LAPACK 3.3.0
|
00001 SUBROUTINE SSPEV( JOBZ, UPLO, N, AP, W, Z, LDZ, WORK, INFO ) 00002 * 00003 * -- LAPACK driver 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 CHARACTER JOBZ, UPLO 00010 INTEGER INFO, LDZ, N 00011 * .. 00012 * .. Array Arguments .. 00013 REAL AP( * ), W( * ), WORK( * ), Z( LDZ, * ) 00014 * .. 00015 * 00016 * Purpose 00017 * ======= 00018 * 00019 * SSPEV computes all the eigenvalues and, optionally, eigenvectors of a 00020 * real symmetric matrix A in packed storage. 00021 * 00022 * Arguments 00023 * ========= 00024 * 00025 * JOBZ (input) CHARACTER*1 00026 * = 'N': Compute eigenvalues only; 00027 * = 'V': Compute eigenvalues and eigenvectors. 00028 * 00029 * UPLO (input) CHARACTER*1 00030 * = 'U': Upper triangle of A is stored; 00031 * = 'L': Lower triangle of A is stored. 00032 * 00033 * N (input) INTEGER 00034 * The order of the matrix A. N >= 0. 00035 * 00036 * AP (input/output) REAL array, dimension (N*(N+1)/2) 00037 * On entry, the upper or lower triangle of the symmetric matrix 00038 * A, packed columnwise in a linear array. The j-th column of A 00039 * is stored in the array AP as follows: 00040 * if UPLO = 'U', AP(i + (j-1)*j/2) = A(i,j) for 1<=i<=j; 00041 * if UPLO = 'L', AP(i + (j-1)*(2*n-j)/2) = A(i,j) for j<=i<=n. 00042 * 00043 * On exit, AP is overwritten by values generated during the 00044 * reduction to tridiagonal form. If UPLO = 'U', the diagonal 00045 * and first superdiagonal of the tridiagonal matrix T overwrite 00046 * the corresponding elements of A, and if UPLO = 'L', the 00047 * diagonal and first subdiagonal of T overwrite the 00048 * corresponding elements of A. 00049 * 00050 * W (output) REAL array, dimension (N) 00051 * If INFO = 0, the eigenvalues in ascending order. 00052 * 00053 * Z (output) REAL array, dimension (LDZ, N) 00054 * If JOBZ = 'V', then if INFO = 0, Z contains the orthonormal 00055 * eigenvectors of the matrix A, with the i-th column of Z 00056 * holding the eigenvector associated with W(i). 00057 * If JOBZ = 'N', then Z is not referenced. 00058 * 00059 * LDZ (input) INTEGER 00060 * The leading dimension of the array Z. LDZ >= 1, and if 00061 * JOBZ = 'V', LDZ >= max(1,N). 00062 * 00063 * WORK (workspace) REAL array, dimension (3*N) 00064 * 00065 * INFO (output) INTEGER 00066 * = 0: successful exit. 00067 * < 0: if INFO = -i, the i-th argument had an illegal value. 00068 * > 0: if INFO = i, the algorithm failed to converge; i 00069 * off-diagonal elements of an intermediate tridiagonal 00070 * form did not converge to zero. 00071 * 00072 * ===================================================================== 00073 * 00074 * .. Parameters .. 00075 REAL ZERO, ONE 00076 PARAMETER ( ZERO = 0.0E0, ONE = 1.0E0 ) 00077 * .. 00078 * .. Local Scalars .. 