This file is indexed.

/usr/lib/petscdir/3.1/include/sieve/ParallelMapping.hh is in libpetsc3.1-dev 3.1.dfsg-11ubuntu1.

This file is owned by root:root, with mode 0o644.

The actual contents of the file can be viewed below.

   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
#ifndef included_ALE_ParallelMapping_hh
#define included_ALE_ParallelMapping_hh

#ifndef  included_ALE_IField_hh
#include <IField.hh>
#endif

#ifndef  included_ALE_Sections_hh
#include <Sections.hh>
#endif

#include <functional>

extern "C" PetscMPIInt Mesh_DelTag(MPI_Comm comm,PetscMPIInt keyval,void* attr_val,void* extra_state);

namespace ALE {
  template<class _Tp>
  struct Identity : public std::unary_function<_Tp,_Tp>
  {
    _Tp& operator()(_Tp& x) const {return x;}
    const _Tp& operator()(const _Tp& x) const {return x;}
  };

  template<class _Tp>
  struct IsEqual : public std::unary_function<_Tp, bool>, public std::binary_function<_Tp, _Tp, bool>
  {
    const _Tp& x;
    IsEqual(const _Tp& x) : x(x) {};
    bool operator()(_Tp& y) const {return x == y;}
    const bool operator()(const _Tp& y) const {return x == y;}
    bool operator()(_Tp& y, _Tp& dummy) const {return x == y;}
    const bool operator()(const _Tp& y, const _Tp& dummy) const {return x == y;}
  };

  // Creates new global point names and renames local points globally
  template<typename Point>
  class PointFactory : ALE::ParallelObject {
  public:
    typedef Point                           point_type;
    typedef std::map<point_type,point_type> renumbering_type;
    typedef std::map<int,std::map<point_type,point_type> > remote_renumbering_type;
  protected:
    point_type       originalMax;
    point_type       currentMax;
    renumbering_type renumbering;
    renumbering_type invRenumbering;
    remote_renumbering_type remoteRenumbering;
  protected:
    PointFactory(MPI_Comm comm, const int debug = 0) : ALE::ParallelObject(comm, debug), originalMax(-1) {};
  public:
    ~PointFactory() {};
  public:
    static PointFactory& singleton(MPI_Comm comm, const point_type& maxPoint, const int debug = 0, bool cleanup = false) {
      static PointFactory *_singleton = NULL;

      if (cleanup) {
        if (debug) {std::cout << "Destroying PointFactory" << std::endl;}
        if (_singleton) {delete _singleton;}
        _singleton = NULL;
      } else if (_singleton == NULL) {
        if (debug) {std::cout << "Creating new PointFactory" << std::endl;}
        _singleton  = new PointFactory(comm, debug);
        _singleton->setMax(maxPoint);
      }
      return *_singleton;
    };
    void setMax(const point_type& maxPoint) {
      PetscErrorCode ierr = MPI_Allreduce((void *) &maxPoint, &this->originalMax, 1, MPI_INT, MPI_MAX, this->comm());CHKERRXX(ierr);
      ++this->originalMax;
      this->currentMax = this->originalMax;
    };
    void clear() {
      this->currentMax = this->originalMax;
      this->renumbering.clear();
      this->invRenumbering.clear();
    };
  public:
    template<typename Iterator>
    void renumberPoints(const Iterator& begin, const Iterator& end) {
      renumberPoints(begin, end, Identity<typename Iterator::value_type>());
    };
    template<typename Iterator, typename KeyExtractor>
    void renumberPoints(const Iterator& begin, const Iterator& end, const KeyExtractor& ex) {
      int numPoints = 0, numGlobalPoints, firstPoint;

      for(Iterator p_iter = begin; p_iter != end; ++p_iter) ++numPoints;
      MPI_Allreduce(&numPoints, &numGlobalPoints, 1, MPI_INT, MPI_SUM, this->comm());
      MPI_Scan(&numPoints, &firstPoint, 1, MPI_INT, MPI_SUM, this->comm());
      firstPoint += this->currentMax - numPoints;
      this->currentMax += numGlobalPoints;
      for(Iterator p_iter = begin; p_iter != end; ++p_iter, ++firstPoint) {
        if (this->debug()) {std::cout << "["<<this->commRank()<<"]: New point " << ex(*p_iter) << " --> " << firstPoint << std::endl;}
        this->renumbering[firstPoint]     = ex(*p_iter);
        this->invRenumbering[ex(*p_iter)] = firstPoint;
      }
    };
  public:
    void constructRemoteRenumbering() {
      const int localSize   = this->renumbering.size();
      int      *remoteSizes = new int[this->commSize()];
      int      *localMap    = new int[localSize*2];
      int      *recvCounts  = new int[this->commSize()];
      int      *displs      = new int[this->commSize()];

      // Populate arrays
      int r = 0;
      for(typename renumbering_type::const_iterator r_iter = renumbering.begin(); r_iter != renumbering.end(); ++r_iter, ++r) {
        localMap[r*2+0] = r_iter->first;
        localMap[r*2+1] = r_iter->second;
      }
      // Communicate renumbering sizes
      MPI_Allgather((void*) &localSize, 1, MPI_INT, remoteSizes, 1, MPI_INT, this->comm());
      for(int p = 0; p < this->commSize(); ++p) {
        recvCounts[p] = remoteSizes[p]*2;
        if (p == 0) {
          displs[p]   = 0;
        } else {
          displs[p]   = displs[p-1] + recvCounts[p-1];
        }
      }
      int *remoteMaps = new int[displs[this->commSize()-1]+recvCounts[this->commSize()-1]];
      // Communicate renumberings
      MPI_Allgatherv(localMap, localSize*2, MPI_INT, remoteMaps, recvCounts, displs, MPI_INT, this->comm());
      // Populate maps
      for(int p = 0; p < this->commSize(); ++p) {
        if (p == this->commRank()) continue;
        int offset = displs[p];

        for(int r = 0; r < remoteSizes[p]; ++r) {
          this->remoteRenumbering[p][remoteMaps[r*2+0+offset]] = remoteMaps[r*2+1+offset];
          if (this->debug()) {std::cout << "["<<this->commRank()<<"]: Remote renumbering["<<p<<"] " << remoteMaps[r*2+0+offset] << " --> " << remoteMaps[r*2+1+offset] << std::endl;}
        }
      }
      // Cleanup
      delete [] recvCounts;
      delete [] displs;
      delete [] localMap;
      delete [] remoteMaps;
      delete [] remoteSizes;
    };
  public:
    // global point --> local point
    renumbering_type& getRenumbering() {
      return this->renumbering;
    };
    // local point --> global point
    renumbering_type& getInvRenumbering() {
      return this->invRenumbering;
    };
    // rank --> global point --> local point
    remote_renumbering_type& getRemoteRenumbering() {
      return this->remoteRenumbering;
    };
  };

  // TODO: Check MPI return values and status of Waits
  template<typename Value_>
  class MPIMover : public ALE::ParallelObject {
  public:
    typedef Value_                                 value_type;
    typedef size_t                                 num_type;
    typedef std::pair<num_type, const value_type*> move_type;
    typedef std::map<int, move_type>               moves_type;
    typedef std::vector<MPI_Request>               requests_type;
  protected:
    bool          _createdType;
    int           _tag;
    MPI_Datatype  _datatype;
    moves_type    _sends;
    moves_type    _recvs;
    requests_type _requests;
  public:
    MPIMover(MPI_Comm comm, const int debug = 0) : ParallelObject(comm, debug), _createdType(0) {
      this->_tag      = this->getNewTag();
      this->_datatype = this->getMPIDatatype();
    };
    MPIMover(MPI_Comm comm, const int tag, const int debug) : ParallelObject(comm, debug), _createdType(0), _tag(tag) {
      this->_datatype = this->getMPIDatatype();
    };
    MPIMover(MPI_Comm comm, const MPI_Datatype datatype, const int tag, const int debug) : ParallelObject(comm, debug), _createdType(0) {
      if (tag == MPI_UNDEFINED) {
        this->_tag      = this->getNewTag();
      } else {
        this->_tag      = tag;
      }
      if (datatype == MPI_DATATYPE_NULL) {
        this->_datatype = this->getMPIDatatype();
      } else {
        this->_datatype = datatype;
      }
    };
    ~MPIMover() {
      if (_createdType) {
        int ierr = MPI_Type_free(&this->_datatype);CHKERRXX(ierr);
      }
    };
  protected:
    // TODO: Can rewrite this with template specialization?
    MPI_Datatype getMPIDatatype() {
      if (sizeof(value_type) == 1) {
        return MPI_BYTE;
      } else if (sizeof(value_type) == 2) {
        return MPI_SHORT;
      } else if (sizeof(value_type) == 4) {
        return MPI_INT;
      } else if (sizeof(value_type) == 8) {
        return MPI_DOUBLE;
      } else if (sizeof(value_type) == 28) {
        int          blen[2], ierr;
        MPI_Aint     indices[2];
        MPI_Datatype oldtypes[2], newtype;
        blen[0] = 1; indices[0] = 0;           oldtypes[0] = MPI_INT;
        blen[1] = 3; indices[1] = sizeof(int); oldtypes[1] = MPI_DOUBLE;
        ierr = MPI_Type_struct(2, blen, indices, oldtypes, &newtype);CHKERRXX(ierr);
        ierr = MPI_Type_commit(&newtype);CHKERRXX(ierr);
        this->_createdType = true;
        return newtype;
      } else if (sizeof(value_type) == 32) {
        int          blen[2], ierr;
        MPI_Aint     indices[2];
        MPI_Datatype oldtypes[2], newtype;
        blen[0] = 1; indices[0] = 0;           oldtypes[0] = MPI_DOUBLE;
        blen[1] = 3; indices[1] = sizeof(int); oldtypes[1] = MPI_DOUBLE;
        ierr = MPI_Type_struct(2, blen, indices, oldtypes, &newtype);CHKERRXX(ierr);
        ierr = MPI_Type_commit(&newtype);CHKERRXX(ierr);
        this->_createdType = true;
        return newtype;
      }
      ostringstream msg;

