/* Subroutine */ int ssyevx_(char *jobz, char *range, char *uplo, integer *n,
real *a, integer *lda, real *vl, real *vu, integer *il, integer *iu,
real *abstol, integer *m, real *w, real *z__, integer *ldz, real *
work, integer *lwork, integer *iwork, integer *ifail, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, z_dim1, z_offset, i__1, i__2;
real r__1, r__2;
/* Local variables */
integer i__, j, nb, jj;
real eps, vll, vuu, tmp1;
integer indd, inde;
real anrm;
integer imax;
real rmin, rmax;
logical test;
integer itmp1, indee;
real sigma;
integer iinfo;
char order[1];
logical lower;
logical wantz, alleig, indeig;
integer iscale, indibl;
logical valeig;
real safmin;
real abstll, bignum;
integer indtau, indisp, indiwo, indwkn;
integer indwrk, lwkmin;
integer llwrkn, llwork, nsplit;
real smlnum;
integer lwkopt;
logical lquery;
/* -- LAPACK driver routine (version 3.2) -- */
/* November 2006 */
/* Purpose */
/* ======= */
/* SSYEVX computes selected eigenvalues and, optionally, eigenvectors */
/* of a real symmetric matrix A. Eigenvalues and eigenvectors can be */
/* selected by specifying either a range of values or a range of indices */
/* for the desired eigenvalues. */
/* Arguments */
/* ========= */
/* JOBZ (input) CHARACTER*1 */
/* = 'N': Compute eigenvalues only; */
/* = 'V': Compute eigenvalues and eigenvectors. */
/* RANGE (input) CHARACTER*1 */
/* = 'A': all eigenvalues will be found. */
/* = 'V': all eigenvalues in the half-open interval (VL,VU] */
/* will be found. */
/* = 'I': the IL-th through IU-th eigenvalues will be found. */
/* UPLO (input) CHARACTER*1 */
/* = 'U': Upper triangle of A is stored; */
/* = 'L': Lower triangle of A is stored. */
/* N (input) INTEGER */
/* The order of the matrix A. N >= 0. */
/* A (input/output) REAL array, dimension (LDA, N) */
/* On entry, the symmetric matrix A. If UPLO = 'U', the */
/* leading N-by-N upper triangular part of A contains the */
/* upper triangular part of the matrix A. If UPLO = 'L', */
/* the leading N-by-N lower triangular part of A contains */
/* the lower triangular part of the matrix A. */
/* On exit, the lower triangle (if UPLO='L') or the upper */
/* triangle (if UPLO='U') of A, including the diagonal, is */
/* destroyed. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,N). */
/* VL (input) REAL */
/* VU (input) REAL */
/* If RANGE='V', the lower and upper bounds of the interval to */
/* be searched for eigenvalues. VL < VU. */
/* Not referenced if RANGE = 'A' or 'I'. */
/* IL (input) INTEGER */
/* IU (input) INTEGER */
/* If RANGE='I', the indices (in ascending order) of the */
/* smallest and largest eigenvalues to be returned. */
/* 1 <= IL <= IU <= N, if N > 0; IL = 1 and IU = 0 if N = 0. */
/* Not referenced if RANGE = 'A' or 'V'. */
/* ABSTOL (input) REAL */
/* The absolute error tolerance for the eigenvalues. */
/* An approximate eigenvalue is accepted as converged */
/* when it is determined to lie in an interval [a,b] */
/* of width less than or equal to */
/* ABSTOL + EPS * max( |a|,|b| ) , */
/* where EPS is the machine precision. If ABSTOL is less than */
/* or equal to zero, then EPS*|T| will be used in its place, */
/* where |T| is the 1-norm of the tridiagonal matrix obtained */
/* by reducing A to tridiagonal form. */
/* Eigenvalues will be computed most accurately when ABSTOL is */
/* set to twice the underflow threshold 2*SLAMCH('S'), not zero. */
/* If this routine returns with INFO>0, indicating that some */
/* eigenvectors did not converge, try setting ABSTOL to */
/* 2*SLAMCH('S'). */
/* See "Computing Small Singular Values of Bidiagonal Matrices */
/* with Guaranteed High Relative Accuracy," by Demmel and */
/* Kahan, LAPACK Working Note #3. */
/* M (output) INTEGER */
/* The total number of eigenvalues found. 0 <= M <= N. */
/* If RANGE = 'A', M = N, and if RANGE = 'I', M = IU-IL+1. */
/* W (output) REAL array, dimension (N) */
/* On normal exit, the first M elements contain the selected */
/* eigenvalues in ascending order. */
/* Z (output) REAL array, dimension (LDZ, max(1,M)) */
/* If JOBZ = 'V', then if INFO = 0, the first M columns of Z */
/* contain the orthonormal eigenvectors of the matrix A */
/* corresponding to the selected eigenvalues, with the i-th */
/* column of Z holding the eigenvector associated with W(i). */
/* If an eigenvector fails to converge, then that column of Z */
/* contains the latest approximation to the eigenvector, and the */
/* index of the eigenvector is returned in IFAIL. */
/* If JOBZ = 'N', then Z is not referenced. */
/* Note: the user must ensure that at least max(1,M) columns are */
/* supplied in the array Z; if RANGE = 'V', the exact value of M */
/* is not known in advance and an upper bound must be used. */
/* LDZ (input) INTEGER */
/* The leading dimension of the array Z. LDZ >= 1, and if */
/* JOBZ = 'V', LDZ >= max(1,N). */
/* WORK (workspace/output) REAL array, dimension (MAX(1,LWORK)) */
/* On exit, if INFO = 0, WORK(1) returns the optimal LWORK. */
/* LWORK (input) INTEGER */
/* The length of the array WORK. LWORK >= 1, when N <= 1; */
/* otherwise 8*N. */
/* For optimal efficiency, LWORK >= (NB+3)*N, */
/* where NB is the max of the blocksize for SSYTRD and SORMTR */
/* returned by ILAENV. */
/* 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. */
/* IWORK (workspace) INTEGER array, dimension (5*N) */
/* IFAIL (output) INTEGER array, dimension (N) */
/* If JOBZ = 'V', then if INFO = 0, the first M elements of */
/* IFAIL are zero. If INFO > 0, then IFAIL contains the */
/* indices of the eigenvectors that failed to converge. */
/* If JOBZ = 'N', then IFAIL is not referenced. */
/* INFO (output) INTEGER */
/* = 0: successful exit */
/* < 0: if INFO = -i, the i-th argument had an illegal value */
/* > 0: if INFO = i, then i eigenvectors failed to converge. */
/* Their indices are stored in array IFAIL. */
/* ===================================================================== */
/* Test the input parameters. */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
--w;
z_dim1 = *ldz;
z_offset = 1 + z_dim1;
z__ -= z_offset;
--work;
--iwork;
--ifail;
/* Function Body */
lower = lsame_(uplo, "L");
wantz = lsame_(jobz, "V");
alleig = lsame_(range, "A");
valeig = lsame_(range, "V");
indeig = lsame_(range, "I");
lquery = *lwork == -1;
*info = 0;
if (! (wantz || lsame_(jobz, "N"))) {
*info = -1;
} else if (! (alleig || valeig || indeig)) {
*info = -2;
} else if (! (lower || lsame_(uplo, "U"))) {
*info = -3;
} else if (*n < 0) {
*info = -4;
} else if (*lda < max(1,*n)) {
*info = -6;
} else {
if (valeig) {
if (*n > 0 && *vu <= *vl) {
*info = -8;
}
} else if (indeig) {
if (*il < 1 || *il > max(1,*n)) {
*info = -9;
} else if (*iu < min(*n,*il) || *iu > *n) {
*info = -10;
}
}
}
if (*info == 0) {
if (*ldz < 1 || wantz && *ldz < *n) {
*info = -15;
}
}
if (*info == 0) {
if (*n <= 1) {
lwkmin = 1;
work[1] = (real) lwkmin;
} else {
lwkmin = *n << 3;
nb = ilaenv_(&c__1, "SSYTRD", uplo, n, &c_n1, &c_n1, &c_n1);
/* Computing MAX */
