|
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 |
SUBROUTINE CBBCSD( JOBU1, JOBU2, JOBV1T, JOBV2T, TRANS, M, P, Q,
$ THETA, PHI, U1, LDU1, U2, LDU2, V1T, LDV1T, $ V2T, LDV2T, B11D, B11E, B12D, B12E, B21D, B21E, $ B22D, B22E, RWORK, LRWORK, INFO ) IMPLICIT NONE * * -- LAPACK routine (version 3.3.0) -- * * -- Contributed by Brian Sutton of the Randolph-Macon College -- * -- November 2010 * * -- LAPACK is a software package provided by Univ. of Tennessee, -- * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- * * .. Scalar Arguments .. CHARACTER JOBU1, JOBU2, JOBV1T, JOBV2T, TRANS INTEGER INFO, LDU1, LDU2, LDV1T, LDV2T, LRWORK, M, P, Q * .. * .. Array Arguments .. REAL B11D( * ), B11E( * ), B12D( * ), B12E( * ), $ B21D( * ), B21E( * ), B22D( * ), B22E( * ), $ PHI( * ), THETA( * ), RWORK( * ) COMPLEX U1( LDU1, * ), U2( LDU2, * ), V1T( LDV1T, * ), $ V2T( LDV2T, * ) * .. * * Purpose * ======= * * CBBCSD computes the CS decomposition of a unitary matrix in * bidiagonal-block form, * * * [ B11 | B12 0 0 ] * [ 0 | 0 -I 0 ] * X = [----------------] * [ B21 | B22 0 0 ] * [ 0 | 0 0 I ] * * [ C | -S 0 0 ] * [ U1 | ] [ 0 | 0 -I 0 ] [ V1 | ]**H * = [---------] [---------------] [---------] . * [ | U2 ] [ S | C 0 0 ] [ | V2 ] * [ 0 | 0 0 I ] * * X is M-by-M, its top-left block is P-by-Q, and Q must be no larger * than P, M-P, or M-Q. (If Q is not the smallest index, then X must be * transposed and/or permuted. This can be done in constant time using * the TRANS and SIGNS options. See CUNCSD for details.) * * The bidiagonal matrices B11, B12, B21, and B22 are represented * implicitly by angles THETA(1:Q) and PHI(1:Q-1). * * The unitary matrices U1, U2, V1T, and V2T are input/output. * The input matrices are pre- or post-multiplied by the appropriate * singular vector matrices. * * Arguments * ========= * * JOBU1 (input) CHARACTER * = 'Y': U1 is updated; * otherwise: U1 is not updated. * * JOBU2 (input) CHARACTER * = 'Y': U2 is updated; * otherwise: U2 is not updated. * * JOBV1T (input) CHARACTER * = 'Y': V1T is updated; * otherwise: V1T is not updated. * * JOBV2T (input) CHARACTER * = 'Y': V2T is updated; * otherwise: V2T is not updated. * * TRANS (input) CHARACTER * = 'T': X, U1, U2, V1T, and V2T are stored in row-major * order; * otherwise: X, U1, U2, V1T, and V2T are stored in column- * major order. * * M (input) INTEGER * The number of rows and columns in X, the unitary matrix in * bidiagonal-block form. * * P (input) INTEGER * The number of rows in the top-left block of X. 0 <= P <= M. * * Q (input) INTEGER * The number of columns in the top-left block of X. * 0 <= Q <= MIN(P,M-P,M-Q). * * THETA (input/output) REAL array, dimension (Q) * On entry, the angles THETA(1),...,THETA(Q) that, along with * PHI(1), ...,PHI(Q-1), define the matrix in bidiagonal-block * form. On exit, the angles whose cosines and sines define the * diagonal blocks in the CS decomposition. * * PHI (input/workspace) REAL array, dimension (Q-1) * The angles PHI(1),...,PHI(Q-1) that, along with THETA(1),..., * THETA(Q), define the matrix in bidiagonal-block form. * * U1 (input/output) COMPLEX array, dimension (LDU1,P) * On entry, an LDU1-by-P