This file is indexed.

/usr/include/trilinos/AnasaziLOBPCG.hpp is in libtrilinos-anasazi-dev 12.12.1-5.

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
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
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
// @HEADER
// ***********************************************************************
//
//                 Anasazi: Block Eigensolvers Package
//                 Copyright 2004 Sandia Corporation
//
// Under terms of Contract DE-AC04-94AL85000 with Sandia Corporation,
// the U.S. Government retains certain rights in this software.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// 1. Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
//
// 2. Redistributions in binary form must reproduce the above copyright
// notice, this list of conditions and the following disclaimer in the
// documentation and/or other materials provided with the distribution.
//
// 3. Neither the name of the Corporation nor the names of the
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY SANDIA CORPORATION "AS IS" AND ANY
// EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
// PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL SANDIA CORPORATION OR THE
// CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
// LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
//
// Questions? Contact Michael A. Heroux (maherou@sandia.gov)
//
// ***********************************************************************
// @HEADER


/*! \file AnasaziLOBPCG.hpp
  \brief Implementation of the locally-optimal block preconditioned conjugate gradient (LOBPCG) method
*/

/*
    LOBPCG contains local storage of up to 10*blockSize_ vectors, representing 10 entities
      X,H,P,R
      KX,KH,KP  (product of K and the above)
      MX,MH,MP  (product of M and the above, not allocated if we don't have an M matrix)
    If full orthogonalization is enabled, one extra multivector of blockSize_ vectors is required to 
    compute the local update of X and P.
    
    A solver is bound to an eigenproblem at declaration.
    Other solver parameters (e.g., block size, auxiliary vectors) can be changed dynamically.
    
    The orthogonalization manager is used to project away from the auxiliary vectors.
    If full orthogonalization is enabled, the orthogonalization manager is also used to construct an M orthonormal basis.
    The orthogonalization manager is subclass of MatOrthoManager, which LOBPCG assumes to be defined by the M inner product.
    LOBPCG will not work correctly if the orthomanager uses a different inner product.
 */


#ifndef ANASAZI_LOBPCG_HPP
#define ANASAZI_LOBPCG_HPP

#include "AnasaziTypes.hpp"

#include "AnasaziEigensolver.hpp"
#include "AnasaziMultiVecTraits.hpp"
#include "AnasaziOperatorTraits.hpp"
#include "Teuchos_ScalarTraits.hpp"

#include "AnasaziMatOrthoManager.hpp"
#include "AnasaziSolverUtils.hpp"

#include "Teuchos_LAPACK.hpp"
#include "Teuchos_BLAS.hpp"
#include "Teuchos_SerialDenseMatrix.hpp"
#include "Teuchos_ParameterList.hpp"
#include "Teuchos_TimeMonitor.hpp"

/*!     \class Anasazi::LOBPCG

        \brief This class provides the Locally Optimal Block Preconditioned Conjugate Gradient (%LOBPCG) iteration, a preconditioned iteration for solving linear Hermitian eigenproblems.

        This implementation is a modification of the one found in 
        A. Knyazev, "Toward the optimal preconditioned eigensolver:
        Locally optimal block preconditioner conjugate gradient method",
        SIAM J. Sci. Comput., vol 23, n 2, pp. 517-541.

        The modification consists of the orthogonalization steps recommended in
        U. Hetmaniuk and R. Lehoucq, "Basis Selection in LOBPCG", Journal of Computational Physics. 

        These modifcation are referred to as full orthogonalization, and consist of also conducting
        the local optimization using an orthonormal basis.

        \ingroup anasazi_solver_framework

        \author Chris Baker, Ulrich Hetmaniuk, Rich Lehoucq, Heidi Thornquist
*/

namespace Anasazi {

  //! @name LOBPCG Structures
  //@{ 

  /** \brief Structure to contain pointers to Anasazi state variables.
   *
   * This struct is utilized by LOBPCG::initialize() and LOBPCG::getState().
   */
  template <class ScalarType, class MultiVector>
  struct LOBPCGState {
    //! The current test basis.
    Teuchos::RCP<const MultiVector> V; 
    //! The image of the current test basis under K.
    Teuchos::RCP<const MultiVector> KV; 
    //! The image of the current test basis under M, or Teuchos::null if M was not specified.
    Teuchos::RCP<const MultiVector> MV;

    //! The current eigenvectors.
    Teuchos::RCP<const MultiVector> X; 
    //! The image of the current eigenvectors under K.
    Teuchos::RCP<const MultiVector> KX; 
    //! The image of the current eigenvectors under M, or Teuchos::null if M was not specified.
    Teuchos::RCP<const MultiVector> MX;

    //! The current search direction.
    Teuchos::RCP<const MultiVector> P; 
    //! The image of the current search direction under K.
    Teuchos::RCP<const MultiVector> KP; 
    //! The image of the current search direction under M, or Teuchos::null if M was not specified.
    Teuchos::RCP<const MultiVector> MP;

    /*! \brief The current preconditioned residual vectors.
     *
     *  H is only useful when LOBPCG::iterate() throw a LOBPCGRitzFailure exception.
     */
    Teuchos::RCP<const MultiVector> H; 
    //! The image of the current preconditioned residual vectors under K.
    Teuchos::RCP<const MultiVector> KH; 
    //! The image of the current preconditioned residual vectors under M, or Teuchos::null if M was not specified.
    Teuchos::RCP<const MultiVector> MH;

    //! The current residual vectors.
    Teuchos::RCP<const MultiVector> R;

    //! The current Ritz values.
    Teuchos::RCP<const std::vector<typename Teuchos::ScalarTraits<ScalarType>::magnitudeType> > T;

    LOBPCGState() : 
      V(Teuchos::null),KV(Teuchos::null),MV(Teuchos::null),
      X(Teuchos::null),KX(Teuchos::null),MX(Teuchos::null),
      P(Teuchos::null),KP(Teuchos::null),MP(Teuchos::null),
      H(Teuchos::null),KH(Teuchos::null),MH(Teuchos::null),
      R(Teuchos::null),T(Teuchos::null) {};
  };

  //@}

  //! @name LOBPCG Exceptions
  //@{ 

  /** \brief LOBPCGRitzFailure is thrown when the LOBPCG solver is unable to
   *  continue a call to LOBPCG::iterate() due to a failure of the algorithm.
   *
   *  This signals that the Rayleigh-Ritz analysis over the subspace \c
   *  colsp([X H P]) detected ill-conditioning of the projected mass matrix
   *  and the inability to generate a set of orthogonal eigenvectors for 
   *  the projected problem.
   *
   *  This exception is only thrown from the LOBPCG::iterate() routine. After
   *  catching this exception, the user can recover the subspace via
   *  LOBPCG::getState(). This information can be used to restart the solver.
   *
   */
  class LOBPCGRitzFailure : public AnasaziError {public:
    LOBPCGRitzFailure(const std::string& what_arg) : AnasaziError(what_arg)
    {}};

  /** \brief LOBPCGInitFailure is thrown when the LOBPCG solver is unable to
   * generate an initial iterate in the LOBPCG::initialize() routine. 
   *
   * This exception is thrown from the LOBPCG::initialize() method, which is
   * called by the user or from the LOBPCG::iterate() method when isInitialized()
   * == \c false.
   *
   * In the case that this exception is thrown, LOBPCG::hasP() and
   * LOBPCG::isInitialized() will be \c false and the user will need to provide
   * a new initial iterate to the solver.
   *
   */
  class LOBPCGInitFailure : public AnasaziError {public:
    LOBPCGInitFailure(const std::string& what_arg) : AnasaziError(what_arg)
    {}};

  /** \brief LOBPCGOrthoFailure is thrown when an orthogonalization attempt 
   * fails.
   *
   * This is thrown in one of two scenarstd::ios. After preconditioning the residual,
   * the orthogonalization manager is asked to orthogonalize the preconditioned
   * residual (H) against the auxiliary vectors. If full orthogonalization
   * is enabled, H is also orthogonalized against X and P and normalized.
   *
   * The second scenario involves the generation of new X and P from the
   * basis [X H P]. When full orthogonalization is enabled, an attempt is
   * made to select coefficients for X and P so that they will be
   * mutually orthogonal and orthonormal.
   *
   * If either of these attempts fail, the solver throws an LOBPCGOrthoFailure
   * exception.
   */
  class LOBPCGOrthoFailure : public AnasaziError {public:
    LOBPCGOrthoFailure(const std::string& what_arg) : AnasaziError(what_arg)
    {}};

  //@}


  template <class ScalarType, class MV, class OP>
  class LOBPCG : public Eigensolver<ScalarType,MV,OP> { 
  public:
    
    //! @name Constructor/Destructor
    //@{ 
    
    /*! \brief %LOBPCG constructor with eigenproblem, solver utilities, and parameter list of solver options.
     *
     * This constructor takes pointers required by the eigensolver, in addition
     * to a parameter list of options for the eigensolver. These options include the following:
     *   - "Block Size" - an \c int specifying the block size used by the algorithm. This can also be specified using the setBlockSize() method.
     *   - "Full Ortho" - a \c bool specifying whether the solver should employ a full orthogonalization technique. This can also be specified using the setFullOrtho() method.
     */
    LOBPCG( const Teuchos::RCP<Eigenproblem<ScalarType,MV,OP> > &problem, 
            const Teuchos::RCP<SortManager<typename Teuchos::ScalarTraits<ScalarType>::magnitudeType> > &sorter,
            const Teuchos::RCP<OutputManager<ScalarType> > &printer,
            const Teuchos::RCP<StatusTest<ScalarType,MV,OP> > &tester,
            const Teuchos::RCP<MatOrthoManager<ScalarType,MV,OP> > &ortho,
            Teuchos::ParameterList &params 
          );
    
    //! %LOBPCG destructor
    virtual ~LOBPCG() {};

    //@}

    //! @name Solver methods
    //@{

    /*! \brief This method performs %LOBPCG iterations until the status test
     * indicates the need to stop or an error occurs (in which case, an
     * exception is thrown).
     *
     * iterate() will first determine whether the solver is initialized; if
     * not, it will call initialize() using default arguments.  After
     * initialization, the solver performs %LOBPCG iterations until the status
     * test evaluates as Passed, at which point the method returns to the
     * caller.
     *
     * The %LOBPCG iteration proceeds as follows:
     * -# The current residual (R) is preconditioned to form H
     * -# H is orthogonalized against the auxiliary vectors and, if full orthogonalization\n
     *    is enabled, against X and P. 
     * -# The basis [X H P] is used to project the problem matrices.
     * -# The projected eigenproblem is solved, and the desired eigenvectors and eigenvalues are selected.
     * -# These are used to form the new eigenvector estimates (X) and the search directions (P).\n
     *    If full orthogonalization is enabled, these are generated to be mutually orthogonal and with orthonormal columns.
     * -# The new residual (R) is formed.
     *
     * The status test is queried at the beginning of the iteration.
     *
     * Possible exceptions thrown include std::logic_error, std::invalid_argument or
     * one of the LOBPCG-specific exceptions.
     *
    */
    void iterate();

    /*! \brief Initialize the solver to an iterate, optionally providing the
     * Ritz values, residual, and search direction.
     *
     * \note LOBPCGState contains fields V, KV and MV: These are ignored by initialize()
     *
     * The %LOBPCG eigensolver contains a certain amount of state relating to
     * the current iterate, including the current residual, the current search
     * direction, and the images of these spaces under the eigenproblem operators.
     *
     * initialize() gives the user the opportunity to manually set these,
     * although this must be done with caution, abiding by the rules
     * given below. All notions of orthogonality and orthonormality are derived
     * from the inner product specified by the orthogonalization manager.
     *
     * \post 
     *   - isInitialized() == true (see post-conditions of isInitialize())
     *   - If newstate.P != Teuchos::null, hasP() == true.\n
     *     Otherwise, hasP() == false
     *
     * The user has the option of specifying any component of the state using
     * initialize(). However, these arguments are assumed to match the
     * post-conditions specified under isInitialized(). Any component of the
     * state (i.e., KX) not given to initialize() will be generated.
     *
     */
    void initialize(LOBPCGState<ScalarType,MV>& newstate);

    /*! \brief Initialize the solver with the initial vectors from the eigenproblem
     *  or random data.
     */
    void initialize();

    /*! \brief Indicates whether the solver has been initialized or not.
     *
     * \return bool indicating the state of the solver.
     * \post
     * If isInitialized() == \c true:
     *   - X is orthogonal to auxiliary vectors and has orthonormal columns
     *   - KX == Op*X
     *   - MX == M*X if M != Teuchos::null\n
     *     Otherwise, MX == Teuchos::null
     *   - getRitzValues() returns the sorted Ritz values with respect to X
     *   - getResNorms(), getRes2Norms(), getRitzResNorms() are correct
     *   - If hasP() == \c true,
     *      - P orthogonal to auxiliary vectors
     *      - If getFullOrtho() == \c true,
     *        - P is orthogonal to X and has orthonormal columns
     *      - KP == Op*P
     *      - MP == M*P if M != Teuchos::null\n
     *        Otherwise, MP == Teuchos::null
     */
    bool isInitialized() const;

    /*! \brief Get the current state of the eigensolver.
     * 
     * The data is only valid if isInitialized() == \c true. The
     * data for the search directions P is only meaningful if hasP() == \c
     * true. Finally, the data for the preconditioned residual (H) is only meaningful in the situation where
     * the solver throws an ::LOBPCGRitzFailure exception during iterate().
     *
     * \returns An LOBPCGState object containing const views to the current
     * solver state.
     */
    LOBPCGState<ScalarType,MV> getState() const;

    //@}

    //! @name Status methods
    //@{

    //! \brief Get the current iteration count.
    int getNumIters() const;

    //! \brief Reset the iteration count.
    void resetNumIters();

    /*! \brief Get the Ritz vectors from the previous iteration.
      