00079 LOGICAL WANTZ 00080 INTEGER IINFO, IMAX, INDE, INDTAU, INDWRK, ISCALE 00081 REAL ANRM, BIGNUM, EPS, RMAX, RMIN, SAFMIN, SIGMA, 00082 $ SMLNUM 00083 * .. 00084 * .. External Functions .. 00085 LOGICAL LSAME 00086 REAL SLAMCH, SLANSP 00087 EXTERNAL LSAME, SLAMCH, SLANSP 00088 * .. 00089 * .. External Subroutines .. 00090 EXTERNAL SOPGTR, SSCAL, SSPTRD, SSTEQR, SSTERF, XERBLA 00091 * .. 00092 * .. Intrinsic Functions .. 00093 INTRINSIC SQRT 00094 * .. 00095 * .. Executable Statements .. 00096 * 00097 * Test the input parameters. 00098 * 00099 WANTZ = LSAME( JOBZ, 'V' ) 00100 * 00101 INFO = 0 00102 IF( .NOT.( WANTZ .OR. LSAME( JOBZ, 'N' ) ) ) THEN 00103 INFO = -1 00104 ELSE IF( .NOT.( LSAME( UPLO, 'U' ) .OR. LSAME( UPLO, 'L' ) ) ) 00105 $ THEN 00106 INFO = -2 00107 ELSE IF( N.LT.0 ) THEN 00108 INFO = -3 00109 ELSE IF( LDZ.LT.1 .OR. ( WANTZ .AND. LDZ.LT.N ) ) THEN 00110 INFO = -7 00111 END IF 00112 * 00113 IF( INFO.NE.0 ) THEN 00114 CALL XERBLA( 'SSPEV ', -INFO ) 00115 RETURN 00116 END IF 00117 * 00118 * Quick return if possible 00119 * 00120 IF( N.EQ.0 ) 00121 $ RETURN 00122 * 00123 IF( N.EQ.1 ) THEN 00124 W( 1 ) = AP( 1 ) 00125 IF( WANTZ ) 00126 $ Z( 1, 1 ) = ONE 00127 RETURN 00128 END IF 00129 * 00130 * Get machine constants. 00131 * 00132 SAFMIN = SLAMCH( 'Safe minimum' ) 00133 EPS = SLAMCH( 'Precision' ) 00134 SMLNUM = SAFMIN / EPS 00135 BIGNUM = ONE / SMLNUM 00136 RMIN = SQRT( SMLNUM ) 00137 RMAX = SQRT( BIGNUM ) 00138 * 00139 * Scale matrix to allowable range, if necessary. 00140 * 00141 ANRM = SLANSP( 'M', UPLO, N, AP, WORK ) 00142 ISCALE = 0 00143 IF( ANRM.GT.ZERO .AND. ANRM.LT.RMIN ) THEN 00144 ISCALE = 1 00145 SIGMA = RMIN / ANRM 00146 ELSE IF( ANRM.GT.RMAX ) THEN 00147 ISCALE = 1 00148 SIGMA = RMAX / ANRM 00149 END IF 00150 IF( ISCALE.EQ.1 ) THEN 00151 CALL SSCAL( ( N*( N+1 ) ) / 2, SIGMA, AP, 1 ) 00152 END IF 00153 * 00154 * Call SSPTRD to reduce symmetric packed matrix to tridiagonal form. 00155 * 00156 INDE = 1 00157 INDTAU = INDE + N 00158 CALL SSPTRD( UPLO, N, AP, W, WORK( INDE ), WORK( INDTAU ), IINFO ) 00159 * 00160 * For eigenvalues only, call SSTERF. For eigenvectors, first call 00161 * SOPGTR to generate the orthogonal matrix, then call SSTEQR. 00162 * 00163 IF( .NOT.WANTZ ) THEN 00164 CALL SSTERF( N, W, WORK( INDE ), INFO ) 00165 ELSE 00166 INDWRK = INDTAU + N 00167 CALL SOPGTR( UPLO, N, AP, WORK( INDTAU ), Z, LDZ, 00168 $ WORK( INDWRK ), IINFO ) 00169 CALL SSTEQR( JOBZ, N, W, WORK( INDE ), Z, LDZ, WORK( INDTAU ), 00170 $ INFO ) 00171 END IF 00172 * 00173 * If matrix was scaled, then rescale eigenvalues appropriately. 00174 * 00175 IF( ISCALE.EQ.1 ) THEN 00176 IF( INFO.EQ.0 ) THEN 00177 IMAX = N 00178 ELSE 00179 IMAX = INFO - 1 00180 END IF 00181 CALL SSCAL( IMAX, ONE / SIGMA, W, 1 ) 00182 END IF 00183 * 00184 RETURN 00185 * 00186 * End of SSPEV 00187 * 00188 END