matrix. On exit, U1 is postmultiplied * by the left singular vector matrix common to [ B11 ; 0 ] and * [ B12 0 0 ; 0 -I 0 0 ]. * * LDU1 (input) INTEGER * The leading dimension of the array U1. * * U2 (input/output) COMPLEX array, dimension (LDU2,M-P) * On entry, an LDU2-by-(M-P) matrix. On exit, U2 is * postmultiplied by the left singular vector matrix common to * [ B21 ; 0 ] and [ B22 0 0 ; 0 0 I ]. * * LDU2 (input) INTEGER * The leading dimension of the array U2. * * V1T (input/output) COMPLEX array, dimension (LDV1T,Q) * On entry, a LDV1T-by-Q matrix. On exit, V1T is premultiplied * by the conjugate transpose of the right singular vector * matrix common to [ B11 ; 0 ] and [ B21 ; 0 ]. * * LDV1T (input) INTEGER * The leading dimension of the array V1T. * * V2T (input/output) COMPLEX array, dimenison (LDV2T,M-Q) * On entry, a LDV2T-by-(M-Q) matrix. On exit, V2T is * premultiplied by the conjugate transpose of the right * singular vector matrix common to [ B12 0 0 ; 0 -I 0 ] and * [ B22 0 0 ; 0 0 I ]. * * LDV2T (input) INTEGER * The leading dimension of the array V2T. * * B11D (output) REAL array, dimension (Q) * When CBBCSD converges, B11D contains the cosines of THETA(1), * ..., THETA(Q). If CBBCSD fails to converge, then B11D * contains the diagonal of the partially reduced top-left * block. * * B11E (output) REAL array, dimension (Q-1) * When CBBCSD converges, B11E contains zeros. If CBBCSD fails * to converge, then B11E contains the superdiagonal of the * partially reduced top-left block. * * B12D (output) REAL array, dimension (Q) * When CBBCSD converges, B12D contains the negative sines of * THETA(1), ..., THETA(Q). If CBBCSD fails to converge, then * B12D contains the diagonal of the partially reduced top-right * block. * * B12E (output) REAL array, dimension (Q-1) * When CBBCSD converges, B12E contains zeros. If CBBCSD fails * to converge, then B12E contains the subdiagonal of the * partially reduced top-right block. * * RWORK (workspace) REAL array, dimension (MAX(1,LWORK)) * On exit, if INFO = 0, WORK(1) returns the optimal LWORK. * * LRWORK (input) INTEGER * The dimension of the array RWORK. LRWORK >= MAX(1,8*Q). * * If LRWORK = -1, then a workspace query is assumed; the * routine only calculates the optimal size of the RWORK array, * returns this value as the first entry of the work array, and * no error message related to LRWORK is issued by XERBLA. * * INFO (output) INTEGER * = 0: successful exit. * < 0: if INFO = -i, the i-th argument had an illegal value. * > 0: if CBBCSD did not converge, INFO specifies the number * of nonzero entries in PHI, and B11D, B11E, etc., * contain the partially reduced matrix. * * Reference * ========= * * [1] Brian D. Sutton. Computing the complete CS decomposition. Numer. * Algorithms, 50(1):33-65, 2009. * * Internal Parameters * =================== * * TOLMUL REAL, default = MAX(10,MIN(100,EPS**(-1/8))) * TOLMUL controls the convergence criterion of the QR loop. * Angles THETA(i), PHI(i) are rounded to 0 or PI/2 when they * are within TOLMUL*EPS of either bound. * * =================================================================== * * .. Parameters .. INTEGER MAXITR