d6/df0/dlaqr1_8f_source.html

00001       SUBROUTINE DLAQR1( N, H, LDH, SR1, SI1, SR2, SI2, V )
00002 *
00003 *  -- LAPACK auxiliary routine (version 3.2) --
00004 *     Univ. of Tennessee, Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..
00005 *     November 2006
00006 *
00007 *     .. Scalar Arguments ..
00008       DOUBLE PRECISION   SI1, SI2, SR1, SR2
00009       INTEGER            LDH, N
00010 *     ..
00011 *     .. Array Arguments ..
00012       DOUBLE PRECISION   H( LDH, * ), V( * )
00013 *     ..
00014 *
00015 *       Given a 2-by-2 or 3-by-3 matrix H, DLAQR1 sets v to a
00016 *       scalar multiple of the first column of the product
00017 *
00018 *       (*)  K = (H - (sr1 + i*si1)*I)*(H - (sr2 + i*si2)*I)
00019 *
00020 *       scaling to avoid overflows and most underflows. It
00021 *       is assumed that either
00022 *
00023 *               1) sr1 = sr2 and si1 = -si2
00024 *           or
00025 *               2) si1 = si2 = 0.
00026 *
00027 *       This is useful for starting double implicit shift bulges
00028 *       in the QR algorithm.
00029 *
00030 *
00031 *       N      (input) integer
00032 *              Order of the matrix H. N must be either 2 or 3.
00033 *
00034 *       H      (input) DOUBLE PRECISION array of dimension (LDH,N)
00035 *              The 2-by-2 or 3-by-3 matrix H in (*).
00036 *
00037 *       LDH    (input) integer
00038 *              The leading dimension of H as declared in
00039 *              the calling procedure.  LDH.GE.N
00040 *
00041 *       SR1    (input) DOUBLE PRECISION
00042 *       SI1    The shifts in (*).
00043 *       SR2
00044 *       SI2
00045 *
00046 *       V      (output) DOUBLE PRECISION array of dimension N
00047 *              A scalar multiple of the first column of the
00048 *              matrix K in (*).
00049 *
00050 *     ================================================================
00051 *     Based on contributions by
00052 *        Karen Braman and Ralph Byers, Department of Mathematics,
00053 *        University of Kansas, USA
00054 *
00055 *     ================================================================
00056 *
00057 *     .. Parameters ..
00058       DOUBLE PRECISION   ZERO
00059       PARAMETER          ( ZERO = 0.0d0 )
00060 *     ..
00061 *     .. Local Scalars ..
00062       DOUBLE PRECISION   H21S, H31S, S
00063 *     ..
00064 *     .. Intrinsic Functions ..
00065       INTRINSIC          ABS
00066 *     ..
00067 *     .. Executable Statements ..
00068       IF( N.EQ.2 ) THEN
00069          S = ABS( H( 1, 1 )-SR2 ) + ABS( SI2 ) + ABS( H( 2, 1 ) )
00070          IF( S.EQ.ZERO ) THEN
00071             V( 1 ) = ZERO
00072             V( 2 ) = ZERO
00073          ELSE
00074             H21S = H( 2, 1 ) / S
00075             V( 1 ) = H21S*H( 1, 2 ) + ( H( 1, 1 )-SR1 )*
00076      $               ( ( H( 1, 1 )-SR2 ) / S ) - SI1*( SI2 / S )
00077             V( 2 ) = H21S*( H( 1, 1 )+H( 2, 2 )-SR1-SR2 )
00078          END IF
00079       ELSE
00080          S = ABS( H( 1, 1 )-SR2 ) + ABS( SI2 ) + ABS( H( 2, 1 ) ) +
00081      $       ABS( H( 3, 1 ) )
00082          IF( S.EQ.ZERO ) THEN
00083             V( 1 ) = ZERO
00084             V( 2 ) = ZERO
00085             V( 3 ) = ZERO
00086          ELSE
00087             H21S = H( 2, 1 ) / S
00088             H31S = H( 3, 1 ) / S
00089             V( 1 ) = ( H( 1, 1 )-SR1 )*( ( H( 1, 1 )-SR2 ) / S ) -
00090      $               SI1*( SI2 / S ) + H( 1, 2 )*H21S + H( 1, 3 )*H31S
00091             V( 2 ) = H21S*( H( 1, 1 )+H( 2, 2 )-SR1-SR2 ) +
00092      $               H( 2, 3 )*H31S
00093             V( 3 ) = H31S*( H( 1, 1 )+H( 3, 3 )-SR1-SR2 ) +
00094      $               H21S*H( 3, 2 )
00095          END IF
00096       END IF
00097       END