i__1 = nb, i__2 = ilaenv_(&c__1, "SORMTR", uplo, n, &c_n1, &c_n1,
&c_n1);
nb = max(i__1,i__2);
/* Computing MAX */
i__1 = lwkmin, i__2 = (nb + 3) * *n;
lwkopt = max(i__1,i__2);
work[1] = (real) lwkopt;
}
if (*lwork < lwkmin && ! lquery) {
*info = -17;
}
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SSYEVX", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
*m = 0;
if (*n == 0) {
return 0;
}
if (*n == 1) {
if (alleig || indeig) {
*m = 1;
w[1] = a[a_dim1 + 1];
} else {
if (*vl < a[a_dim1 + 1] && *vu >= a[a_dim1 + 1]) {
*m = 1;
w[1] = a[a_dim1 + 1];
}
}
if (wantz) {
z__[z_dim1 + 1] = 1.f;
}
return 0;
}
/* Get machine constants. */
safmin = slamch_("Safe minimum");
eps = slamch_("Precision");
smlnum = safmin / eps;
bignum = 1.f / smlnum;
rmin = sqrt(smlnum);
/* Computing MIN */
r__1 = sqrt(bignum), r__2 = 1.f / sqrt(sqrt(safmin));
rmax = dmin(r__1,r__2);
/* Scale matrix to allowable range, if necessary. */
iscale = 0;
abstll = *abstol;
if (valeig) {
vll = *vl;
vuu = *vu;
}
anrm = slansy_("M", uplo, n, &a[a_offset], lda, &work[1]);
if (anrm > 0.f && anrm < rmin) {
iscale = 1;
sigma = rmin / anrm;
} else if (anrm > rmax) {
iscale = 1;
sigma = rmax / anrm;
}
if (iscale == 1) {
if (lower) {
i__1 = *n;
for (j = 1; j <= i__1; ++j) {
i__2 = *n - j + 1;
sscal_(&i__2, &sigma, &a[j + j * a_dim1], &c__1);
}
} else {
i__1 = *n;
for (j = 1; j <= i__1; ++j) {
sscal_(&j, &sigma, &a[j * a_dim1 + 1], &c__1);
}
}
if (*abstol > 0.f) {
abstll = *abstol * sigma;
}
if (valeig) {
vll = *vl * sigma;
vuu = *vu * sigma;
}
}
/* Call SSYTRD to reduce symmetric matrix to tridiagonal form. */
indtau = 1;
inde = indtau + *n;
indd = inde + *n;
indwrk = indd + *n;
llwork = *lwork - indwrk + 1;
ssytrd_(uplo, n, &a[a_offset], lda, &work[indd], &work[inde], &work[
indtau], &work[indwrk], &llwork, &iinfo);
/* If all eigenvalues are desired and ABSTOL is less than or equal to */
/* zero, then call SSTERF or SORGTR and SSTEQR. If this fails for */
/* some eigenvalue, then try SSTEBZ. */
test = FALSE_;
if (indeig) {
if (*il == 1 && *iu == *n) {
test = TRUE_;
}
}
if ((alleig || test) && *abstol <= 0.f) {
scopy_(n, &work[indd], &c__1, &w[1], &c__1);
indee = indwrk + (*n << 1);
if (! wantz) {
i__1 = *n - 1;
scopy_(&i__1, &work[inde], &c__1, &work[indee], &c__1);
ssterf_(n, &w[1], &work[indee], info);
} else {
slacpy_("A", n, n, &a[a_offset], lda, &z__[z_offset], ldz);
sorgtr_(uplo, n, &z__[z_offset], ldz, &work[indtau], &work[indwrk]
, &llwork, &iinfo);
i__1 = *n - 1;
scopy_(&i__1, &work[inde], &c__1, &work[indee], &c__1);
ssteqr_(jobz, n, &w[1], &work[indee], &z__[z_offset], ldz, &work[
indwrk], info);
if (*info == 0) {
i__1 = *n;
for (i__ = 1; i__ <= i__1; ++i__) {
ifail[i__] = 0;
}
}
}
if (*info == 0) {
*m = *n;
goto L40;
}
*info = 0;
}
/* Otherwise, call SSTEBZ and, if eigenvectors are desired, SSTEIN. */
if (wantz) {
*(unsigned char *)order = 'B';
} else {
*(unsigned char *)order = 'E';
}
indibl = 1;
indisp = indibl + *n;
indiwo = indisp + *n;
sstebz_(range, order, n, &vll, &vuu, il, iu, &abstll, &work[indd], &work[
inde], m, &nsplit, &w[1], &iwork[indibl], &iwork[indisp], &work[
indwrk], &iwork[indiwo], info);
if (wantz) {
sstein_(n, &work[indd], &work[inde], m, &w[1], &iwork[indibl], &iwork[
indisp], &z__[z_offset], ldz, &work[indwrk], &iwork[indiwo], &
ifail[1], info);
/* Apply orthogonal matrix used in reduction to tridiagonal */
/* form to eigenvectors returned by SSTEIN. */
indwkn = inde;
llwrkn = *lwork - indwkn + 1;
sormtr_("L", uplo, "N", n, m, &a[a_offset], lda, &work[indtau], &z__[
z_offset], ldz, &work[indwkn], &llwrkn, &iinfo);
}
/* If matrix was scaled, then rescale eigenvalues appropriately. */
L40:
if (iscale == 1) {
if (*info == 0) {
imax = *m;
} else {
imax = *info - 1;
}
r__1 = 1.f / sigma;
sscal_(&imax, &r__1, &w[1], &c__1);
}
/* If eigenvalues are not in order, then sort them, along with */
/* eigenvectors. */
if (wantz) {
i__1 = *m - 1;
for (j = 1; j <= i__1; ++j) {
i__ = 0;
tmp1 = w[j];
i__2 = *m;
for (jj = j + 1; jj <= i__2; ++jj) {
if (w[jj] < tmp1) {
i__ = jj;
tmp1 = w[jj];
}
}
if (i__ != 0) {
itmp1 = iwork[indibl + i__ - 1];
w[i__] = w[j];
iwork[indibl + i__ - 1] = iwork[indibl + j - 1];
w[j] = tmp1;
iwork[indibl + j - 1] = itmp1;
sswap_(n, &z__[i__ * z_dim1 + 1], &c__1, &z__[j * z_dim1 + 1],
&c__1);
if (*info != 0) {
itmp1 = ifail[i__];
ifail[i__] = ifail[j];
ifail[j] = itmp1;
}
}
}
}
/* Set WORK(1) to optimal workspace size. */
work[1] = (real) lwkopt;
return 0;
/* End of SSYEVX */
} /* ssyevx_ */
/* Subroutine */ int ssyevr_(char *jobz, char *range, char *uplo, integer *n,
real *a, integer *lda, real *vl, real *vu, integer *il, integer *iu,
real *abstol, integer *m, real *w, real *z__, integer *ldz, integer *
isuppz, real *work, integer *lwork, integer *iwork, integer *liwork,
integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, z_dim1, z_offset, i__1, i__2;
real r__1, r__2;
/* Builtin functions */
double sqrt(doublereal);
/* Local variables */
integer i__, j, nb, jj;
real eps, vll, vuu, tmp1;
integer indd, inde;
real anrm;
integer imax;
real rmin, rmax;
logical test;
integer inddd, indee;
real sigma;
extern logical lsame_(char *, char *);
integer iinfo;
extern /* Subroutine */ int sscal_(integer *, real *, real *, integer *);
char order[1];
integer indwk, lwmin;
logical lower;
extern /* Subroutine */ int scopy_(integer *, real *, integer *, real *,
integer *), sswap_(integer *, real *, integer *, real *, integer *
);
logical wantz, alleig, indeig;
integer iscale, ieeeok, indibl, indifl;
logical valeig;
extern doublereal slamch_(char *);
real safmin;
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *);
extern /* Subroutine */ int xerbla_(char *, integer *);
real abstll, bignum;
integer indtau, indisp, indiwo, indwkn, liwmin;
logical tryrac;
extern /* Subroutine */ int sstein_(integer *, real *, real *, integer *,
real *, integer *, integer *, real *, integer *, real *, integer *
, integer *, integer *), ssterf_(integer *, real *, real *,
integer *);
integer llwrkn, llwork, nsplit;
real smlnum;
extern doublereal slansy_(char *, char *, integer *, real *, integer *,
real *);
extern /* Subroutine */ int sstebz_(char *, char *, integer *, real *,
real *, integer *, integer *, real *, real *, real *, integer *,
integer *, real *, integer *, integer *, real *, integer *,
integer *), sstemr_(char *, char *, integer *,
real *, real *, real *, real *, integer *, integer *, integer *,
real *, real *, integer *, integer *, integer *, logical *, real *
, integer *, integer *, integer *, integer *);
integer lwkopt;
logical lquery;
extern /* Subroutine */ int sormtr_(char *, char *, char *, integer *,
integer *, real *, integer *, real *, real *, integer *, real *,
integer *, integer *), ssytrd_(char *,
integer *, real *, integer *, real *, real *, real *, real *,
integer *, integer *);
/* -- LAPACK driver routine (version 3.1) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* SSYEVR computes selected eigenvalues and, optionally, eigenvectors */
/* of a real symmetric matrix A. Eigenvalues and eigenvectors can be */