PARAMETER ( MAXITR = 6 ) REAL HUNDRED, MEIGHTH, ONE, PIOVER2, TEN, ZERO PARAMETER ( HUNDRED = 100.0E0, MEIGHTH = -0.125E0, $ ONE = 1.0E0, PIOVER2 = 1.57079632679489662E0, $ TEN = 10.0E0, ZERO = 0.0E0 ) COMPLEX NEGONECOMPLEX PARAMETER ( NEGONECOMPLEX = (-1.0E0,0.0E0) ) * .. * .. Local Scalars .. LOGICAL COLMAJOR, LQUERY, RESTART11, RESTART12, $ RESTART21, RESTART22, WANTU1, WANTU2, WANTV1T, $ WANTV2T INTEGER I, IMIN, IMAX, ITER, IU1CS, IU1SN, IU2CS, $ IU2SN, IV1TCS, IV1TSN, IV2TCS, IV2TSN, J, $ LRWORKMIN, LRWORKOPT, MAXIT, MINI REAL B11BULGE, B12BULGE, B21BULGE, B22BULGE, DUMMY, $ EPS, MU, NU, R, SIGMA11, SIGMA21, $ TEMP, THETAMAX, THETAMIN, THRESH, TOL, TOLMUL, $ UNFL, X1, X2, Y1, Y2 * * .. External Subroutines .. EXTERNAL CLASR, CSCAL, CSWAP, SLARTGP, SLARTGS, SLAS2, $ XERBLA * .. * .. External Functions .. REAL SLAMCH LOGICAL LSAME EXTERNAL LSAME, SLAMCH * .. * .. Intrinsic Functions .. INTRINSIC ABS, ATAN2, COS, MAX, MIN, SIN, SQRT * .. * .. Executable Statements .. * * Test input arguments * INFO = 0 LQUERY = LRWORK .EQ. -1 WANTU1 = LSAME( JOBU1, 'Y' ) WANTU2 = LSAME( JOBU2, 'Y' ) WANTV1T = LSAME( JOBV1T, 'Y' ) WANTV2T = LSAME( JOBV2T, 'Y' ) COLMAJOR = .NOT. LSAME( TRANS, 'T' ) * IF( M .LT. 0 ) THEN INFO = -6 ELSE IF( P .LT. 0 .OR. P .GT. M ) THEN INFO = -7 ELSE IF( Q .LT. 0 .OR. Q .GT. M ) THEN INFO = -8 ELSE IF( Q .GT. P .OR. Q .GT. M-P .OR. Q .GT. M-Q ) THEN INFO = -8 ELSE IF( WANTU1 .AND. LDU1 .LT. P ) THEN INFO = -12 ELSE IF( WANTU2 .AND. LDU2 .LT. M-P ) THEN INFO = -14 ELSE IF( WANTV1T .AND. LDV1T .LT. Q ) THEN INFO = -16 ELSE IF( WANTV2T .AND. LDV2T .LT. M-Q ) THEN INFO = -18 END IF * * Quick return if Q = 0 * IF( INFO .EQ. 0 .AND. Q .EQ. 0 ) THEN LRWORKMIN = 1 RWORK(1) = LRWORKMIN RETURN END IF * * Compute workspace * IF( INFO .EQ. 0 ) THEN IU1CS = 1 IU1SN = IU1CS + Q IU2CS = IU1SN + Q IU2SN = IU2CS + Q IV1TCS = IU2SN + Q IV1TSN = IV1TCS + Q IV2TCS = IV1TSN + Q IV2TSN = IV2TCS + Q LRWORKOPT = IV2TSN + Q - 1 LRWORKMIN = LRWORKOPT RWORK(1) = LRWORKOPT IF( LRWORK .LT. LRWORKMIN .AND. .NOT. LQUERY ) THEN INFO = -28 END IF END IF * IF( INFO .NE. 0 ) THEN CALL XERBLA( 'CBBCSD', -INFO ) RETURN ELSE IF( LQUERY ) THEN RETURN END IF * * Get machine constants * EPS = SLAMCH( 'Epsilon' ) UNFL = SLAMCH( 'Safe minimum' ) TOLMUL = MAX( TEN, MIN( HUNDRED, EPS**MEIGHTH ) ) TOL = TOLMUL*EPS THRESH = MAX( TOL, MAXITR*Q*Q*UNFL ) * * Test for negligible sines or cosines * DO I = 1, Q IF( THETA(I) .LT. THRESH ) THEN THETA(I) = ZERO ELSE IF( THETA(I) .GT. PIOVER2-THRESH ) THEN THETA(I) = PIOVER2 END IF END DO DO I = 1, Q-1 IF( PHI(I) .LT. THRESH ) THEN PHI(I) = ZERO ELSE IF( PHI(I) .GT. PIOVER2-THRESH ) THEN PHI(I) = PIOVER2 END IF END DO * * Initial deflation * IMAX = Q DO WHILE( ( IMAX .GT. 1 ) .AND. ( PHI(IMAX-1) .EQ. ZERO ) ) IMAX = IMAX - 1 END DO IMIN = IMAX - 1 IF ( IMIN .GT. 1 ) THEN DO WHILE( PHI(IMIN-1) .NE. ZERO ) IMIN = IMIN - 1 IF ( IMIN .LE. 1 ) EXIT END DO END IF * * Initialize iteration counter * MAXIT = MAXITR*Q*Q ITER = 0 * * Begin main iteration loop * DO WHILE( IMAX .GT. 1 ) * * Compute the matrix entries * B11D(IMIN) = COS( THETA(IMIN) ) B21D(IMIN) = -SIN( THETA(IMIN) ) DO I = IMIN, IMAX - 1 