SUBROUTINE DHSEQR( JOB, COMPZ, N, ILO, IHI, H, LDH, WR, WI, Z, $ LDZ, WORK, LWORK, INFO ) * * -- LAPACK routine (instrumented to count operations, version 3.0) -- * Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd., * Courant Institute, Argonne National Lab, and Rice University * June 30, 1999 * * .. Scalar Arguments .. CHARACTER COMPZ, JOB INTEGER IHI, ILO, INFO, LDH, LDZ, LWORK, N * .. * .. Array Arguments .. DOUBLE PRECISION H( LDH, * ), WI( * ), WORK( * ), WR( * ), $ Z( LDZ, * ) * .. * Common block to return operation count. * .. Common blocks .. COMMON / LATIME / OPS, ITCNT * .. * .. Scalars in Common .. DOUBLE PRECISION ITCNT, OPS * .. * * Purpose * ======= * * DHSEQR computes the eigenvalues of a real upper Hessenberg matrix H * and, optionally, the matrices T and Z from the Schur decomposition * H = Z T Z**T, where T is an upper quasi-triangular matrix (the Schur * form), and Z is the orthogonal matrix of Schur vectors. * * Optionally Z may be postmultiplied into an input orthogonal matrix Q, * so that this routine can give the Schur factorization of a matrix A * which has been reduced to the Hessenberg form H by the orthogonal * matrix Q: A = Q*H*Q**T = (QZ)*T*(QZ)**T. * * Arguments * ========= * * JOB (input) CHARACTER*1 * = 'E': compute eigenvalues only; * = 'S': compute eigenvalues and the Schur form T. * * COMPZ (input) CHARACTER*1 * = 'N': no Schur vectors are computed; * = 'I': Z is initialized to the unit matrix and the matrix Z * of Schur vectors of H is returned; * = 'V': Z must contain an orthogonal matrix Q on entry, and * the product Q*Z is returned. * * N (input) INTEGER * The order of the matrix H. N >= 0. * * ILO (input) INTEGER * IHI (input) INTEGER * It is assumed that H is already upper triangular in rows * and columns 1:ILO-1 and IHI+1:N. ILO and IHI are normally * set by a previous call to DGEBAL, and then passed to SGEHRD * when the matrix output by DGEBAL is reduced to Hessenberg * form. Otherwise ILO and IHI should be set to 1 and N * respectively. * 1 <= ILO <= IHI <= N, if N > 0; ILO=1 and IHI=0, if N=0. * * H (input/output) DOUBLE PRECISION array, dimension (LDH,N) * On entry, the upper Hessenberg matrix H. * On exit, if JOB = 'S', H contains the upper quasi-triangular * matrix T from the Schur decomposition (the Schur form); * 2-by-2 diagonal blocks (corresponding to complex conjugate * pairs of eigenvalues) are returned in standard form, with * H(i,i) = H(i+1,i+1) and H(i+1,i)*H(i,i+1) < 0. If JOB = 'E', * the contents of H are unspecified on exit. * * LDH (input) INTEGER * The leading dimension of the array H. LDH >= max(1,N). * * WR (output) DOUBLE PRECISION array, dimension (N) * WI (output) DOUBLE PRECISION array, dimension (N) * The real and imaginary parts, respectively, of the computed * eigenvalues. If two eigenvalues are computed as a complex * conjugate pair, they are stored in consecutive elements of * WR and WI, say the i-th and (i+1)th, with WI(i) > 0 and * WI(i+1) < 0. If JOB = 'S', the eigenvalues are stored in the * same order as on the diagonal of the Schur form returned in * H, with WR(i) = H(i,i) and, if H(i:i+1,i:i+1) is a 2-by-2 * diagonal block, WI(i) = sqrt(H(i+1,i)*H(i,i+1)) and * WI(i+1) = -WI(i). * * Z (input/output) DOUBLE PRECISION array, dimension (LDZ,N) * If COMPZ = 'N': Z is not referenced. * If COMPZ = 'I': on entry, Z need not be set, and on exit, Z * contains