/* Subroutine */ int sgelsx_(integer *m, integer *n, integer *nrhs, real *a,
integer *lda, real *b, integer *ldb, integer *jpvt, real *rcond,
integer *rank, real *work, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, i__1, i__2;
real r__1;
/* Local variables */
integer i__, j, k;
real c1, c2, s1, s2, t1, t2;
integer mn;
real anrm, bnrm, smin, smax;
integer iascl, ibscl, ismin, ismax;
extern /* Subroutine */ int strsm_(char *, char *, char *, char *,
integer *, integer *, real *, real *, integer *, real *, integer *
), slaic1_(integer *, integer *,
real *, real *, real *, real *, real *, real *, real *), sorm2r_(
char *, char *, integer *, integer *, integer *, real *, integer *
, real *, real *, integer *, real *, integer *),
slabad_(real *, real *);
extern doublereal slamch_(char *), slange_(char *, integer *,
integer *, real *, integer *, real *);
extern /* Subroutine */ int xerbla_(char *, integer *);
real bignum;
extern /* Subroutine */ int slascl_(char *, integer *, integer *, real *,
real *, integer *, integer *, real *, integer *, integer *), sgeqpf_(integer *, integer *, real *, integer *, integer
*, real *, real *, integer *), slaset_(char *, integer *, integer
*, real *, real *, real *, integer *);
real sminpr, smaxpr, smlnum;
extern /* Subroutine */ int slatzm_(char *, integer *, integer *, real *,
integer *, real *, real *, real *, integer *, real *),
stzrqf_(integer *, integer *, real *, integer *, real *, integer *
);
/* -- LAPACK driver routine (version 3.2) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* This routine is deprecated and has been replaced by routine SGELSY. */
/* SGELSX computes the minimum-norm solution to a real linear least */
/* squares problem: */
/* minimize || A * X - B || */
/* using a complete orthogonal factorization of A. A is an M-by-N */
/* matrix which may be rank-deficient. */
/* Several right hand side vectors b and solution vectors x can be */
/* handled in a single call; they are stored as the columns of the */
/* M-by-NRHS right hand side matrix B and the N-by-NRHS solution */
/* matrix X. */
/* The routine first computes a QR factorization with column pivoting: */
/* A * P = Q * [ R11 R12 ] */
/* [ 0 R22 ] */
/* with R11 defined as the largest leading submatrix whose estimated */
/* condition number is less than 1/RCOND. The order of R11, RANK, */
/* is the effective rank of A. */
/* Then, R22 is considered to be negligible, and R12 is annihilated */
/* by orthogonal transformations from the right, arriving at the */
/* complete orthogonal factorization: */
/* A * P = Q * [ T11 0 ] * Z */
/* [ 0 0 ] */
/* The minimum-norm solution is then */
/* X = P * Z' [ inv(T11)*Q1'*B ] */
/* [ 0 ] */
/* where Q1 consists of the first RANK columns of Q. */
/* Arguments */
/* ========= */
/* M (input) INTEGER */
/* The number of rows of the matrix A. M >= 0. */
/* N (input) INTEGER */
/* The number of columns of the matrix A. N >= 0. */
/* NRHS (input) INTEGER */
/* The number of right hand sides, i.e., the number of */
/* columns of matrices B and X. NRHS >= 0. */
/* A (input/output) REAL array, dimension (LDA,N) */
/* On entry, the M-by-N matrix A. */
/* On exit, A has been overwritten by details of its */
/* complete orthogonal factorization. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,M). */
/* B (input/output) REAL array, dimension (LDB,NRHS) */
/* On entry, the M-by-NRHS right hand side matrix B. */
/* On exit, the N-by-NRHS solution matrix X. */
/* If m >= n and RANK = n, the residual sum-of-squares for */
/* the solution in the i-th column is given by the sum of */
/* squares of elements N+1:M in that column. */
/* LDB (input) INTEGER */
/* The leading dimension of the array B. LDB >= max(1,M,N). */
/* JPVT (input/output) INTEGER array, dimension (N) */
/* On entry, if JPVT(i) .ne. 0, the i-th column of A is an */
/* initial column, otherwise it is a free column. Before */
/* the QR factorization of A, all initial columns are */
/* permuted to the leading positions; only the remaining */
/* free columns are moved as a result of column pivoting */
/* during the factorization. */
/* On exit, if JPVT(i) = k, then the i-th column of A*P */
/* was the k-th column of A. */
/* RCOND (input) REAL */