        \return A multivector with getBlockSize() vectors containing 
        the sorted Ritz vectors corresponding to the most significant Ritz values.
        The i-th vector of the return corresponds to the i-th Ritz vector; there is no need to use
        getRitzIndex().
     */
    Teuchos::RCP<const MV> getRitzVectors();

    /*! \brief Get the Ritz values from the previous iteration.
     *
     *  \return A vector of length getCurSubspaceDim() containing the Ritz values from the
     *  previous projected eigensolve.
     */
    std::vector<Value<ScalarType> > getRitzValues();

    /*! \brief Get the index used for extracting Ritz vectors from getRitzVectors().
     *
     * Because BlockDavidson is a Hermitian solver, all Ritz values are real and all Ritz vectors can be represented in a 
     * single column of a multivector. Therefore, getRitzIndex() is not needed when using the output from getRitzVectors().
     *
     * \return An \c int vector of size getCurSubspaceDim() composed of zeros.
     */
    std::vector<int> getRitzIndex();


    /*! \brief Get the current residual norms
     *
     *  \return A vector of length getBlockSize() containing the norms of the
     *  residuals, with respect to the orthogonalization manager norm() method.
     */
    std::vector<typename Teuchos::ScalarTraits<ScalarType>::magnitudeType> getResNorms();


    /*! \brief Get the current residual 2-norms
     *
     *  \return A vector of length getBlockSize() containing the 2-norms of the
     *  residuals. 
     */
    std::vector<typename Teuchos::ScalarTraits<ScalarType>::magnitudeType> getRes2Norms();


    /*! \brief Get the 2-norms of the residuals.
     * 
     * The Ritz residuals are not defined for the %LOBPCG iteration. Hence, this method returns the 
     * 2-norms of the direct residuals, and is equivalent to calling getRes2Norms().
     *
     *  \return A vector of length getBlockSize() containing the 2-norms of the direct residuals.
     */
    std::vector<typename Teuchos::ScalarTraits<ScalarType>::magnitudeType> getRitzRes2Norms();


    /*! \brief Get the dimension of the search subspace used to generate the current eigenvectors and eigenvalues.
     *
     *  %LOBPCG employs a sequential subspace iteration, maintaining a fixed-rank basis, as opposed to an expanding subspace
     *  mechanism employed by Krylov-subspace solvers like BlockKrylovSchur and BlockDavidson.
     *  
     *  \return An integer specifying the rank of the subspace generated by the eigensolver. If isInitialized() == \c false, 
     *  the return is 0. Otherwise, the return will be 2*getBlockSize() or 3*getBlockSize().
     */
    int getCurSubspaceDim() const;

    /*! \brief Get the maximum dimension allocated for the search subspace. For %LOBPCG, this always returns 3*getBlockSize(), the dimension of the 
     *   subspace colspan([X H P]).
     */
    int getMaxSubspaceDim() const;

    //@}

    //!  @name Accessor routines from Eigensolver
    //@{

    //! Set a new StatusTest for the solver.
    void setStatusTest(Teuchos::RCP<StatusTest<ScalarType,MV,OP> > test);

    //! Get the current StatusTest used by the solver.
    Teuchos::RCP<StatusTest<ScalarType,MV,OP> > getStatusTest() const;

    //! Get a constant reference to the eigenvalue problem.
    const Eigenproblem<ScalarType,MV,OP>& getProblem() const;


    /*! \brief Set the blocksize to be used by the iterative solver in solving
     * this eigenproblem.
     *  
     *  If the block size is reduced, then the new iterate (and residual and
     *  search direction) are chosen as the subset of the current iterate
     *  preferred by the sort manager.  Otherwise, the solver state is set to
     *  uninitialized.
     */
    void setBlockSize(int blockSize);


    //! Get the blocksize to be used by the iterative solver in solving this eigenproblem.
    int getBlockSize() const;


    /*! \brief Set the auxiliary vectors for the solver.
     *
     *  Because the current iterate X and search direction P cannot be assumed
     *  orthogonal to the new auxiliary vectors, a call to setAuxVecs() with a
     *  non-empty argument will reset the solver to the uninitialized state.
     *
     *  In order to preserve the current state, the user will need to extract
     *  it from the solver using getState(), orthogonalize it against the new
     *  auxiliary vectors, and manually reinitialize the solver using
     *  initialize().
     */
    void setAuxVecs(const Teuchos::Array<Teuchos::RCP<const MV> > &auxvecs);

    //! Get the current auxiliary vectors.
    Teuchos::Array<Teuchos::RCP<const MV> > getAuxVecs() const;

    //@}

    //!  @name %LOBPCG-specific accessor routines
    //@{

    /*! \brief Instruct the LOBPCG iteration to use full orthogonality.
     *
     *  If the getFullOrtho() == \c false and isInitialized() == \c true and hasP() == \c true, then
     *  P will be invalidated by setting full orthogonalization to \c true.
     */
    void setFullOrtho(bool fullOrtho);

    //! Determine if the LOBPCG iteration is using full orthogonality.
    bool getFullOrtho() const;
    
    //! Indicates whether the search direction given by getState() is valid.
    bool hasP();

    //@}
    
    //!  @name Output methods
    //@{

    //! This method requests that the solver print out its current status to screen.
    void currentStatus(std::ostream &os);

    //@}

  private:
    //
    //
    //
    void setupViews();
    //
    // Convenience typedefs
    //
    typedef SolverUtils<ScalarType,MV,OP> Utils;
    typedef MultiVecTraits<ScalarType,MV> MVT;
    typedef OperatorTraits<ScalarType,MV,OP> OPT;
    typedef Teuchos::ScalarTraits<ScalarType> SCT;
    typedef typename SCT::magnitudeType MagnitudeType;
    const MagnitudeType ONE;  
    const MagnitudeType ZERO; 
    const MagnitudeType NANVAL;
    //
    // Internal structs
    //
    struct CheckList {
      bool checkX, checkMX, checkKX;
      bool checkH, checkMH;
      bool checkP, checkMP, checkKP;
      bool checkR, checkQ;
      CheckList() : checkX(false),checkMX(false),checkKX(false),
                    checkH(false),checkMH(false),
                    checkP(false),checkMP(false),checkKP(false),
                    checkR(false),checkQ(false) {};
    };
    //
    // Internal methods
    //
    std::string accuracyCheck(const CheckList &chk, const std::string &where) const;
    //
    // Classes inputed through constructor that define the eigenproblem to be solved.
    //
    const Teuchos::RCP<Eigenproblem<ScalarType,MV,OP> >     problem_;
    const Teuchos::RCP<SortManager<typename Teuchos::ScalarTraits<ScalarType>::magnitudeType> > sm_;
    const Teuchos::RCP<OutputManager<ScalarType> >          om_;
    Teuchos::RCP<StatusTest<ScalarType,MV,OP> >       tester_;
    const Teuchos::RCP<MatOrthoManager<ScalarType,MV,OP> >  orthman_;
    //
    // Information obtained from the eigenproblem
    //
    Teuchos::RCP<const OP> Op_;
    Teuchos::RCP<const OP> MOp_;
    Teuchos::RCP<const OP> Prec_;
    bool hasM_;
    //
    // Internal timers
    //
    Teuchos::RCP<Teuchos::Time> timerOp_, timerMOp_, timerPrec_,
                                timerSort_, 
                                timerLocalProj_, timerDS_,
                                timerLocalUpdate_, timerCompRes_,
                                timerOrtho_, timerInit_;
    //
    // Counters
    //
    // Number of operator applications
    int count_ApplyOp_, count_ApplyM_, count_ApplyPrec_;

    //
    // Algorithmic parameters.
    //
    // blockSize_ is the solver block size
    int blockSize_;
    //
    // fullOrtho_ dictates whether the orthogonalization procedures specified by Hetmaniuk and Lehoucq should
    // be activated (see citations at the top of this file)
    bool fullOrtho_;

    //
    // Current solver state
    //
    // initialized_ specifies that the basis vectors have been initialized and the iterate() routine
    // is capable of running; _initialize is controlled  by the initialize() member method
    // For the implications of the state of initialized_, please see documentation for initialize()
    bool initialized_;
    //
    // nevLocal_ reflects how much of the current basis is valid (0 <= nevLocal_ <= 3*blockSize_)
    // this tells us how many of the values in theta_ are valid Ritz values
    int nevLocal_;
    //
    // hasP_ tells us whether there is valid data in P (and KP,MP)
    bool hasP_;
    //
    // State Multivecs
    // V_, KV_ MV_  and  R_ are primary pointers to allocated multivectors
    Teuchos::RCP<MV> V_, KV_, MV_, R_;
    // the rest are multivector views into V_, KV_ and MV_
    Teuchos::RCP<MV> X_, KX_, MX_,
                     H_, KH_, MH_,
                     P_, KP_, MP_;

    //
    // if fullOrtho_ == true, then we must produce the following on every iteration:
    // [newX newP] = [X H P] [CX;CP]
    // the structure of [CX;CP] when using full orthogonalization does not allow us to 
    // do this in situ, and R_ does not have enough storage for newX and newP. therefore, 
    // we must allocate additional storage for this.
    // otherwise, when not using full orthogonalization, the structure
    // [newX newP] = [X H P] [CX1  0 ]
    //                       [CX2 CP2]  allows us to work using only R as work space
    //                       [CX3 CP3] 
    Teuchos::RCP<MV> tmpmvec_;        
    // 
    // auxiliary vectors
    Teuchos::Array<Teuchos::RCP<const MV> > auxVecs_;
    int numAuxVecs_;
    //
    // Number of iterations that have been performed.
    int iter_;
    // 
    // Current eigenvalues, residual norms
    std::vector<MagnitudeType> theta_, Rnorms_, R2norms_;
    // 
    // are the residual norms current with the residual?
    bool Rnorms_current_, R2norms_current_;

  };