the orthogonal matrix Z of the Schur vectors of H. * If COMPZ = 'V': on entry Z must contain an N-by-N matrix Q, * which is assumed to be equal to the unit matrix except for * the submatrix Z(ILO:IHI,ILO:IHI); on exit Z contains Q*Z. * Normally Q is the orthogonal matrix generated by DORGHR after * the call to DGEHRD which formed the Hessenberg matrix H. * * LDZ (input) INTEGER * The leading dimension of the array Z. * LDZ >= max(1,N) if COMPZ = 'I' or 'V'; LDZ >= 1 otherwise. * * WORK (workspace/output) DOUBLE PRECISION array, dimension (LWORK) * On exit, if INFO = 0, WORK(1) returns the optimal LWORK. * * LWORK (input) INTEGER * The dimension of the array WORK. LWORK >= max(1,N). * * If LWORK = -1, then a workspace query is assumed; the routine * only calculates the optimal size of the WORK array, returns * this value as the first entry of the WORK array, and no error * message related to LWORK is issued by XERBLA. * * INFO (output) INTEGER * = 0: successful exit * < 0: if INFO = -i, the i-th argument had an illegal value * > 0: if INFO = i, DHSEQR failed to compute all of the * eigenvalues in a total of 30*(IHI-ILO+1) iterations; * elements 1:ilo-1 and i+1:n of WR and WI contain those * eigenvalues which have been successfully computed. * * ===================================================================== * * .. Parameters .. DOUBLE PRECISION ZERO, ONE, TWO PARAMETER ( ZERO = 0.0D+0, ONE = 1.0D+0, TWO = 2.0D+0 ) DOUBLE PRECISION CONST PARAMETER ( CONST = 1.5D+0 ) INTEGER NSMAX, LDS PARAMETER ( NSMAX = 15, LDS = NSMAX ) * .. * .. Local Scalars .. LOGICAL INITZ, LQUERY, WANTT, WANTZ INTEGER I, I1, I2, IERR, II, ITEMP, ITN, ITS, J, K, L, $ MAXB, NH, NR, NS, NV DOUBLE PRECISION ABSW, OPST, OVFL, SMLNUM, TAU, TEMP, TST1, ULP, $ UNFL * .. * .. Local Arrays .. DOUBLE PRECISION S( LDS, NSMAX ), V( NSMAX+1 ), VV( NSMAX+1 ) * .. * .. External Functions .. LOGICAL LSAME INTEGER IDAMAX, ILAENV DOUBLE PRECISION DLAMCH, DLANHS, DLAPY2 EXTERNAL LSAME, IDAMAX, ILAENV, DLAMCH, DLANHS, DLAPY2 * .. * .. External Subroutines .. EXTERNAL DCOPY, DGEMV, DLABAD, DLACPY, DLAHQR, DLARFG, $ DLARFX, DLASET, DSCAL, XERBLA * .. * .. Intrinsic Functions .. INTRINSIC ABS, MAX, MIN * .. * .. Executable Statements .. * * Decode and test the input parameters * WANTT = LSAME( JOB, 'S' ) INITZ = LSAME( COMPZ, 'I' ) WANTZ = INITZ .OR. LSAME( COMPZ, 'V' ) * INFO = 0 WORK( 1 ) = MAX( 1, N ) LQUERY = ( LWORK.EQ.-1 ) IF( .NOT.LSAME( JOB, 'E' ) .AND. .NOT.WANTT ) THEN INFO = -1 ELSE IF( .NOT.LSAME( COMPZ, 'N' ) .AND. .NOT.WANTZ ) THEN INFO = -2 ELSE IF( N.LT.0 ) THEN INFO = -3 ELSE IF( ILO.LT.1 .OR. ILO.GT.MAX( 1, N ) ) THEN INFO = -4 ELSE IF( IHI.LT.MIN( ILO, N ) .OR. IHI.GT.N ) THEN INFO = -5 ELSE IF( LDH.LT.MAX( 1, N ) ) THEN INFO = -7 ELSE IF( LDZ.LT.1 .OR. WANTZ .AND. LDZ.LT.MAX( 1, N ) ) THEN INFO = -11 ELSE IF( LWORK.LT.MAX( 1, N ) .AND. .NOT.LQUERY ) THEN INFO = -13 END IF IF( INFO.NE.0 ) THEN CALL XERBLA( 'DHSEQR', -INFO ) RETURN ELSE IF( LQUERY ) THEN RETURN END IF *** * Initialize OPST = 0 *** * * Initialize Z, if necessary * IF( INITZ ) $ CALL DLASET( 'Full', N, N, ZERO, ONE, Z, LDZ ) * * Store the eigenvalues isolated by DGEBAL. * DO 10 I = 1, ILO - 1 WR( I ) = H( I, I ) WI( I ) = ZERO 10 CONTINUE DO 20 I = IHI + 1, N WR( I ) = H( I, I ) WI( I ) = ZERO 20 CONTINUE * * Quick return if possible. * IF( N.EQ.0 ) $ RETURN IF( ILO.EQ.IHI ) THEN WR( ILO ) = H( ILO, ILO ) WI( ILO ) = ZERO RETURN END IF * * Set rows and columns ILO to IHI to zero below the first * subdiagonal. * DO 40 J = ILO, IHI - 2 DO 30 I = J + 2, N H( I, J ) = ZERO 30 CONTINUE 40 CONTINUE NH = IHI - ILO + 1 * * Determine the order of the multi-shift QR algorithm to be used. * NS = ILAENV( 4, 'DHSEQR', JOB // COMPZ, N, ILO, IHI, -1 ) MAXB = ILAENV( 8, 'DHSEQR', JOB // COMPZ, N, ILO, IHI, -1 ) IF( NS.LE.2 .OR. NS.GT.NH .OR. MAXB.GE.NH ) THEN * * Use the standard double-shift algorithm * CALL DLAHQR( WANTT, WANTZ, N, ILO, IHI, H, LDH, WR, WI, ILO, $ IHI, Z, LDZ, INFO ) RETURN END IF MAXB = MAX( 3, MAXB ) NS = MIN( NS, MAXB, NSMAX ) * * Now 2 < NS <= MAXB < NH. * * Set machine-dependent constants for the stopping criterion. * If norm(H) <= sqrt(OVFL), overflow should not occur. * UNFL = DLAMCH( 'Safe minimum' ) OVFL = ONE / UNFL CALL DLABAD( UNFL, OVFL ) ULP = DLAMCH( 'Precision' ) SMLNUM = UNFL*( NH / ULP ) * * I1 and I2 are the indices of the first row and last column of H * to which transformations must be applied. If eigenvalues only are * being computed, I1 and I2 are set inside the main loop. * IF( WANTT ) THEN I1 = 1 I2 = N END IF * * ITN is the total number of multiple-shift QR iterations allowed. * ITN = 30*NH * * The main loop begins here. I is the loop index and decreases from * IHI to ILO in steps of at most MAXB. Each iteration of the loop * works with the active submatrix in rows and columns L to I. * Eigenvalues I+1 to IHI have already converged. Either L = ILO or * H(L,L-1) is negligible so that the matrix splits. * I = IHI 50 CONTINUE L = ILO IF( I.LT.ILO ) $ GO TO 170 * * Perform multiple-shift QR iterations on rows and columns ILO to I * until a submatrix of order at most MAXB splits off at the bottom * because a subdiagonal element has become negligible. * DO 150 ITS = 0, ITN * * Look for a single small subdiagonal element. * DO 60 K = I, L + 1, -1 TST1 = ABS( H( K-1, K-1 ) ) + ABS( H( K, K ) ) IF( TST1.EQ.ZERO ) THEN TST1 = DLANHS( '1', I-L+1, H( L, L ), LDH, WORK ) *** * Increment op count OPS = OPS + ( I-L+1 )*( I-L+2 ) / 2 *** END IF IF( ABS( H( K, K-1 ) ).LE.MAX( ULP*TST1, SMLNUM ) ) $ GO TO 70 60 CONTINUE 70 CONTINUE L = K *** * Increment op count OPST = OPST + 3*( I-L+1 ) *** IF( L.GT.ILO ) THEN * * H(L,L-1) is negligible. * H( L, L-1 ) = ZERO END IF * * Exit from loop if a submatrix of order <= MAXB has split off. * IF( L.GE.I-MAXB+1 ) $ GO TO 160 * * Now the active submatrix is in rows and columns L to I. If * eigenvalues only are being computed, only the active submatrix * need be transformed. * IF( .NOT.WANTT ) THEN I1 = L I2 = I END IF * IF( ITS.EQ.20 .OR. ITS.EQ.30 ) THEN * * Exceptional shifts. * DO 80 II = I - NS + 1, I WR( II ) = CONST*( ABS( H( II, II-1 ) )+ $ ABS( H( II, II ) ) ) WI( II ) = ZERO 80 CONTINUE *** * Increment op count OPST = OPST + 2*NS *** ELSE * * Use eigenvalues of trailing submatrix of order NS as shifts. * CALL DLACPY( 'Full', NS, NS, H( I-NS+1, I-NS+1 ), LDH, S, $ LDS ) CALL DLAHQR( .FALSE., .FALSE., NS, 1, NS, S, LDS, $ WR( I-NS+1 ), WI( I-NS+1 ), 1, NS, Z, LDZ, $ IERR ) IF( IERR.GT.0 ) THEN * * If DLAHQR failed to compute all NS eigenvalues, use the * unconverged diagonal elements as the remaining shifts. * DO 90 II = 1, IERR WR( I-NS+II ) = S( II, II ) WI( I-NS+II ) = ZERO 90 CONTINUE END IF END IF * * Form the first column of (G-w(1)) (G-w(2)) . . . (G-w(ns)) * where