      msg << "Cannot determine MPI type for value type with size " << sizeof(value_type);
      throw PETSc::Exception(msg.str().c_str());
    };
    int getNewTag() const {
      static int tagKeyval = MPI_KEYVAL_INVALID;
      int *tagvalp = NULL, *maxval, flg, ierr;

      if (tagKeyval == MPI_KEYVAL_INVALID) {
        tagvalp = (int *) malloc(sizeof(int));
        ierr = MPI_Keyval_create(MPI_NULL_COPY_FN, Mesh_DelTag, &tagKeyval, (void *) NULL);CHKERRXX(ierr);
        ierr = MPI_Attr_put(this->comm(), tagKeyval, tagvalp);CHKERRXX(ierr);
        tagvalp[0] = 0;
      }
      ierr = MPI_Attr_get(this->comm(), tagKeyval, (void **) &tagvalp, &flg);CHKERRXX(ierr);
      if (tagvalp[0] < 1) {
        ierr = MPI_Attr_get(MPI_COMM_WORLD, MPI_TAG_UB, (void **) &maxval, &flg);CHKERRXX(ierr);
        tagvalp[0] = *maxval - 128; // hope that any still active tags were issued right at the beginning of the run
      }
      if (this->debug()) {
        std::cout << "[" << this->commRank() << "]Got new tag " << tagvalp[0] << std::endl;
      }
      return tagvalp[0]--;
    };
    void constructRequests() {
      this->_requests.clear();

      for(typename moves_type::const_iterator s_iter = this->_sends.begin(); s_iter != this->_sends.end(); ++s_iter) {
        const int   rank = s_iter->first;
        const int   num  = s_iter->second.first;
        void       *data = (void *) s_iter->second.second;
        MPI_Request request;
        int         ierr;

        if (this->_debug) {std::cout <<"["<<this->commRank()<<"] Sending data (" << num << ") to " << rank << " tag " << this->_tag << std::endl;}
        ierr = MPI_Send_init(data, num, this->_datatype, rank, this->_tag, this->comm(), &request);CHKERRXX(ierr);
        this->_requests.push_back(request);
#if defined(PETSC_USE_LOG)
        // PETSc logging
        isend_ct++;
        TypeSize(&isend_len, num, this->_datatype);
#endif
      }
      for(typename moves_type::const_iterator r_iter = this->_recvs.begin(); r_iter != this->_recvs.end(); ++r_iter) {
        const int   rank = r_iter->first;
        const int   num  = r_iter->second.first;
        void       *data = (void *) r_iter->second.second;
        MPI_Request request;
        int         ierr;

        if (this->_debug) {std::cout <<"["<<this->commRank()<<"] Receiving data (" << num << ") from " << rank << " tag " << this->_tag << std::endl;}
        ierr = MPI_Recv_init(data, num, this->_datatype, rank, this->_tag, this->comm(), &request);CHKERRXX(ierr);
        this->_requests.push_back(request);
#if defined(PETSC_USE_LOG)
        // PETSc logging
        irecv_ct++;
        TypeSize(&irecv_len, num, this->_datatype);
#endif
      }
    };
  public:
    void send(const int rank, const int num, const value_type *data) {
      this->_sends[rank] = move_type(num, data);
    };
    void recv(const int rank, const int num, const value_type *data) {
      this->_recvs[rank] = move_type(num, data);
    };
    void start() {
      this->constructRequests();
      for(typename requests_type::const_iterator r_iter = this->_requests.begin(); r_iter != this->_requests.end(); ++r_iter) {
        MPI_Request request = *r_iter;

        int ierr = MPI_Start(&request);CHKERRXX(ierr);
      }
    };
    void end() {
      MPI_Status status;

      for(typename requests_type::const_iterator r_iter = this->_requests.begin(); r_iter != this->_requests.end(); ++r_iter) {
        MPI_Request request = *r_iter;

        int ierr = MPI_Wait(&request, &status);CHKERRXX(ierr);
      }
      for(typename requests_type::const_iterator r_iter = this->_requests.begin(); r_iter != this->_requests.end(); ++r_iter) {
        MPI_Request request = *r_iter;

        int ierr = MPI_Request_free(&request);CHKERRXX(ierr);
      }
    };
  };
  template<typename Alloc_ = malloc_allocator<int> >
  class OverlapBuilder {
  public:
    typedef Alloc_ alloc_type;
  protected:
    template<typename T>
    struct lessPair: public std::binary_function<std::pair<T,T>, std::pair<T,T>, bool> {
      bool operator()(const std::pair<T,T>& x, const std::pair<T,T>& y) const {
        return x.first < y.first;
      }
    };
    template<typename T>
    struct mergePair: public std::binary_function<std::pair<T,T>, std::pair<T,T>, bool> {
      std::pair<T,std::pair<T,T> > operator()(const std::pair<T,T>& x, const std::pair<T,T>& y) const {
        return std::pair<T,std::pair<T,T> >(x.first, std::pair<T,T>(x.second, y.second));
      }
    };
    template<typename _InputIterator1, typename _InputIterator2, typename _OutputIterator, typename _Compare, typename _Merge>
    static _OutputIterator set_intersection_merge(_InputIterator1 __first1, _InputIterator1 __last1,
                                           _InputIterator2 __first2, _InputIterator2 __last2,
                                           _OutputIterator __result, _Compare __comp, _Merge __merge)
    {
      while(__first1 != __last1 && __first2 != __last2) {
        if (__comp(*__first1, *__first2))
          ++__first1;
        else if (__comp(*__first2, *__first1))
          ++__first2;
        else
        {
          *__result = __merge(*__first1, *__first2);
          ++__first1;
          ++__first2;
          ++__result;
        }
      }
      return __result;
    };
  public:
    template<typename Sequence, typename Renumbering, typename SendOverlap, typename RecvOverlap>
    static void constructOverlap(const Sequence& points, Renumbering& renumbering, const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap) {
      typedef typename SendOverlap::source_type point_type;
      typedef std::pair<point_type,point_type>  pointPair;
      typedef std::pair<point_type,pointPair>   pointTriple;
      alloc_type allocator;
      typename alloc_type::template rebind<point_type>::other point_allocator;
      typename alloc_type::template rebind<pointPair>::other  pointPair_allocator;
      const MPI_Comm comm     = sendOverlap->comm();
      const int      commSize = sendOverlap->commSize();
      const int      commRank = sendOverlap->commRank();
      point_type    *sendBuf  = point_allocator.allocate(points.size()*2);
      for(size_t i = 0; i < points.size()*2; ++i) {point_allocator.construct(sendBuf+i, point_type());}
      int            size     = 0;
      const int      debug    = sendOverlap->debug();
      for(typename Sequence::const_iterator l_iter = points.begin(); l_iter != points.end(); ++l_iter) {
        if (debug) {std::cout << "["<<commRank<<"]Send point["<<size<<"]: " << *l_iter << " " << renumbering[*l_iter] << std::endl;}
        sendBuf[size++] = *l_iter;
        sendBuf[size++] = renumbering[*l_iter];
      }
      int *sizes = allocator.allocate(commSize*3+2); // [size]   The number of points coming from each process
      for(int i = 0; i < commSize*3+2; ++i) {allocator.construct(sizes+i, 0);}
      int *offsets = sizes+commSize;                 // [size+1] Prefix sums for sizes
      int *oldOffs = offsets+commSize+1;             // [size+1] Temporary storage
      pointPair  *remotePoints = NULL;               // The points from each process
      int        *remoteRanks  = NULL;               // The rank and number of overlap points of each process that overlaps another
      int         numRemotePoints = 0;
      int         numRemoteRanks  = 0;

      // Change to Allgather() for the correct binning algorithm
      MPI_Gather(&size, 1, MPI_INT, sizes, 1, MPI_INT, 0, comm);
      if (commRank == 0) {
        offsets[0] = 0;
        for(int p = 1; p <= commSize; p++) {
          offsets[p] = offsets[p-1] + sizes[p-1];
        }
        numRemotePoints = offsets[commSize];
        remotePoints    = pointPair_allocator.allocate(numRemotePoints/2);
        for(int i = 0; i < numRemotePoints/2; ++i) {pointPair_allocator.construct(remotePoints+i, pointPair());}
      }
      MPI_Gatherv(sendBuf, size, MPI_INT, remotePoints, sizes, offsets, MPI_INT, 0, comm);
      for(size_t i = 0; i < points.size(); ++i) {point_allocator.destroy(sendBuf+i);}
      point_allocator.deallocate(sendBuf, points.size());
      std::map<int, std::map<int, std::set<pointTriple> > > overlapInfo; // Maps (p,q) to their set of overlap points