/* selected by specifying either a range of values or a range of */
/* indices for the desired eigenvalues. */
/* SSYEVR first reduces the matrix A to tridiagonal form T with a call */
/* to SSYTRD. Then, whenever possible, SSYEVR calls SSTEMR to compute */
/* the eigenspectrum using Relatively Robust Representations. SSTEMR */
/* computes eigenvalues by the dqds algorithm, while orthogonal */
/* eigenvectors are computed from various "good" L D L^T representations */
/* (also known as Relatively Robust Representations). Gram-Schmidt */
/* orthogonalization is avoided as far as possible. More specifically, */
/* the various steps of the algorithm are as follows. */
/* For each unreduced block (submatrix) of T, */
/* (a) Compute T - sigma I = L D L^T, so that L and D */
/* define all the wanted eigenvalues to high relative accuracy. */
/* This means that small relative changes in the entries of D and L */
/* cause only small relative changes in the eigenvalues and */
/* eigenvectors. The standard (unfactored) representation of the */
/* tridiagonal matrix T does not have this property in general. */
/* (b) Compute the eigenvalues to suitable accuracy. */
/* If the eigenvectors are desired, the algorithm attains full */
/* accuracy of the computed eigenvalues only right before */
/* the corresponding vectors have to be computed, see steps c) and d). */
/* (c) For each cluster of close eigenvalues, select a new */
/* shift close to the cluster, find a new factorization, and refine */
/* the shifted eigenvalues to suitable accuracy. */
/* (d) For each eigenvalue with a large enough relative separation compute */
/* the corresponding eigenvector by forming a rank revealing twisted */
/* factorization. Go back to (c) for any clusters that remain. */
/* The desired accuracy of the output can be specified by the input */
/* parameter ABSTOL. */
/* For more details, see SSTEMR's documentation and: */
/* - Inderjit S. Dhillon and Beresford N. Parlett: "Multiple representations */
/* to compute orthogonal eigenvectors of symmetric tridiagonal matrices," */
/* Linear Algebra and its Applications, 387(1), pp. 1-28, August 2004. */
/* - Inderjit Dhillon and Beresford Parlett: "Orthogonal Eigenvectors and */
/* Relative Gaps," SIAM Journal on Matrix Analysis and Applications, Vol. 25, */
/* 2004. Also LAPACK Working Note 154. */
/* - Inderjit Dhillon: "A new O(n^2) algorithm for the symmetric */
/* tridiagonal eigenvalue/eigenvector problem", */
/* Computer Science Division Technical Report No. UCB/CSD-97-971, */
/* UC Berkeley, May 1997. */
/* Note 1 : SSYEVR calls SSTEMR when the full spectrum is requested */
/* on machines which conform to the ieee-754 floating point standard. */
/* SSYEVR calls SSTEBZ and SSTEIN on non-ieee machines and */
/* when partial spectrum requests are made. */
/* Normal execution of SSTEMR may create NaNs and infinities and */
/* hence may abort due to a floating point exception in environments */
/* which do not handle NaNs and infinities in the ieee standard default */
/* manner. */
/* Arguments */
/* ========= */
/* JOBZ (input) CHARACTER*1 */
/* = 'N': Compute eigenvalues only; */
/* = 'V': Compute eigenvalues and eigenvectors. */
/* RANGE (input) CHARACTER*1 */
/* = 'A': all eigenvalues will be found. */
/* = 'V': all eigenvalues in the half-open interval (VL,VU] */
/* will be found. */
/* = 'I': the IL-th through IU-th eigenvalues will be found. */
/* ********* For RANGE = 'V' or 'I' and IU - IL < N - 1, SSTEBZ and */
/* ********* SSTEIN are called */
/* UPLO (input) CHARACTER*1 */
/* = 'U': Upper triangle of A is stored; */
/* = 'L': Lower triangle of A is stored. */
/* N (input) INTEGER */
/* The order of the matrix A. N >= 0. */
/* A (input/output) REAL array, dimension (LDA, N) */
/* On entry, the symmetric matrix A. If UPLO = 'U', the */
/* leading N-by-N upper triangular part of A contains the */
/* upper triangular part of the matrix A. If UPLO = 'L', */
/* the leading N-by-N lower triangular part of A contains */
/* the lower triangular part of the matrix A. */
/* On exit, the lower triangle (if UPLO='L') or the upper */
/* triangle (if UPLO='U') of A, including the diagonal, is */
/* destroyed. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,N). */
/* VL (input) REAL */
/* VU (input) REAL */
/* If RANGE='V', the lower and upper bounds of the interval to */
/* be searched for eigenvalues. VL < VU. */
/* Not referenced if RANGE = 'A' or 'I'. */
/* IL (input) INTEGER */
/* IU (input) INTEGER */
/* If RANGE='I', the indices (in ascending order) of the */
/* smallest and largest eigenvalues to be returned. */
/* 1 <= IL <= IU <= N, if N > 0; IL = 1 and IU = 0 if N = 0. */
/* Not referenced if RANGE = 'A' or 'V'. */
/* ABSTOL (input) REAL */
/* The absolute error tolerance for the eigenvalues. */
/* An approximate eigenvalue is accepted as converged */
/* when it is determined to lie in an interval [a,b] */
/* of width less than or equal to */
/* ABSTOL + EPS * max( |a|,|b| ) , */
/* where EPS is the machine precision. If ABSTOL is less than */
/* or equal to zero, then EPS*|T| will be used in its place, */
/* where |T| is the 1-norm of the tridiagonal matrix obtained */
/* by reducing A to tridiagonal form. */
/* See "Computing Small Singular Values of Bidiagonal Matrices */
/* with Guaranteed High Relative Accuracy," by Demmel and */
/* Kahan, LAPACK Working Note #3. */
/* If high relative accuracy is important, set ABSTOL to */
/* SLAMCH( 'Safe minimum' ). Doing so will guarantee that */
/* eigenvalues are computed to high relative accuracy when */
/* possible in future releases. The current code does not */
/* make any guarantees about high relative accuracy, but */
/* future releases will. See J. Barlow and J. Demmel, */
/* "Computing Accurate Eigensystems of Scaled Diagonally */
/* Dominant Matrices", LAPACK Working Note #7, for a discussion */
/* of which matrices define their eigenvalues to high relative */
/* accuracy. */
/* M (output) INTEGER */
/* The total number of eigenvalues found. 0 <= M <= N. */
/* If RANGE = 'A', M = N, and if RANGE = 'I', M = IU-IL+1. */
/* W (output) REAL array, dimension (N) */
/* The first M elements contain the selected eigenvalues in */
/* ascending order. */
/* Z (output) REAL array, dimension (LDZ, max(1,M)) */
/* If JOBZ = 'V', then if INFO = 0, the first M columns of Z */
/* contain the orthonormal eigenvectors of the matrix A */
/* corresponding to the selected eigenvalues, with the i-th */
/* column of Z holding the eigenvector associated with W(i). */
/* If JOBZ = 'N', then Z is not referenced. */
/* Note: the user must ensure that at least max(1,M) columns are */
/* supplied in the array Z; if RANGE = 'V', the exact value of M */
/* is not known in advance and an upper bound must be used. */
/* Supplying N columns is always safe. */
/* LDZ (input) INTEGER */
/* The leading dimension of the array Z. LDZ >= 1, and if */
/* JOBZ = 'V', LDZ >= max(1,N). */