  //////////////////////////////////////////////////////////////////////////////////////////////////
  // Constructor
  template <class ScalarType, class MV, class OP>
  LOBPCG<ScalarType,MV,OP>::LOBPCG(
        const Teuchos::RCP<Eigenproblem<ScalarType,MV,OP> > &problem, 
        const Teuchos::RCP<SortManager<typename Teuchos::ScalarTraits<ScalarType>::magnitudeType> > &sorter,
        const Teuchos::RCP<OutputManager<ScalarType> > &printer,
        const Teuchos::RCP<StatusTest<ScalarType,MV,OP> > &tester,
        const Teuchos::RCP<MatOrthoManager<ScalarType,MV,OP> > &ortho,
        Teuchos::ParameterList &params
        ) :
    ONE(Teuchos::ScalarTraits<MagnitudeType>::one()),
    ZERO(Teuchos::ScalarTraits<MagnitudeType>::zero()),
    NANVAL(Teuchos::ScalarTraits<MagnitudeType>::nan()),
    // problem, tools
    problem_(problem), 
    sm_(sorter),
    om_(printer),
    tester_(tester),
    orthman_(ortho),
    // timers, counters
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
    timerOp_(Teuchos::TimeMonitor::getNewTimer("Anasazi: LOBPCG::Operation Op*x")),
    timerMOp_(Teuchos::TimeMonitor::getNewTimer("Anasazi: LOBPCG::Operation M*x")),
    timerPrec_(Teuchos::TimeMonitor::getNewTimer("Anasazi: LOBPCG::Operation Prec*x")),
    timerSort_(Teuchos::TimeMonitor::getNewTimer("Anasazi: LOBPCG::Sorting eigenvalues")),
    timerLocalProj_(Teuchos::TimeMonitor::getNewTimer("Anasazi: LOBPCG::Local projection")),
    timerDS_(Teuchos::TimeMonitor::getNewTimer("Anasazi: LOBPCG::Direct solve")),
    timerLocalUpdate_(Teuchos::TimeMonitor::getNewTimer("Anasazi: LOBPCG::Local update")),
    timerCompRes_(Teuchos::TimeMonitor::getNewTimer("Anasazi: LOBPCG::Computing residuals")),
    timerOrtho_(Teuchos::TimeMonitor::getNewTimer("Anasazi: LOBPCG::Orthogonalization")),
    timerInit_(Teuchos::TimeMonitor::getNewTimer("Anasazi: LOBPCG::Initialization")),
#endif
    count_ApplyOp_(0),
    count_ApplyM_(0),
    count_ApplyPrec_(0),
    // internal data
    blockSize_(0),
    fullOrtho_(params.get("Full Ortho", true)),
    initialized_(false),
    nevLocal_(0),
    hasP_(false),
    auxVecs_( Teuchos::Array<Teuchos::RCP<const MV> >(0) ), 
    numAuxVecs_(0),
    iter_(0),
    Rnorms_current_(false),
    R2norms_current_(false)
  {     
    TEUCHOS_TEST_FOR_EXCEPTION(problem_ == Teuchos::null,std::invalid_argument,
                       "Anasazi::LOBPCG::constructor: user passed null problem pointer.");
    TEUCHOS_TEST_FOR_EXCEPTION(sm_ == Teuchos::null,std::invalid_argument,
                       "Anasazi::LOBPCG::constructor: user passed null sort manager pointer.");
    TEUCHOS_TEST_FOR_EXCEPTION(om_ == Teuchos::null,std::invalid_argument,
                       "Anasazi::LOBPCG::constructor: user passed null output manager pointer.");
    TEUCHOS_TEST_FOR_EXCEPTION(tester_ == Teuchos::null,std::invalid_argument,
                       "Anasazi::LOBPCG::constructor: user passed null status test pointer.");
    TEUCHOS_TEST_FOR_EXCEPTION(orthman_ == Teuchos::null,std::invalid_argument,
                       "Anasazi::LOBPCG::constructor: user passed null orthogonalization manager pointer.");
    TEUCHOS_TEST_FOR_EXCEPTION(problem_->isProblemSet() == false, std::invalid_argument,
                       "Anasazi::LOBPCG::constructor: problem is not set.");
    TEUCHOS_TEST_FOR_EXCEPTION(problem_->isHermitian() == false, std::invalid_argument,
                       "Anasazi::LOBPCG::constructor: problem is not Hermitian; LOBPCG requires Hermitian problem.");

    // get the problem operators
    Op_   = problem_->getOperator();
    TEUCHOS_TEST_FOR_EXCEPTION(Op_ == Teuchos::null, std::invalid_argument,
                       "Anasazi::LOBPCG::constructor: problem provides no operator.");
    MOp_  = problem_->getM();
    Prec_ = problem_->getPrec();
    hasM_ = (MOp_ != Teuchos::null);

    // set the block size and allocate data
    int bs = params.get("Block Size", problem_->getNEV());
    setBlockSize(bs);
  }


  //////////////////////////////////////////////////////////////////////////////////////////////////
  // Set the block size and make necessary adjustments.
  template <class ScalarType, class MV, class OP>
  void LOBPCG<ScalarType,MV,OP>::setBlockSize (int newBS) 
  {
    // time spent here counts towards timerInit_
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
    Teuchos::TimeMonitor lcltimer( *timerInit_ );
#endif

    // This routine only allocates space; it doesn't not perform any computation
    // if size is decreased, take the first newBS vectors of all and leave state as is
    // otherwise, grow/allocate space and set solver to unitialized

    Teuchos::RCP<const MV> tmp;
    // grab some Multivector to Clone
    // in practice, getInitVec() should always provide this, but it is possible to use a 
    // Eigenproblem with nothing in getInitVec() by manually initializing with initialize(); 
    // in case of that strange scenario, we will try to Clone from R_ because it is smaller 
    // than V_, and we don't want to keep V_ around longer than necessary
    if (blockSize_ > 0) {
      tmp = R_;
    }
    else {
      tmp = problem_->getInitVec();
      TEUCHOS_TEST_FOR_EXCEPTION(tmp == Teuchos::null,std::logic_error,
                         "Anasazi::LOBPCG::setBlockSize(): eigenproblem did not specify initial vectors to clone from.");
    }
    
    TEUCHOS_TEST_FOR_EXCEPTION(newBS <= 0 || static_cast<ptrdiff_t>(newBS) > MVT::GetGlobalLength(*tmp), 
                         std::invalid_argument, "Anasazi::LOBPCG::setBlockSize(): block size must be strictly positive.");
    if (newBS == blockSize_) {
      // do nothing
      return;
    }
    else if (newBS < blockSize_ && initialized_) {
      //
      // shrink vectors

      // release views so we can modify the bases
      X_ = Teuchos::null;
      KX_ = Teuchos::null;
      MX_ = Teuchos::null;
      H_ = Teuchos::null;
      KH_ = Teuchos::null;
      MH_ = Teuchos::null;
      P_ = Teuchos::null;
      KP_ = Teuchos::null;
      MP_ = Teuchos::null;

      // make new indices vectors
      std::vector<int> newind(newBS), oldind(newBS);
      for (int i=0; i<newBS; i++) {
        newind[i] = i;
        oldind[i] = i;
      }

      Teuchos::RCP<MV> newV, newMV, newKV, newR;
      Teuchos::RCP<const MV> src;
      // allocate R and newV
      newR = MVT::Clone(*tmp,newBS);
      newV = MVT::Clone(*tmp,newBS*3);
      newKV = MVT::Clone(*tmp,newBS*3);
      if (hasM_) {
        newMV = MVT::Clone(*tmp,newBS*3);
      }

      //
      // if we are initialized, we want to pull the data from V_ into newV:
      //           bs  |  bs  | bs 
      // newV  = [newX | **** |newP ]
      // newKV = [newKX| **** |newKP]
      // newMV = [newMX| **** |newMP]
      // where 
      //          oldbs   |  oldbs  |   oldbs   
      //  V_ = [newX  *** | ******* | newP  ***]
      // KV_ = [newKX *** | ******* | newKP ***]
      // MV_ = [newMX *** | ******* | newMP ***]
      //
      // we don't care to copy the data corresponding to H
      // we will not copy the M data if !hasM_, because it doesn't exist
      //

      // these are shrink operations which preserve their data
      theta_.resize(3*newBS);
      Rnorms_.resize(newBS);
      R2norms_.resize(newBS);

      // copy residual vectors: oldind,newind currently contains [0,...,newBS-1]
      src = MVT::CloneView(*R_,newind);
      MVT::SetBlock(*src,newind,*newR);
      // free old memory and point to new memory
      R_ = newR;

      // copy in order: newX newKX newMX, then newP newKP newMP
      // for  X: [0,bs-1] <-- [0,bs-1] 
      src = MVT::CloneView(*V_,oldind);
      MVT::SetBlock(*src,newind,*newV);
      src = MVT::CloneView(*KV_,oldind);
      MVT::SetBlock(*src,newind,*newKV);
      if (hasM_) {
        src = MVT::CloneView(*MV_,oldind);
        MVT::SetBlock(*src,newind,*newMV);
      }
      // for  P: [2*bs, 3*bs-1] <-- [2*oldbs, 2*oldbs+bs-1] 
      for (int i=0; i<newBS; i++) {
        newind[i] += 2*newBS;
        oldind[i] += 2*blockSize_;
      }
      src = MVT::CloneView(*V_,oldind);
      MVT::SetBlock(*src,newind,*newV);
      src = MVT::CloneView(*KV_,oldind);
      MVT::SetBlock(*src,newind,*newKV);
      if (hasM_) {
        src = MVT::CloneView(*MV_,oldind);
        MVT::SetBlock(*src,newind,*newMV);
      }

      // release temp view
      src = Teuchos::null;

      // release old allocations and point at new memory
      V_ = newV;
      KV_ = newKV;
      if (hasM_) {
        MV_ = newMV;
      }
      else {
        MV_ = V_;
      }
    }
    else {  
      // newBS > blockSize_  or  not initialized
      // this is also the scenario for our initial call to setBlockSize(), in the constructor
      initialized_ = false;
      hasP_ = false;

      // release views
      X_ = Teuchos::null;
      KX_ = Teuchos::null;
      MX_ = Teuchos::null;
      H_ = Teuchos::null;
      KH_ = Teuchos::null;
      MH_ = Teuchos::null;
      P_ = Teuchos::null;
      KP_ = Teuchos::null;
      MP_ = Teuchos::null;

      // free allocated storage
      R_ = Teuchos::null;
      V_ = Teuchos::null;

      // allocate scalar vectors
      theta_.resize(3*newBS,NANVAL);
      Rnorms_.resize(newBS,NANVAL);
      R2norms_.resize(newBS,NANVAL);
      
      // clone multivectors off of tmp
      R_ = MVT::Clone(*tmp,newBS);
      V_ = MVT::Clone(*tmp,newBS*3);
      KV_ = MVT::Clone(*tmp,newBS*3);
      if (hasM_) {
        MV_ = MVT::Clone(*tmp,newBS*3);
      }
      else {
        MV_ = V_;
      }
    }

    // allocate tmp space
    tmpmvec_ = Teuchos::null;
    if (fullOrtho_) {
      tmpmvec_ = MVT::Clone(*tmp,newBS);
    }

    // set new block size
    blockSize_ = newBS;

    // setup new views
    setupViews();
  }


  //////////////////////////////////////////////////////////////////////////////////////////////////
  // Setup views into V,KV,MV
  template <class ScalarType, class MV, class OP>
  void LOBPCG<ScalarType,MV,OP>::setupViews() 
  {
    std::vector<int> ind(blockSize_);

    for (int i=0; i<blockSize_; i++) {
      ind[i] = i;
    }
    X_  = MVT::CloneViewNonConst(*V_,ind);
    KX_ = MVT::CloneViewNonConst(*KV_,ind);
    if (hasM_) {
      MX_ = MVT::CloneViewNonConst(*MV_,ind);
    }
    else {
      MX_ = X_;
    }

    for (int i=0; i<blockSize_; i++) {
      ind[i] += blockSize_;
    }
    H_  = MVT::CloneViewNonConst(*V_,ind);
    KH_ = MVT::CloneViewNonConst(*KV_,ind);
    if (hasM_) {
      MH_ = MVT::CloneViewNonConst(*MV_,ind);
    }
    else {
      MH_ = H_;
    }

    for (int i=0; i<blockSize_; i++) {
      ind[i] += blockSize_;
    }
    P_  = MVT::CloneViewNonConst(*V_,ind);
    KP_ = MVT::CloneViewNonConst(*KV_,ind);
    if (hasM_) {
      MP_ = MVT::CloneViewNonConst(*MV_,ind);
    }
    else {
      MP_ = P_;
    }
  }