      if (commRank == 0) {
        for(int p = 0; p <= commSize; p++) {
          offsets[p] /= 2;
        }
        for(int p = 0; p < commSize; p++) {
          std::sort(&remotePoints[offsets[p]], &remotePoints[offsets[p+1]], lessPair<point_type>());
        }
        for(int p = 0; p <= commSize; p++) {
          oldOffs[p] = offsets[p];
        }
        for(int p = 0; p < commSize; p++) {
          for(int q = 0; q < commSize; q++) {
            if (p == q) continue;
            set_intersection_merge(&remotePoints[oldOffs[p]], &remotePoints[oldOffs[p+1]],
                                   &remotePoints[oldOffs[q]], &remotePoints[oldOffs[q+1]],
                                   std::insert_iterator<std::set<pointTriple> >(overlapInfo[p][q], overlapInfo[p][q].begin()),
                                   lessPair<point_type>(), mergePair<point_type>());
          }
          sizes[p]     = overlapInfo[p].size()*2;
          offsets[p+1] = offsets[p] + sizes[p];
        }
        numRemoteRanks = offsets[commSize];
        remoteRanks    = allocator.allocate(numRemoteRanks);
        for(int i = 0; i < numRemoteRanks; ++i) {allocator.construct(remoteRanks+i, 0);}
        int     k = 0;
        for(int p = 0; p < commSize; p++) {
          for(typename std::map<int, std::set<pointTriple> >::iterator r_iter = overlapInfo[p].begin(); r_iter != overlapInfo[p].end(); ++r_iter) {
            remoteRanks[k*2]   = r_iter->first;
            remoteRanks[k*2+1] = r_iter->second.size();
            k++;
          }
        }
      }
      int numOverlaps;                          // The number of processes overlapping this process
      MPI_Scatter(sizes, 1, MPI_INT, &numOverlaps, 1, MPI_INT, 0, comm);
      int *overlapRanks = allocator.allocate(numOverlaps); // The rank and overlap size for each overlapping process
      for(int i = 0; i < numOverlaps; ++i) {allocator.construct(overlapRanks+i, 0);}
      MPI_Scatterv(remoteRanks, sizes, offsets, MPI_INT, overlapRanks, numOverlaps, MPI_INT, 0, comm);
      point_type *sendPoints    = NULL;         // The points to send to each process
      int         numSendPoints = 0;
      if (commRank == 0) {
        for(int p = 0, k = 0; p < commSize; p++) {
          sizes[p] = 0;
          for(int r = 0; r < (int) overlapInfo[p].size(); r++) {
            sizes[p] += remoteRanks[k*2+1]*2;
            k++;
          }
          offsets[p+1] = offsets[p] + sizes[p];
        }
        numSendPoints = offsets[commSize];
        sendPoints    = point_allocator.allocate(numSendPoints*2);
        for(int i = 0; i < numSendPoints*2; ++i) {point_allocator.construct(sendPoints+i, point_type());}
        for(int p = 0, k = 0; p < commSize; p++) {
          for(typename std::map<int, std::set<pointTriple> >::const_iterator r_iter = overlapInfo[p].begin(); r_iter != overlapInfo[p].end(); ++r_iter) {
            int rank = r_iter->first;
            for(typename std::set<pointTriple>::const_iterator p_iter = (overlapInfo[p][rank]).begin(); p_iter != (overlapInfo[p][rank]).end(); ++p_iter) {
              sendPoints[k++] = p_iter->first;
              sendPoints[k++] = p_iter->second.second;
              if (debug) {std::cout << "["<<commRank<<"]Sending points " << p_iter->first << " " << p_iter->second.second << " to rank " << rank << std::endl;}
            }
          }
        }
      }
      int numOverlapPoints = 0;
      for(int r = 0; r < numOverlaps/2; r++) {
        numOverlapPoints += overlapRanks[r*2+1];
      }
      point_type *overlapPoints = point_allocator.allocate(numOverlapPoints*2);
      for(int i = 0; i < numOverlapPoints*2; ++i) {point_allocator.construct(overlapPoints+i, point_type());}
      MPI_Scatterv(sendPoints, sizes, offsets, MPI_INT, overlapPoints, numOverlapPoints*2, MPI_INT, 0, comm);

      for(int r = 0, k = 0; r < numOverlaps/2; r++) {
        int rank = overlapRanks[r*2];

        for(int p = 0; p < overlapRanks[r*2+1]; p++) {
          point_type point       = overlapPoints[k++];
          point_type remotePoint = overlapPoints[k++];

          if (debug) {std::cout << "["<<commRank<<"]Matched up remote point " << remotePoint << "("<<point<<") to local " << renumbering[point] << std::endl;}
          sendOverlap->addArrow(renumbering[point], rank, remotePoint);
          recvOverlap->addArrow(rank, renumbering[point], remotePoint);
        }
      }

      for(int i = 0; i < numOverlapPoints; ++i) {point_allocator.destroy(overlapPoints+i);}
      point_allocator.deallocate(overlapPoints, numOverlapPoints);
      for(int i = 0; i < numOverlaps; ++i) {allocator.destroy(overlapRanks+i);}
      allocator.deallocate(overlapRanks, numOverlaps);
      for(int i = 0; i < commSize*3+2; ++i) {allocator.destroy(sizes+i);}
      allocator.deallocate(sizes, commSize*3+2);
      if (commRank == 0) {
        for(int i = 0; i < numRemoteRanks; ++i) {allocator.destroy(remoteRanks+i);}
        allocator.deallocate(remoteRanks, numRemoteRanks);
        for(int i = 0; i < numRemotePoints; ++i) {pointPair_allocator.destroy(remotePoints+i);}
        pointPair_allocator.deallocate(remotePoints, numRemotePoints);
        for(int i = 0; i < numSendPoints; ++i) {point_allocator.destroy(sendPoints+i);}
        point_allocator.deallocate(sendPoints, numSendPoints);
      }
    };
  };
  namespace Pullback {
    class SimpleCopy {
    public:
      // Copy the overlap section to the related processes
      //   This version is for Constant sections, meaning the same, single value over all points
      template<typename SendOverlap, typename RecvOverlap, typename SendSection, typename RecvSection>
      static void copyConstant(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<SendSection>& sendSection, const Obj<RecvSection>& recvSection) {
        MPIMover<char>                             pMover(sendSection->comm(), sendSection->debug());
        MPIMover<typename SendSection::value_type> vMover(sendSection->comm(), sendSection->debug());
        std::map<int, char *>                      sendPoints;
        std::map<int, char *>                      recvPoints;
        typename SendSection::alloc_type::template rebind<char>::other sendAllocator;
        typename RecvSection::alloc_type::template rebind<char>::other recvAllocator;

        const Obj<typename SendOverlap::traits::baseSequence>      sRanks  = sendOverlap->base();
        const typename SendOverlap::traits::baseSequence::iterator sEnd    = sRanks->end();
        const typename SendSection::value_type                    *sValues = sendSection->restrictSpace();

        for(typename SendOverlap::traits::baseSequence::iterator r_iter = sRanks->begin(); r_iter != sEnd; ++r_iter) {
          const Obj<typename SendOverlap::coneSequence>&     points = sendOverlap->cone(*r_iter);
          const int                                          pSize  = points->size();
          const typename SendOverlap::coneSequence::iterator pEnd   = points->end();
          char                                              *v      = sendAllocator.allocate(points->size());
          int                                                k      = 0;

          for(int i = 0; i < pSize; ++i) {sendAllocator.construct(v+i, 0);}
          for(typename SendOverlap::coneSequence::iterator p_iter = points->begin(); p_iter != pEnd; ++p_iter, ++k) {
            v[k] = (char) sendSection->hasPoint(*p_iter);
          }
          sendPoints[*r_iter] = v;
          pMover.send(*r_iter, pSize, sendPoints[*r_iter]);
          vMover.send(*r_iter, 2, sValues);
        }
        const Obj<typename RecvOverlap::traits::capSequence>      rRanks  = recvOverlap->cap();
        const typename RecvOverlap::traits::capSequence::iterator rEnd    = rRanks->end();
        const typename RecvSection::value_type                   *rValues = recvSection->restrictSpace();

        for(typename RecvOverlap::traits::capSequence::iterator r_iter = rRanks->begin(); r_iter != rEnd; ++r_iter) {
          const Obj<typename RecvOverlap::traits::supportSequence>& points = recvOverlap->support(*r_iter);
          const int                                                 pSize  = points->size();
          char                                                     *v      = recvAllocator.allocate(pSize);

          for(int i = 0; i < pSize; ++i) {recvAllocator.construct(v+i, 0);}
          recvPoints[*r_iter] = v;
          pMover.recv(*r_iter, pSize, recvPoints[*r_iter]);
          vMover.recv(*r_iter, 2, rValues);
        }
        pMover.start();
        pMover.end();
        vMover.start();
        vMover.end();
        for(typename RecvOverlap::traits::capSequence::iterator r_iter = rRanks->begin(); r_iter != rEnd; ++r_iter) {
          const Obj<typename RecvOverlap::traits::supportSequence>&     points = recvOverlap->support(*r_iter);
          const typename RecvOverlap::traits::supportSequence::iterator pEnd   = points->end();
          const char                                                   *v      = recvPoints[*r_iter];
          int                                                           k      = 0;

          for(typename RecvOverlap::traits::supportSequence::iterator s_iter = points->begin(); s_iter != pEnd; ++s_iter, ++k) {
            if (v[k]) {recvSection->addPoint(typename RecvSection::point_type(*r_iter, s_iter.color()));}
          }
        }
        for(typename SendOverlap::traits::baseSequence::iterator r_iter = sRanks->begin(); r_iter != sEnd; ++r_iter) {
          sendAllocator.deallocate(sendPoints[*r_iter], sendOverlap->cone(*r_iter)->size());
        }
        for(typename RecvOverlap::traits::capSequence::iterator r_iter = rRanks->begin(); r_iter != rEnd; ++r_iter) {
          recvAllocator.deallocate(recvPoints[*r_iter], recvOverlap->support(*r_iter)->size());
        }
      };
      // Specialize to ArrowSection
      template<typename SendOverlap, typename RecvOverlap, typename SendSection, typename RecvSection>
      static void copyConstantArrow(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<SendSection>& sendSection, const Obj<RecvSection>& recvSection) {
        MPIMover<char>                             pMover(sendSection->comm(), sendSection->debug());
        MPIMover<typename SendSection::value_type> vMover(sendSection->comm(), sendSection->debug());
        std::map<int, char *>                      sendPoints;
        std::map<int, char *>                      recvPoints;
        typename SendSection::alloc_type::template rebind<char>::other sendAllocator;
        typename RecvSection::alloc_type::template rebind<char>::other recvAllocator;

        const Obj<typename SendOverlap::traits::baseSequence>      sRanks  = sendOverlap->base();
        const typename SendOverlap::traits::baseSequence::iterator sEnd    = sRanks->end();
        const typename SendSection::value_type                    *sValues = sendSection->restrictSpace();