/* ISUPPZ (output) INTEGER array, dimension ( 2*max(1,M) ) */
/* The support of the eigenvectors in Z, i.e., the indices */
/* indicating the nonzero elements in Z. The i-th eigenvector */
/* is nonzero only in elements ISUPPZ( 2*i-1 ) through */
/* ISUPPZ( 2*i ). */
/* ********* Implemented only for RANGE = 'A' or 'I' and IU - IL = N - 1 */
/* WORK (workspace/output) REAL array, dimension (MAX(1,LWORK)) */
/* On exit, if INFO = 0, WORK(1) returns the optimal LWORK. */
/* LWORK (input) INTEGER */
/* The dimension of the array WORK. LWORK >= max(1,26*N). */
/* For optimal efficiency, LWORK >= (NB+6)*N, */
/* where NB is the max of the blocksize for SSYTRD and SORMTR */
/* returned by ILAENV. */
/* If LWORK = -1, then a workspace query is assumed; the routine */
/* only calculates the optimal sizes of the WORK and IWORK */
/* arrays, returns these values as the first entries of the WORK */
/* and IWORK arrays, and no error message related to LWORK or */
/* LIWORK is issued by XERBLA. */
/* IWORK (workspace/output) INTEGER array, dimension (MAX(1,LIWORK)) */
/* On exit, if INFO = 0, IWORK(1) returns the optimal LWORK. */
/* LIWORK (input) INTEGER */
/* The dimension of the array IWORK. LIWORK >= max(1,10*N). */
/* If LIWORK = -1, then a workspace query is assumed; the */
/* routine only calculates the optimal sizes of the WORK and */
/* IWORK arrays, returns these values as the first entries of */
/* the WORK and IWORK arrays, and no error message related to */
/* LWORK or LIWORK is issued by XERBLA. */
/* INFO (output) INTEGER */
/* = 0: successful exit */
/* < 0: if INFO = -i, the i-th argument had an illegal value */
/* > 0: Internal error */
/* Further Details */
/* =============== */
/* Based on contributions by */
/* Inderjit Dhillon, IBM Almaden, USA */
/* Osni Marques, LBNL/NERSC, USA */
/* Ken Stanley, Computer Science Division, University of */
/* California at Berkeley, USA */
/* Jason Riedy, Computer Science Division, University of */
/* California at Berkeley, USA */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Test the input parameters. */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
--w;
z_dim1 = *ldz;
z_offset = 1 + z_dim1;
z__ -= z_offset;
--isuppz;
--work;
--iwork;
/* Function Body */
ieeeok = ilaenv_(&c__10, "SSYEVR", "N", &c__1, &c__2, &c__3, &c__4);
lower = lsame_(uplo, "L");
wantz = lsame_(jobz, "V");
alleig = lsame_(range, "A");
valeig = lsame_(range, "V");
indeig = lsame_(range, "I");
lquery = *lwork == -1 || *liwork == -1;
/* Computing MAX */
i__1 = 1, i__2 = *n * 26;
lwmin = max(i__1,i__2);
/* Computing MAX */
i__1 = 1, i__2 = *n * 10;
liwmin = max(i__1,i__2);
*info = 0;
if (! (wantz || lsame_(jobz, "N"))) {
*info = -1;
} else if (! (alleig || valeig || indeig)) {
*info = -2;
} else if (! (lower || lsame_(uplo, "U"))) {
*info = -3;
} else if (*n < 0) {
*info = -4;
} else if (*lda < max(1,*n)) {
*info = -6;
} else {
if (valeig) {
if (*n > 0 && *vu <= *vl) {
*info = -8;
}
} else if (indeig) {
if (*il < 1 || *il > max(1,*n)) {
*info = -9;
} else if (*iu < min(*n,*il) || *iu > *n) {
*info = -10;
}
}
}
if (*info == 0) {
if (*ldz < 1 || wantz && *ldz < *n) {
*info = -15;
}
}
if (*info == 0) {
nb = ilaenv_(&c__1, "SSYTRD", uplo, n, &c_n1, &c_n1, &c_n1);
/* Computing MAX */
i__1 = nb, i__2 = ilaenv_(&c__1, "SORMTR", uplo, n, &c_n1, &c_n1, &
c_n1);
nb = max(i__1,i__2);
/* Computing MAX */
i__1 = (nb + 1) * *n;
lwkopt = max(i__1,lwmin);
work[1] = (real) lwkopt;
iwork[1] = liwmin;
if (*lwork < lwmin && ! lquery) {
*info = -18;
} else if (*liwork < liwmin && ! lquery) {
*info = -20;
}
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SSYEVR", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
*m = 0;
if (*n == 0) {
work[1] = 1.f;
return 0;
}
if (*n == 1) {
work[1] = 26.f;
if (alleig || indeig) {
*m = 1;
w[1] = a[a_dim1 + 1];
} else {
if (*vl < a[a_dim1 + 1] && *vu >= a[a_dim1 + 1]) {
*m = 1;
w[1] = a[a_dim1 + 1];
}
}
if (wantz) {
z__[z_dim1 + 1] = 1.f;
}
return 0;
}
/* Get machine constants. */
safmin = slamch_("Safe minimum");
eps = slamch_("Precision");
smlnum = safmin / eps;
bignum = 1.f / smlnum;
rmin = sqrt(smlnum);
/* Computing MIN */
r__1 = sqrt(bignum), r__2 = 1.f / sqrt(sqrt(safmin));
rmax = dmin(r__1,r__2);
/* Scale matrix to allowable range, if necessary. */
iscale = 0;
abstll = *abstol;
if (valeig) {
vll = *vl;
vuu = *vu;
}
anrm = slansy_("M", uplo, n, &a[a_offset], lda, &work[1]);
if (anrm > 0.f && anrm < rmin) {
iscale = 1;
sigma = rmin / anrm;
} else if (anrm > rmax) {
iscale = 1;
sigma = rmax / anrm;
}
if (iscale == 1) {
if (lower) {
i__1 = *n;
for (j = 1; j <= i__1; ++j) {
i__2 = *n - j + 1;
sscal_(&i__2, &sigma, &a[j + j * a_dim1], &c__1);
/* L10: */
}
} else {
i__1 = *n;
for (j = 1; j <= i__1; ++j) {
sscal_(&j, &sigma, &a[j * a_dim1 + 1], &c__1);
/* L20: */
}
}
if (*abstol > 0.f) {
abstll = *abstol * sigma;
}
if (valeig) {
vll = *vl * sigma;
vuu = *vu * sigma;
}
}
/* Initialize indices into workspaces. Note: The IWORK indices are */
/* used only if SSTERF or SSTEMR fail. */
/* WORK(INDTAU:INDTAU+N-1) stores the scalar factors of the */
/* elementary reflectors used in SSYTRD. */
indtau = 1;
/* WORK(INDD:INDD+N-1) stores the tridiagonal's diagonal entries. */
indd = indtau + *n;
/* WORK(INDE:INDE+N-1) stores the off-diagonal entries of the */
/* tridiagonal matrix from SSYTRD. */
inde = indd + *n;
/* WORK(INDDD:INDDD+N-1) is a copy of the diagonal entries over */
/* -written by SSTEMR (the SSTERF path copies the diagonal to W). */
inddd = inde + *n;
/* WORK(INDEE:INDEE+N-1) is a copy of the off-diagonal entries over */
/* -written while computing the eigenvalues in SSTERF and SSTEMR. */
indee = inddd + *n;
/* INDWK is the starting offset of the left-over workspace, and */
/* LLWORK is the remaining workspace size. */
indwk = indee + *n;
llwork = *lwork - indwk + 1;
/* IWORK(INDIBL:INDIBL+M-1) corresponds to IBLOCK in SSTEBZ and */
/* stores the block indices of each of the M<=N eigenvalues. */
indibl = 1;
/* IWORK(INDISP:INDISP+NSPLIT-1) corresponds to ISPLIT in SSTEBZ and */
/* stores the starting and finishing indices of each block. */
indisp = indibl + *n;
/* IWORK(INDIFL:INDIFL+N-1) stores the indices of eigenvectors */
/* that corresponding to eigenvectors that fail to converge in */
/* SSTEIN. This information is discarded; if any fail, the driver */
/* returns INFO > 0. */
indifl = indisp + *n;
/* INDIWO is the offset of the remaining integer workspace. */
indiwo = indisp + *n;
/* Call SSYTRD to reduce symmetric matrix to tridiagonal form. */
ssytrd_(uplo, n, &a[a_offset], lda, &work[indd], &work[inde], &work[
indtau], &work[indwk], &llwork, &iinfo);
/* If all eigenvalues are desired */
/* then call SSTERF or SSTEMR and SORMTR. */
test = FALSE_;
if (indeig) {
if (*il == 1 && *iu == *n) {
test = TRUE_;
}
}
if ((alleig || test) && ieeeok == 1) {
if (! wantz) {
scopy_(n, &work[indd], &c__1, &w[1], &c__1);
i__1 = *n - 1;
scopy_(&i__1, &work[inde], &c__1, &work[indee], &c__1);
ssterf_(n, &w[1], &work[indee], info);
} else {
i__1 = *n - 1;