  //////////////////////////////////////////////////////////////////////////////////////////////////
  // Set the auxiliary vectors
  template <class ScalarType, class MV, class OP>
  void LOBPCG<ScalarType,MV,OP>::setAuxVecs(const Teuchos::Array<Teuchos::RCP<const MV> > &auxvecs) {
    typedef typename Teuchos::Array<Teuchos::RCP<const MV> >::iterator tarcpmv;

    // set new auxiliary vectors
    auxVecs_ = auxvecs;
    
    numAuxVecs_ = 0;
    for (tarcpmv i=auxVecs_.begin(); i != auxVecs_.end(); i++) {
      numAuxVecs_ += MVT::GetNumberVecs(**i);
    }
    
    // If the solver has been initialized, X and P are not necessarily orthogonal to new auxiliary vectors
    if (numAuxVecs_ > 0 && initialized_) {
      initialized_ = false;
      hasP_ = false;
    }

    if (om_->isVerbosity( Debug ) ) {
      // Check almost everything here
      CheckList chk;
      chk.checkQ = true;
      om_->print( Debug, accuracyCheck(chk, ": in setAuxVecs()") );
    }
  }


  //////////////////////////////////////////////////////////////////////////////////////////////////
  /* Initialize the state of the solver
   * 
   * POST-CONDITIONS:
   *
   * initialized_ == true
   * X is orthonormal, orthogonal to auxVecs_
   * KX = Op*X
   * MX = M*X if hasM_
   * theta_ contains Ritz values of X
   * R = KX - MX*diag(theta_)
   * if hasP() == true,
   *   P orthogonal to auxVecs_
   *   if fullOrtho_ == true,
   *     P orthonormal and orthogonal to X
   *   KP = Op*P
   *   MP = M*P
   */
  template <class ScalarType, class MV, class OP>
  void LOBPCG<ScalarType,MV,OP>::initialize(LOBPCGState<ScalarType,MV>& newstate)
  {
    // NOTE: memory has been allocated by setBlockSize(). Use SetBlock below; do not Clone
    // NOTE: Overall time spent in this routine is counted to timerInit_; portions will also be counted towards other primitives

#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
    Teuchos::TimeMonitor inittimer( *timerInit_ );
#endif

    std::vector<int> bsind(blockSize_);
    for (int i=0; i<blockSize_; i++) bsind[i] = i;

    // in LOBPCG, X (the subspace iterate) is primary
    // the order of dependence follows like so.
    // --init->                 X
    //    --op apply->          MX,KX
    //       --ritz analysis->  theta
    //          --optional->    P,MP,KP
    // 
    // if the user specifies all data for a level, we will accept it.
    // otherwise, we will generate the whole level, and all subsequent levels.
    //
    // the data members are ordered based on dependence, and the levels are
    // partitioned according to the amount of work required to produce the
    // items in a level.
    //
    // inconsitent multivectors widths and lengths will not be tolerated, and
    // will be treated with exceptions.

    // set up X, KX, MX: get them from "state" if user specified them

    //----------------------------------------
    // set up X, MX, KX
    //----------------------------------------
    if (newstate.X != Teuchos::null) {
      TEUCHOS_TEST_FOR_EXCEPTION( MVT::GetGlobalLength(*newstate.X) != MVT::GetGlobalLength(*X_),
                          std::invalid_argument, "Anasazi::LOBPCG::initialize(newstate): vector length of newstate.X not correct." );
      // newstate.X must have blockSize_ vectors; any more will be ignored
      TEUCHOS_TEST_FOR_EXCEPTION( MVT::GetNumberVecs(*newstate.X) < blockSize_,
                          std::invalid_argument, "Anasazi::LOBPCG::initialize(newstate): newstate.X must have at least block size vectors.");

      // put X data in X_
      MVT::SetBlock(*newstate.X,bsind,*X_);

      // put MX data in MX_
      if (hasM_) {
        if (newstate.MX != Teuchos::null) {
          TEUCHOS_TEST_FOR_EXCEPTION( MVT::GetGlobalLength(*newstate.MX) != MVT::GetGlobalLength(*MX_),
                              std::invalid_argument, "Anasazi::LOBPCG::initialize(newstate): vector length of newstate.MX not correct." );
          // newstate.MX must have blockSize_ vectors; any more will be ignored
          TEUCHOS_TEST_FOR_EXCEPTION( MVT::GetNumberVecs(*newstate.MX) < blockSize_,
                              std::invalid_argument, "Anasazi::LOBPCG::initialize(newstate): newstate.MX must have at least block size vectors.");
          MVT::SetBlock(*newstate.MX,bsind,*MX_);
        }
        else {
          // user didn't specify MX, compute it
          {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
            Teuchos::TimeMonitor lcltimer( *timerMOp_ );
#endif
            OPT::Apply(*MOp_,*X_,*MX_);
            count_ApplyM_ += blockSize_;
          }
          // we generated MX; we will generate R as well
          newstate.R = Teuchos::null;
        }
      }
  
      // put data in KX
      if (newstate.KX != Teuchos::null) {
        TEUCHOS_TEST_FOR_EXCEPTION( MVT::GetGlobalLength(*newstate.KX) != MVT::GetGlobalLength(*KX_),
                            std::invalid_argument, "Anasazi::LOBPCG::initialize(newstate): vector length of newstate.KX not correct." );
        // newstate.KX must have blockSize_ vectors; any more will be ignored
        TEUCHOS_TEST_FOR_EXCEPTION( MVT::GetNumberVecs(*newstate.KX) < blockSize_,
                            std::invalid_argument, "Anasazi::LOBPCG::initialize(newstate): newstate.KX must have at least block size vectors.");
        MVT::SetBlock(*newstate.KX,bsind,*KX_);
      }
      else {
        // user didn't specify KX, compute it
        {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
          Teuchos::TimeMonitor lcltimer( *timerOp_ );
#endif
          OPT::Apply(*Op_,*X_,*KX_);
          count_ApplyOp_ += blockSize_;
        }
        // we generated KX; we will generate R as well
        newstate.R = Teuchos::null;
      }
    }
    else {
      // user did not specify X
      // we will initialize X, compute KX and MX, and compute R
      //
      // clear state so we won't use any data from it below
      newstate.P  = Teuchos::null;
      newstate.KP = Teuchos::null;
      newstate.MP = Teuchos::null;
      newstate.R  = Teuchos::null;
      newstate.T  = Teuchos::null;

      // generate a basis and projectAndNormalize
      Teuchos::RCP<const MV> ivec = problem_->getInitVec();
      TEUCHOS_TEST_FOR_EXCEPTION(ivec == Teuchos::null,std::logic_error,
                         "Anasazi::LOBPCG::initialize(): Eigenproblem did not specify initial vectors to clone from.");

      int initSize = MVT::GetNumberVecs(*ivec);
      if (initSize > blockSize_) {
        // we need only the first blockSize_ vectors from ivec; get a view of them
        initSize = blockSize_;
        std::vector<int> ind(blockSize_);
        for (int i=0; i<blockSize_; i++) ind[i] = i;
        ivec = MVT::CloneView(*ivec,ind);
      }

      // assign ivec to first part of X_ 
      if (initSize > 0) {
        std::vector<int> ind(initSize);
        for (int i=0; i<initSize; i++) ind[i] = i;
        MVT::SetBlock(*ivec,ind,*X_);
      }
      // fill the rest of X_ with random
      if (blockSize_ > initSize) {
        std::vector<int> ind(blockSize_ - initSize);
        for (int i=0; i<blockSize_ - initSize; i++) ind[i] = initSize + i;
        Teuchos::RCP<MV> rX = MVT::CloneViewNonConst(*X_,ind);
        MVT::MvRandom(*rX);
        rX = Teuchos::null;
      }

      // put data in MX
      if (hasM_) {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
        Teuchos::TimeMonitor lcltimer( *timerMOp_ );
#endif
        OPT::Apply(*MOp_,*X_,*MX_);
        count_ApplyM_ += blockSize_;
      }
  
      // remove auxVecs from X_ and normalize it
      if (numAuxVecs_ > 0) {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
        Teuchos::TimeMonitor lcltimer( *timerOrtho_ );
#endif
        Teuchos::Array<Teuchos::RCP<Teuchos::SerialDenseMatrix<int,ScalarType> > > dummy;
        int rank = orthman_->projectAndNormalizeMat(*X_,auxVecs_,dummy,Teuchos::null,MX_);
        TEUCHOS_TEST_FOR_EXCEPTION(rank != blockSize_, LOBPCGInitFailure,
                           "Anasazi::LOBPCG::initialize(): Couldn't generate initial basis of full rank.");
      }
      else {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
        Teuchos::TimeMonitor lcltimer( *timerOrtho_ );
#endif
        int rank = orthman_->normalizeMat(*X_,Teuchos::null,MX_);
        TEUCHOS_TEST_FOR_EXCEPTION(rank != blockSize_, LOBPCGInitFailure,
                           "Anasazi::LOBPCG::initialize(): Couldn't generate initial basis of full rank.");
      }

      // put data in KX
      {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
        Teuchos::TimeMonitor lcltimer( *timerOp_ );
#endif
        OPT::Apply(*Op_,*X_,*KX_);
        count_ApplyOp_ += blockSize_;
      }
    } // end if (newstate.X != Teuchos::null)


    //----------------------------------------
    // set up Ritz values
    //----------------------------------------
    theta_.resize(3*blockSize_,NANVAL);
    if (newstate.T != Teuchos::null) {
      TEUCHOS_TEST_FOR_EXCEPTION( (signed int)(newstate.T->size()) < blockSize_,
                          std::invalid_argument, "Anasazi::LOBPCG::initialize(newstate): newstate.T must contain at least block size Ritz values.");
      for (int i=0; i<blockSize_; i++) {
        theta_[i] = (*newstate.T)[i];
      }
      nevLocal_ = blockSize_;
    }
    else {
      // get ritz vecs/vals
      Teuchos::SerialDenseMatrix<int,ScalarType> KK(blockSize_,blockSize_),
                                                 MM(blockSize_,blockSize_),
                                                  S(blockSize_,blockSize_);
      {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
        Teuchos::TimeMonitor lcltimer( *timerLocalProj_ );
#endif
        // project K
        MVT::MvTransMv(ONE,*X_,*KX_,KK);
        // project M
        MVT::MvTransMv(ONE,*X_,*MX_,MM);
        nevLocal_ = blockSize_;
      }

      // solve the projected problem
      {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
        Teuchos::TimeMonitor lcltimer( *timerDS_ );
#endif
        Utils::directSolver(blockSize_, KK, Teuchos::rcpFromRef(MM), S, theta_, nevLocal_, 1);
        TEUCHOS_TEST_FOR_EXCEPTION(nevLocal_ != blockSize_,LOBPCGInitFailure,
                           "Anasazi::LOBPCG::initialize(): Initial Ritz analysis did not produce enough Ritz pairs to initialize algorithm.");
      }

      // We only have blockSize_ ritz pairs, ergo we do not need to select.
      // However, we still require them to be ordered correctly
      {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
        Teuchos::TimeMonitor lcltimer( *timerSort_ );
#endif

        std::vector<int> order(blockSize_);
        // 
        // sort the first blockSize_ values in theta_
        sm_->sort(theta_, Teuchos::rcpFromRef(order), blockSize_);   // don't catch exception
        //
        // apply the same ordering to the primitive ritz vectors
        Utils::permuteVectors(order,S);
      }

      // update the solution, use R for storage
      {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
        Teuchos::TimeMonitor lcltimer( *timerLocalUpdate_ );
#endif
        // X <- X*S
        MVT::MvAddMv( ONE, *X_, ZERO, *X_, *R_ );        
        MVT::MvTimesMatAddMv( ONE, *R_, S, ZERO, *X_ );
        // KX <- KX*S
        MVT::MvAddMv( ONE, *KX_, ZERO, *KX_, *R_ );        
        MVT::MvTimesMatAddMv( ONE, *R_, S, ZERO, *KX_ );
        if (hasM_) {
          // MX <- MX*S
          MVT::MvAddMv( ONE, *MX_, ZERO, *MX_, *R_ );        
          MVT::MvTimesMatAddMv( ONE, *R_, S, ZERO, *MX_ );
        }
      }
    }