B11E(I) = -SIN( THETA(I) ) * SIN( PHI(I) ) B11D(I+1) = COS( THETA(I+1) ) * COS( PHI(I) ) B12D(I) = SIN( THETA(I) ) * COS( PHI(I) ) B12E(I) = COS( THETA(I+1) ) * SIN( PHI(I) ) B21E(I) = -COS( THETA(I) ) * SIN( PHI(I) ) B21D(I+1) = -SIN( THETA(I+1) ) * COS( PHI(I) ) B22D(I) = COS( THETA(I) ) * COS( PHI(I) ) B22E(I) = -SIN( THETA(I+1) ) * SIN( PHI(I) ) END DO B12D(IMAX) = SIN( THETA(IMAX) ) B22D(IMAX) = COS( THETA(IMAX) ) * * Abort if not converging; otherwise, increment ITER * IF( ITER .GT. MAXIT ) THEN INFO = 0 DO I = 1, Q IF( PHI(I) .NE. ZERO ) $ INFO = INFO + 1 END DO RETURN END IF * ITER = ITER + IMAX - IMIN * * Compute shifts * THETAMAX = THETA(IMIN) THETAMIN = THETA(IMIN) DO I = IMIN+1, IMAX IF( THETA(I) > THETAMAX ) $ THETAMAX = THETA(I) IF( THETA(I) < THETAMIN ) $ THETAMIN = THETA(I) END DO * IF( THETAMAX .GT. PIOVER2 - THRESH ) THEN * * Zero on diagonals of B11 and B22; induce deflation with a * zero shift * MU = ZERO NU = ONE * ELSE IF( THETAMIN .LT. THRESH ) THEN * * Zero on diagonals of B12 and B22; induce deflation with a * zero shift * MU = ONE NU = ZERO * ELSE * * Compute shifts for B11 and B21 and use the lesser * CALL SLAS2( B11D(IMAX-1), B11E(IMAX-1), B11D(IMAX), SIGMA11, $ DUMMY ) CALL SLAS2( B21D(IMAX-1), B21E(IMAX-1), B21D(IMAX), SIGMA21, $ DUMMY ) * IF( SIGMA11 .LE. SIGMA21 ) THEN MU = SIGMA11 NU = SQRT( ONE - MU**2 ) IF( MU .LT. THRESH ) THEN MU = ZERO NU = ONE END IF ELSE NU = SIGMA21 MU = SQRT( 1.0 - NU**2 ) IF( NU .LT. THRESH ) THEN MU = ONE NU = ZERO END IF END IF END IF * * Rotate to produce bulges in B11 and B21 * IF( MU .LE. NU ) THEN CALL SLARTGS( B11D(IMIN), B11E(IMIN), MU, $ RWORK(IV1TCS+IMIN-1), RWORK(IV1TSN+IMIN-1) ) ELSE CALL SLARTGS( B21D(IMIN), B21E(IMIN), NU, $ RWORK(IV1TCS+IMIN-1), RWORK(IV1TSN+IMIN-1) ) END IF * TEMP = RWORK(IV1TCS+IMIN-1)*B11D(IMIN) + $ RWORK(IV1TSN+IMIN-1)*B11E(IMIN) B11E(IMIN) = RWORK(IV1TCS+IMIN-1)*B11E(IMIN) - $ RWORK(IV1TSN+IMIN-1)*B11D(IMIN) B11D(IMIN) = TEMP B11BULGE = RWORK(IV1TSN+IMIN-1)*B11D(IMIN+1) B11D(IMIN+1) = RWORK(IV1TCS+IMIN-1)*B11D(IMIN+1) TEMP = RWORK(IV1TCS+IMIN-1)*B21D(IMIN) + $ RWORK(IV1TSN+IMIN-1)*B21E(IMIN) B21E(IMIN) = RWORK(IV1TCS+IMIN-1)*B21E(IMIN) - $ RWORK(IV1TSN+IMIN-1)*B21D(IMIN) B21D(IMIN) = TEMP B21BULGE = RWORK(IV1TSN+IMIN-1)*B21D(IMIN+1) B21D(IMIN+1) = RWORK(IV1TCS+IMIN-1)*B21D(IMIN+1) * * Compute THETA(IMIN) * THETA( IMIN ) = ATAN2( SQRT( B21D(IMIN)**2+B21BULGE**2 ), $ SQRT( B11D(IMIN)**2+B11BULGE**2 ) ) * * Chase the bulges in B11(IMIN+1,IMIN) and B21(IMIN+1,IMIN) * IF( B11D(IMIN)**2+B11BULGE**2 .GT. THRESH**2 ) THEN CALL SLARTGP( B11BULGE, B11D(IMIN), RWORK(IU1SN+IMIN-1), $ RWORK(IU1CS+IMIN-1), R ) ELSE IF( MU .LE. NU ) THEN CALL SLARTGS( B11E( IMIN ), B11D( IMIN + 1 ), MU, $ RWORK(IU1CS+IMIN-1), RWORK(IU1SN+IMIN-1) ) ELSE CALL SLARTGS( B12D( IMIN ), B12E( IMIN ), NU, $ RWORK(IU1CS+IMIN-1), RWORK(IU1SN+IMIN-1) ) END IF IF( B21D(IMIN)**2+B21BULGE**2 .GT. THRESH**2 ) THEN CALL SLARTGP( B21BULGE, B21D(IMIN), RWORK(IU2SN+IMIN-1), $ RWORK(IU2CS+IMIN-1), R ) ELSE IF( NU .LT. MU ) THEN CALL SLARTGS( B21E( IMIN ), B21D( IMIN + 1 ), NU, $ RWORK(IU2CS+IMIN-1), RWORK(IU2SN+IMIN-1) ) ELSE CALL SLARTGS( B22D(IMIN), B22E(IMIN), MU, $ RWORK(IU2CS+IMIN-1), RWORK(IU2SN+IMIN-1) ) END IF RWORK(IU2CS+IMIN-1) = -RWORK(IU2CS+IMIN-1) RWORK(IU2SN+IMIN-1) = -RWORK(IU2SN+IMIN-1) * TEMP = RWORK(IU1CS+IMIN-1)*B11E(IMIN) + $ RWORK(IU1SN+IMIN-1)*B11D(IMIN+1) B11D(IMIN+1) = RWORK(IU1CS+IMIN-1)*B11D(IMIN+1) - $ RWORK(IU1SN+IMIN-1)*B11E(IMIN) B11E(IMIN) = TEMP IF( IMAX .GT. IMIN+1 ) THEN B11BULGE = RWORK(IU1SN+IMIN-1)*B11E(IMIN+1) B11E(IMIN+1) = RWORK(IU1CS+IMIN-1)*B11E(IMIN+1) END IF TEMP = RWORK(IU1CS+IMIN-1)*B12D(IMIN) + $ RWORK(IU1SN+IMIN-1)*B12E(IMIN) B12E(IMIN) = RWORK(IU1CS+IMIN-1)*B12E(IMIN) - $ RWORK(IU1SN+IMIN-1)*B12D(IMIN) B12D(IMIN) = TEMP B12BULGE = RWORK(IU1SN+IMIN-1)*B12D(IMIN+1) B12D(IMIN+1) = RWORK(IU1CS+IMIN-1)*B12D(IMIN+1) TEMP = RWORK(IU2CS+IMIN-1)*B21E(IMIN) + $ RWORK(IU2SN+IMIN-1)*B21D(IMIN+1) B21D(IMIN+1) = RWORK(IU2CS+IMIN-1)*B21D(IMIN+1) - $ RWORK(IU2SN+IMIN-1)*B21E(IMIN) B21E(IMIN) = TEMP IF( IMAX .GT. IMIN+1 ) THEN B21BULGE = RWORK(IU2SN+IMIN-1)*B21E(IMIN+1) B21E(IMIN+1) = RWORK(IU2CS+IMIN-1)*B21E(IMIN+1) END IF TEMP = RWORK(IU2CS+IMIN-1)*B22D(IMIN) + $ RWORK(IU2SN+IMIN-1)*B22E(IMIN) B22E(IMIN) = RWORK(IU2CS+IMIN-1)*B22E(IMIN) - $ RWORK(IU2SN+IMIN-1)*B22D(IMIN) B22D(IMIN) = TEMP B22BULGE = RWORK(IU2SN+IMIN-1)*B22D(IMIN+1) B22D(IMIN+1) = RWORK(IU2CS+IMIN-1)*B22D(IMIN+1) * * Inner loop: chase bulges from B11(IMIN,IMIN+2), * B12(IMIN,IMIN+1), B21(IMIN,IMIN+2), and B22(IMIN,IMIN+1) to * bottom-right * DO I = IMIN+1, IMAX-1 * * Compute PHI(I-1) * X1 = SIN(THETA(I-1))*B11E(I-1) + COS(THETA(I-1))*B21E(I-1) X2 = SIN(THETA(I-1))*B11BULGE + COS(THETA(I-1))*B21BULGE Y1 = SIN(THETA(I-1))*B12D(I-1) + COS(THETA(I-1))*B22D(I-1) Y2 = SIN(THETA(I-1))*B12BULGE + COS(THETA(I-1))*B22BULGE * PHI(I-1) = ATAN2( SQRT(X1**2+X2**2), SQRT(Y1**2+Y2**2) ) * * Determine if there are bulges to chase or if a new direct * summand has been reached * RESTART11 = B11E(I-1)**2 + B11BULGE**2 .LE. THRESH**2 RESTART21 = B21E(I-1)**2 + B21BULGE**2 .LE. THRESH**2 RESTART12 = B12D(I-1)**2 + B12BULGE**2 .LE. THRESH**2 RESTART22 = B22D(I-1)**2 + B22BULGE**2 .LE. THRESH**2 * * If possible, chase bulges from B11(I-1,I+1), B12(I-1,I), * B21(I-1,I+1), and B22(I-1,I). If necessary, restart bulge- * chasing by applying the original shift again. * IF( .NOT. RESTART11 .AND. .NOT. RESTART21 ) THEN CALL SLARTGP( X2, X1, RWORK(IV1TSN+I-1), $ RWORK(IV1TCS+I-1), R ) ELSE IF( .NOT. RESTART11 .AND. RESTART21 ) THEN CALL SLARTGP( B11BULGE, B11E(I-1), RWORK(IV1TSN+I-1), $ RWORK(IV1TCS+I-1), R ) ELSE IF( RESTART11 .AND. .NOT. RESTART21 ) THEN CALL SLARTGP( B21BULGE, B21E(I-1), RWORK(IV1TSN+I-1), $ RWORK(IV1TCS+I-1), R ) ELSE IF( MU .LE. NU ) THEN CALL SLARTGS( B11D(I), B11E(I), MU, RWORK(IV1TCS+I-1), $ RWORK(IV1TSN+I-1) ) ELSE CALL SLARTGS( B21D(I), B21E(I), NU, RWORK(IV1TCS+I-1), $ RWORK(IV1TSN+I-1) ) END IF RWORK(IV1TCS+I-1) = -RWORK(IV1TCS+I-1) RWORK(IV1TSN+I-1) = -RWORK(IV1TSN+I-1) IF( .NOT. RESTART12 .AND. .NOT. RESTART22 ) THEN CALL SLARTGP( Y2, Y1, RWORK(IV2TSN+I-1-1), $ RWORK(IV2TCS+I-1-1), R ) ELSE IF( .NOT. RESTART12 .AND. RESTART22 ) THEN CALL SLARTGP( B12BULGE, B12D(I-1), RWORK(IV2TSN+I-1-1), $ RWORK(IV2TCS+I-1-1), R ) ELSE IF( RESTART12 .AND. .NOT. RESTART22 ) THEN CALL SLARTGP( B22BULGE, B22D(I-1), RWORK(IV2TSN+I-1-1), $ RWORK(IV2TCS+I-1-1), R ) ELSE IF( NU .LT. MU ) THEN CALL SLARTGS( B12E(I-1), B12D(I), NU, $ RWORK(IV2TCS+I-1-1), RWORK(IV2TSN+I-1-1) ) ELSE CALL SLARTGS( B22E(I-1), B22D(I), MU, $ RWORK(IV2TCS+I-1-1), RWORK(IV2TSN+I-1-1) ) END IF * TEMP = RWORK(IV1TCS+I-1)*B11D(I) + RWORK(IV1TSN+I-1)*B11E(I) B11E(I) = RWORK(IV1TCS+I-1)*B11E(I) - $ RWORK(IV1TSN+I-1)*B11D(I) B11D(I) = TEMP B11BULGE = RWORK(IV1TSN+I-1)*B11D(I+1) B11D(I+1) = RWORK(IV1TCS+I-1)*B11D(I+1) TEMP = RWORK(IV1TCS+I-1)*B21D(I) + RWORK(IV1TSN+I-1)*B21E(I) B21E(I) = RWORK(IV1TCS+I-1)*B21E(I) - $ RWORK(IV1TSN+I-1)*B21D(I) B21D(I) = TEMP B21BULGE = RWORK(IV1TSN+I-1)*B21D(I+1) B21D(I+1) = RWORK(IV1TCS+I-1)*B21D(I+1) TEMP = RWORK(IV2TCS+I-1-1)*B12E(I-1) + $ RWORK(IV2TSN+I-1-1)*B12D(I) B12D(I) = RWORK(IV2TCS+I-1-1)*B12D(I) - $ RWORK(IV2TSN+I-1-1)*B12E(I-1) B12E(I-1) = TEMP B12BULGE = RWORK(IV2TSN+I-1-1)*B12E(I) B12E(I) = RWORK(IV2TCS+I-1-1)*B12E(I) TEMP = RWORK(IV2TCS+I-1-1)*B22E(I-1) + $ RWORK(IV2TSN+I-1-1)*B22D(I) B22D(I) = RWORK(IV2TCS+I-1-1)*B22D(I) - $ RWORK(IV2TSN+I-1-1)*B22E(I-1) B22E(I-1) = TEMP B22BULGE = RWORK(IV2TSN+I-1-1)*B22E(I) B22E(I) = RWORK(IV2TCS+I-1-1)*B22E(I) * * Compute THETA(I) * X1 = COS(PHI(I-1))*B11D(I) + SIN(PHI(I-1))*B12E(I-1) X2 = COS(PHI(I-1))*B11BULGE + SIN(PHI(I-1))*B12BULGE Y1 = COS(PHI(I-1))*B21D(I) + SIN(PHI(I-1))*B22E(I-1) Y2 = COS(PHI(I-1))*B21BULGE + SIN(PHI(I-1))*B22BULGE * THETA(I) = ATAN2( SQRT(Y1**2+Y2**2), SQRT(X1**2+X2**2) ) * * Determine if there are bulges to chase or if a new direct * summand has been reached * RESTART11 = B11D(I)**2 + B11BULGE**2 .LE. THRESH**2 RESTART12 = B12E(I-1)**2 + B12BULGE**2 .LE. THRESH**2 RESTART21 = B21D(I)**2 + B21BULGE**2 .LE. THRESH**2 RESTART22 = B22E(I-1)**2 + B22BULGE**2 .LE. THRESH**2 * * If possible, chase bulges from B11(I+1,I), B12(I+1,I-1), * B21(I+1,I), and B22(I+1,I-1). If necessary, restart bulge- * chasing by applying the original shift again. * IF( .NOT. RESTART11 .AND. .NOT. RESTART12 ) THEN CALL SLARTGP( X2, X1, RWORK(IU1SN+I-1), RWORK(IU1CS+I-1), $ R ) ELSE IF( .NOT. RESTART11 .AND. RESTART12 ) THEN CALL SLARTGP( B11BULGE, B11D(I), RWORK(IU1SN+I-1), $ RWORK(IU1CS+I-1), R ) ELSE IF( RESTART11 .AND. .NOT. RESTART12 ) THEN CALL SLARTGP( B12BULGE, B12E(I-1), RWORK(IU1SN+I-1), $ RWORK(IU1CS+I-1), R ) ELSE IF( MU .LE. NU ) THEN CALL SLARTGS( B11E(I), B11D(I+1), MU, RWORK(IU1CS+I-1), $ RWORK(IU1SN+I-1) ) ELSE CALL SLARTGS( B12D(I), B12E(I), NU, RWORK(IU1CS+I-1), $ RWORK(IU1SN+I-1) ) END IF IF( .NOT. RESTART21 .AND. .NOT. RESTART22 ) THEN CALL SLARTGP( Y2, Y1, RWORK(IU2SN+I-1), RWORK(IU2CS+I-1), $ R ) ELSE IF( .NOT. RESTART21 .AND. RESTART22 ) THEN CALL SLARTGP( B21BULGE, B21D(I), RWORK(IU2SN+I-1), $ RWORK(IU2CS+I-1), R ) ELSE IF( RESTART21 .AND. .NOT. RESTART22 ) THEN CALL SLARTGP( B22BULGE, B22E(I-1), RWORK(IU2SN+I-1), $ RWORK(IU2CS+I-1), R ) ELSE IF( NU .LT. MU ) THEN CALL SLARTGS( B21E(I), B21E(I+1), NU, RWORK(IU2CS+I-1), $ RWORK(IU2SN+I-1) ) ELSE CALL SLARTGS( B22D(I), B22E(I), MU, RWORK(IU2CS+I-1), $ RWORK(IU2SN+I-1) ) END IF RWORK(IU2CS+I-1) = -RWORK(IU2CS+I-1) RWORK(IU2SN+I-1) = -RWORK(IU2SN+I-1) * TEMP = RWORK(IU1CS+I-1)*B11E(I) + RWORK(IU1SN+I-1)*B11D(I+1) B11D(I+1) = RWORK(IU1CS+I-1)*B11D(I+1) - $ RWORK(IU1SN+I-1)*B11E(I) B11E(I) = TEMP IF( I .LT. IMAX - 1 ) THEN B11BULGE = RWORK(IU1SN+I-1)*B11E(I+1) B11E(I+1) = RWORK(IU1CS+I-1)*B11E(I+1) END IF TEMP = RWORK(IU2CS+I-1)*B21E(I) + RWORK(IU2SN+I-1)*B21D(I+1) B21D(I+1) = RWORK(IU2CS+I-1)*B21D(I+1) - $ RWORK(IU2SN+I-1)*B21E(I) B21E(I) = TEMP IF( I .LT. IMAX - 1 ) THEN B21BULGE = RWORK(IU2SN+I-1)*B21E(I+1) B21E(I+1) = RWORK(IU2CS+I-1)*B21E(I+1) END IF TEMP = RWORK(IU1CS+I-1)*B12D(I) + RWORK(IU1SN+I-1)*B12E(I) B12E(I) = RWORK(IU1CS+I-1)*B12E(I) - $ RWORK(IU1SN+I-1)*B12D(I) B12D(I) = TEMP B12BULGE = RWORK(IU1SN+I-1)*B12D(I+1) B12D(I+1) = RWORK(IU1CS+I-1)*B12D(I+1) TEMP = RWORK(IU2CS+I-1)*B22D(I) + RWORK(IU2SN+I-1)*B22E(I) B22E(I) = RWORK(IU2CS+I-1)*B22E(I) - $ RWORK(IU2SN+I-1)*B22D(I) B22D(I) = TEMP B22BULGE = RWORK(IU2SN+I-1)*B22D(I+1) B22D(I+1) = RWORK(IU2CS+I-1)*B22D(I+1) * END DO * * Compute PHI(IMAX-1) * X1 = SIN(THETA(IMAX-1))*B11E(IMAX-1) + $ COS(THETA(IMAX-1))*B21E(IMAX-1) Y1 = SIN(THETA(IMAX-1))*B12D(IMAX-1) + $ COS(THETA(IMAX-1))*B22D(IMAX-1) Y2 = SIN(THETA(IMAX-1))*B12BULGE + COS(THETA(IMAX-1))*B22BULGE * PHI(IMAX-1) = ATAN2( ABS(X1), SQRT(Y1**2+Y2**2) ) * * Chase bulges from B12(IMAX-1,IMAX) and B22(IMAX-1,IMAX) * RESTART12 = B12D(IMAX-1)**2 + B12BULGE**2 .LE. THRESH**2 RESTART22 = B22D(IMAX-1)**2 + B22BULGE**2 .LE. THRESH**2 * IF( .NOT. RESTART12 .AND. .NOT. RESTART22 ) THEN CALL SLARTGP( Y2, Y1, RWORK(IV2TSN+IMAX-1-1), $ RWORK(IV2TCS+IMAX-1-1), R ) ELSE IF( .NOT. RESTART12 .AND. RESTART22 ) THEN CALL SLARTGP( B12BULGE, B12D(IMAX-1), $ RWORK(IV2TSN+IMAX-1-1), $ RWORK(IV2TCS+IMAX-1-1), R ) ELSE IF( RESTART12 .AND. .NOT. RESTART22 ) THEN CALL SLARTGP( B22BULGE, B22D(IMAX-1), $ RWORK(IV2TSN+IMAX-1-1), $ RWORK(IV2TCS+IMAX-1-1), R ) ELSE IF( NU .LT. MU ) THEN CALL SLARTGS( B12E(IMAX-1), B12D(IMAX), NU, $ RWORK(IV2TCS+IMAX-1-1), $ RWORK(IV2TSN+IMAX-1-1) ) ELSE CALL SLARTGS( B22E(IMAX-1), B22D(IMAX), MU, $ RWORK(IV2TCS+IMAX-1-1), $ RWORK(IV2TSN+IMAX-1-1) ) END IF * TEMP = RWORK(IV2TCS+IMAX-1-1)*B12E(IMAX-1) + $ RWORK(IV2TSN+IMAX-1-1)*B12D(IMAX) B12D(IMAX) = RWORK(IV2TCS+IMAX-1-1)*B12D(IMAX) - $ RWORK(IV2TSN+IMAX-1-1)*B12E(IMAX-1) B12E(IMAX-1) = TEMP TEMP = RWORK(IV2TCS+IMAX-1-1)*B22E(IMAX-1) + $ RWORK(IV2TSN+IMAX-1-1)*B22D(IMAX) B22D(IMAX) = RWORK(IV2TCS+IMAX-1-1)*B22D(IMAX) - $ RWORK(IV2TSN+IMAX-1-1)*B22E(IMAX-1) B22E(IMAX-1) = TEMP * * Update singular vectors * IF( WANTU1 ) THEN IF( COLMAJOR ) THEN CALL CLASR( 'R', 'V', 'F', P, IMAX-IMIN+1, $ RWORK(IU1CS+IMIN-1), RWORK(IU1SN+IMIN-1), $ U1(1,IMIN), LDU1 ) ELSE CALL CLASR( 'L', 'V', 'F', IMAX-IMIN+1, P, $ RWORK(IU1CS+IMIN-1), RWORK(IU1SN+IMIN-1), $ U1(IMIN,1), LDU1 ) END IF END IF IF( WANTU2 ) THEN IF( COLMAJOR ) THEN CALL CLASR( 'R', 'V', 'F', M-P, IMAX-IMIN+1, $ RWORK(IU2CS+IMIN-1), RWORK(IU2SN+IMIN-1), $ U2(1,IMIN), LDU2 ) ELSE CALL CLASR( 'L', 'V', 'F', IMAX-IMIN+1, M-P, $ RWORK(IU2CS+IMIN-1), RWORK(IU2SN+IMIN-1), $ U2(IMIN,1), LDU2 ) END IF END IF IF( WANTV1T ) THEN IF( COLMAJOR ) THEN CALL CLASR( 'L', 'V', 'F', IMAX-IMIN+1, Q, $ RWORK(IV1TCS+IMIN-1), RWORK(IV1TSN+IMIN-1), $ V1T(IMIN,1), LDV1T ) ELSE CALL CLASR( 'R', 'V', 'F', Q, IMAX-IMIN+1, $ RWORK(IV1TCS+IMIN-1), RWORK(IV1TSN+IMIN-1), $ V1T(1,IMIN), LDV1T ) END IF END IF IF( WANTV2T ) THEN IF( COLMAJOR ) THEN CALL CLASR( 'L', 'V', 'F', IMAX-IMIN+1, M-Q, $ RWORK(IV2TCS+IMIN-1), RWORK(IV2TSN+IMIN-1), $ V2T(IMIN,1), LDV2T ) ELSE CALL CLASR( 'R', 'V', 'F', M-Q, IMAX-IMIN+1, $ RWORK(IV2TCS+IMIN-1), RWORK(IV2TSN+IMIN-1), $ V2T(1,IMIN), LDV2T ) END IF END IF * * Fix signs on B11(IMAX-1,IMAX) and B21(IMAX-1,IMAX) * IF( B11E(IMAX-1)+B21E(IMAX-1) .GT. 0 ) THEN B11D(IMAX) = -B11D(IMAX) B21D(IMAX) = -B21D(IMAX) IF( WANTV1T ) THEN IF( COLMAJOR ) THEN CALL CSCAL( Q, NEGONECOMPLEX, V1T(IMAX,1), LDV1T ) ELSE CALL CSCAL( Q, NEGONECOMPLEX, V1T(1,IMAX), 1 ) END IF END IF END IF * * Compute THETA(IMAX) * X1 = COS(PHI(IMAX-1))*B11D(IMAX) + $ SIN(PHI(IMAX-1))*B12E(IMAX-1) Y1 = COS(PHI(IMAX-1))*B21D(IMAX) + $ SIN(PHI(IMAX-1))*B22E(IMAX-1) * THETA(IMAX) = ATAN2( ABS(Y1), ABS(X1) ) * * Fix signs on B11(IMAX,IMAX), B12(IMAX,IMAX-1), B21(IMAX,IMAX), * and B22(IMAX,IMAX-1) * IF( B11D(IMAX)+B12E(IMAX-1) .LT. 0 ) THEN B12D(IMAX) = -B12D(IMAX) IF( WANTU1 ) THEN IF( COLMAJOR ) THEN CALL CSCAL( P, NEGONECOMPLEX, U1(1,IMAX), 1 ) ELSE CALL CSCAL( P, NEGONECOMPLEX, U1(IMAX,1), LDU1 ) END IF END IF END IF IF( B21D(IMAX)+B22E(IMAX-1) .GT. 0 ) THEN B22D(IMAX) = -B22D(IMAX) IF( WANTU2 ) THEN IF( COLMAJOR ) THEN CALL CSCAL( M-P, NEGONECOMPLEX, U2(1,IMAX), 1 ) ELSE CALL CSCAL( M-P, NEGONECOMPLEX, U2(IMAX,1), LDU2 ) END IF END IF END IF * * Fix signs on B12(IMAX,IMAX) and B22(IMAX,IMAX) * IF( B12D(IMAX)+B22D(IMAX) .LT. 0 ) THEN IF( WANTV2T ) THEN IF( COLMAJOR ) THEN CALL CSCAL( M-Q, NEGONECOMPLEX, V2T(IMAX,1), LDV2T ) ELSE CALL CSCAL( M-Q, NEGONECOMPLEX, V2T(1,IMAX), 1 ) END IF END IF END IF * * Test for negligible sines or cosines * DO I = IMIN, IMAX IF( THETA(I) .LT. THRESH ) THEN THETA(I) = ZERO ELSE IF( THETA(I) .GT. PIOVER2-THRESH ) THEN THETA(I) = PIOVER2 END IF END DO DO I = IMIN, IMAX-1 IF( PHI(I) .LT. THRESH ) THEN PHI(I) = ZERO ELSE IF( PHI(I) .GT. PIOVER2-THRESH ) THEN PHI(I) = PIOVER2 END IF END DO * * Deflate * IF (IMAX .GT. 1) THEN DO WHILE( PHI(IMAX-1) .EQ. ZERO ) IMAX = IMAX - 1 IF (IMAX .LE. 1) EXIT END DO END IF IF( IMIN .GT. IMAX - 1 ) $ IMIN = IMAX - 1 IF (IMIN .GT. 1) THEN DO WHILE (PHI(IMIN-1) .NE. ZERO) IMIN = IMIN - 1 IF (IMIN .LE. 1) EXIT END DO END IF * * Repeat main iteration loop * END DO * * Postprocessing: order THETA from least to greatest * DO I = 1, Q * MINI = I THETAMIN = THETA(I) DO J = I+1, Q IF( THETA(J) .LT. THETAMIN ) THEN MINI = J THETAMIN = THETA(J) END IF END DO * IF( MINI .NE. I ) THEN THETA(MINI) = THETA(I) THETA(I) = THETAMIN IF( COLMAJOR ) THEN IF( WANTU1 ) $ CALL CSWAP( P, U1(1,I), 1, U1(1,MINI), 1 ) IF( WANTU2 ) $ CALL CSWAP( M-P, U2(1,I), 1, U2(1,MINI), 1 ) IF( WANTV1T ) $ CALL CSWAP( Q, V1T(I,1), LDV1T, V1T(MINI,1), LDV1T ) IF( WANTV2T ) $ CALL CSWAP( M-Q, V2T(I,1), LDV2T, V2T(MINI,1), $ LDV2T ) ELSE IF( WANTU1 ) $ CALL CSWAP( P, U1(I,1), LDU1, U1(MINI,1), LDU1 ) IF( WANTU2 ) $ CALL CSWAP( M-P, U2(I,1), LDU2, U2(MINI,1), LDU2 ) IF( WANTV1T ) $ CALL CSWAP( Q, V1T(1,I), 1, V1T(1,MINI), 1 ) IF( WANTV2T ) $ CALL CSWAP( M-Q, V2T(1,I), 1, V2T(1,MINI), 1 ) END IF END IF * END DO * RETURN * * End of CBBCSD * END |