scopy_(&i__1, &work[inde], &c__1, &work[indee], &c__1);
scopy_(n, &work[indd], &c__1, &work[inddd], &c__1);
if (*abstol <= *n * 2.f * eps) {
tryrac = TRUE_;
} else {
tryrac = FALSE_;
}
sstemr_(jobz, "A", n, &work[inddd], &work[indee], vl, vu, il, iu,
m, &w[1], &z__[z_offset], ldz, n, &isuppz[1], &tryrac, &
work[indwk], lwork, &iwork[1], liwork, info);
/* Apply orthogonal matrix used in reduction to tridiagonal */
/* form to eigenvectors returned by SSTEIN. */
if (wantz && *info == 0) {
indwkn = inde;
llwrkn = *lwork - indwkn + 1;
sormtr_("L", uplo, "N", n, m, &a[a_offset], lda, &work[indtau]
, &z__[z_offset], ldz, &work[indwkn], &llwrkn, &iinfo);
}
}
if (*info == 0) {
/* Everything worked. Skip SSTEBZ/SSTEIN. IWORK(:) are */
/* undefined. */
*m = *n;
goto L30;
}
*info = 0;
}
/* Otherwise, call SSTEBZ and, if eigenvectors are desired, SSTEIN. */
/* Also call SSTEBZ and SSTEIN if SSTEMR fails. */
if (wantz) {
*(unsigned char *)order = 'B';
} else {
*(unsigned char *)order = 'E';
}
sstebz_(range, order, n, &vll, &vuu, il, iu, &abstll, &work[indd], &work[
inde], m, &nsplit, &w[1], &iwork[indibl], &iwork[indisp], &work[
indwk], &iwork[indiwo], info);
if (wantz) {
sstein_(n, &work[indd], &work[inde], m, &w[1], &iwork[indibl], &iwork[
indisp], &z__[z_offset], ldz, &work[indwk], &iwork[indiwo], &
iwork[indifl], info);
/* Apply orthogonal matrix used in reduction to tridiagonal */
/* form to eigenvectors returned by SSTEIN. */
indwkn = inde;
llwrkn = *lwork - indwkn + 1;
sormtr_("L", uplo, "N", n, m, &a[a_offset], lda, &work[indtau], &z__[
z_offset], ldz, &work[indwkn], &llwrkn, &iinfo);
}
/* If matrix was scaled, then rescale eigenvalues appropriately. */
/* Jump here if SSTEMR/SSTEIN succeeded. */
L30:
if (iscale == 1) {
if (*info == 0) {
imax = *m;
} else {
imax = *info - 1;
}
r__1 = 1.f / sigma;
sscal_(&imax, &r__1, &w[1], &c__1);
}
/* If eigenvalues are not in order, then sort them, along with */
/* eigenvectors. Note: We do not sort the IFAIL portion of IWORK. */
/* It may not be initialized (if SSTEMR/SSTEIN succeeded), and we do */
/* not return this detailed information to the user. */
if (wantz) {
i__1 = *m - 1;
for (j = 1; j <= i__1; ++j) {
i__ = 0;
tmp1 = w[j];
i__2 = *m;
for (jj = j + 1; jj <= i__2; ++jj) {
if (w[jj] < tmp1) {
i__ = jj;
tmp1 = w[jj];
}
/* L40: */
}
if (i__ != 0) {
w[i__] = w[j];
w[j] = tmp1;
sswap_(n, &z__[i__ * z_dim1 + 1], &c__1, &z__[j * z_dim1 + 1],
&c__1);
}
/* L50: */
}
}
/* Set WORK(1) to optimal workspace size. */
work[1] = (real) lwkopt;
iwork[1] = liwmin;
return 0;
/* End of SSYEVR */
} /* ssyevr_ */
/* Subroutine */ int stimtd_(char *line, integer *nm, integer *mval, integer *
nn, integer *nval, integer *nnb, integer *nbval, integer *nxval,
integer *nlda, integer *ldaval, real *timmin, real *a, real *b, real *
d__, real *tau, real *work, real *reslts, integer *ldr1, integer *
ldr2, integer *ldr3, integer *nout, ftnlen line_len)
{
/* Initialized data */
static char subnam[6*4] = "SSYTRD" "TRED1 " "SORGTR" "SORMTR";
static char sides[1*2] = "L" "R";
static char transs[1*2] = "N" "T";
static char uplos[1*2] = "U" "L";
static integer iseed[4] = { 0,0,0,1 };
/* Format strings */
static char fmt_9999[] = "(1x,a6,\002 timing run not attempted\002,/)";
static char fmt_9998[] = "(/\002 *** Speed of \002,a6,\002 in megaflops "
"*** \002)";
static char fmt_9997[] = "(5x,\002line \002,i2,\002 with LDA = \002,i5)";
static char fmt_9996[] = "(/5x,a6,\002 with UPLO = '\002,a1,\002'\002,/)";
static char fmt_9995[] = "(/5x,a6,\002 with SIDE = '\002,a1,\002', UPLO "
"= '\002,a1,\002', TRANS = '\002,a1,\002', \002,a1,\002 =\002,i6,"
"/)";
/* System generated locals */
integer reslts_dim1, reslts_dim2, reslts_dim3, reslts_offset, i__1, i__2,
i__3, i__4, i__5, i__6;
/* Builtin functions
Subroutine */ int s_copy(char *, char *, ftnlen, ftnlen);
integer s_wsfe(cilist *), do_fio(integer *, char *, ftnlen), e_wsfe(void),
s_wsle(cilist *), e_wsle(void);
/* Local variables */
static integer ilda;
static char side[1];
static integer info;
static char path[3];
static real time;
static integer isub;
static char uplo[1];
extern /* Subroutine */ int tred1_(integer *, integer *, real *, real *,
real *, real *);
static integer i__, m, n;
static char cname[6];
static integer iside, itoff, itran;
extern doublereal sopla_(char *, integer *, integer *, integer *, integer
*, integer *);
extern /* Subroutine */ int icopy_(integer *, integer *, integer *,
integer *, integer *);
static char trans[1];
static integer iuplo, i3, i4, m1, n1;
static real s1, s2;
static integer ic;
extern /* Subroutine */ int sprtb3_(char *, char *, integer *, integer *,
integer *, integer *, integer *, integer *, real *, integer *,
integer *, integer *, ftnlen, ftnlen);
static integer nb, im, in, lw, nx, reseed[4];
extern /* Subroutine */ int atimck_(integer *, char *, integer *, integer
*, integer *, integer *, integer *, integer *, ftnlen);
extern doublereal second_(void);
extern /* Subroutine */ int atimin_(char *, char *, integer *, char *,
logical *, integer *, integer *, ftnlen, ftnlen, ftnlen), slacpy_(
char *, integer *, integer *, real *, integer *, real *, integer *
), xlaenv_(integer *, integer *);
extern doublereal smflop_(real *, real *, integer *);
static real untime;
extern /* Subroutine */ int stimmg_(integer *, integer *, integer *, real
*, integer *, integer *, integer *);
static logical timsub[4];
extern /* Subroutine */ int slatms_(integer *, integer *, char *, integer
*, char *, real *, integer *, real *, real *, integer *, integer *
, char *, real *, integer *, real *, integer *), sprtbl_(char *, char *, integer *, integer *, integer *,
integer *, integer *, real *, integer *, integer *, integer *,
ftnlen, ftnlen), sorgtr_(char *, integer *, real *, integer *,
real *, real *, integer *, integer *), sormtr_(char *,
char *, char *, integer *, integer *, real *, integer *, real *,
real *, integer *, real *, integer *, integer *), ssytrd_(char *, integer *, real *, integer *, real *,
real *, real *, real *, integer *, integer *);
static integer lda, icl, inb;
static real ops;
static char lab1[1], lab2[1];
/* Fortran I/O blocks */
static cilist io___10 = { 0, 0, 0, fmt_9999, 0 };
static cilist io___11 = { 0, 0, 0, fmt_9999, 0 };
static cilist io___42 = { 0, 0, 0, fmt_9998, 0 };
static cilist io___44 = { 0, 0, 0, fmt_9997, 0 };
static cilist io___45 = { 0, 0, 0, 0, 0 };
static cilist io___46 = { 0, 0, 0, fmt_9996, 0 };
static cilist io___49 = { 0, 0, 0, fmt_9998, 0 };
static cilist io___50 = { 0, 0, 0, fmt_9997, 0 };
static cilist io___51 = { 0, 0, 0, fmt_9995, 0 };
#define subnam_ref(a_0,a_1) &subnam[(a_1)*6 + a_0 - 6]
#define reslts_ref(a_1,a_2,a_3,a_4) reslts[(((a_4)*reslts_dim3 + (a_3))*\
reslts_dim2 + (a_2))*reslts_dim1 + a_1]
/* -- LAPACK timing routine (version 3.0) --
Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd.,
Courant Institute, Argonne National Lab, and Rice University
March 31, 1993
Purpose
=======
STIMTD times the LAPACK routines SSYTRD, SORGTR, and SORMTR and the
EISPACK routine TRED1.