        for(typename SendOverlap::traits::baseSequence::iterator r_iter = sRanks->begin(); r_iter != sEnd; ++r_iter) {
          const Obj<typename SendOverlap::coneSequence>&     points = sendOverlap->cone(*r_iter);
          const int                                          pSize  = points->size();
          const typename SendOverlap::coneSequence::iterator pBegin = points->begin();
          const typename SendOverlap::coneSequence::iterator pEnd   = points->end();
          char                                              *v      = sendAllocator.allocate(pSize*pSize);
          int                                                k      = 0;

          for(size_t i = 0; i < pSize*pSize; ++i) {sendAllocator.construct(v+i, 0);}
          for(typename SendOverlap::coneSequence::iterator p_iter = pBegin; p_iter != pEnd; ++p_iter) {
            for(typename SendOverlap::coneSequence::iterator q_iter = pBegin; q_iter != pEnd; ++q_iter, ++k) {
              v[k] = (char) sendSection->hasPoint(typename SendSection::point_type(*p_iter,*q_iter));
            }
          }
          sendPoints[*r_iter] = v;
          pMover.send(*r_iter, pSize*pSize, sendPoints[*r_iter]);
          vMover.send(*r_iter, 2, sValues);
        }
        const Obj<typename RecvOverlap::traits::capSequence>      rRanks  = recvOverlap->cap();
        const typename RecvOverlap::traits::capSequence::iterator rEnd    = rRanks->end();
        const typename RecvSection::value_type                   *rValues = recvSection->restrictSpace();

        for(typename RecvOverlap::traits::capSequence::iterator r_iter = rRanks->begin(); r_iter != rEnd; ++r_iter) {
          const Obj<typename RecvOverlap::traits::supportSequence>& points = recvOverlap->support(*r_iter);
          const int                                                 pSize  = points->size();
          char                                                     *v      = recvAllocator.allocate(points->size()*points->size());

          for(size_t i = 0; i < pSize*pSize; ++i) {recvAllocator.construct(v+i, 0);}
          recvPoints[*r_iter] = v;
          pMover.recv(*r_iter, pSize*pSize, recvPoints[*r_iter]);
          vMover.recv(*r_iter, 2, rValues);
        }
        pMover.start();
        pMover.end();
        vMover.start();
        vMover.end();
        for(typename RecvOverlap::traits::capSequence::iterator r_iter = rRanks->begin(); r_iter != rEnd; ++r_iter) {
          const Obj<typename RecvOverlap::traits::supportSequence>&     points = recvOverlap->support(*r_iter);
          const typename RecvOverlap::traits::supportSequence::iterator pBegin = points->begin();
          const typename RecvOverlap::traits::supportSequence::iterator pEnd   = points->end();
          const char                                                   *v      = recvPoints[*r_iter];
          int                                                           k      = 0;

          for(typename RecvOverlap::traits::supportSequence::iterator s_iter = pBegin; s_iter != pEnd; ++s_iter) {
            for(typename RecvOverlap::traits::supportSequence::iterator t_iter = pBegin; t_iter != pEnd; ++t_iter, ++k) {
              if (v[k]) {recvSection->addPoint(typename RecvSection::point_type(s_iter.color(),t_iter.color()));}
            }
          }
        }
        for(typename SendOverlap::traits::baseSequence::iterator r_iter = sRanks->begin(); r_iter != sEnd; ++r_iter) {
          sendAllocator.deallocate(sendPoints[*r_iter], sendOverlap->cone(*r_iter)->size()*sendOverlap->cone(*r_iter)->size());
        }
        for(typename RecvOverlap::traits::capSequence::iterator r_iter = rRanks->begin(); r_iter != rEnd; ++r_iter) {
          recvAllocator.deallocate(recvPoints[*r_iter], recvOverlap->support(*r_iter)->size()*recvOverlap->support(*r_iter)->size());
        }
      };
      // Copy the overlap section to the related processes
      //   This version is for IConstant sections, meaning the same, single value over all points
      template<typename SendOverlap, typename RecvOverlap, typename SendSection, typename RecvSection>
      static void copyIConstant(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<SendSection>& sendSection, const Obj<RecvSection>& recvSection) {
        MPIMover<typename SendSection::point_type> pMover(sendSection->comm(), sendSection->debug());
        MPIMover<typename SendSection::value_type> vMover(sendSection->comm(), sendSection->debug());
        std::map<int, typename SendSection::point_type *> sendPoints;
        std::map<int, typename SendSection::point_type *> recvPoints;
        typename SendSection::alloc_type::template rebind<typename SendSection::point_type>::other sendAllocator;
        typename RecvSection::alloc_type::template rebind<typename SendSection::point_type>::other recvAllocator;

        const Obj<typename SendOverlap::traits::baseSequence>      sRanks  = sendOverlap->base();
        const typename SendOverlap::traits::baseSequence::iterator sEnd    = sRanks->end();
        const typename SendSection::value_type                    *sValues = sendSection->restrictSpace();

        for(typename SendOverlap::traits::baseSequence::iterator r_iter = sRanks->begin(); r_iter != sEnd; ++r_iter) {
          typename SendSection::point_type *v = sendAllocator.allocate(2);

          for(size_t i = 0; i < 2; ++i) {sendAllocator.construct(v+i, 0);}
          v[0] = sendSection->getChart().min();
          v[1] = sendSection->getChart().max();
          sendPoints[*r_iter] = v;
          pMover.send(*r_iter, 2, sendPoints[*r_iter]);
          vMover.send(*r_iter, 2, sValues);
          ///std::cout << "["<<sendOverlap->commRank()<<"]Sending chart (" << v[0] << ", " << v[1] << ") with values (" << sValues[0] << ", " << sValues[1] << ") to process " << *r_iter << std::endl;
        }
        const Obj<typename RecvOverlap::traits::capSequence>      rRanks  = recvOverlap->cap();
        const typename RecvOverlap::traits::capSequence::iterator rEnd    = rRanks->end();
        const typename RecvSection::value_type                   *rValues = recvSection->restrictSpace();

        for(typename RecvOverlap::traits::capSequence::iterator r_iter = rRanks->begin(); r_iter != rEnd; ++r_iter) {
          typename SendSection::point_type *v = recvAllocator.allocate(2);

          for(size_t i = 0; i < 2; ++i) {recvAllocator.construct(v+i, 0);}
          recvPoints[*r_iter] = v;
          pMover.recv(*r_iter, 2, recvPoints[*r_iter]);
          vMover.recv(*r_iter, 2, rValues);
        }
        pMover.start();
        pMover.end();
        vMover.start();
        vMover.end();

        typename SendSection::point_type min = -1;
        typename SendSection::point_type max = -1;

        for(typename RecvOverlap::traits::capSequence::iterator r_iter = rRanks->begin(); r_iter != rEnd; ++r_iter) {
          const typename RecvSection::point_type *v = recvPoints[*r_iter];
          typename SendSection::point_type        newMin = v[0];
          typename SendSection::point_type        newMax = v[1]-1;
          //int                                     pSize  = 0;

          ///std::cout << "["<<recvOverlap->commRank()<<"]Received chart (" << v[0] << ", " << v[1] << ") from process " << *r_iter << std::endl;
#if 0
          // Translate to local numbering
          if (recvOverlap->support(*r_iter)->size()) {
            while(!pSize) {
              const Obj<typename RecvOverlap::supportSequence>& points = recvOverlap->support(*r_iter, newMin);
              pSize = points->size();
              if (pSize) {
                newMin = *points->begin();
              } else {
                newMin++;
              }
            }
            pSize  = 0;
            while(!pSize) {
              const Obj<typename RecvOverlap::supportSequence>& points = recvOverlap->support(*r_iter, newMax);
              pSize = points->size();
              if (pSize) {
                newMax = *points->begin();
              } else {
                newMax--;
              }
            }
          }
          std::cout << "["<<recvOverlap->commRank()<<"]Translated to chart (" << newMin << ", " << newMax+1 << ") from process " << *r_iter << std::endl;
#endif
          // Update chart
          if (min < 0) {
            min = newMin;
            max = newMax+1;
          } else {
            min = std::min(min, newMin);
            max = std::max(max, (typename SendSection::point_type) (newMax+1));
          }
        }
        if (!rRanks->size()) {min = max = 0;}
        recvSection->setChart(typename RecvSection::chart_type(min, max));
        for(typename SendOverlap::traits::baseSequence::iterator r_iter = sRanks->begin(); r_iter != sEnd; ++r_iter) {
          sendAllocator.deallocate(sendPoints[*r_iter], 2);
        }
        for(typename RecvOverlap::traits::capSequence::iterator r_iter = rRanks->begin(); r_iter != rEnd; ++r_iter) {
          recvAllocator.deallocate(recvPoints[*r_iter], 2);
        }
      };
      // Copy the overlap section to the related processes
      //   This version is for different sections, possibly with different data types
      // TODO: Can cache MPIMover objects (like a VecScatter)
      template<typename SendOverlap, typename RecvOverlap, typename SendSection, typename RecvSection>
      static void copy(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<SendSection>& sendSection, const Obj<RecvSection>& recvSection, const MPI_Datatype datatype = MPI_DATATYPE_NULL) {
        typedef std::pair<int, typename SendSection::value_type *> allocPair;
        typedef typename RecvSection::point_type                   recv_point_type;
        const Obj<typename SendSection::atlas_type>& sendAtlas = sendSection->getAtlas();
        const Obj<typename RecvSection::atlas_type>& recvAtlas = recvSection->getAtlas();
        MPIMover<typename SendSection::value_type>   vMover(sendSection->comm(), datatype, MPI_UNDEFINED, sendSection->debug());
        std::map<int, allocPair>                     sendValues;
        std::map<int, allocPair>                     recvValues;
        typename SendSection::alloc_type             sendAllocator;
        typename RecvSection::alloc_type             recvAllocator;

        copy(sendOverlap, recvOverlap, sendAtlas, recvAtlas);
        const Obj<typename SendOverlap::traits::baseSequence>      sRanks = sendOverlap->base();
        const typename SendOverlap::traits::baseSequence::iterator sEnd   = sRanks->end();

