LAPACK: SRC/dlaev2.f Source File

Go to the documentation of this file.
00001       SUBROUTINE DLAEV2( A, B, C, RT1, RT2, CS1, SN1 )
00002 *
00003 *  -- LAPACK auxiliary routine (version 3.2) --
00004 *  -- LAPACK is a software package provided by Univ. of Tennessee,    --
00005 *  -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
00006 *     November 2006
00007 *
00008 *     .. Scalar Arguments ..
00009       DOUBLE PRECISION   A, B, C, CS1, RT1, RT2, SN1
00010 *     ..
00011 *
00012 *  Purpose
00013 *  =======
00014 *
00015 *  DLAEV2 computes the eigendecomposition of a 2-by-2 symmetric matrix
00016 *     [  A   B  ]
00017 *     [  B   C  ].
00018 *  On return, RT1 is the eigenvalue of larger absolute value, RT2 is the
00019 *  eigenvalue of smaller absolute value, and (CS1,SN1) is the unit right
00020 *  eigenvector for RT1, giving the decomposition
00021 *
00022 *     [ CS1  SN1 ] [  A   B  ] [ CS1 -SN1 ]  =  [ RT1  0  ]
00023 *     [-SN1  CS1 ] [  B   C  ] [ SN1  CS1 ]     [  0  RT2 ].
00024 *
00025 *  Arguments
00026 *  =========
00027 *
00028 *  A       (input) DOUBLE PRECISION
00029 *          The (1,1) element of the 2-by-2 matrix.
00030 *
00031 *  B       (input) DOUBLE PRECISION
00032 *          The (1,2) element and the conjugate of the (2,1) element of
00033 *          the 2-by-2 matrix.
00034 *
00035 *  C       (input) DOUBLE PRECISION
00036 *          The (2,2) element of the 2-by-2 matrix.
00037 *
00038 *  RT1     (output) DOUBLE PRECISION
00039 *          The eigenvalue of larger absolute value.
00040 *
00041 *  RT2     (output) DOUBLE PRECISION
00042 *          The eigenvalue of smaller absolute value.
00043 *
00044 *  CS1     (output) DOUBLE PRECISION
00045 *  SN1     (output) DOUBLE PRECISION
00046 *          The vector (CS1, SN1) is a unit right eigenvector for RT1.
00047 *
00048 *  Further Details
00049 *  ===============
00050 *
00051 *  RT1 is accurate to a few ulps barring over/underflow.
00052 *
00053 *  RT2 may be inaccurate if there is massive cancellation in the
00054 *  determinant A*C-B*B; higher precision or correctly rounded or
00055 *  correctly truncated arithmetic would be needed to compute RT2
00056 *  accurately in all cases.
00057 *
00058 *  CS1 and SN1 are accurate to a few ulps barring over/underflow.
00059 *
00060 *  Overflow is possible only if RT1 is within a factor of 5 of overflow.
00061 *  Underflow is harmless if the input data is 0 or exceeds
00062 *     underflow_threshold / macheps.
00063 *
00064 * =====================================================================
00065 *
00066 *     .. Parameters ..
00067       DOUBLE PRECISION   ONE
00068       PARAMETER          ( ONE = 1.0D0 )
00069       DOUBLE PRECISION   TWO
00070       PARAMETER          ( TWO = 2.0D0 )
00071       DOUBLE PRECISION   ZERO
00072       PARAMETER          ( ZERO = 0.0D0 )
00073       DOUBLE PRECISION   HALF
00074       PARAMETER          ( HALF = 0.5D0 )
00075 *     ..
00076 *     .. Local Scalars ..
00077       INTEGER            SGN1, SGN2
00078       DOUBLE PRECISION   AB, ACMN, ACMX, ACS, ADF, CS, CT, DF, RT, SM,
00079      $                   TB, TN
00080 *     ..
00081 *     .. Intrinsic Functions ..
00082       INTRINSIC          ABS, SQRT
00083 *     ..
00084 *     .. Executable Statements ..
00085 *
00086 *     Compute the eigenvalues
00087 *
00088       SM = A + C
00089       DF = A - C
00090       ADF = ABS( DF )
00091       TB = B + B
00092       AB = ABS( TB )
00093       IF( ABS( A ).GT.ABS( C ) ) THEN
00094          ACMX = A
00095          ACMN = C
00096       ELSE
00097          ACMX = C
00098          ACMN = A
00099       END IF
00100       IF( ADF.GT.AB ) THEN
00101          RT = ADF*SQRT( ONE+( AB / ADF )**2 )
00102       ELSE IF( ADF.LT.AB ) THEN
00103          RT = AB*SQRT( ONE+( ADF / AB )**2 )
00104       ELSE
00105 *
00106 *        Includes case AB=ADF=0
00107 *
00108          RT = AB*SQRT( TWO )
00109       END IF
00110       IF( SM.LT.ZERO ) THEN
00111          RT1 = HALF*( SM-RT )
00112          SGN1 = -1
00113 *
00114 *        Order of execution important.
00115 *        To get fully accurate smaller eigenvalue,
00116 *        next line needs to be executed in higher precision.
00117 *
00118          RT2 = ( ACMX / RT1 )*ACMN - ( B / RT1 )*B
00119       ELSE IF( SM.GT.ZERO ) THEN
00120          RT1 = HALF*( SM+RT )
00121          SGN1 = 1
00122 *
00123 *        Order of execution important.
00124 *        To get fully accurate smaller eigenvalue,
00125 *        next line needs to be executed in higher precision.
00126 *
00127          RT2 = ( ACMX / RT1 )*ACMN - ( B / RT1 )*B
00128       ELSE
00129 *
00130 *        Includes case RT1 = RT2 = 0
00131 *
00132          RT1 = HALF*RT
00133          RT2 = -HALF*RT
00134          SGN1 = 1
00135       END IF
00136 *
00137 *     Compute the eigenvector
00138 *
00139       IF( DF.GE.ZERO ) THEN
00140          CS = DF + RT
00141          SGN2 = 1
00142       ELSE
00143          CS = DF - RT
00144          SGN2 = -1
00145       END IF
00146       ACS = ABS( CS )
00147       IF( ACS.GT.AB ) THEN
00148          CT = -TB / CS
00149          SN1 = ONE / SQRT( ONE+CT*CT )
00150          CS1 = CT*SN1
00151       ELSE
00152          IF( AB.EQ.ZERO ) THEN
00153             CS1 = ONE
00154             SN1 = ZERO
00155          ELSE
00156             TN = -CS / TB
00157             CS1 = ONE / SQRT( ONE+TN*TN )
00158             SN1 = TN*CS1
00159          END IF
00160       END IF
00161       IF( SGN1.EQ.SGN2 ) THEN
00162          TN = CS1
00163          CS1 = -SN1
00164          SN1 = TN
00165       END IF
00166       RETURN
00167 *
00168 *     End of DLAEV2
00169 *
00170       END