Arguments
=========
LINE (input) CHARACTER*80
The input line that requested this routine. The first six
characters contain either the name of a subroutine or a
generic path name. The remaining characters may be used to
specify the individual routines to be timed. See ATIMIN for
a full description of the format of the input line.
NM (input) INTEGER
The number of values of M contained in the vector MVAL.
MVAL (input) INTEGER array, dimension (NM)
The values of the matrix size M.
NN (input) INTEGER
The number of values of N contained in the vector NVAL.
NVAL (input) INTEGER array, dimension (NN)
The values of the matrix column dimension N.
NNB (input) INTEGER
The number of values of NB and NX contained in the
vectors NBVAL and NXVAL. The blocking parameters are used
in pairs (NB,NX).
NBVAL (input) INTEGER array, dimension (NNB)
The values of the blocksize NB.
NXVAL (input) INTEGER array, dimension (NNB)
The values of the crossover point NX.
NLDA (input) INTEGER
The number of values of LDA contained in the vector LDAVAL.
LDAVAL (input) INTEGER array, dimension (NLDA)
The values of the leading dimension of the array A.
TIMMIN (input) REAL
The minimum time a subroutine will be timed.
A (workspace) REAL array, dimension (LDAMAX*NMAX)
where LDAMAX and NMAX are the maximum values of LDA and N.
B (workspace) REAL array, dimension (LDAMAX*NMAX)
D (workspace) REAL array, dimension (2*NMAX-1)
TAU (workspace) REAL array, dimension (NMAX)
WORK (workspace) REAL array, dimension (NMAX*NBMAX)
where NBMAX is the maximum value of NB.
RESLTS (workspace) REAL array, dimension
(LDR1,LDR2,LDR3,4*NN+3)
The timing results for each subroutine over the relevant
values of M, (NB,NX), LDA, and N.
LDR1 (input) INTEGER
The first dimension of RESLTS. LDR1 >= max(1,NNB).
LDR2 (input) INTEGER
The second dimension of RESLTS. LDR2 >= max(1,NM).
LDR3 (input) INTEGER
The third dimension of RESLTS. LDR3 >= max(1,2*NLDA).
NOUT (input) INTEGER
The unit number for output.
Internal Parameters
===================
MODE INTEGER
The matrix type. MODE = 3 is a geometric distribution of
eigenvalues. See SLATMS for further details.
COND REAL
The condition number of the matrix. The singular values are
set to values from DMAX to DMAX/COND.
DMAX REAL
The magnitude of the largest singular value.
=====================================================================
Parameter adjustments */
--mval;
--nval;
--nbval;
--nxval;
--ldaval;
--a;
--b;
--d__;
--tau;
--work;
reslts_dim1 = *ldr1;
reslts_dim2 = *ldr2;
reslts_dim3 = *ldr3;
reslts_offset = 1 + reslts_dim1 * (1 + reslts_dim2 * (1 + reslts_dim3 * 1)
);
reslts -= reslts_offset;
/* Function Body
Extract the timing request from the input line. */
s_copy(path, "Single precision", (ftnlen)1, (ftnlen)16);
s_copy(path + 1, "TD", (ftnlen)2, (ftnlen)2);
atimin_(path, line, &c__4, subnam, timsub, nout, &info, (ftnlen)3, (
ftnlen)80, (ftnlen)6);
if (info != 0) {
goto L240;
}
/* Check that M <= LDA for the input values. */
s_copy(cname, line, (ftnlen)6, (ftnlen)6);
atimck_(&c__2, cname, nm, &mval[1], nlda, &ldaval[1], nout, &info, (
ftnlen)6);
if (info > 0) {
io___10.ciunit = *nout;
s_wsfe(&io___10);
do_fio(&c__1, cname, (ftnlen)6);
e_wsfe();
goto L240;
}
/* Check that K <= LDA for SORMTR */
if (timsub[3]) {
atimck_(&c__3, cname, nn, &nval[1], nlda, &ldaval[1], nout, &info, (
ftnlen)6);
if (info > 0) {
io___11.ciunit = *nout;
s_wsfe(&io___11);
do_fio(&c__1, subnam_ref(0, 4), (ftnlen)6);
e_wsfe();
timsub[3] = FALSE_;
}
}
/* Do first for UPLO = 'U', then for UPLO = 'L' */
for (iuplo = 1; iuplo <= 2; ++iuplo) {
*(unsigned char *)uplo = *(unsigned char *)&uplos[iuplo - 1];
/* Do for each value of M: */
i__1 = *nm;
for (im = 1; im <= i__1; ++im) {
m = mval[im];
icopy_(&c__4, iseed, &c__1, reseed, &c__1);
/* Do for each value of LDA: */
i__2 = *nlda;
for (ilda = 1; ilda <= i__2; ++ilda) {
lda = ldaval[ilda];
i3 = (iuplo - 1) * *nlda + ilda;
/* Do for each pair of values (NB, NX) in NBVAL and NXVAL. */
i__3 = *nnb;
for (inb = 1; inb <= i__3; ++inb) {
nb = nbval[inb];
xlaenv_(&c__1, &nb);
nx = nxval[inb];
xlaenv_(&c__3, &nx);
/* Computing MAX */
i__4 = 1, i__5 = m * max(1,nb);
lw = max(i__4,i__5);
/* Generate a test matrix of order M. */
icopy_(&c__4, reseed, &c__1, iseed, &c__1);
slatms_(&m, &m, "Uniform", iseed, "Symmetric", &tau[1], &
c__3, &c_b27, &c_b28, &m, &m, "No packing", &b[1],
&lda, &work[1], &info);
if (timsub[1] && inb == 1 && iuplo == 2) {
/* TRED1: Eispack reduction using orthogonal
transformations. */
slacpy_(uplo, &m, &m, &b[1], &lda, &a[1], &lda);
ic = 0;
s1 = second_();
L10:
tred1_(&lda, &m, &a[1], &d__[1], &d__[m + 1], &d__[m
+ 1]);
s2 = second_();
time = s2 - s1;
++ic;
if (time < *timmin) {
slacpy_(uplo, &m, &m, &b[1], &lda, &a[1], &lda);
goto L10;
}
/* Subtract the time used in SLACPY. */
icl = 1;
s1 = second_();
L20:
s2 = second_();
untime = s2 - s1;
++icl;
if (icl <= ic) {
slacpy_(uplo, &m, &m, &b[1], &lda, &a[1], &lda);
goto L20;
}
time = (time - untime) / (real) ic;
ops = sopla_("SSYTRD", &m, &m, &c_n1, &c_n1, &nb);
reslts_ref(inb, im, ilda, 2) = smflop_(&ops, &time, &
info);
}
if (timsub[0]) {
/* SSYTRD: Reduction to tridiagonal form */
slacpy_(uplo, &m, &m, &b[1], &lda, &a[1], &lda);
ic = 0;
s1 = second_();
L30:
ssytrd_(uplo, &m, &a[1], &lda, &d__[1], &d__[m + 1], &
tau[1], &work[1], &lw, &info);
s2 = second_();
time = s2 - s1;
++ic;
if (time < *timmin) {
slacpy_(uplo, &m, &m, &b[1], &lda, &a[1], &lda);
goto L30;
}
/* Subtract the time used in SLACPY. */
icl = 1;
s1 = second_();
L40:
s2 = second_();
untime = s2 - s1;
++icl;
if (icl <= ic) {
slacpy_(uplo, &m, &m, &a[1], &lda, &b[1], &lda);
goto L40;
}
time = (time - untime) / (real) ic;
ops = sopla_("SSYTRD", &m, &m, &c_n1, &c_n1, &nb);
reslts_ref(inb, im, i3, 1) = smflop_(&ops, &time, &
info);
} else {
/* If SSYTRD was not timed, generate a matrix and
factor it using SSYTRD anyway so that the factored
form of the matrix can be used in timing the other
routines. */
