution, converging to steady state
    for (std::size_t i=0; i<m_u.nunk(); ++i)
      for (ncomp_t c=0; c<m_u.nprop(); ++c)
        m_u(i,c,0) = m_un(i,c,0) + rkcoef[m_stage] * m_dtp[i] * m_rhs(i,c,0);

  } else {

    auto adt = rkcoef[m_stage] * d->Dt();

    // Advance unsteady solution
    m_u = m_un + adt * m_rhs;

    // Advance mesh if ALE is enabled
    if (g_inputdeck.get< tag::ale, tag::ale >()) {
      auto& coord = d->Coord();
      const auto& w = d->meshvel();
      for (auto j : g_inputdeck.get< tag::ale, tag::mesh_motion >())
        for (std::size_t i=0; i<coord[j].size(); ++i)
          coord[j][i] = d->Coordn()[j][i] + adt * w(i,j,0);
    }

  }

  m_newmesh = 0;  // recompute normals after ALE (if enabled)
  // Activate SDAG waits
  thisProxy[ thisIndex ].wait4norm();
  thisProxy[ thisIndex ].wait4mesh();

  //! [Continue after solve]
  // Recompute mesh volumes if ALE is enabled
  if (g_inputdeck.get< tag::ale, tag::ale >()) {

    transfer_complete();
    // Save nodal volumes at previous time step stage
    d->Voln() = d->Vol();
    // Prepare for recomputing the nodal volumes
    d->startvol();
    auto meshid = d->MeshId();
    contribute( sizeof(std::size_t), &meshid, CkReduction::nop,
                CkCallback(CkReductionTarget(Transporter,resized), d->Tr()) );

  } else {

    norm_complete();
    resized();

  }
  //! [Continue after solve]
}

void
ALECG::ale()
// *****************************************************************************
//  Continue after computing the new mesh velocity for ALE
// *****************************************************************************
{
  auto d = Disc();

  if (m_stage < 2) {

    // Activate SDAG wait for next time step stage
    thisProxy[ thisIndex ].wait4grad();
    thisProxy[ thisIndex ].wait4rhs();

    // continue to mesh-to-mesh transfer (if coupled)
    transfer();

  } else {

    // Ensure new field output file if mesh moved if ALE is enabled
    if (g_inputdeck.get< tag::ale, tag::ale >()) {
      d->Itf() = 0;  // Zero field output iteration count if mesh moved
      ++d->Itr();    // Increase number of iterations with a change in the mesh
    }

    // Compute diagnostics, e.g., residuals
    conserved( m_u, Disc()->Vol() );
    conserved( m_un, Disc()->Voln() );
    auto diag_computed = m_diag.compute( *d, m_u, m_un, m_bnorm,
                                         m_symbcnodes, m_farfieldbcnodes );
    volumetric( m_u, Disc()->Vol() );
    volumetric( m_un, Disc()->Voln() );
    // Increase number of iterations and physical time
    d->next();
    // Advance physical time for local time stepping
    if (g_inputdeck.get< tag::discr, tag::steady_state >())
      for (std::size_t i=0; i<m_u.nunk(); ++i) m_tp[i] += m_dtp[i];
    // Continue to mesh refinement (if configured)
    if (!diag_computed) refine( std::vector< tk::real >( m_u.nprop(), 1.0 ) );

  }
}

//! [Refine]
void
ALECG::refine( const std::vector< tk::real >& l2res )
// *****************************************************************************
// Optionally refine/derefine mesh
//! \param[in] l2res L2-norms of the residual for each scalar component
//!   computed across the whole problem
// *****************************************************************************
{
  auto d = Disc();

  const auto steady = g_inputdeck.get< tag::discr, tag::steady_state >();
  const auto residual = g_inputdeck.get< tag::discr, tag::residual >();
  const auto rc = g_inputdeck.get< tag::discr, tag::rescomp >() - 1;

  if (steady) {

    // this is the last time step if max time of max number of time steps
    // reached or the residual has reached its convergence criterion
    if (d->finished() or l2res[rc] < residual) m_finished = 1;

  } else {

    // this is the last time step if max time or max iterations reached
    if (d->finished()) m_finished = 1;

  }

  auto dtref = g_inputdeck.get< tag::amr, tag::dtref >();
  auto dtfreq = g_inputdeck.get< tag::amr, tag::dtfreq >();

  // Activate SDAG waits for re-computing the normals
  m_newmesh = 1;  // recompute normals after AMR (if enabled)
  thisProxy[ thisIndex ].wait4norm();
  thisProxy[ thisIndex ].wait4mesh();

  // if t>0 refinement enabled and we hit the frequency
  if (dtref && !(d->It() % dtfreq)) {   // refine

    d->startvol();
    d->Ref()->dtref( {}, m_bnode, {} );
    d->refined() = 1;