    //----------------------------------------
    // compute R
    //----------------------------------------
    if (newstate.R != Teuchos::null) {
      TEUCHOS_TEST_FOR_EXCEPTION( MVT::GetGlobalLength(*newstate.R) != MVT::GetGlobalLength(*R_),
                          std::invalid_argument, "Anasazi::LOBPCG::initialize(newstate): vector length of newstate.R not correct." );
      TEUCHOS_TEST_FOR_EXCEPTION( MVT::GetNumberVecs(*newstate.R) < blockSize_,
                          std::invalid_argument, "Anasazi::LOBPCG::initialize(newstate): newstate.R must have blockSize number of vectors." );
      MVT::SetBlock(*newstate.R,bsind,*R_);
    }
    else {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
      Teuchos::TimeMonitor lcltimer( *timerCompRes_ );
#endif
      // form R <- KX - MX*T
      MVT::MvAddMv(ZERO,*KX_,ONE,*KX_,*R_);
      Teuchos::SerialDenseMatrix<int,ScalarType> T(blockSize_,blockSize_);
      for (int i=0; i<blockSize_; i++) T(i,i) = theta_[i];
      MVT::MvTimesMatAddMv(-ONE,*MX_,T,ONE,*R_);
    }

    // R has been updated; mark the norms as out-of-date
    Rnorms_current_ = false;
    R2norms_current_ = false;
  
    // put data in P,KP,MP: P is not used to set theta
    if (newstate.P != Teuchos::null) {
      TEUCHOS_TEST_FOR_EXCEPTION( MVT::GetNumberVecs(*newstate.P) < blockSize_ ,
                          std::invalid_argument, "Anasazi::LOBPCG::initialize(newstate): newstate.P must have blockSize number of vectors." );
      TEUCHOS_TEST_FOR_EXCEPTION( MVT::GetGlobalLength(*newstate.P) != MVT::GetGlobalLength(*P_),
                          std::invalid_argument, "Anasazi::LOBPCG::initialize(newstate): vector length of newstate.P not correct." );
      hasP_ = true;

      // set P_
      MVT::SetBlock(*newstate.P,bsind,*P_);

      // set/compute KP_
      if (newstate.KP != Teuchos::null) {
        TEUCHOS_TEST_FOR_EXCEPTION( MVT::GetNumberVecs(*newstate.KP) < blockSize_,
                            std::invalid_argument, "Anasazi::LOBPCG::initialize(newstate): newstate.KP must have blockSize number of vectors." );
        TEUCHOS_TEST_FOR_EXCEPTION( MVT::GetGlobalLength(*newstate.KP) != MVT::GetGlobalLength(*KP_),
                            std::invalid_argument, "Anasazi::LOBPCG::initialize(newstate): vector length of newstate.KP not correct." );
        MVT::SetBlock(*newstate.KP,bsind,*KP_);
      }
      else {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
        Teuchos::TimeMonitor lcltimer( *timerOp_ );
#endif
        OPT::Apply(*Op_,*P_,*KP_);
        count_ApplyOp_ += blockSize_;
      }

      // set/compute MP_
      if (hasM_) {
        if (newstate.MP != Teuchos::null) {
          TEUCHOS_TEST_FOR_EXCEPTION( MVT::GetNumberVecs(*newstate.MP) < blockSize_,
                              std::invalid_argument, "Anasazi::LOBPCG::initialize(newstate): newstate.MP must have blockSize number of vectors." );
          TEUCHOS_TEST_FOR_EXCEPTION( MVT::GetGlobalLength(*newstate.MP) != MVT::GetGlobalLength(*MP_),
                              std::invalid_argument, "Anasazi::LOBPCG::initialize(newstate): vector length of newstate.MP not correct." );
          MVT::SetBlock(*newstate.MP,bsind,*MP_);
        }
        else {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
          Teuchos::TimeMonitor lcltimer( *timerMOp_ );
#endif
          OPT::Apply(*MOp_,*P_,*MP_);
          count_ApplyM_ += blockSize_;
        }
      }
    }
    else {
      hasP_ = false;
    }

    // finally, we are initialized
    initialized_ = true;

    if (om_->isVerbosity( Debug ) ) {
      // Check almost everything here
      CheckList chk;
      chk.checkX = true;
      chk.checkKX = true;
      chk.checkMX = true;
      chk.checkP = true;
      chk.checkKP = true;
      chk.checkMP = true;
      chk.checkR = true;
      chk.checkQ = true;
      om_->print( Debug, accuracyCheck(chk, ": after initialize()") );
    }

  }

  template <class ScalarType, class MV, class OP>
  void LOBPCG<ScalarType,MV,OP>::initialize()
  {
    LOBPCGState<ScalarType,MV> empty;
    initialize(empty);
  }


  //////////////////////////////////////////////////////////////////////////////////////////////////
  // Instruct the solver to use full orthogonalization
  template <class ScalarType, class MV, class OP>
  void LOBPCG<ScalarType,MV,OP>::setFullOrtho (bool fullOrtho)
  {
    if ( fullOrtho_ == true || initialized_ == false || fullOrtho == fullOrtho_ ) {
      // state is already orthogonalized or solver is not initialized
      fullOrtho_ = fullOrtho;
    }
    else {
      // solver is initialized, state is not fully orthogonalized, and user has requested full orthogonalization
      // ergo, we must throw away data in P
      fullOrtho_ = true;
      hasP_ = false;
    }

    // the user has called setFullOrtho, so the class has been instantiated
    // ergo, the data has already been allocated, i.e., setBlockSize() has been called
    // if it is already allocated, it should be the proper size
    if (fullOrtho_ && tmpmvec_ == Teuchos::null) {
      // allocated the workspace
      tmpmvec_ = MVT::Clone(*X_,blockSize_);
    }
    else if (fullOrtho_==false) {
      // free the workspace
      tmpmvec_ = Teuchos::null;
    }
  }


  //////////////////////////////////////////////////////////////////////////////////////////////////
  // Perform LOBPCG iterations until the StatusTest tells us to stop.
  template <class ScalarType, class MV, class OP>
  void LOBPCG<ScalarType,MV,OP>::iterate () 
  {
    //
    // Allocate/initialize data structures
    //
    if (initialized_ == false) {
      initialize();
    }

    //
    // Miscellaneous definitions
    const int oneBlock    =   blockSize_;
    const int twoBlocks   = 2*blockSize_;
    const int threeBlocks = 3*blockSize_;

    std::vector<int> indblock1(blockSize_), indblock2(blockSize_), indblock3(blockSize_);
    for (int i=0; i<blockSize_; i++) {
      indblock1[i] = i;
      indblock2[i] = i +  blockSize_;
      indblock3[i] = i + 2*blockSize_;
    }

    //
    // Define dense projected/local matrices
    //   KK = Local stiffness matrix               (size: 3*blockSize_ x 3*blockSize_)
    //   MM = Local mass matrix                    (size: 3*blockSize_ x 3*blockSize_)
    //    S = Local eigenvectors                   (size: 3*blockSize_ x 3*blockSize_)
    Teuchos::SerialDenseMatrix<int,ScalarType> KK( threeBlocks, threeBlocks ), 
                                               MM( threeBlocks, threeBlocks ),
                                                S( threeBlocks, threeBlocks );

    while (tester_->checkStatus(this) != Passed) {

      // Print information on current status
      if (om_->isVerbosity(Debug)) {
        currentStatus( om_->stream(Debug) );
      }
      else if (om_->isVerbosity(IterationDetails)) {
        currentStatus( om_->stream(IterationDetails) );
      }

      // increment iteration counter
      iter_++;

      // Apply the preconditioner on the residuals: H <- Prec*R
      if (Prec_ != Teuchos::null) {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
        Teuchos::TimeMonitor lcltimer( *timerPrec_ );
#endif
        OPT::Apply( *Prec_, *R_, *H_ );   // don't catch the exception
        count_ApplyPrec_ += blockSize_;
      }
      else {
        MVT::MvAddMv(ONE,*R_,ZERO,*R_,*H_);
      }

      // Apply the mass matrix on H
      if (hasM_) {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
        Teuchos::TimeMonitor lcltimer( *timerMOp_ );
#endif
        OPT::Apply( *MOp_, *H_, *MH_);    // don't catch the exception
        count_ApplyM_ += blockSize_;
      }

      // orthogonalize H against the auxiliary vectors
      // optionally: orthogonalize H against X and P ([X P] is already orthonormal)
      Teuchos::Array<Teuchos::RCP<const MV> > Q;
      Q = auxVecs_;
      if (fullOrtho_) {
        // X and P are not contiguous, so there is not much point in putting them under 
        // a single multivector view
        Q.push_back(X_);
        if (hasP_) {
          Q.push_back(P_);
        }
      }
      {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
        Teuchos::TimeMonitor lcltimer( *timerOrtho_ );
#endif
        Teuchos::Array<Teuchos::RCP<Teuchos::SerialDenseMatrix<int,ScalarType> > > dummyC = 
          Teuchos::tuple<Teuchos::RCP<Teuchos::SerialDenseMatrix<int,ScalarType> > >(Teuchos::null);
        int rank = orthman_->projectAndNormalizeMat(*H_,Q,dummyC,Teuchos::null,MH_);
        // our views are currently in place; it is safe to throw an exception
        TEUCHOS_TEST_FOR_EXCEPTION(rank != blockSize_,LOBPCGOrthoFailure,
                           "Anasazi::LOBPCG::iterate(): unable to compute orthonormal basis for H.");
      }

      if (om_->isVerbosity( Debug ) ) {
        CheckList chk;
        chk.checkH = true;
        chk.checkMH = true;
        om_->print( Debug, accuracyCheck(chk, ": after ortho H") );
      }
      else if (om_->isVerbosity( OrthoDetails ) ) {
        CheckList chk;
        chk.checkH = true;
        chk.checkMH = true;
        om_->print( OrthoDetails, accuracyCheck(chk,": after ortho H") );
      }

      // Apply the stiffness matrix to H
      {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
        Teuchos::TimeMonitor lcltimer( *timerOp_ );
#endif
        OPT::Apply( *Op_, *H_, *KH_);   // don't catch the exception
        count_ApplyOp_ += blockSize_;
      }

      if (hasP_) {
        nevLocal_ = threeBlocks;
      }
      else {
        nevLocal_ = twoBlocks;
      }

      //
      // we need bases: [X H P] and [H P] (only need the latter if fullOrtho == false)
      // we need to perform the following operations:
      //    X' [KX KH KP]
      //    X' [MX MH MP]
      //    H' [KH KP]
      //    H' [MH MP]
      //    P' [KP]
      //    P' [MP]
      //    [X H P] CX
      //    [X H P] CP    if  fullOrtho
      //      [H P] CP    if !fullOrtho
      //
      // since M[X H P] is potentially the same memory as [X H P], and 
      // because we are not allowed to have overlapping non-const views of 
      // a multivector, we will now abandon our non-const views in favor of 
      // const views
      //
      X_ = Teuchos::null;
      KX_ = Teuchos::null;
      MX_ = Teuchos::null;
      H_ = Teuchos::null;
      KH_ = Teuchos::null;
      MH_ = Teuchos::null;
      P_ = Teuchos::null;
      KP_ = Teuchos::null;
      MP_ = Teuchos::null;
      Teuchos::RCP<const MV> cX, cH, cXHP, cHP, cK_XHP, cK_HP, cM_XHP, cM_HP, cP, cK_P, cM_P;
      {
        cX = MVT::CloneView(*Teuchos::rcp_implicit_cast<const MV>(V_),indblock1);
        cH = MVT::CloneView(*Teuchos::rcp_implicit_cast<const MV>(V_),indblock2);

        std::vector<int> indXHP(nevLocal_);
        for (int i=0; i<nevLocal_; i++) {
          indXHP[i] = i;
        }
        cXHP = MVT::CloneView(*Teuchos::rcp_implicit_cast<const MV>(V_),indXHP);
        cK_XHP = MVT::CloneView(*Teuchos::rcp_implicit_cast<const MV>(KV_),indXHP);
        if (hasM_) {
          cM_XHP = MVT::CloneView(*Teuchos::rcp_implicit_cast<const MV>(MV_),indXHP);
        }
        else {
          cM_XHP = cXHP;
        }

        std::vector<int> indHP(nevLocal_-blockSize_);
        for (int i=blockSize_; i<nevLocal_; i++) {
          indHP[i-blockSize_] = i;
        }
        cHP = MVT::CloneView(*Teuchos::rcp_implicit_cast<const MV>(V_),indHP);
        cK_HP = MVT::CloneView(*Teuchos::rcp_implicit_cast<const MV>(KV_),indHP);
        if (hasM_) {
          cM_HP = MVT::CloneView(*Teuchos::rcp_implicit_cast<const MV>(MV_),indHP);
        }
        else {
          cM_HP = cHP;
        }

        if (nevLocal_ == threeBlocks) {
          cP = MVT::CloneView(*Teuchos::rcp_implicit_cast<const MV>(V_),indblock3);
          cK_P = MVT::CloneView(*Teuchos::rcp_implicit_cast<const MV>(KV_),indblock3);
          if (hasM_) {
            cM_P = MVT::CloneView(*Teuchos::rcp_implicit_cast<const MV>(MV_),indblock3);
          }
          else {
            cM_P = cP;
          }
        }
      }

