LAPACK 3.3.1
Linear Algebra PACKage
|
00001 SUBROUTINE DLANV2( A, B, C, D, RT1R, RT1I, RT2R, RT2I, CS, SN ) 00002 * 00003 * -- LAPACK auxiliary routine (version 3.2.2) -- 00004 * -- LAPACK is a software package provided by Univ. of Tennessee, -- 00005 * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- 00006 * June 2010 00007 * 00008 * .. Scalar Arguments .. 00009 DOUBLE PRECISION A, B, C, CS, D, RT1I, RT1R, RT2I, RT2R, SN 00010 * .. 00011 * 00012 * Purpose 00013 * ======= 00014 * 00015 * DLANV2 computes the Schur factorization of a real 2-by-2 nonsymmetric 00016 * matrix in standard form: 00017 * 00018 * [ A B ] = [ CS -SN ] [ AA BB ] [ CS SN ] 00019 * [ C D ] [ SN CS ] [ CC DD ] [-SN CS ] 00020 * 00021 * where either 00022 * 1) CC = 0 so that AA and DD are real eigenvalues of the matrix, or 00023 * 2) AA = DD and BB*CC < 0, so that AA + or - sqrt(BB*CC) are complex 00024 * conjugate eigenvalues. 00025 * 00026 * Arguments 00027 * ========= 00028 * 00029 * A (input/output) DOUBLE PRECISION 00030 * B (input/output) DOUBLE PRECISION 00031 * C (input/output) DOUBLE PRECISION 00032 * D (input/output) DOUBLE PRECISION 00033 * On entry, the elements of the input matrix. 00034 * On exit, they are overwritten by the elements of the 00035 * standardised Schur form. 00036 * 00037 * RT1R (output) DOUBLE PRECISION 00038 * RT1I (output) DOUBLE PRECISION 00039 * RT2R (output) DOUBLE PRECISION 00040 * RT2I (output) DOUBLE PRECISION 00041 * The real and imaginary parts of the eigenvalues. If the 00042 * eigenvalues are a complex conjugate pair, RT1I > 0. 00043 * 00044 * CS (output) DOUBLE PRECISION 00045 * SN (output) DOUBLE PRECISION 00046 * Parameters of the rotation matrix. 00047 * 00048 * Further Details 00049 * =============== 00050 * 00051 * Modified by V. Sima, Research Institute for Informatics, Bucharest, 00052 * Romania, to reduce the risk of cancellation errors, 00053 * when computing real eigenvalues, and to ensure, if possible, that 00054 * abs(RT1R) >= abs(RT2R). 00055 * 00056 * ===================================================================== 00057 * 00058 * .. Parameters .. 00059 DOUBLE PRECISION ZERO, HALF, ONE 00060 PARAMETER ( ZERO = 0.0D+0, HALF = 0.5D+0, ONE = 1.0D+0 ) 00061 DOUBLE PRECISION MULTPL 00062 PARAMETER ( MULTPL = 4.0D+0 ) 00063 * .. 00064 * .. Local Scalars .. 00065 DOUBLE PRECISION AA, BB, BCMAX, BCMIS, CC, CS1, DD, EPS, P, SAB, 00066 $ SAC, SCALE, SIGMA, SN1, TAU, TEMP, Z 00067 * .. 00068 * .. External Functions .. 00069 DOUBLE PRECISION DLAMCH, DLAPY2 00070 EXTERNAL DLAMCH, DLAPY2 00071 * .. 00072 * .. Intrinsic Functions .. 00073 INTRINSIC ABS, MAX, MIN, SIGN, SQRT 00074 * .. 00075 * .. Executable Statements .. 