  } else {      // do not refine

    d->refined() = 0;
    norm_complete();
    resized();

  }
}
//! [Refine]

//! [Resize]
void
ALECG::resizePostAMR(
  const std::vector< std::size_t >& /*ginpoel*/,
  const tk::UnsMesh::Chunk& chunk,
  const tk::UnsMesh::Coords& coord,
  const std::unordered_map< std::size_t, tk::UnsMesh::Edge >& addedNodes,
  const std::unordered_map< std::size_t, std::size_t >& /*addedTets*/,
  const std::set< std::size_t >& removedNodes,
  const tk::NodeCommMap& nodeCommMap,
  const std::map< int, std::vector< std::size_t > >& bface,
  const std::map< int, std::vector< std::size_t > >& bnode,
  const std::vector< std::size_t >& triinpoel )
// *****************************************************************************
//  Receive new mesh from Refiner
//! \param[in] ginpoel Mesh connectivity with global node ids
//! \param[in] chunk New mesh chunk (connectivity and global<->local id maps)
//! \param[in] coord New mesh node coordinates
//! \param[in] addedNodes Newly added mesh nodes and their parents (local ids)
//! \param[in] addedTets Newly added mesh cells and their parents (local ids)
//! \param[in] removedNodes Newly removed mesh nodes (local ids)
//! \param[in] nodeCommMap New node communication map
//! \param[in] bface Boundary-faces mapped to side set ids
//! \param[in] bnode Boundary-node lists mapped to side set ids
//! \param[in] triinpoel Boundary-face connectivity
// *****************************************************************************
{
  auto d = Disc();

  d->Itf() = 0;  // Zero field output iteration count if AMR
  ++d->Itr();    // Increase number of iterations with a change in the mesh

  // Resize mesh data structures after mesh refinement
  d->resizePostAMR( chunk, coord, nodeCommMap );

  // Remove newly removed nodes from solution vectors
  m_u.rm(removedNodes);
  m_un.rm(removedNodes);
  m_rhs.rm(removedNodes);

  // Resize auxiliary solution vectors
  auto npoin = coord[0].size();
  auto nprop = m_u.nprop();
  m_u.resize( npoin );
  m_un.resize( npoin );
  m_rhs.resize( npoin );
  m_chBndGrad.resize( d->Bid().size() );

  // Update solution on new mesh
  for (const auto& n : addedNodes)
    for (std::size_t c=0; c<nprop; ++c)
      m_u(n.first,c,0) = (m_u(n.second[0],c,0) + m_u(n.second[1],c,0))/2.0;

  // Update physical-boundary node-, face-, and element lists
  m_bnode = bnode;
  m_bface = bface;
  m_triinpoel = tk::remap( triinpoel, d->Lid() );

  auto meshid = d->MeshId();
  contribute( sizeof(std::size_t), &meshid, CkReduction::nop,
              CkCallback(CkReductionTarget(Transporter,resized), d->Tr()) );
}
//! [Resize]

void
ALECG::resized()
// *****************************************************************************
// Resizing data sutrctures after mesh refinement has been completed
// *****************************************************************************
{
  resize_complete();
}

void
ALECG::transfer()
// *****************************************************************************
// Transfer solution to other solver and mesh if coupled
// *****************************************************************************
{
  // Initiate solution transfer (if coupled)
  //Disc()->transfer( m_u,
  //  CkCallback(CkIndex_ALECG::stage(), thisProxy[thisIndex]) );
  thisProxy[thisIndex].stage();
}

//! [stage]
void
ALECG::stage()
// *****************************************************************************
// Evaluate whether to continue with next time step stage
// *****************************************************************************
{
  // Increment Runge-Kutta stage counter
  ++m_stage;

  // if not all Runge-Kutta stages complete, continue to next time stage,
  // otherwise output field data to file(s)
  if (m_stage < 3) chBndGrad(); else out();
}
//! [stage]

void
ALECG::writeFields( CkCallback c )
// *****************************************************************************
// Output mesh-based fields to file
//! \param[in] c Function to continue with after the write
// *****************************************************************************
{
  if (g_inputdeck.get< tag::cmd, tag::benchmark >()) {

    c.send();

  } else {

    auto d = Disc();
    const auto& coord = d->Coord();

    // Query fields names requested by user
    auto nodefieldnames = numericFieldNames( tk::Centering::NODE );
    // Collect field output from numerical solution requested by user
    conserved( m_u, Disc()->Vol() );
    auto nodefields = numericFieldOutput( m_u, tk::Centering::NODE );
    volumetric( m_u, Disc()->Vol() );

    //! Lambda to put in a field for output if not empty
    auto add_node_field = [&]( const auto& name, const auto& field ){
      if (not field.empty()) {
        nodefieldnames.push_back( name );
        nodefields.push_back( field );
      }
    };