slacpy_(uplo, &m, &m, &b[1], &lda, &a[1], &lda);
ssytrd_(uplo, &m, &a[1], &lda, &d__[1], &d__[m + 1], &
tau[1], &work[1], &lw, &info);
}
if (timsub[2]) {
/* SORGTR: Generate the orthogonal matrix Q from the
reduction to Hessenberg form A = Q*H*Q' */
slacpy_(uplo, &m, &m, &a[1], &lda, &b[1], &lda);
ic = 0;
s1 = second_();
L50:
sorgtr_(uplo, &m, &b[1], &lda, &tau[1], &work[1], &lw,
&info);
s2 = second_();
time = s2 - s1;
++ic;
if (time < *timmin) {
slacpy_(uplo, &m, &m, &a[1], &lda, &b[1], &lda);
goto L50;
}
/* Subtract the time used in SLACPY. */
icl = 1;
s1 = second_();
L60:
s2 = second_();
untime = s2 - s1;
++icl;
if (icl <= ic) {
slacpy_(uplo, &m, &m, &a[1], &lda, &b[1], &lda);
goto L60;
}
time = (time - untime) / (real) ic;
/* Op count for SORGTR: same as
SORGQR( N-1, N-1, N-1, ... ) */
i__4 = m - 1;
i__5 = m - 1;
i__6 = m - 1;
ops = sopla_("SORGQR", &i__4, &i__5, &i__6, &c_n1, &
nb);
reslts_ref(inb, im, i3, 3) = smflop_(&ops, &time, &
info);
}
if (timsub[3]) {
/* SORMTR: Multiply by Q stored as a product of
elementary transformations */
i4 = 3;
for (iside = 1; iside <= 2; ++iside) {
*(unsigned char *)side = *(unsigned char *)&sides[
iside - 1];
i__4 = *nn;
for (in = 1; in <= i__4; ++in) {
n = nval[in];
/* Computing MAX */
i__5 = 1, i__6 = max(1,nb) * n;
lw = max(i__5,i__6);
if (iside == 1) {
m1 = m;
n1 = n;
} else {
m1 = n;
n1 = m;
}
itoff = 0;
for (itran = 1; itran <= 2; ++itran) {
*(unsigned char *)trans = *(unsigned char
*)&transs[itran - 1];
stimmg_(&c__0, &m1, &n1, &b[1], &lda, &
c__0, &c__0);
ic = 0;
s1 = second_();
L70:
sormtr_(side, uplo, trans, &m1, &n1, &a[1]
, &lda, &tau[1], &b[1], &lda, &
work[1], &lw, &info);
s2 = second_();
time = s2 - s1;
++ic;
if (time < *timmin) {
stimmg_(&c__0, &m1, &n1, &b[1], &lda,
&c__0, &c__0);
goto L70;
}
/* Subtract the time used in STIMMG. */
icl = 1;
s1 = second_();
L80:
s2 = second_();
untime = s2 - s1;
++icl;
if (icl <= ic) {
stimmg_(&c__0, &m1, &n1, &b[1], &lda,
&c__0, &c__0);
goto L80;
}
time = (time - untime) / (real) ic;
/* Op count for SORMTR, SIDE='L': same as
SORMQR( 'L', TRANS, M-1, N, M-1, ...)
Op count for SORMTR, SIDE='R': same as
SORMQR( 'R', TRANS, M, N-1, N-1, ...) */
if (iside == 1) {
i__5 = m1 - 1;
i__6 = m1 - 1;
ops = sopla_("SORMQR", &i__5, &n1, &
i__6, &c_n1, &nb);
} else {
i__5 = n1 - 1;
i__6 = n1 - 1;
ops = sopla_("SORMQR", &m1, &i__5, &
i__6, &c__1, &nb);
}
reslts_ref(inb, im, i3, i4 + itoff + in) =
smflop_(&ops, &time, &info);
itoff = *nn;
/* L90: */
}
/* L100: */
}
i4 += *nn << 1;
/* L110: */
}
}
/* L120: */
}
/* L130: */
}
/* L140: */
}
/* L150: */
}
/* Print tables of results for SSYTRD, TRED1, and SORGTR */
for (isub = 1; isub <= 3; ++isub) {
if (! timsub[isub - 1]) {
goto L180;
}
io___42.ciunit = *nout;
s_wsfe(&io___42);
do_fio(&c__1, subnam_ref(0, isub), (ftnlen)6);
e_wsfe();
if (*nlda > 1) {
i__1 = *nlda;
for (i__ = 1; i__ <= i__1; ++i__) {
io___44.ciunit = *nout;
s_wsfe(&io___44);
do_fio(&c__1, (char *)&i__, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&ldaval[i__], (ftnlen)sizeof(integer));
e_wsfe();
/* L160: */
}
}
if (isub == 2) {
io___45.ciunit = *nout;
s_wsle(&io___45);
e_wsle();
sprtb3_(" ", "N", &c__1, &nbval[1], &nxval[1], nm, &mval[1], nlda,
&reslts_ref(1, 1, 1, isub), ldr1, ldr2, nout, (ftnlen)1,
(ftnlen)1);
} else {
i3 = 1;
for (iuplo = 1; iuplo <= 2; ++iuplo) {
io___46.ciunit = *nout;
s_wsfe(&io___46);
do_fio(&c__1, subnam_ref(0, isub), (ftnlen)6);
do_fio(&c__1, uplos + (iuplo - 1), (ftnlen)1);
e_wsfe();
sprtb3_("( NB, NX)", "N", nnb, &nbval[1], &nxval[1], nm, &
mval[1], nlda, &reslts_ref(1, 1, i3, isub), ldr1,
ldr2, nout, (ftnlen)11, (ftnlen)1);
i3 += *nlda;
/* L170: */
}
}
L180:
;
}
/* Print tables of results for SORMTR */
isub = 4;
if (timsub[isub - 1]) {
i4 = 3;
for (iside = 1; iside <= 2; ++iside) {
if (iside == 1) {
*(unsigned char *)lab1 = 'M';
*(unsigned char *)lab2 = 'N';
if (*nlda > 1) {
io___49.ciunit = *nout;
s_wsfe(&io___49);
do_fio(&c__1, subnam_ref(0, isub), (ftnlen)6);
e_wsfe();
i__1 = *nlda;
for (i__ = 1; i__ <= i__1; ++i__) {
io___50.ciunit = *nout;
s_wsfe(&io___50);
do_fio(&c__1, (char *)&i__, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&ldaval[i__], (ftnlen)sizeof(
integer));
e_wsfe();
/* L190: */
}
}
} else {
*(unsigned char *)lab1 = 'N';
*(unsigned char *)lab2 = 'M';
}
for (itran = 1; itran <= 2; ++itran) {
i__1 = *nn;
for (in = 1; in <= i__1; ++in) {
i3 = 1;
for (iuplo = 1; iuplo <= 2; ++iuplo) {
io___51.ciunit = *nout;
s_wsfe(&io___51);
do_fio(&c__1, subnam_ref(0, isub), (ftnlen)6);
do_fio(&c__1, sides + (iside - 1), (ftnlen)1);
do_fio(&c__1, uplos + (iuplo - 1), (ftnlen)1);
do_fio(&c__1, transs + (itran - 1), (ftnlen)1);
do_fio(&c__1, lab2, (ftnlen)1);
do_fio(&c__1, (char *)&nval[in], (ftnlen)sizeof(
integer));
e_wsfe();
sprtbl_("NB", lab1, nnb, &nbval[1], nm, &mval[1],
nlda, &reslts_ref(1, 1, i3, i4 + in), ldr1,
ldr2, nout, (ftnlen)2, (ftnlen)1);
i3 += *nlda;
/* L200: */
}
/* L210: */
}
i4 += *nn;
/* L220: */
}
/* L230: */
}
}
L240:
/* Print a table of results for each timed routine. */
return 0;
/* End of STIMTD */
} /* stimtd_ */
/* Subroutine */ int ssyevd_(char *jobz, char *uplo, integer *n, real *a,
integer *lda, real *w, real *work, integer *lwork, integer *iwork,
integer *liwork, integer *info)
{
/* -- LAPACK driver routine (version 3.0) --
Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd.,
Courant Institute, Argonne National Lab, and Rice University
June 30, 1999
Purpose
=======
SSYEVD computes all eigenvalues and, optionally, eigenvectors of a
real symmetric matrix A. If eigenvectors are desired, it uses a
divide and conquer algorithm.
The divide and conquer algorithm makes very mild assumptions about
floating point arithmetic. It will work on machines with a guard
digit in add/subtract, or on those binary machines without guard
digits which subtract like the Cray X-MP, Cray Y-MP, Cray C-90, or
Cray-2. It could conceivably fail on hexadecimal or decimal machines
without guard digits, but we know of none.