      //
      //----------------------------------------
      // Form "local" mass and stiffness matrices
      //----------------------------------------
      {
        // We will form only the block upper triangular part of 
        // [X H P]' K [X H P]  and  [X H P]' M [X H P]
        // Get the necessary views into KK and MM:
        //      [--K1--]        [--M1--]
        // KK = [  -K2-]   MM = [  -M2-]
        //      [    K3]        [    M3]
        // 
        // It is okay to declare a zero-area view of a Teuchos::SerialDenseMatrix
        //
        Teuchos::SerialDenseMatrix<int,ScalarType> 
          K1(Teuchos::View,KK,blockSize_,nevLocal_             ,0*blockSize_,0*blockSize_),
          K2(Teuchos::View,KK,blockSize_,nevLocal_-1*blockSize_,1*blockSize_,1*blockSize_),
          K3(Teuchos::View,KK,blockSize_,nevLocal_-2*blockSize_,2*blockSize_,2*blockSize_),
          M1(Teuchos::View,MM,blockSize_,nevLocal_             ,0*blockSize_,0*blockSize_),
          M2(Teuchos::View,MM,blockSize_,nevLocal_-1*blockSize_,1*blockSize_,1*blockSize_),
          M3(Teuchos::View,MM,blockSize_,nevLocal_-2*blockSize_,2*blockSize_,2*blockSize_);
        {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
          Teuchos::TimeMonitor lcltimer( *timerLocalProj_ );
#endif
          MVT::MvTransMv( ONE, *cX, *cK_XHP, K1 );
          MVT::MvTransMv( ONE, *cX, *cM_XHP, M1 );
          MVT::MvTransMv( ONE, *cH, *cK_HP , K2 );
          MVT::MvTransMv( ONE, *cH, *cM_HP , M2 );
          if (nevLocal_ == threeBlocks) {
            MVT::MvTransMv( ONE, *cP, *cK_P, K3 );
            MVT::MvTransMv( ONE, *cP, *cM_P, M3 );
          }
        }
      }
      // below, we only need bases [X H P] and [H P] and friends
      // furthermore, we only need [H P] and friends if fullOrtho == false
      // clear the others now
      cX   = Teuchos::null;
      cH   = Teuchos::null;
      cP   = Teuchos::null;
      cK_P = Teuchos::null;
      cM_P = Teuchos::null;
      if (fullOrtho_ == true) {
        cHP = Teuchos::null;
        cK_HP = Teuchos::null;
        cM_HP = Teuchos::null;
      }

      //
      //---------------------------------------------------
      // Perform a spectral decomposition of (KK,MM)
      //---------------------------------------------------
      //
      // Get pointers to relevant part of KK, MM and S for Rayleigh-Ritz analysis
      Teuchos::SerialDenseMatrix<int,ScalarType> lclKK(Teuchos::View,KK,nevLocal_,nevLocal_), 
                                                 lclMM(Teuchos::View,MM,nevLocal_,nevLocal_),
                                                  lclS(Teuchos::View, S,nevLocal_,nevLocal_);
      {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
        Teuchos::TimeMonitor lcltimer( *timerDS_ );
#endif
        int localSize = nevLocal_;
        Utils::directSolver(localSize, lclKK, Teuchos::rcpFromRef(lclMM), lclS, theta_, nevLocal_, 0);
        // localSize tells directSolver() how big KK,MM are
        // however, directSolver() may choose to use only the principle submatrices of KK,MM 
        // because of loss of MM-orthogonality in the projected eigenvectors
        // nevLocal_ tells us how much it used, telling us the effective localSize 
        // (i.e., how much of KK,MM used by directSolver)
        // we will not tolerate any indefiniteness, and will throw an exception if it was 
        // detected by directSolver
        //
        if (nevLocal_ != localSize) {
          // before throwing the exception, and thereby leaving iterate(), setup the views again
          // first, clear the const views
          cXHP   = Teuchos::null;
          cK_XHP = Teuchos::null;
          cM_XHP = Teuchos::null;
          cHP    = Teuchos::null;
          cK_HP  = Teuchos::null;
          cM_HP  = Teuchos::null;
          setupViews();
        }
        TEUCHOS_TEST_FOR_EXCEPTION(nevLocal_ != localSize, LOBPCGRitzFailure, 
            "Anasazi::LOBPCG::iterate(): indefiniteness detected in projected mass matrix." );
      }

      //
      //---------------------------------------------------
      // Sort the ritz values using the sort manager
      //---------------------------------------------------
      Teuchos::LAPACK<int,ScalarType> lapack;
      Teuchos::BLAS<int,ScalarType> blas;
      {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
        Teuchos::TimeMonitor lcltimer( *timerSort_ );
#endif

        std::vector<int> order(nevLocal_);
        // 
        // Sort the first nevLocal_ values in theta_
        sm_->sort(theta_, Teuchos::rcpFromRef(order), nevLocal_);   // don't catch exception
        //
        // Sort the primitive ritz vectors
        Utils::permuteVectors(order,lclS);
      }

      //
      //----------------------------------------
      // Compute coefficients for X and P under [X H P]
      //----------------------------------------
      // Before computing X,P, optionally perform orthogonalization per Hetmaniuk,Lehoucq paper
      // CX will be the coefficients of [X,H,P] for new X, CP for new P
      // The paper suggests orthogonalizing CP against CX and orthonormalizing CP, w.r.t. MM
      // Here, we will also orthonormalize CX.
      // This is accomplished using the Cholesky factorization of [CX CP]^H lclMM [CX CP]
      Teuchos::RCP<Teuchos::SerialDenseMatrix<int,ScalarType> > CX, CP;
      if (fullOrtho_) {
        // build orthonormal basis for (  0  ) that is MM orthogonal to ( S11 )
        //                             ( S21 )                          ( S21 )
        //                             ( S31 )                          ( S31 )
        // Do this using Cholesky factorization of ( S11  0  )
        //                                         ( S21 S21 )
        //                                         ( S31 S31 )
        //           ( S11  0  )
        // Build C = ( S21 S21 )
        //           ( S31 S31 )
        Teuchos::SerialDenseMatrix<int,ScalarType> C(nevLocal_,twoBlocks),
                                                MMC(nevLocal_,twoBlocks),
                                                CMMC(twoBlocks  ,twoBlocks);

        // first block of rows: ( S11 0 )
        for (int j=0; j<oneBlock; j++) {
          for (int i=0; i<oneBlock; i++) {
            // CX
            C(i,j) = lclS(i,j);
            // CP
            C(i,j+oneBlock) = ZERO;
          }
        }
        // second block of rows: (S21 S21)
        for (int j=0; j<oneBlock; j++) {
          for (int i=oneBlock; i<twoBlocks; i++) {
            // CX
            C(i,j)          = lclS(i,j);
            // CP
            C(i,j+oneBlock) = lclS(i,j);
          }
        }
        // third block of rows
        if (nevLocal_ == threeBlocks) {
          for (int j=0; j<oneBlock; j++) {
            for (int i=twoBlocks; i<threeBlocks; i++) {
              // CX
              C(i,j)          = lclS(i,j);
              // CP
              C(i,j+oneBlock) = lclS(i,j);
            }
          }
        }

        // compute MMC = lclMM*C
        {
          int teuchosret;
          teuchosret = MMC.multiply(Teuchos::NO_TRANS,Teuchos::NO_TRANS,ONE,lclMM,C,ZERO);
          TEUCHOS_TEST_FOR_EXCEPTION(teuchosret != 0,std::logic_error,
              "Anasazi::LOBPCG::iterate(): Logic error calling SerialDenseMatrix::multiply");
          // compute CMMC = C^H*MMC == C^H*lclMM*C
          teuchosret = CMMC.multiply(Teuchos::CONJ_TRANS,Teuchos::NO_TRANS,ONE,C,MMC,ZERO);
          TEUCHOS_TEST_FOR_EXCEPTION(teuchosret != 0,std::logic_error,
              "Anasazi::LOBPCG::iterate(): Logic error calling SerialDenseMatrix::multiply");
        }

        // compute R (cholesky) of CMMC
        {
          int info;
          lapack.POTRF('U',twoBlocks,CMMC.values(),CMMC.stride(),&info);
          // our views ARE NOT currently in place; we must reestablish them before throwing an exception
          if (info != 0) {
            // Check symmetry of H'*K*H, this might be the first place a non-Hermitian operator will cause a problem.
            Teuchos::SerialDenseMatrix<int,ScalarType> K22(Teuchos::View,KK,blockSize_,blockSize_,1*blockSize_,1*blockSize_);
            Teuchos::SerialDenseMatrix<int,ScalarType> K22_trans( K22, Teuchos::CONJ_TRANS );
            K22_trans -= K22;
            MagnitudeType symNorm = K22_trans.normOne();
            MagnitudeType tol = SCT::magnitude(SCT::squareroot(SCT::eps()));

            cXHP = Teuchos::null;
            cHP = Teuchos::null;
            cK_XHP = Teuchos::null;
            cK_HP = Teuchos::null;
            cM_XHP = Teuchos::null;
            cM_HP = Teuchos::null;
            setupViews();
           
            std::string errMsg = "Anasazi::LOBPCG::iterate(): Cholesky factorization failed during full orthogonalization.";
            if ( symNorm < tol )
            {
              TEUCHOS_TEST_FOR_EXCEPTION(info != 0, LOBPCGOrthoFailure, errMsg );
            }
            else
            {
              errMsg += " Check the operator A (or K), it may not be Hermitian!";
              TEUCHOS_TEST_FOR_EXCEPTION(info != 0, LOBPCGOrthoFailure, errMsg );
            }
          }
        }
        // compute C = C inv(R)
        blas.TRSM(Teuchos::RIGHT_SIDE,Teuchos::UPPER_TRI,Teuchos::NO_TRANS,Teuchos::NON_UNIT_DIAG,
                  nevLocal_,twoBlocks,ONE,CMMC.values(),CMMC.stride(),C.values(),C.stride());
        // put C(:,0:oneBlock-1) into CX
        CX = Teuchos::rcp( new Teuchos::SerialDenseMatrix<int,ScalarType>(Teuchos::Copy,C,nevLocal_,oneBlock,0,0) );
        // put C(:,oneBlock:twoBlocks-1) into CP
        CP = Teuchos::rcp( new Teuchos::SerialDenseMatrix<int,ScalarType>(Teuchos::Copy,C,nevLocal_,oneBlock,0,oneBlock) );

        // check the results
        if (om_->isVerbosity( Debug ) ) {
          Teuchos::SerialDenseMatrix<int,ScalarType> tmp1(nevLocal_,oneBlock),
                                                     tmp2(oneBlock,oneBlock);
          MagnitudeType tmp;
          int teuchosret;
          std::stringstream os;
          os.precision(2);
          os.setf(std::ios::scientific, std::ios::floatfield);

          os << " Checking Full Ortho: iteration " << iter_ << std::endl;