G is the Hessenberg submatrix H(L:I,L:I) and w is * the vector of shifts (stored in WR and WI). The result is * stored in the local array V. * V( 1 ) = ONE DO 100 II = 2, NS + 1 V( II ) = ZERO 100 CONTINUE NV = 1 DO 120 J = I - NS + 1, I IF( WI( J ).GE.ZERO ) THEN IF( WI( J ).EQ.ZERO ) THEN * * real shift * CALL DCOPY( NV+1, V, 1, VV, 1 ) CALL DGEMV( 'No transpose', NV+1, NV, ONE, H( L, L ), $ LDH, VV, 1, -WR( J ), V, 1 ) NV = NV + 1 *** * Increment op count OPST = OPST + 2*NV*( NV+1 ) + NV + 1 *** ELSE IF( WI( J ).GT.ZERO ) THEN * * complex conjugate pair of shifts * CALL DCOPY( NV+1, V, 1, VV, 1 ) CALL DGEMV( 'No transpose', NV+1, NV, ONE, H( L, L ), $ LDH, V, 1, -TWO*WR( J ), VV, 1 ) ITEMP = IDAMAX( NV+1, VV, 1 ) TEMP = ONE / MAX( ABS( VV( ITEMP ) ), SMLNUM ) CALL DSCAL( NV+1, TEMP, VV, 1 ) ABSW = DLAPY2( WR( J ), WI( J ) ) TEMP = ( TEMP*ABSW )*ABSW CALL DGEMV( 'No transpose', NV+2, NV+1, ONE, $ H( L, L ), LDH, VV, 1, TEMP, V, 1 ) NV = NV + 2 *** * Increment op count OPST = OPST + 4*( NV+1 )**2 + 4*NV + 9 *** END IF * * Scale V(1:NV) so that max(abs(V(i))) = 1. If V is zero, * reset it to the unit vector. * ITEMP = IDAMAX( NV, V, 1 ) *** * Increment op count OPST = OPST + NV *** TEMP = ABS( V( ITEMP ) ) IF( TEMP.EQ.ZERO ) THEN V( 1 ) = ONE DO 110 II = 2, NV V( II ) = ZERO 110 CONTINUE ELSE TEMP = MAX( TEMP, SMLNUM ) CALL DSCAL( NV, ONE / TEMP, V, 1 ) *** * Increment op count OPST = OPST + NV *** END IF END IF 120 CONTINUE * * Multiple-shift QR step * DO 140 K = L, I - 1 * * The first iteration of this loop determines a reflection G * from the vector V and applies it from left and right to H, * thus creating a nonzero bulge below the subdiagonal. * * Each subsequent iteration determines a reflection G to * restore the Hessenberg form in the (K-1)th column, and thus * chases the bulge one step toward the bottom of the active * submatrix. NR is the order of G. * NR = MIN( NS+1, I-K+1 ) IF( K.GT.L ) $ CALL DCOPY( NR, H( K, K-1 ), 1, V, 1 ) CALL DLARFG( NR, V( 1 ), V( 2 ), 1, TAU ) *** * Increment op count OPST = OPST + 3*NR + 9 *** IF( K.GT.L ) THEN H( K, K-1 ) = V( 1 ) DO 130 II = K + 1, I H( II, K-1 ) = ZERO 130 CONTINUE END IF V( 1 ) = ONE * * Apply G from the left to transform the rows of the matrix in * columns K to I2. * CALL DLARFX( 'Left', NR, I2-K+1, V, TAU, H( K, K ), LDH, $ WORK ) * * Apply G from the right to transform the columns of the * matrix in rows I1 to min(K+NR,I). * CALL DLARFX( 'Right', MIN( K+NR, I )-I1+1, NR, V, TAU, $ H( I1, K ), LDH, WORK ) *** * Increment op count OPS = OPS + ( 4*NR-2 )*( I2-I1+2+MIN( NR, I-K ) ) *** * IF( WANTZ ) THEN * * Accumulate transformations in the matrix Z * CALL DLARFX( 'Right', NH, NR, V, TAU, Z( ILO, K ), LDZ, $ WORK ) *** * Increment op count OPS = OPS + ( 4*NR-2 )*NH *** END IF 140 CONTINUE * 150 CONTINUE * * Failure to converge in remaining number of iterations * INFO = I RETURN * 160 CONTINUE * * A submatrix of order <= MAXB in rows and columns L to I has split * off. Use the double-shift QR algorithm to handle it. * CALL DLAHQR( WANTT, WANTZ, N, L, I, H, LDH, WR, WI, ILO, IHI, Z, $ LDZ, INFO ) IF( INFO.GT.0 ) $ RETURN * * Decrement number of remaining iterations, and return to start of * the main loop with a new value of I. * ITN = ITN - ITS I = L - 1 GO TO 50 * 170 CONTINUE *** * Compute final op count OPS = OPS + OPST *** WORK( 1 ) = MAX( 1, N ) RETURN * * End of DHSEQR * END