00076 * 00077 EPS = DLAMCH( 'P' ) 00078 IF( C.EQ.ZERO ) THEN 00079 CS = ONE 00080 SN = ZERO 00081 GO TO 10 00082 * 00083 ELSE IF( B.EQ.ZERO ) THEN 00084 * 00085 * Swap rows and columns 00086 * 00087 CS = ZERO 00088 SN = ONE 00089 TEMP = D 00090 D = A 00091 A = TEMP 00092 B = -C 00093 C = ZERO 00094 GO TO 10 00095 ELSE IF( ( A-D ).EQ.ZERO .AND. SIGN( ONE, B ).NE.SIGN( ONE, C ) ) 00096 $ THEN 00097 CS = ONE 00098 SN = ZERO 00099 GO TO 10 00100 ELSE 00101 * 00102 TEMP = A - D 00103 P = HALF*TEMP 00104 BCMAX = MAX( ABS( B ), ABS( C ) ) 00105 BCMIS = MIN( ABS( B ), ABS( C ) )*SIGN( ONE, B )*SIGN( ONE, C ) 00106 SCALE = MAX( ABS( P ), BCMAX ) 00107 Z = ( P / SCALE )*P + ( BCMAX / SCALE )*BCMIS 00108 * 00109 * If Z is of the order of the machine accuracy, postpone the 00110 * decision on the nature of eigenvalues 00111 * 00112 IF( Z.GE.MULTPL*EPS ) THEN 00113 * 00114 * Real eigenvalues. Compute A and D. 00115 * 00116 Z = P + SIGN( SQRT( SCALE )*SQRT( Z ), P ) 00117 A = D + Z 00118 D = D - ( BCMAX / Z )*BCMIS 00119 * 00120 * Compute B and the rotation matrix 00121 * 00122 TAU = DLAPY2( C, Z ) 00123 CS = Z / TAU 00124 SN = C / TAU 00125 B = B - C 00126 C = ZERO 00127 ELSE 00128 * 00129 * Complex eigenvalues, or real (almost) equal eigenvalues. 00130 * Make diagonal elements equal. 00131 * 00132 SIGMA = B + C 00133 TAU = DLAPY2( SIGMA, TEMP ) 00134 CS = SQRT( HALF*( ONE+ABS( SIGMA ) / TAU ) ) 00135 SN = -( P / ( TAU*CS ) )*SIGN( ONE, SIGMA ) 00136 * 00137 * Compute [ AA BB ] = [ A B ] [ CS -SN ] 00138 * [ CC DD ] [ C D ] [ SN CS ] 00139 * 00140 AA = A*CS + B*SN 00141 BB = -A*SN + B*CS 00142 CC = C*CS + D*SN 00143 DD = -C*SN + D*CS 00144 * 00145 * Compute [ A B ] = [ CS SN ] [ AA BB ] 00146 * [ C D ] [-SN CS ] [ CC DD ] 00147 * 00148 A = AA*CS + CC*SN 00149 B = BB*CS + DD*SN 00150 C = -AA*SN + CC*CS 00151 D = -BB*SN + DD*CS 00152 * 00153 TEMP = HALF*( A+D ) 00154 A = TEMP 00155 D = TEMP 00156 * 00157 IF( C.NE.ZERO ) THEN 00158 IF( B.NE.ZERO ) THEN 00159 IF( SIGN( ONE, B ).EQ.SIGN( ONE, C ) ) THEN 00160 * 00161 * Real eigenvalues: reduce to upper triangular form 00162 * 00163 SAB = SQRT( ABS( B ) ) 00164 SAC = SQRT( ABS( C ) ) 00165 P = SIGN( SAB*SAC, C ) 00166 TAU = ONE / SQRT( ABS( B+C ) ) 00167 A = TEMP + P 00168 D = TEMP - P 00169 B = B - C 00170 C = ZERO 00171 CS1 = SAB*TAU 00172 SN1 = SAC*TAU 00173 TEMP = CS*CS1 - SN*SN1 00174 SN = CS*SN1 + SN*CS1 00175 CS = TEMP 00176 END IF 00177 ELSE 00178 B = -C 00179 C = ZERO 00180 TEMP = CS 00181 CS = -SN 00182 SN = TEMP 00183 END IF 00184 END IF 00185 END IF 00186 * 00187 END IF 00188 * 00189 10 CONTINUE 00190 * 00191 * Store eigenvalues in (RT1R,RT1I) and (RT2R,RT2I). 00192 * 00193 RT1R = A 00194 RT2R = D 00195 IF( C.EQ.ZERO ) THEN 00196 RT1I = ZERO 00197 RT2I = ZERO 00198 ELSE 00199 RT1I = SQRT( ABS( B ) )*SQRT( ABS( C ) ) 00200 RT2I = -RT1I 00201 END IF 00202 RETURN 00203 * 00204 * End of DLANV2 00205 * 00206 END