        // TODO: This should be const_iterator, but Sifter sucks
        for(typename SendOverlap::traits::baseSequence::iterator r_iter = sRanks->begin(); r_iter != sEnd; ++r_iter) {
          const Obj<typename SendOverlap::coneSequence>&     points  = sendOverlap->cone(*r_iter);
          const typename SendOverlap::coneSequence::iterator pEnd    = points->end();
          int                                                numVals = 0;

          // TODO: This should be const_iterator, but Sifter sucks
          for(typename SendOverlap::coneSequence::iterator c_iter = points->begin(); c_iter != pEnd; ++c_iter) {
            numVals += sendSection->getFiberDimension(*c_iter);
          }
          typename SendSection::value_type *v = sendAllocator.allocate(numVals);
          int                               k = 0;

          for(int i = 0; i < numVals; ++i) {sendAllocator.construct(v+i, 0);}
          for(typename SendOverlap::coneSequence::iterator c_iter = points->begin(); c_iter != pEnd; ++c_iter) {
            const typename SendSection::value_type *vals = sendSection->restrictPoint(*c_iter);

            for(int i = 0; i < sendSection->getFiberDimension(*c_iter); ++i, ++k) v[k] = vals[i];
          }
          sendValues[*r_iter] = allocPair(numVals, v);
          vMover.send(*r_iter, numVals, v);
        }
        const Obj<typename RecvOverlap::traits::capSequence>      rRanks = recvOverlap->cap();
        const typename RecvOverlap::traits::capSequence::iterator rEnd   = rRanks->end();

        recvSection->allocatePoint();
        // TODO: This should be const_iterator, but Sifter sucks
        int maxVals = 0;
        for(typename RecvOverlap::traits::capSequence::iterator r_iter = rRanks->begin(); r_iter != rEnd; ++r_iter) {
          const Obj<typename RecvOverlap::supportSequence>&     points  = recvOverlap->support(*r_iter);
          const typename RecvOverlap::supportSequence::iterator pEnd    = points->end();
          int                                                   numVals = 0;

          // TODO: This should be const_iterator, but Sifter sucks
          for(typename RecvOverlap::supportSequence::iterator s_iter = points->begin(); s_iter != pEnd; ++s_iter) {
            numVals += recvSection->getFiberDimension(recv_point_type(*r_iter, s_iter.color()));
          }
          typename SendSection::value_type *v = sendAllocator.allocate(numVals);

          for(int i = 0; i < numVals; ++i) {sendAllocator.construct(v+i, 0);}
          recvValues[*r_iter] = allocPair(numVals, v);
          vMover.recv(*r_iter, numVals, v);
          maxVals = std::max(maxVals, numVals);
        }
        vMover.start();
        vMover.end();
        typename RecvSection::value_type *convertedValues = recvAllocator.allocate(maxVals);
        for(int i = 0; i < maxVals; ++i) {recvAllocator.construct(convertedValues+i, 0);}
        for(typename RecvOverlap::traits::capSequence::iterator r_iter = rRanks->begin(); r_iter != rEnd; ++r_iter) {
          const Obj<typename RecvOverlap::supportSequence>&     points  = recvOverlap->support(*r_iter);
          const typename RecvOverlap::supportSequence::iterator pEnd    = points->end();
          typename SendSection::value_type                     *v       = recvValues[*r_iter].second;
          const int                                             numVals = recvValues[*r_iter].first;
          int                                                   k       = 0;

          for(typename RecvOverlap::supportSequence::iterator s_iter = points->begin(); s_iter != pEnd; ++s_iter) {
            const int size = recvSection->getFiberDimension(recv_point_type(*r_iter, s_iter.color()));

            for(int i = 0; i < size; ++i) {convertedValues[i] = (typename RecvSection::value_type) v[k+i];}
            if (size) {recvSection->updatePoint(recv_point_type(*r_iter, s_iter.color()), convertedValues);}
            k += size;
          }
          for(int i = 0; i < numVals; ++i) {sendAllocator.destroy(v+i);}
        }
        for(int i = 0; i < maxVals; ++i) {recvAllocator.destroy(convertedValues+i);}
        recvAllocator.deallocate(convertedValues, maxVals);
        for(typename SendOverlap::traits::baseSequence::iterator r_iter = sRanks->begin(); r_iter != sEnd; ++r_iter) {
          typename SendSection::value_type *v       = sendValues[*r_iter].second;
          const int                         numVals = sendValues[*r_iter].first;

          for(int i = 0; i < numVals; ++i) {sendAllocator.destroy(v+i);}
          sendAllocator.deallocate(v, numVals);
        }
        for(typename RecvOverlap::traits::capSequence::iterator r_iter = rRanks->begin(); r_iter != rEnd; ++r_iter) {
          typename SendSection::value_type *v       = recvValues[*r_iter].second;
          const int                         numVals = recvValues[*r_iter].first;

          for(int i = 0; i < numVals; ++i) {sendAllocator.destroy(v+i);}
          sendAllocator.deallocate(v, numVals);
        }
        //recvSection->view("After copy");
      };
      // Copy the overlap section to the related processes
      //   This version is for sections with the same type
      template<typename SendOverlap, typename RecvOverlap, typename Section>
      static void copy(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<Section>& sendSection, const Obj<Section>& recvSection, const MPI_Datatype datatype = MPI_DATATYPE_NULL) {
        typedef std::pair<int, typename Section::value_type *> allocPair;
        const Obj<typename Section::atlas_type>& sendAtlas = sendSection->getAtlas();
        const Obj<typename Section::atlas_type>& recvAtlas = recvSection->getAtlas();
        MPIMover<typename Section::value_type>   vMover(sendSection->comm(), datatype, MPI_UNDEFINED, sendSection->debug());
        std::map<int, allocPair>                 sendValues;
        std::map<int, allocPair>                 recvValues;
        typename Section::alloc_type             allocator;

        ///sendAtlas->view("Send Atlas in same type copy()");
        copy(sendOverlap, recvOverlap, sendAtlas, recvAtlas);
        ///recvAtlas->view("Recv Atlas after same type copy()");
        const Obj<typename SendOverlap::traits::baseSequence>      sRanks = sendOverlap->base();
        const typename SendOverlap::traits::baseSequence::iterator sEnd   = sRanks->end();

        // TODO: This should be const_iterator, but Sifter sucks
        for(typename SendOverlap::traits::baseSequence::iterator r_iter = sRanks->begin(); r_iter != sEnd; ++r_iter) {
          const Obj<typename SendOverlap::coneSequence>&     points  = sendOverlap->cone(*r_iter);
          const typename SendOverlap::coneSequence::iterator pEnd    = points->end();
          int                                                numVals = 0;

          // TODO: This should be const_iterator, but Sifter sucks
          for(typename SendOverlap::coneSequence::iterator c_iter = points->begin(); c_iter != pEnd; ++c_iter) {
            numVals += sendSection->getFiberDimension(*c_iter);
          }
          typename Section::value_type *v = allocator.allocate(numVals);
          int                           k = 0;

          for(int i = 0; i < numVals; ++i) {allocator.construct(v+i, 0);}
          for(typename SendOverlap::coneSequence::iterator c_iter = points->begin(); c_iter != pEnd; ++c_iter) {
            const typename Section::value_type *vals = sendSection->restrictPoint(*c_iter);
            const int                           dim  = sendSection->getFiberDimension(*c_iter);

            for(int i = 0; i < dim; ++i, ++k) v[k] = vals[i];
          }
          sendValues[*r_iter] = allocPair(numVals, v);
          vMover.send(*r_iter, numVals, v);
        }
        const Obj<typename RecvOverlap::traits::capSequence>      rRanks = recvOverlap->cap();
        const typename RecvOverlap::traits::capSequence::iterator rEnd   = rRanks->end();

        recvSection->allocatePoint();
        ///recvSection->view("Recv Section after same type copy() allocation");
        // TODO: This should be const_iterator, but Sifter sucks
        for(typename RecvOverlap::traits::capSequence::iterator r_iter = rRanks->begin(); r_iter != rEnd; ++r_iter) {
          const Obj<typename RecvOverlap::supportSequence>&     points  = recvOverlap->support(*r_iter);
          const typename RecvOverlap::supportSequence::iterator pEnd    = points->end();
          int                                                   numVals = 0;