    // Output mesh velocity if ALE is enabled
    if (g_inputdeck.get< tag::ale, tag::ale >()) {
      const auto& w = d->meshvel();
      add_node_field( "x-mesh-velocity", w.extract(0,0) );
      add_node_field( "y-mesh-velocity", w.extract(1,0) );
      add_node_field( "z-mesh-velocity", w.extract(2,0) );
      add_node_field( "volume", d->Vol() );
    }

    // Collect field output names for analytical solutions
    for (const auto& eq : g_cgpde)
      analyticFieldNames( eq, tk::Centering::NODE, nodefieldnames );

    // Collect field output from analytical solutions (if exist)
    for (const auto& eq : g_cgpde)
      analyticFieldOutput( eq, tk::Centering::NODE, coord[0], coord[1],
                           coord[2], d->T(), nodefields );

    // Query and collect block and surface field names from PDEs integrated
    std::vector< std::string > nodesurfnames;
    for (const auto& eq : g_cgpde) {
      auto s = eq.surfNames();
      nodesurfnames.insert( end(nodesurfnames), begin(s), end(s) );
    }

    // Collect node block and surface field solution
    std::vector< std::vector< tk::real > > nodesurfs;
    conserved( m_u, Disc()->Vol() );
    for (const auto& eq : g_cgpde) {
      auto s = eq.surfOutput( tk::bfacenodes(m_bface,m_triinpoel), m_u );
      nodesurfs.insert( end(nodesurfs), begin(s), end(s) );
    }
    volumetric( m_u, Disc()->Vol() );

    Assert( nodefieldnames.size() == nodefields.size(), "Size mismatch" );

    // Send mesh and fields data (solution dump) for output to file
    d->write( d->Inpoel(), coord, m_bface, tk::remap(m_bnode,d->Lid()),
              m_triinpoel, {}, nodefieldnames, nodesurfnames, {}, nodefields,
              nodesurfs, c );

  }
}

void
ALECG::out()
// *****************************************************************************
// Output mesh field data
// *****************************************************************************
{
  auto d = Disc();

  // Output time history
  if (d->histiter() or d->histtime() or d->histrange()) {
    std::vector< std::vector< tk::real > > hist;
    conserved( m_u, Disc()->Vol() );
    for (const auto& eq : g_cgpde) {
      auto h = eq.histOutput( d->Hist(), d->Inpoel(), m_u );
      hist.insert( end(hist), begin(h), end(h) );
    }
    volumetric( m_u, Disc()->Vol() );
    d->history( std::move(hist) );
  }

  // Output field data
  if (d->fielditer() or d->fieldtime() or d->fieldrange() or m_finished)
    writeFields( CkCallback(CkIndex_ALECG::step(), thisProxy[thisIndex]) );
  else
    step();
}

void
ALECG::evalLB( int nrestart )
// *****************************************************************************
// Evaluate whether to do load balancing
//! \param[in] nrestart Number of times restarted
// *****************************************************************************
{
  auto d = Disc();

  // Detect if just returned from a checkpoint and if so, zero timers and
  // finished flag
  if (d->restarted( nrestart )) m_finished = 0;

  const auto lbfreq = g_inputdeck.get< tag::cmd, tag::lbfreq >();
  const auto nonblocking = g_inputdeck.get< tag::cmd, tag::nonblocking >();

  // Load balancing if user frequency is reached or after the second time-step
  if ( (d->It()) % lbfreq == 0 || d->It() == 2 ) {

    AtSync();
    if (nonblocking) next();

  } else {

    next();

  }
}

void
ALECG::evalRestart()
// *****************************************************************************
// Evaluate whether to save checkpoint/restart
// *****************************************************************************
{
  auto d = Disc();

  const auto rsfreq = g_inputdeck.get< tag::cmd, tag::rsfreq >();
  const auto benchmark = g_inputdeck.get< tag::cmd, tag::benchmark >();

  if (not benchmark and not (d->It() % rsfreq)) {

    std::vector< std::size_t > meshdata{ /* finished = */ 0, d->MeshId() };
    contribute( meshdata, CkReduction::nop,
      CkCallback(CkReductionTarget(Transporter,checkpoint), d->Tr()) );

  } else {

    evalLB( /* nrestart = */ -1 );

  }
}

void
ALECG::step()
// *****************************************************************************
// Evaluate whether to continue with next time step
// *****************************************************************************
{
  auto d = Disc();

  // Output one-liner status report to screen
  d->status();
  // Reset Runge-Kutta stage counter
  m_stage = 0;

  if (not m_finished) {

    evalRestart();

  } else {

    auto meshid = d->MeshId();
    d->contribute( sizeof(std::size_t), &meshid, CkReduction::nop,
                   CkCallback(CkReductionTarget(Transporter,finish), d->Tr()) );

  }
}

#include "NoWarning/alecg.def.h"