          // check CX^T MM CX == I
          // compute tmp1 = MM*CX
          teuchosret = tmp1.multiply(Teuchos::NO_TRANS,Teuchos::NO_TRANS,ONE,lclMM,*CX,ZERO);
          TEUCHOS_TEST_FOR_EXCEPTION(teuchosret != 0,std::logic_error,
              "Anasazi::LOBPCG::iterate(): Logic error calling SerialDenseMatrix::multiply");
          // compute tmp2 = CX^H*tmp1 == CX^H*MM*CX
          teuchosret = tmp2.multiply(Teuchos::CONJ_TRANS,Teuchos::NO_TRANS,ONE,*CX,tmp1,ZERO);
          TEUCHOS_TEST_FOR_EXCEPTION(teuchosret != 0,std::logic_error,
              "Anasazi::LOBPCG::iterate(): Logic error calling SerialDenseMatrix::multiply");
          // subtrace tmp2 - I == CX^H * MM * CX - I
          for (int i=0; i<oneBlock; i++) tmp2(i,i) -= ONE;
          tmp = tmp2.normFrobenius();          
          os << " >> Error in CX^H MM CX == I : " << tmp << std::endl;

          // check CP^T MM CP == I
          // compute tmp1 = MM*CP
          teuchosret = tmp1.multiply(Teuchos::NO_TRANS,Teuchos::NO_TRANS,ONE,lclMM,*CP,ZERO);
          TEUCHOS_TEST_FOR_EXCEPTION(teuchosret != 0,std::logic_error,
              "Anasazi::LOBPCG::iterate(): Logic error calling SerialDenseMatrix::multiply");
          // compute tmp2 = CP^H*tmp1 == CP^H*MM*CP
          teuchosret = tmp2.multiply(Teuchos::CONJ_TRANS,Teuchos::NO_TRANS,ONE,*CP,tmp1,ZERO);
          TEUCHOS_TEST_FOR_EXCEPTION(teuchosret != 0,std::logic_error,
              "Anasazi::LOBPCG::iterate(): Logic error calling SerialDenseMatrix::multiply");
          // subtrace tmp2 - I == CP^H * MM * CP - I
          for (int i=0; i<oneBlock; i++) tmp2(i,i) -= ONE;
          tmp = tmp2.normFrobenius();          
          os << " >> Error in CP^H MM CP == I : " << tmp << std::endl;

          // check CX^T MM CP == 0
          // compute tmp1 = MM*CP
          teuchosret = tmp1.multiply(Teuchos::NO_TRANS,Teuchos::NO_TRANS,ONE,lclMM,*CP,ZERO);
          TEUCHOS_TEST_FOR_EXCEPTION(teuchosret != 0,std::logic_error,"Anasazi::LOBPCG::iterate(): Logic error calling SerialDenseMatrix::multiply");
          // compute tmp2 = CX^H*tmp1 == CX^H*MM*CP
          teuchosret = tmp2.multiply(Teuchos::CONJ_TRANS,Teuchos::NO_TRANS,ONE,*CX,tmp1,ZERO);
          TEUCHOS_TEST_FOR_EXCEPTION(teuchosret != 0,std::logic_error,"Anasazi::LOBPCG::iterate(): Logic error calling SerialDenseMatrix::multiply");
          // subtrace tmp2 == CX^H * MM * CP
          tmp = tmp2.normFrobenius();          
          os << " >> Error in CX^H MM CP == 0 : " << tmp << std::endl;

          os << std::endl;
          om_->print(Debug,os.str());
        }
      }
      else {
        //     [S11 ... ...]
        // S = [S21 ... ...]
        //     [S31 ... ...]
        //
        // CX = [S11]
        //      [S21]
        //      [S31]   ->  X = [X H P] CX
        //      
        // CP = [S21]   ->  P =   [H P] CP
        //      [S31]
        //
        CX = Teuchos::rcp( new Teuchos::SerialDenseMatrix<int,ScalarType>(Teuchos::Copy,lclS,nevLocal_         ,oneBlock,0       ,0) );
        CP = Teuchos::rcp( new Teuchos::SerialDenseMatrix<int,ScalarType>(Teuchos::Copy,lclS,nevLocal_-oneBlock,oneBlock,oneBlock,0) );
      }

      //
      //----------------------------------------
      // Compute new X and new P
      //----------------------------------------
      // Note: Use R as a temporary work space and (if full ortho) tmpMV as well
      {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
        Teuchos::TimeMonitor lcltimer( *timerLocalUpdate_ );
#endif

        // if full ortho, then CX and CP are dense
        // we multiply [X H P]*CX into tmpMV
        //             [X H P]*CP into R
        // then put V(:,firstblock) <- tmpMV
        //          V(:,thirdblock) <- R
        //
        // if no full ortho, then [H P]*CP doesn't reference first block (X) 
        // of V, so that we can modify it before computing P
        // so we multiply [X H P]*CX into R
        //                V(:,firstblock) <- R
        //       multiply [H P]*CP into R
        //                V(:,thirdblock) <- R
        //
        // mutatis mutandis for K[XP] and M[XP]
        //
        // use SetBlock to do the assignments into V_
        //
        // in either case, views are only allowed to be overlapping
        // if they are const, and it should be assume that SetBlock
        // creates a view of the associated part
        //
        // we have from above const-pointers to [KM]XHP, [KM]HP and (if hasP) [KM]P
        //
        if (fullOrtho_) {
          // X,P
          MVT::MvTimesMatAddMv(ONE,*cXHP,*CX,ZERO,*tmpmvec_);
          MVT::MvTimesMatAddMv(ONE,*cXHP,*CP,ZERO,*R_);
          cXHP = Teuchos::null;
          MVT::SetBlock(*tmpmvec_,indblock1,*V_);
          MVT::SetBlock(*R_      ,indblock3,*V_);
          // KX,KP
          MVT::MvTimesMatAddMv(ONE,*cK_XHP,*CX,ZERO,*tmpmvec_);
          MVT::MvTimesMatAddMv(ONE,*cK_XHP,*CP,ZERO,*R_);
          cK_XHP = Teuchos::null;
          MVT::SetBlock(*tmpmvec_,indblock1,*KV_);
          MVT::SetBlock(*R_      ,indblock3,*KV_);
          // MX,MP
          if (hasM_) {
            MVT::MvTimesMatAddMv(ONE,*cM_XHP,*CX,ZERO,*tmpmvec_);
            MVT::MvTimesMatAddMv(ONE,*cM_XHP,*CP,ZERO,*R_);
            cM_XHP = Teuchos::null;
            MVT::SetBlock(*tmpmvec_,indblock1,*MV_);
            MVT::SetBlock(*R_      ,indblock3,*MV_);
          }
          else {
            cM_XHP = Teuchos::null;
          }
        }
        else {
          // X,P
          MVT::MvTimesMatAddMv(ONE,*cXHP,*CX,ZERO,*R_);
          cXHP = Teuchos::null;
          MVT::SetBlock(*R_,indblock1,*V_);
          MVT::MvTimesMatAddMv(ONE,*cHP,*CP,ZERO,*R_);
          cHP = Teuchos::null;
          MVT::SetBlock(*R_,indblock3,*V_);
          // KX,KP
          MVT::MvTimesMatAddMv(ONE,*cK_XHP,*CX,ZERO,*R_);
          cK_XHP = Teuchos::null;
          MVT::SetBlock(*R_,indblock1,*KV_);
          MVT::MvTimesMatAddMv(ONE,*cK_HP,*CP,ZERO,*R_);
          cK_HP = Teuchos::null;
          MVT::SetBlock(*R_,indblock3,*KV_);
          // MX,MP
          if (hasM_) {
            MVT::MvTimesMatAddMv(ONE,*cM_XHP,*CX,ZERO,*R_);
            cM_XHP = Teuchos::null;
            MVT::SetBlock(*R_,indblock1,*MV_);
            MVT::MvTimesMatAddMv(ONE,*cM_HP,*CP,ZERO,*R_);
            cM_HP = Teuchos::null;
            MVT::SetBlock(*R_,indblock3,*MV_);
          }
          else {
            cM_XHP = Teuchos::null;
            cM_HP = Teuchos::null;
          }
        }
      } // end timing block
      // done with coefficient matrices
      CX = Teuchos::null;
      CP = Teuchos::null;

      //
      // we now have a P direction
      hasP_ = true;

      // debugging check: all of our const views should have been cleared by now
      // if not, we have a logic error above
      TEUCHOS_TEST_FOR_EXCEPTION(   cXHP != Teuchos::null || cK_XHP != Teuchos::null || cM_XHP != Teuchos::null
                          || cHP != Teuchos::null ||  cK_HP != Teuchos::null || cM_HP  != Teuchos::null
                          ||  cP != Teuchos::null ||   cK_P != Teuchos::null || cM_P   != Teuchos::null,
                          std::logic_error,
                          "Anasazi::BlockKrylovSchur::iterate(): const views were not all cleared! Something went wrong!" );

      //
      // recreate our const MV views of X,H,P and friends
      setupViews();

      //
      // Compute the new residuals, explicitly
      {
#ifdef ANASAZI_TEUCHOS_TIME_MONITOR
        Teuchos::TimeMonitor lcltimer( *timerCompRes_ );
#endif
        MVT::MvAddMv( ONE, *KX_, ZERO, *KX_, *R_ );
        Teuchos::SerialDenseMatrix<int,ScalarType> T( blockSize_, blockSize_ );
        for (int i = 0; i < blockSize_; i++) {
          T(i,i) = theta_[i];
        }
        MVT::MvTimesMatAddMv( -ONE, *MX_, T, ONE, *R_ );
      }

      // R has been updated; mark the norms as out-of-date
      Rnorms_current_ = false;
      R2norms_current_ = false;

      // When required, monitor some orthogonalities
      if (om_->isVerbosity( Debug ) ) {
        // Check almost everything here
        CheckList chk;
        chk.checkX = true;
        chk.checkKX = true;
        chk.checkMX = true;
        chk.checkP = true;
        chk.checkKP = true;
        chk.checkMP = true;
        chk.checkR = true;
        om_->print( Debug, accuracyCheck(chk, ": after local update") );
      }
      else if (om_->isVerbosity( OrthoDetails )) {
        CheckList chk;
        chk.checkX = true;
        chk.checkP = true;
        chk.checkR = true;
        om_->print( OrthoDetails, accuracyCheck(chk, ": after local update") );
      }
    } // end while (statusTest == false)
  }


  //////////////////////////////////////////////////////////////////////////////////////////////////
  // compute/return residual M-norms
  template <class ScalarType, class MV, class OP>
  std::vector<typename Teuchos::ScalarTraits<ScalarType>::magnitudeType> 
  LOBPCG<ScalarType,MV,OP>::getResNorms() {
    if (Rnorms_current_ == false) {
      // Update the residual norms
      orthman_->norm(*R_,Rnorms_);
      Rnorms_current_ = true;
    }
    return Rnorms_;
  }

  
  //////////////////////////////////////////////////////////////////////////////////////////////////
  // compute/return residual 2-norms
  template <class ScalarType, class MV, class OP>
  std::vector<typename Teuchos::ScalarTraits<ScalarType>::magnitudeType> 
  LOBPCG<ScalarType,MV,OP>::getRes2Norms() {
    if (R2norms_current_ == false) {
      // Update the residual 2-norms 
      MVT::MvNorm(*R_,R2norms_);
      R2norms_current_ = true;
    }
    return R2norms_;
  }




  //////////////////////////////////////////////////////////////////////////////////////////////////
  // Check accuracy, orthogonality, and other debugging stuff
  // 
  // bools specify which tests we want to run (instead of running more than we actually care about)
  //
  // we don't bother checking the following because they are computed explicitly:
  //    H == Prec*R
  //   KH == K*H
  //
  // 
  // checkX : X orthonormal
  //          orthogonal to auxvecs
  // checkMX: check MX == M*X
  // checkKX: check KX == K*X
  // checkP : if fullortho P orthonormal and orthogonal to X
  //          orthogonal to auxvecs
  // checkMP: check MP == M*P
  // checkKP: check KP == K*P
  // checkH : if fullortho H orthonormal and orthogonal to X and P
  //          orthogonal to auxvecs
  // checkMH: check MH == M*H
  // checkR : check R orthogonal to X
  // checkQ : check that auxiliary vectors are actually orthonormal
  //
  // TODO: 
  //  add checkTheta 
  //
  template <class ScalarType, class MV, class OP>
  std::string LOBPCG<ScalarType,MV,OP>::accuracyCheck( const CheckList &chk, const std::string &where ) const 
  {
    using std::endl;

    std::stringstream os;
    os.precision(2);
    os.setf(std::ios::scientific, std::ios::floatfield);
    MagnitudeType tmp;

    os << " Debugging checks: iteration " << iter_ << where << endl;