          // TODO: This should be const_iterator, but Sifter sucks
          for(typename RecvOverlap::supportSequence::iterator s_iter = points->begin(); s_iter != pEnd; ++s_iter) {
            numVals += recvSection->getFiberDimension(s_iter.color());
          }
          typename Section::value_type *v = allocator.allocate(numVals);

          recvValues[*r_iter] = allocPair(numVals, v);
          for(int i = 0; i < numVals; ++i) {allocator.construct(v+i, 0);}
          vMover.recv(*r_iter, numVals, v);
        }
        vMover.start();
        vMover.end();
        for(typename RecvOverlap::traits::capSequence::iterator r_iter = rRanks->begin(); r_iter != rEnd; ++r_iter) {
          const Obj<typename RecvOverlap::supportSequence>&     points  = recvOverlap->support(*r_iter);
          const typename RecvOverlap::supportSequence::iterator pEnd    = points->end();
          typename Section::value_type                         *v       = recvValues[*r_iter].second;
          const int                                             numVals = recvValues[*r_iter].first;
          int                                                   k       = 0;

          for(typename RecvOverlap::supportSequence::iterator s_iter = points->begin(); s_iter != pEnd; ++s_iter) {
            const int size = recvSection->getFiberDimension(s_iter.color());

            if (size) {recvSection->updatePoint(s_iter.color(), &v[k]);}
            k += size;
          }
          for(int i = 0; i < numVals; ++i) {allocator.destroy(v+i);}
          allocator.deallocate(v, numVals);
        }
        for(typename SendOverlap::traits::baseSequence::iterator r_iter = sRanks->begin(); r_iter != sEnd; ++r_iter) {
          typename Section::value_type *v       = sendValues[*r_iter].second;
          const int                     numVals = sendValues[*r_iter].first;

          for(int i = 0; i < numVals; ++i) {allocator.destroy(v+i);}
          allocator.deallocate(v, numVals);
        }
        //recvSection->view("After copy");
      };
      // Specialize to ArrowSection
      template<typename SendOverlap, typename RecvOverlap>
      static void copy(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<UniformSection<MinimalArrow<int,int>,int> >& sendSection, const Obj<UniformSection<MinimalArrow<int,int>,int> >& recvSection, const MPI_Datatype datatype = MPI_DATATYPE_NULL) {
        typedef UniformSection<MinimalArrow<int,int>,int>      Section;
        typedef std::pair<int, typename Section::value_type *> allocPair;
        const Obj<typename Section::atlas_type>& sendAtlas = sendSection->getAtlas();
        const Obj<typename Section::atlas_type>& recvAtlas = recvSection->getAtlas();
        MPIMover<typename Section::value_type>   vMover(sendSection->comm(), datatype, MPI_UNDEFINED, sendSection->debug());
        std::map<int, allocPair>                 sendValues;
        std::map<int, allocPair>                 recvValues;
        typename Section::alloc_type             allocator;

        copy(sendOverlap, recvOverlap, sendAtlas, recvAtlas);
        const Obj<typename SendOverlap::traits::baseSequence>      sRanks = sendOverlap->base();
        const typename SendOverlap::traits::baseSequence::iterator sEnd   = sRanks->end();

        // TODO: This should be const_iterator, but Sifter sucks
        for(typename SendOverlap::traits::baseSequence::iterator r_iter = sRanks->begin(); r_iter != sEnd; ++r_iter) {
          const Obj<typename SendOverlap::coneSequence>&     points  = sendOverlap->cone(*r_iter);
          const typename SendOverlap::coneSequence::iterator pBegin  = points->begin();
          const typename SendOverlap::coneSequence::iterator pEnd    = points->end();
          int                                                numVals = 0;

          // TODO: This should be const_iterator, but Sifter sucks
          for(typename SendOverlap::coneSequence::iterator c_iter = pBegin; c_iter != pEnd; ++c_iter) {
            for(typename SendOverlap::coneSequence::iterator d_iter = pBegin; d_iter != pEnd; ++d_iter) {
              numVals += sendSection->getFiberDimension(typename Section::point_type(*c_iter, *d_iter));
            }
          }
          typename Section::value_type *v = allocator.allocate(numVals);
          int                           k = 0;

          for(int i = 0; i < numVals; ++i) {allocator.construct(v+i, 0);}
          for(typename SendOverlap::coneSequence::iterator c_iter = pBegin; c_iter != pEnd; ++c_iter) {
            for(typename SendOverlap::coneSequence::iterator d_iter = pBegin; d_iter != pEnd; ++d_iter) {
              const typename Section::point_type  arrow(*c_iter, *d_iter);
              const typename Section::value_type *vals = sendSection->restrictPoint(arrow);
              const int                           dim  = sendSection->getFiberDimension(arrow);

              for(int i = 0; i < dim; ++i, ++k) v[k] = vals[i];
            }
          }
          sendValues[*r_iter] = allocPair(numVals, v);
          vMover.send(*r_iter, numVals, v);
        }
        const Obj<typename RecvOverlap::traits::capSequence>      rRanks = recvOverlap->cap();
        const typename RecvOverlap::traits::capSequence::iterator rEnd   = rRanks->end();

        recvSection->allocatePoint();
        // TODO: This should be const_iterator, but Sifter sucks
        for(typename RecvOverlap::traits::capSequence::iterator r_iter = rRanks->begin(); r_iter != rEnd; ++r_iter) {
          const Obj<typename RecvOverlap::supportSequence>&     points  = recvOverlap->support(*r_iter);
          const typename RecvOverlap::supportSequence::iterator pBegin  = points->begin();
          const typename RecvOverlap::supportSequence::iterator pEnd    = points->end();
          int                                                   numVals = 0;

          // TODO: This should be const_iterator, but Sifter sucks
          for(typename RecvOverlap::supportSequence::iterator s_iter = pBegin; s_iter != pEnd; ++s_iter) {
            for(typename RecvOverlap::supportSequence::iterator t_iter = pBegin; t_iter != pEnd; ++t_iter) {
              numVals += recvSection->getFiberDimension(typename Section::point_type(s_iter.color(), t_iter.color()));
            }
          }
          typename Section::value_type *v = allocator.allocate(numVals);

          recvValues[*r_iter] = allocPair(numVals, v);
          for(int i = 0; i < numVals; ++i) {allocator.construct(v+i, 0);}
          vMover.recv(*r_iter, numVals, v);
        }
        vMover.start();
        vMover.end();
        for(typename RecvOverlap::traits::capSequence::iterator r_iter = rRanks->begin(); r_iter != rEnd; ++r_iter) {
          const Obj<typename RecvOverlap::supportSequence>&     points  = recvOverlap->support(*r_iter);
          const typename RecvOverlap::supportSequence::iterator pBegin  = points->begin();
          const typename RecvOverlap::supportSequence::iterator pEnd    = points->end();
          typename Section::value_type                         *v       = recvValues[*r_iter].second;
          const int                                             numVals = recvValues[*r_iter].first;
          int                                                   k       = 0;

          for(typename RecvOverlap::supportSequence::iterator s_iter = pBegin; s_iter != pEnd; ++s_iter) {
            for(typename RecvOverlap::supportSequence::iterator t_iter = pBegin; t_iter != pEnd; ++t_iter) {
              const typename Section::point_type arrow(s_iter.color(), t_iter.color());
              const int size = recvSection->getFiberDimension(arrow);

              if (size) {recvSection->updatePoint(arrow, &v[k]);}
              k += size;
            }
          }
          for(int i = 0; i < numVals; ++i) {allocator.destroy(v+i);}
          allocator.deallocate(v, numVals);
        }
        for(typename SendOverlap::traits::baseSequence::iterator r_iter = sRanks->begin(); r_iter != sEnd; ++r_iter) {
          typename Section::value_type *v       = sendValues[*r_iter].second;
          const int                     numVals = sendValues[*r_iter].first;

          for(int i = 0; i < numVals; ++i) {allocator.destroy(v+i);}
          allocator.deallocate(v, numVals);
        }
        //recvSection->view("After copy");
      };
      // Specialize to a ConstantSection
#if 0
      template<typename SendOverlap, typename RecvOverlap, typename Value>
      static void copy(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<ConstantSection<typename SendOverlap::source_type, Value> >& sendSection, const Obj<ConstantSection<typename SendOverlap::source_type, Value> >& recvSection) {
        copyConstant(sendOverlap, recvOverlap, sendSection, recvSection);
      };
      template<typename SendOverlap, typename RecvOverlap, typename Value>
      static void copy(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<IConstantSection<typename SendOverlap::source_type, Value> >& sendSection, const Obj<ConstantSection<typename SendOverlap::source_type, Value> >& recvSection) {
        copyConstant(sendOverlap, recvOverlap, sendSection, recvSection);
      };
#else
      template<typename SendOverlap, typename RecvOverlap, typename Value>
      static void copy(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<ConstantSection<typename SendOverlap::source_type, Value> >& sendSection, const Obj<ConstantSection<ALE::Pair<int, typename SendOverlap::source_type>, Value> >& recvSection) {
        copyConstant(sendOverlap, recvOverlap, sendSection, recvSection);
      };
      template<typename SendOverlap, typename RecvOverlap, typename Value>
      static void copy(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<IConstantSection<typename SendOverlap::source_type, Value> >& sendSection, const Obj<ConstantSection<ALE::Pair<int, typename SendOverlap::source_type>, Value> >& recvSection) {
        copyConstant(sendOverlap, recvOverlap, sendSection, recvSection);
      };
#endif
      // Specialize to an IConstantSection
      template<typename SendOverlap, typename RecvOverlap, typename Value>
      static void copy(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<IConstantSection<typename SendOverlap::source_type, Value> >& sendSection, const Obj<IConstantSection<typename SendOverlap::source_type, Value> >& recvSection) {
        // Why doesn't this work?
        //   This supposed to be a copy, BUT filtered through the sendOverlap
        //   However, an IConstant section does not allow filtration of its
        //   chart. Therefore, you end up with either
        //
        //   a) Too many items in the chart, copied from the remote sendSection
        //   b) A chart mapped to the local numbering, which we do not want
        copyIConstant(sendOverlap, recvOverlap, sendSection, recvSection);
      };
      // Specialize to an BaseSection/ConstantSection pair
#if 0
      template<typename SendOverlap, typename RecvOverlap, typename Sieve_>
      static void copy(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<BaseSection<Sieve_> >& sendSection, const Obj<ConstantSection<typename SendOverlap::source_type, int> >& recvSection) {
        copyConstant(sendOverlap, recvOverlap, sendSection, recvSection);
      };
#else
      template<typename SendOverlap, typename RecvOverlap, typename Sieve_>
      static void copy(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<BaseSection<Sieve_> >& sendSection, const Obj<ConstantSection<ALE::Pair<int, typename SendOverlap::source_type>, int> >& recvSection) {
        copyConstant(sendOverlap, recvOverlap, sendSection, recvSection);
      };
#endif
      // Specialize to an BaseSectionV/ConstantSection pair
#if 0
      template<typename SendOverlap, typename RecvOverlap, typename Sieve_>
      static void copy(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<BaseSectionV<Sieve_> >& sendSection, const Obj<ConstantSection<typename SendOverlap::source_type, int> >& recvSection) {
        copyConstant(sendOverlap, recvOverlap, sendSection, recvSection);
      };
#else
      template<typename SendOverlap, typename RecvOverlap, typename Sieve_>
      static void copy(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<BaseSectionV<Sieve_> >& sendSection, const Obj<ConstantSection<ALE::Pair<int, typename SendOverlap::source_type>, int> >& recvSection) {
        copyConstant(sendOverlap, recvOverlap, sendSection, recvSection);
      };
#endif
      // Specialize to an LabelBaseSection/ConstantSection pair
#if 0
      template<typename SendOverlap, typename RecvOverlap, typename Sieve_, typename Label_>
      static void copy(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<LabelBaseSection<Sieve_, Label_> >& sendSection, const Obj<ConstantSection<typename SendOverlap::source_type, int> >& recvSection) {
        copyConstant(sendOverlap, recvOverlap, sendSection, recvSection);
      };
#else
      template<typename SendOverlap, typename RecvOverlap, typename Sieve_, typename Label_>
      static void copy(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<LabelBaseSection<Sieve_, Label_> >& sendSection, const Obj<ConstantSection<ALE::Pair<int, typename SendOverlap::source_type>, int> >& recvSection) {
        copyConstant(sendOverlap, recvOverlap, sendSection, recvSection);
      };
#endif
      template<typename SendOverlap, typename RecvOverlap, typename Sieve_, typename Label_>
      static void copy(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<LabelBaseSectionV<Sieve_, Label_> >& sendSection, const Obj<ConstantSection<ALE::Pair<int, typename SendOverlap::source_type>, int> >& recvSection) {
        copyConstant(sendOverlap, recvOverlap, sendSection, recvSection);
      };
      // Specialize to a ConstantSection for ArrowSection
      template<typename SendOverlap, typename RecvOverlap, typename Value>
      static void copy(const Obj<SendOverlap>& sendOverlap, const Obj<RecvOverlap>& recvOverlap, const Obj<ConstantSection<MinimalArrow<typename SendOverlap::source_type,typename SendOverlap::source_type>, Value> >& sendSection, const Obj<ConstantSection<MinimalArrow<typename SendOverlap::source_type,typename SendOverlap::source_type>, Value> >& recvSection) {
        copyConstantArrow(sendOverlap, recvOverlap, sendSection, recvSection);
      };
    };
    class BinaryFusion {
    public:
      template<typename OverlapSection, typename RecvOverlap, typename Section, typename Function>
      static void fuse(const Obj<OverlapSection>& overlapSection, const Obj<RecvOverlap>& recvOverlap, const Obj<Section>& section, Function binaryOp) {
        const Obj<typename RecvOverlap::traits::baseSequence>      rPoints = recvOverlap->base();
        const typename RecvOverlap::traits::baseSequence::iterator rEnd    = rPoints->end();