Because of large use of BLAS of level 3, SSYEVD needs N**2 more
workspace than SSYEVX.
Arguments
=========
JOBZ (input) CHARACTER*1
= 'N': Compute eigenvalues only;
= 'V': Compute eigenvalues and eigenvectors.
UPLO (input) CHARACTER*1
= 'U': Upper triangle of A is stored;
= 'L': Lower triangle of A is stored.
N (input) INTEGER
The order of the matrix A. N >= 0.
A (input/output) REAL array, dimension (LDA, N)
On entry, the symmetric matrix A. If UPLO = 'U', the
leading N-by-N upper triangular part of A contains the
upper triangular part of the matrix A. If UPLO = 'L',
the leading N-by-N lower triangular part of A contains
the lower triangular part of the matrix A.
On exit, if JOBZ = 'V', then if INFO = 0, A contains the
orthonormal eigenvectors of the matrix A.
If JOBZ = 'N', then on exit the lower triangle (if UPLO='L')
or the upper triangle (if UPLO='U') of A, including the
diagonal, is destroyed.
LDA (input) INTEGER
The leading dimension of the array A. LDA >= max(1,N).
W (output) REAL array, dimension (N)
If INFO = 0, the eigenvalues in ascending order.
WORK (workspace/output) REAL array,
dimension (LWORK)
On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
LWORK (input) INTEGER
The dimension of the array WORK.
If N <= 1, LWORK must be at least 1.
If JOBZ = 'N' and N > 1, LWORK must be at least 2*N+1.
If JOBZ = 'V' and N > 1, LWORK must be at least
1 + 6*N + 2*N**2.
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.
IWORK (workspace/output) INTEGER array, dimension (LIWORK)
On exit, if INFO = 0, IWORK(1) returns the optimal LIWORK.
LIWORK (input) INTEGER
The dimension of the array IWORK.
If N <= 1, LIWORK must be at least 1.
If JOBZ = 'N' and N > 1, LIWORK must be at least 1.
If JOBZ = 'V' and N > 1, LIWORK must be at least 3 + 5*N.
If LIWORK = -1, then a workspace query is assumed; the
routine only calculates the optimal size of the IWORK array,
returns this value as the first entry of the IWORK array, and
no error message related to LIWORK 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, the algorithm failed to converge; i
off-diagonal elements of an intermediate tridiagonal
form did not converge to zero.
Further Details
===============
Based on contributions by
Jeff Rutter, Computer Science Division, University of California
at Berkeley, USA
Modified by Francoise Tisseur, University of Tennessee.
=====================================================================
Test the input parameters.
Parameter adjustments */
/* Table of constant values */
static integer c__0 = 0;
static real c_b12 = 1.f;
static integer c__1 = 1;
/* System generated locals */
integer a_dim1, a_offset, i__1, i__2, i__3;
real r__1;
/* Builtin functions */
double sqrt(doublereal);
/* Local variables */
static integer inde;
static real anrm, rmin, rmax;
static integer lopt;
static real sigma;
extern logical lsame_(char *, char *);
static integer iinfo;
extern /* Subroutine */ int sscal_(integer *, real *, real *, integer *);
static integer lwmin, liopt;
static logical lower, wantz;
static integer indwk2, llwrk2, iscale;
extern doublereal slamch_(char *);
static real safmin;
extern /* Subroutine */ int xerbla_(char *, integer *);
static real bignum;
extern /* Subroutine */ int slascl_(char *, integer *, integer *, real *,
real *, integer *, integer *, real *, integer *, integer *);
static integer indtau;
extern /* Subroutine */ int sstedc_(char *, integer *, real *, real *,
real *, integer *, real *, integer *, integer *, integer *,
integer *), slacpy_(char *, integer *, integer *, real *,
integer *, real *, integer *);
static integer indwrk, liwmin;
extern /* Subroutine */ int ssterf_(integer *, real *, real *, integer *);
extern doublereal slansy_(char *, char *, integer *, real *, integer *,
real *);
static integer llwork;
static real smlnum;
static logical lquery;
extern /* Subroutine */ int sormtr_(char *, char *, char *, integer *,
integer *, real *, integer *, real *, real *, integer *, real *,
integer *, integer *), ssytrd_(char *,
integer *, real *, integer *, real *, real *, real *, real *,
integer *, integer *);
static real eps;
#define a_ref(a_1,a_2) a[(a_2)*a_dim1 + a_1]
a_dim1 = *lda;
a_offset = 1 + a_dim1 * 1;
a -= a_offset;
--w;
--work;
--iwork;
/* Function Body */
wantz = lsame_(jobz, "V");
lower = lsame_(uplo, "L");
lquery = *lwork == -1 || *liwork == -1;
*info = 0;
if (*n <= 1) {
liwmin = 1;
lwmin = 1;
lopt = lwmin;
liopt = liwmin;
} else {
if (wantz) {
liwmin = *n * 5 + 3;
/* Computing 2nd power */
i__1 = *n;
lwmin = *n * 6 + 1 + (i__1 * i__1 << 1);
} else {
liwmin = 1;
lwmin = (*n << 1) + 1;
}
lopt = lwmin;
liopt = liwmin;
}
if (! (wantz || lsame_(jobz, "N"))) {
*info = -1;
} else if (! (lower || lsame_(uplo, "U"))) {
*info = -2;
} else if (*n < 0) {
*info = -3;
} else if (*lda < max(1,*n)) {
*info = -5;
} else if (*lwork < lwmin && ! lquery) {
*info = -8;
} else if (*liwork < liwmin && ! lquery) {
*info = -10;
}
if (*info == 0) {
work[1] = (real) lopt;
iwork[1] = liopt;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SSYEVD", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
if (*n == 0) {
return 0;
}
if (*n == 1) {
w[1] = a_ref(1, 1);
if (wantz) {
a_ref(1, 1) = 1.f;
}
return 0;
}
/* Get machine constants. */
safmin = slamch_("Safe minimum");
eps = slamch_("Precision");
smlnum = safmin / eps;
bignum = 1.f / smlnum;
rmin = sqrt(smlnum);
rmax = sqrt(bignum);
/* Scale matrix to allowable range, if necessary. */
anrm = slansy_("M", uplo, n, &a[a_offset], lda, &work[1]);
iscale = 0;
if (anrm > 0.f && anrm < rmin) {
iscale = 1;
sigma = rmin / anrm;
} else if (anrm > rmax) {
iscale = 1;
sigma = rmax / anrm;
}
if (iscale == 1) {
slascl_(uplo, &c__0, &c__0, &c_b12, &sigma, n, n, &a[a_offset], lda,
info);
}
/* Call SSYTRD to reduce symmetric matrix to tridiagonal form. */
inde = 1;
indtau = inde + *n;
indwrk = indtau + *n;
llwork = *lwork - indwrk + 1;
indwk2 = indwrk + *n * *n;
llwrk2 = *lwork - indwk2 + 1;
ssytrd_(uplo, n, &a[a_offset], lda, &w[1], &work[inde], &work[indtau], &
work[indwrk], &llwork, &iinfo);
lopt = (*n << 1) + work[indwrk];
/* For eigenvalues only, call SSTERF. For eigenvectors, first call
SSTEDC to generate the eigenvector matrix, WORK(INDWRK), of the
tridiagonal matrix, then call SORMTR to multiply it by the
Householder transformations stored in A. */
if (! wantz) {
ssterf_(n, &w[1], &work[inde], info);
} else {
sstedc_("I", n, &w[1], &work[inde], &work[indwrk], n, &work[indwk2], &
llwrk2, &iwork[1], liwork, info);
sormtr_("L", uplo, "N", n, n, &a[a_offset], lda, &work[indtau], &work[
indwrk], n, &work[indwk2], &llwrk2, &iinfo);
slacpy_("A", n, n, &work[indwrk], n, &a[a_offset], lda);
/* Computing MAX
Computing 2nd power */
i__3 = *n;
i__1 = lopt, i__2 = *n * 6 + 1 + (i__3 * i__3 << 1);
lopt = max(i__1,i__2);
}
/* If matrix was scaled, then rescale eigenvalues appropriately. */
if (iscale == 1) {
r__1 = 1.f / sigma;
sscal_(n, &r__1, &w[1], &c__1);
}
work[1] = (real) lopt;
iwork[1] = liopt;
return 0;
/* End of SSYEVD */
} /* ssyevd_ */