    // X and friends
    if (chk.checkX && initialized_) {
      tmp = orthman_->orthonormError(*X_);
      os << " >> Error in X^H M X == I : " << tmp << endl;
      for (Array_size_type i=0; i<auxVecs_.size(); i++) {
        tmp = orthman_->orthogError(*X_,*auxVecs_[i]);
        os << " >> Error in X^H M Q[" << i << "] == 0 : " << tmp << endl;
      }
    }
    if (chk.checkMX && hasM_ && initialized_) {
      tmp = Utils::errorEquality(*X_, *MX_, MOp_);
      os << " >> Error in MX == M*X    : " << tmp << endl;
    }
    if (chk.checkKX && initialized_) {
      tmp = Utils::errorEquality(*X_, *KX_, Op_);
      os << " >> Error in KX == K*X    : " << tmp << endl;
    }

    // P and friends
    if (chk.checkP && hasP_ && initialized_) {
      if (fullOrtho_) {
        tmp = orthman_->orthonormError(*P_);
        os << " >> Error in P^H M P == I : " << tmp << endl;
        tmp = orthman_->orthogError(*P_,*X_);
        os << " >> Error in P^H M X == 0 : " << tmp << endl;
      }
      for (Array_size_type i=0; i<auxVecs_.size(); i++) {
        tmp = orthman_->orthogError(*P_,*auxVecs_[i]);
        os << " >> Error in P^H M Q[" << i << "] == 0 : " << tmp << endl;
      }
    }
    if (chk.checkMP && hasM_ && hasP_ && initialized_) {
      tmp = Utils::errorEquality(*P_, *MP_, MOp_);
      os << " >> Error in MP == M*P    : " << tmp << endl;
    }
    if (chk.checkKP && hasP_ && initialized_) {
      tmp = Utils::errorEquality(*P_, *KP_, Op_);
      os << " >> Error in KP == K*P    : " << tmp << endl;
    }

    // H and friends
    if (chk.checkH && initialized_) {
      if (fullOrtho_) {
        tmp = orthman_->orthonormError(*H_);
        os << " >> Error in H^H M H == I : " << tmp << endl;
        tmp = orthman_->orthogError(*H_,*X_);
        os << " >> Error in H^H M X == 0 : " << tmp << endl;
        if (hasP_) {
          tmp = orthman_->orthogError(*H_,*P_);
          os << " >> Error in H^H M P == 0 : " << tmp << endl;
        }
      }
      for (Array_size_type i=0; i<auxVecs_.size(); i++) {
        tmp = orthman_->orthogError(*H_,*auxVecs_[i]);
        os << " >> Error in H^H M Q[" << i << "] == 0 : " << tmp << endl;
      }
    }
    if (chk.checkMH && hasM_ && initialized_) {
      tmp = Utils::errorEquality(*H_, *MH_, MOp_);
      os << " >> Error in MH == M*H    : " << tmp << endl;
    }

    // R: this is not M-orthogonality, but standard euclidean orthogonality
    if (chk.checkR && initialized_) {
      Teuchos::SerialDenseMatrix<int,ScalarType> xTx(blockSize_,blockSize_);
      MVT::MvTransMv(ONE,*X_,*R_,xTx);
      tmp = xTx.normFrobenius();
      MVT::MvTransMv(ONE,*R_,*R_,xTx);
      double normR = xTx.normFrobenius();
      os << " >> RelError in X^H R == 0: " << tmp/normR << endl;
    }

    // Q
    if (chk.checkQ) {
      for (Array_size_type i=0; i<auxVecs_.size(); i++) {
        tmp = orthman_->orthonormError(*auxVecs_[i]);
        os << " >> Error in Q[" << i << "]^H M Q[" << i << "] == I : " << tmp << endl;
        for (Array_size_type j=i+1; j<auxVecs_.size(); j++) {
          tmp = orthman_->orthogError(*auxVecs_[i],*auxVecs_[j]);
          os << " >> Error in Q[" << i << "]^H M Q[" << j << "] == 0 : " << tmp << endl;
        }
      }
    }

    os << endl;

    return os.str();
  }


  //////////////////////////////////////////////////////////////////////////////////////////////////
  // Print the current status of the solver
  template <class ScalarType, class MV, class OP>
  void 
  LOBPCG<ScalarType,MV,OP>::currentStatus(std::ostream &os) 
  {
    using std::endl;

    os.setf(std::ios::scientific, std::ios::floatfield);  
    os.precision(6);
    os <<endl;
    os <<"================================================================================" << endl;
    os << endl;
    os <<"                              LOBPCG Solver Status" << endl;
    os << endl;
    os <<"The solver is "<<(initialized_ ? "initialized." : "not initialized.") << endl;
    os <<"The number of iterations performed is " << iter_       << endl;
    os <<"The current block size is             " << blockSize_  << endl;
    os <<"The number of auxiliary vectors is    " << numAuxVecs_ << endl;
    os <<"The number of operations Op*x   is " << count_ApplyOp_   << endl;
    os <<"The number of operations M*x    is " << count_ApplyM_    << endl;
    os <<"The number of operations Prec*x is " << count_ApplyPrec_ << endl;

    os.setf(std::ios_base::right, std::ios_base::adjustfield);

    if (initialized_) {
      os << endl;
      os <<"CURRENT EIGENVALUE ESTIMATES             "<<endl;
      os << std::setw(20) << "Eigenvalue" 
         << std::setw(20) << "Residual(M)"
         << std::setw(20) << "Residual(2)"
         << endl;
      os <<"--------------------------------------------------------------------------------"<<endl;
      for (int i=0; i<blockSize_; i++) {
        os << std::setw(20) << theta_[i];
        if (Rnorms_current_) os << std::setw(20) << Rnorms_[i];
        else os << std::setw(20) << "not current";
        if (R2norms_current_) os << std::setw(20) << R2norms_[i];
        else os << std::setw(20) << "not current";
        os << endl;
      }
    }
    os <<"================================================================================" << endl;
    os << endl;
  }

  //////////////////////////////////////////////////////////////////////////////////////////////////
  // are we initialized or not?
  template <class ScalarType, class MV, class OP>
  bool LOBPCG<ScalarType,MV,OP>::isInitialized() const { 
    return initialized_; 
  }


  //////////////////////////////////////////////////////////////////////////////////////////////////
  // is P valid or not?
  template <class ScalarType, class MV, class OP>
  bool LOBPCG<ScalarType,MV,OP>::hasP() {
    return hasP_;
  }
  
  //////////////////////////////////////////////////////////////////////////////////////////////////
  // is full orthogonalization enabled or not?
  template <class ScalarType, class MV, class OP>
  bool LOBPCG<ScalarType,MV,OP>::getFullOrtho() const { 
    return(fullOrtho_); 
  }

  
  //////////////////////////////////////////////////////////////////////////////////////////////////
  // return the current auxilliary vectors
  template <class ScalarType, class MV, class OP>
  Teuchos::Array<Teuchos::RCP<const MV> > LOBPCG<ScalarType,MV,OP>::getAuxVecs() const {
    return auxVecs_;
  }

  //////////////////////////////////////////////////////////////////////////////////////////////////
  // return the current block size
  template <class ScalarType, class MV, class OP>
  int LOBPCG<ScalarType,MV,OP>::getBlockSize() const {
    return(blockSize_); 
  }

  //////////////////////////////////////////////////////////////////////////////////////////////////
  // return the current eigenproblem
  template <class ScalarType, class MV, class OP>
  const Eigenproblem<ScalarType,MV,OP>& LOBPCG<ScalarType,MV,OP>::getProblem() const { 
    return(*problem_); 
  }

  
  //////////////////////////////////////////////////////////////////////////////////////////////////
  // return the max subspace dimension
  template <class ScalarType, class MV, class OP>
  int LOBPCG<ScalarType,MV,OP>::getMaxSubspaceDim() const {
    return 3*blockSize_;
  }

  //////////////////////////////////////////////////////////////////////////////////////////////////
  // return the current subspace dimension
  template <class ScalarType, class MV, class OP>
  int LOBPCG<ScalarType,MV,OP>::getCurSubspaceDim() const {
    if (!initialized_) return 0;
    return nevLocal_;
  }

  
  //////////////////////////////////////////////////////////////////////////////////////////////////
  // return the current ritz residual norms
  template <class ScalarType, class MV, class OP>
  std::vector<typename Teuchos::ScalarTraits<ScalarType>::magnitudeType> 
  LOBPCG<ScalarType,MV,OP>::getRitzRes2Norms() 
  {
    return this->getRes2Norms();
  }

  
  //////////////////////////////////////////////////////////////////////////////////////////////////
  // return the current compression indices
  template <class ScalarType, class MV, class OP>
  std::vector<int> LOBPCG<ScalarType,MV,OP>::getRitzIndex() {
    std::vector<int> ret(nevLocal_,0);
    return ret;
  }

  
  //////////////////////////////////////////////////////////////////////////////////////////////////
  // return the current ritz values
  template <class ScalarType, class MV, class OP>
  std::vector<Value<ScalarType> > LOBPCG<ScalarType,MV,OP>::getRitzValues() { 
    std::vector<Value<ScalarType> > ret(nevLocal_);
    for (int i=0; i<nevLocal_; i++) {
      ret[i].realpart = theta_[i];
      ret[i].imagpart = ZERO;
    }
    return ret;
  }

  //////////////////////////////////////////////////////////////////////////////////////////////////
  // Set a new StatusTest for the solver.
  template <class ScalarType, class MV, class OP>
  void LOBPCG<ScalarType,MV,OP>::setStatusTest(Teuchos::RCP<StatusTest<ScalarType,MV,OP> > test) {
    TEUCHOS_TEST_FOR_EXCEPTION(test == Teuchos::null,std::invalid_argument,
        "Anasazi::LOBPCG::setStatusTest() was passed a null StatusTest.");
    tester_ = test;
  }

  //////////////////////////////////////////////////////////////////////////////////////////////////
  // Get the current StatusTest used by the solver.
  template <class ScalarType, class MV, class OP>
  Teuchos::RCP<StatusTest<ScalarType,MV,OP> > LOBPCG<ScalarType,MV,OP>::getStatusTest() const {
    return tester_;
  }
  
  //////////////////////////////////////////////////////////////////////////////////////////////////
  // return the current ritz vectors
  template <class ScalarType, class MV, class OP>
  Teuchos::RCP<const MV> LOBPCG<ScalarType,MV,OP>::getRitzVectors() {
    return X_;
  }

  
  //////////////////////////////////////////////////////////////////////////////////////////////////
  // reset the iteration counter
  template <class ScalarType, class MV, class OP>
  void LOBPCG<ScalarType,MV,OP>::resetNumIters() {
    iter_=0; 
  }

  
  //////////////////////////////////////////////////////////////////////////////////////////////////
  // return the number of iterations
  template <class ScalarType, class MV, class OP>
  int LOBPCG<ScalarType,MV,OP>::getNumIters() const { 
    return(iter_); 
  }

  
  //////////////////////////////////////////////////////////////////////////////////////////////////
  // return the state
  template <class ScalarType, class MV, class OP>
  LOBPCGState<ScalarType,MV> LOBPCG<ScalarType,MV,OP>::getState() const {
    LOBPCGState<ScalarType,MV> state;
    state.V = V_;
    state.KV = KV_;
    state.X = X_;
    state.KX = KX_;
    state.P = P_;
    state.KP = KP_;
    state.H = H_;
    state.KH = KH_;
    state.R = R_;
    state.T = Teuchos::rcp(new std::vector<MagnitudeType>(theta_));
    if (hasM_) {
      state.MV = MV_;
      state.MX = MX_;
      state.MP = MP_;
      state.MH = MH_;
    }
    else {
      state.MX = Teuchos::null;
      state.MP = Teuchos::null;
      state.MH = Teuchos::null;
    }
    return state;
  }

} // end Anasazi namespace

#endif // ANASAZI_LOBPCG_HPP