        for(typename RecvOverlap::traits::baseSequence::iterator p_iter = rPoints->begin(); p_iter != rEnd; ++p_iter) {
          // TODO: This must become a more general iterator over colors
          const Obj<typename RecvOverlap::coneSequence>& points = recvOverlap->cone(*p_iter);
          // Just taking the first value
          const typename Section::point_type&        localPoint    = *p_iter;
          const typename OverlapSection::point_type& remotePoint   = points->begin().color();
          const typename OverlapSection::value_type *overlapValues = overlapSection->restrictPoint(remotePoint);
          const typename Section::value_type        *localValues   = section->restrictPoint(localPoint);
          const int                                  dim           = section->getFiberDimension(localPoint);
          // TODO: optimize allocation
          typename Section::value_type              *values        = new typename Section::value_type[dim];

          for(int d = 0; d < dim; ++d) {
            values[d] = binaryOp(overlapValues[d], localValues[d]);
          }
          section->updatePoint(localPoint, values);
          delete [] values;
        }
      };
    };
    class ReplacementBinaryFusion {
    public:
      template<typename OverlapSection, typename RecvOverlap, typename Section>
      static void fuse(const Obj<OverlapSection>& overlapSection, const Obj<RecvOverlap>& recvOverlap, const Obj<Section>& section) {
        const Obj<typename RecvOverlap::traits::baseSequence>      rPoints = recvOverlap->base();
        const typename RecvOverlap::traits::baseSequence::iterator rEnd    = rPoints->end();

        for(typename RecvOverlap::traits::baseSequence::iterator p_iter = rPoints->begin(); p_iter != rEnd; ++p_iter) {
          // TODO: This must become a more general iterator over colors
          const Obj<typename RecvOverlap::coneSequence>& points      = recvOverlap->cone(*p_iter);
          // Just taking the first value
          const typename Section::point_type&            localPoint  = *p_iter;
          const typename OverlapSection::point_type&     remotePoint = points->begin().color();

          section->update(localPoint, overlapSection->restrictPoint(remotePoint));
        }
      };
    };
    class AdditiveBinaryFusion {
    public:
      template<typename OverlapSection, typename RecvOverlap, typename Section>
      static void fuse(const Obj<OverlapSection>& overlapSection, const Obj<RecvOverlap>& recvOverlap, const Obj<Section>& section) {
        typedef typename Section::point_type        point_type;
        typedef typename OverlapSection::point_type overlap_point_type;
        const Obj<typename RecvOverlap::traits::baseSequence>      rPoints = recvOverlap->base();
        const typename RecvOverlap::traits::baseSequence::iterator rEnd    = rPoints->end();

        for(typename RecvOverlap::traits::baseSequence::iterator p_iter = rPoints->begin(); p_iter != rEnd; ++p_iter) {
          // TODO: This must become a more general iterator over colors
          const typename Section::point_type&                localPoint = *p_iter;
          const Obj<typename RecvOverlap::coneSequence>&     points     = recvOverlap->cone(*p_iter);
          const typename RecvOverlap::coneSequence::iterator cEnd       = points->end();

          for(typename RecvOverlap::coneSequence::iterator c_iter = points->begin(); c_iter != cEnd; ++c_iter) {
            const int         rank        = *c_iter;
            const point_type& remotePoint = c_iter.color();

            section->updateAddPoint(localPoint, overlapSection->restrictPoint(overlap_point_type(rank, remotePoint)));
          }
        }
      };
    };
    class InsertionBinaryFusion {
    public:
      // Insert the overlapSection values into section along recvOverlap
      template<typename OverlapSection, typename RecvOverlap, typename Section>
      static void fuse(const Obj<OverlapSection>& overlapSection, const Obj<RecvOverlap>& recvOverlap, const Obj<Section>& section) {
        typedef typename Section::point_type        point_type;
        typedef typename OverlapSection::point_type overlap_point_type;
        const Obj<typename RecvOverlap::traits::baseSequence>      rPoints = recvOverlap->base();
        const typename RecvOverlap::traits::baseSequence::iterator rEnd    = rPoints->end();

        for(typename RecvOverlap::traits::baseSequence::iterator p_iter = rPoints->begin(); p_iter != rEnd; ++p_iter) {
          const Obj<typename RecvOverlap::coneSequence>& points      = recvOverlap->cone(*p_iter);
          const point_type&                              localPoint  = *p_iter;
          const int                                      rank        = *points->begin();
          const point_type&                              remotePoint = points->begin().color();

          if (overlapSection->hasPoint(overlap_point_type(rank, remotePoint))) {
            if (!section->hasPoint(localPoint)) {
              std::cout <<"["<<section->commRank()<<"]: Destination section does not have local point " << localPoint << " remote point " << remotePoint << " fiber dim " << overlapSection->getFiberDimension(overlap_point_type(rank, remotePoint)) << std::endl;
            }
            section->setFiberDimension(localPoint, overlapSection->getFiberDimension(overlap_point_type(rank, remotePoint)));
          }
        }
        if (rPoints->size()) {section->allocatePoint();}
        for(typename RecvOverlap::traits::baseSequence::iterator p_iter = rPoints->begin(); p_iter != rEnd; ++p_iter) {
          const Obj<typename RecvOverlap::coneSequence>& points      = recvOverlap->cone(*p_iter);
          const point_type&                              localPoint  = *p_iter;
          const int                                      rank        = *points->begin();
          const point_type&                              remotePoint = points->begin().color();

          if (overlapSection->hasPoint(overlap_point_type(rank, remotePoint))) {
            section->updatePoint(localPoint, overlapSection->restrictPoint(overlap_point_type(rank, remotePoint)));
          }
        }
      };
      // Specialize to ArrowSection
      template<typename OverlapSection, typename RecvOverlap>
      static void fuse(const Obj<OverlapSection>& overlapSection, const Obj<RecvOverlap>& recvOverlap, const Obj<UniformSection<MinimalArrow<int,int>,int> >& section) {
        typedef UniformSection<MinimalArrow<int,int>,int> Section;
        const Obj<typename RecvOverlap::traits::baseSequence>      rPoints = recvOverlap->base();
        const typename RecvOverlap::traits::baseSequence::iterator rBegin  = rPoints->begin();
        const typename RecvOverlap::traits::baseSequence::iterator rEnd    = rPoints->end();

        for(typename RecvOverlap::traits::baseSequence::iterator p_iter = rBegin; p_iter != rEnd; ++p_iter) {
          const Obj<typename RecvOverlap::coneSequence>& sources      = recvOverlap->cone(*p_iter);
          const typename RecvOverlap::target_type        localSource  = *p_iter;
          const typename RecvOverlap::target_type        remoteSource = sources->begin().color();

          for(typename RecvOverlap::traits::baseSequence::iterator q_iter = rBegin; q_iter != rEnd; ++q_iter) {
            const Obj<typename RecvOverlap::coneSequence>& targets      = recvOverlap->cone(*q_iter);
            const typename RecvOverlap::target_type        localTarget  = *q_iter;
            const typename RecvOverlap::target_type        remoteTarget = targets->begin().color();
            const typename Section::point_type             localPoint(localSource, localTarget);
            const typename OverlapSection::point_type      remotePoint(remoteSource, remoteTarget);

            if (overlapSection->hasPoint(remotePoint)) {section->setFiberDimension(localPoint, overlapSection->getFiberDimension(remotePoint));}
          }
        }
        if (rPoints->size()) {section->allocatePoint();}
        for(typename RecvOverlap::traits::baseSequence::iterator p_iter = rBegin; p_iter != rEnd; ++p_iter) {
          const Obj<typename RecvOverlap::coneSequence>& sources      = recvOverlap->cone(*p_iter);
          const typename RecvOverlap::target_type        localSource  = *p_iter;
          const typename RecvOverlap::target_type        remoteSource = sources->begin().color();

          for(typename RecvOverlap::traits::baseSequence::iterator q_iter = rBegin; q_iter != rEnd; ++q_iter) {
            const Obj<typename RecvOverlap::coneSequence>& targets      = recvOverlap->cone(*q_iter);
            const typename RecvOverlap::target_type        localTarget  = *q_iter;
            const typename RecvOverlap::target_type        remoteTarget = targets->begin().color();
            const typename Section::point_type             localPoint(localSource, localTarget);
            const typename OverlapSection::point_type      remotePoint(remoteSource, remoteTarget);
              
            if (overlapSection->hasPoint(remotePoint)) {section->updatePoint(localPoint, overlapSection->restrictPoint(remotePoint));}
          }
        }
      };
      // Specialize to the Sieve
      template<typename OverlapSection, typename RecvOverlap, typename Renumbering, typename Point>
      static void fuse(const Obj<OverlapSection>& overlapSection, const Obj<RecvOverlap>& recvOverlap, Renumbering& renumbering, const Obj<Sieve<Point,Point,int> >& sieve) {
        typedef typename OverlapSection::point_type overlap_point_type;
        const Obj<typename RecvOverlap::traits::baseSequence>      rPoints = recvOverlap->base();
        const typename RecvOverlap::traits::baseSequence::iterator rEnd    = rPoints->end();

        for(typename RecvOverlap::traits::baseSequence::iterator p_iter = rPoints->begin(); p_iter != rEnd; ++p_iter) {
          const Obj<typename RecvOverlap::coneSequence>& points      = recvOverlap->cone(*p_iter);
          const Point&                                   localPoint  = *p_iter;
          const int                                      rank        = *points->begin();
          const Point&                                   remotePoint = points->begin().color();
          const int                                      size        = overlapSection->getFiberDimension(overlap_point_type(rank, remotePoint));
          const typename OverlapSection::value_type     *values      = overlapSection->restrictPoint(overlap_point_type(rank, remotePoint));
          int                                            c           = 0;

          sieve->clearCone(localPoint);
          for(int i = 0; i < size; ++i, ++c) {sieve->addCone(renumbering[values[i]], localPoint, c);}
        }
      };
      // Specialize to the ISieve
      template<typename OverlapSection, typename RecvOverlap, typename Renumbering, typename Point>
      static void fuse(const Obj<OverlapSection>& overlapSection, const Obj<RecvOverlap>& recvOverlap, Renumbering& renumbering, const Obj<IFSieve<Point> >& sieve) {
        typedef typename OverlapSection::point_type overlap_point_type;
        const Obj<typename RecvOverlap::traits::baseSequence>      rPoints = recvOverlap->base();
        const typename RecvOverlap::traits::baseSequence::iterator rEnd    = rPoints->end();
        int                                                        maxSize = 0;

        for(typename RecvOverlap::traits::baseSequence::iterator p_iter = rPoints->begin(); p_iter != rEnd; ++p_iter) {
          const Obj<typename RecvOverlap::coneSequence>& points      = recvOverlap->cone(*p_iter);
          const Point&                                   localPoint  = *p_iter;
          const int                                      rank        = *points->begin();
          const Point&                                   remotePoint = points->begin().color();
          const int                                      size        = overlapSection->getFiberDimension(overlap_point_type(rank, remotePoint));
          const typename OverlapSection::value_type     *values      = overlapSection->restrictPoint(overlap_point_type(rank, remotePoint));

          sieve->setConeSize(localPoint, size);
          ///for(int i = 0; i < size; ++i) {sieve->addSupportSize(renumbering[values[i]], 1);}
          for(int i = 0; i < size; ++i) {sieve->addSupportSize(renumbering[values[i].first], 1);}
          maxSize = std::max(maxSize, size);
        }
        sieve->allocate();
        ///typename OverlapSection::value_type *localValues = new typename OverlapSection::value_type[maxSize];
        typename OverlapSection::value_type::first_type  *localValues      = new typename OverlapSection::value_type::first_type[maxSize];
        typename OverlapSection::value_type::second_type *localOrientation = new typename OverlapSection::value_type::second_type[maxSize];

        for(typename RecvOverlap::traits::baseSequence::iterator p_iter = rPoints->begin(); p_iter != rEnd; ++p_iter) {
          const Obj<typename RecvOverlap::coneSequence>& points      = recvOverlap->cone(*p_iter);
          const Point&                                   localPoint  = *p_iter;
          const int                                      rank        = *points->begin();
          const Point&                                   remotePoint = points->begin().color();
          const int                                      size        = overlapSection->getFiberDimension(overlap_point_type(rank, remotePoint));
          const typename OverlapSection::value_type     *values      = overlapSection->restrictPoint(overlap_point_type(rank, remotePoint));

          ///for(int i = 0; i < size; ++i) {localValues[i] = renumbering[values[i]];}
          for(int i = 0; i < size; ++i) {
            localValues[i]      = renumbering[values[i].first];
            localOrientation[i] = values[i].second;
          }
          sieve->setCone(localValues, localPoint);
          sieve->setConeOrientation(localOrientation, localPoint);
        }
        delete [] localValues;
        delete [] localOrientation;
      };
      template<typename OverlapSection, typename RecvOverlap, typename Point>
      static void fuse(const Obj<OverlapSection>& overlapSection, const Obj<RecvOverlap>& recvOverlap, const Obj<IFSieve<Point> >& sieve) {
        typedef typename OverlapSection::point_type overlap_point_type;
        const Obj<typename RecvOverlap::traits::baseSequence>      rPoints = recvOverlap->base();
        const typename RecvOverlap::traits::baseSequence::iterator rEnd    = rPoints->end();
        int                                                        maxSize = 0;

        for(typename RecvOverlap::traits::baseSequence::iterator p_iter = rPoints->begin(); p_iter != rEnd; ++p_iter) {
          const Obj<typename RecvOverlap::coneSequence>& points      = recvOverlap->cone(*p_iter);
          const Point&                                   localPoint  = *p_iter;
          const int                                      rank        = *points->begin();
          const Point&                                   remotePoint = points->begin().color();
          const int                                      size        = overlapSection->getFiberDimension(overlap_point_type(rank, remotePoint));
          const typename OverlapSection::value_type     *values      = overlapSection->restrictPoint(overlap_point_type(rank, remotePoint));

          sieve->setConeSize(localPoint, size);
          for(int i = 0; i < size; ++i) {sieve->addSupportSize(values[i].first, 1);}
          maxSize = std::max(maxSize, size);
        }
        sieve->allocate();
        typename OverlapSection::value_type::first_type  *localValues      = new typename OverlapSection::value_type::first_type[maxSize];
        typename OverlapSection::value_type::second_type *localOrientation = new typename OverlapSection::value_type::second_type[maxSize];

        for(typename RecvOverlap::traits::baseSequence::iterator p_iter = rPoints->begin(); p_iter != rEnd; ++p_iter) {
          const Obj<typename RecvOverlap::coneSequence>& points      = recvOverlap->cone(*p_iter);
          const Point&                                   localPoint  = *p_iter;
          const int                                      rank        = *points->begin();
          const Point&                                   remotePoint = points->begin().color();
          const int                                      size        = overlapSection->getFiberDimension(overlap_point_type(rank, remotePoint));
          const typename OverlapSection::value_type     *values      = overlapSection->restrictPoint(overlap_point_type(rank, remotePoint));

          for(int i = 0; i < size; ++i) {
            localValues[i]      = values[i].first;
            localOrientation[i] = values[i].second;
          }
          sieve->setCone(localValues, localPoint);
          sieve->setConeOrientation(localOrientation, localPoint);
        }
        delete [] localValues;
        delete [] localOrientation;
      };
      // Generic
      template<typename OverlapSection, typename RecvOverlap, typename Section, typename Bundle>
      static void fuse(const Obj<OverlapSection>& overlapSection, const Obj<RecvOverlap>& recvOverlap, const Obj<Section>& section, const Obj<Bundle>& bundle) {
        typedef typename OverlapSection::point_type overlap_point_type;
        const Obj<typename RecvOverlap::traits::baseSequence>      rPoints = recvOverlap->base();
        const typename RecvOverlap::traits::baseSequence::iterator rEnd    = rPoints->end();

        for(typename RecvOverlap::traits::baseSequence::iterator p_iter = rPoints->begin(); p_iter != rEnd; ++p_iter) {
          const Obj<typename RecvOverlap::coneSequence>& points      = recvOverlap->cone(*p_iter);
          const typename Section::point_type&            localPoint  = *p_iter;
          const int                                      rank        = *points->begin();
          const typename OverlapSection::point_type&     remotePoint = points->begin().color();

          section->setFiberDimension(localPoint, overlapSection->getFiberDimension(overlap_point_type(rank, remotePoint)));
        }
        bundle->allocate(section);
        for(typename RecvOverlap::traits::baseSequence::iterator p_iter = rPoints->begin(); p_iter != rEnd; ++p_iter) {
          const Obj<typename RecvOverlap::coneSequence>& points      = recvOverlap->cone(*p_iter);
          const typename Section::point_type&            localPoint  = *p_iter;
          const int                                      rank        = *points->begin();
          const typename OverlapSection::point_type&     remotePoint = points->begin().color();

          section->update(localPoint, overlapSection->restrictPoint(overlap_point_type(rank, remotePoint)));
        }
      };
    };
  }
}

#endif