/* RCOND is used to determine the effective rank of A, which */
/* is defined as the order of the largest leading triangular */
/* submatrix R11 in the QR factorization with pivoting of A, */
/* whose estimated condition number < 1/RCOND. */
/* RANK (output) INTEGER */
/* The effective rank of A, i.e., the order of the submatrix */
/* R11. This is the same as the order of the submatrix T11 */
/* in the complete orthogonal factorization of A. */
/* WORK (workspace) REAL array, dimension */
/* (max( min(M,N)+3*N, 2*min(M,N)+NRHS )), */
/* INFO (output) INTEGER */
/* = 0: successful exit */
/* < 0: if INFO = -i, the i-th argument had an illegal value */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
b_dim1 = *ldb;
b_offset = 1 + b_dim1;
b -= b_offset;
--jpvt;
--work;
/* Function Body */
mn = min(*m,*n);
ismin = mn + 1;
ismax = (mn << 1) + 1;
/* Test the input arguments. */
*info = 0;
if (*m < 0) {
*info = -1;
} else if (*n < 0) {
*info = -2;
} else if (*nrhs < 0) {
*info = -3;
} else if (*lda < max(1,*m)) {
*info = -5;
} else /* if(complicated condition) */ {
/* Computing MAX */
i__1 = max(1,*m);
if (*ldb < max(i__1,*n)) {
*info = -7;
}
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SGELSX", &i__1);
return 0;
}
/* Quick return if possible */
/* Computing MIN */
i__1 = min(*m,*n);
if (min(i__1,*nrhs) == 0) {
*rank = 0;
return 0;
}
/* Get machine parameters */
smlnum = slamch_("S") / slamch_("P");
bignum = 1.f / smlnum;
slabad_(&smlnum, &bignum);
/* Scale A, B if max elements outside range [SMLNUM,BIGNUM] */
anrm = slange_("M", m, n, &a[a_offset], lda, &work[1]);
iascl = 0;
if (anrm > 0.f && anrm < smlnum) {
/* Scale matrix norm up to SMLNUM */
slascl_("G", &c__0, &c__0, &anrm, &smlnum, m, n, &a[a_offset], lda,
info);
iascl = 1;
} else if (anrm > bignum) {
/* Scale matrix norm down to BIGNUM */
slascl_("G", &c__0, &c__0, &anrm, &bignum, m, n, &a[a_offset], lda,
info);
iascl = 2;
} else if (anrm == 0.f) {
/* Matrix all zero. Return zero solution. */
i__1 = max(*m,*n);
slaset_("F", &i__1, nrhs, &c_b13, &c_b13, &b[b_offset], ldb);
*rank = 0;
goto L100;
}
bnrm = slange_("M", m, nrhs, &b[b_offset], ldb, &work[1]);
ibscl = 0;
if (bnrm > 0.f && bnrm < smlnum) {
/* Scale matrix norm up to SMLNUM */
slascl_("G", &c__0, &c__0, &bnrm, &smlnum, m, nrhs, &b[b_offset], ldb,
info);
ibscl = 1;
} else if (bnrm > bignum) {
/* Scale matrix norm down to BIGNUM */
slascl_("G", &c__0, &c__0, &bnrm, &bignum, m, nrhs, &b[b_offset], ldb,
info);
ibscl = 2;
}
/* Compute QR factorization with column pivoting of A: */
/* A * P = Q * R */
sgeqpf_(m, n, &a[a_offset], lda, &jpvt[1], &work[1], &work[mn + 1], info);
/* workspace 3*N. Details of Householder rotations stored */
/* in WORK(1:MN). */
/* Determine RANK using incremental condition estimation */
work[ismin] = 1.f;
work[ismax] = 1.f;
smax = (r__1 = a[a_dim1 + 1], dabs(r__1));
smin = smax;
if ((r__1 = a[a_dim1 + 1], dabs(r__1)) == 0.f) {
*rank = 0;
i__1 = max(*m,*n);
slaset_("F", &i__1, nrhs, &c_b13, &c_b13, &b[b_offset], ldb);
goto L100;
} else {
*rank = 1;
}
L10:
if (*rank < mn) {
i__ = *rank + 1;
slaic1_(&c__2, rank, &work[ismin], &smin, &a[i__ * a_dim1 + 1], &a[
i__ + i__ * a_dim1], &sminpr, &s1, &c1);
slaic1_(&c__1, rank, &work[ismax], &smax, &a[i__ * a_dim1 + 1], &a[
i__ + i__ * a_dim1], &smaxpr, &s2, &c2);
if (smaxpr * *rcond <= sminpr) {
i__1 = *rank;
for (i__ = 1; i__ <= i__1; ++i__) {
work[ismin + i__ - 1] = s1 * work[ismin + i__ - 1];
work[ismax + i__ - 1] = s2 * work[ismax + i__ - 1];
/* L20: */
}
work[ismin + *rank] = c1;
work[ismax + *rank] = c2;
smin = sminpr;
smax = smaxpr;
++(*rank);
goto L10;
}
}
/* Logically partition R = [ R11 R12 ] */
/* [ 0 R22 ] */
/* where R11 = R(1:RANK,1:RANK) */
/* [R11,R12] = [ T11, 0 ] * Y */
if (*rank < *n) {
stzrqf_(rank, n, &a[a_offset], lda, &work[mn + 1], info);
}
/* Details of Householder rotations stored in WORK(MN+1:2*MN) */
/* B(1:M,1:NRHS) := Q' * B(1:M,1:NRHS) */
sorm2r_("Left", "Transpose", m, nrhs, &mn, &a[a_offset], lda, &work[1], &
b[b_offset], ldb, &work[(mn << 1) + 1], info);
/* workspace NRHS */
/* B(1:RANK,1:NRHS) := inv(T11) * B(1:RANK,1:NRHS) */
strsm_("Left", "Upper", "No transpose", "Non-unit", rank, nrhs, &c_b36, &
a[a_offset], lda, &b[b_offset], ldb);
i__1 = *n;
for (i__ = *rank + 1; i__ <= i__1; ++i__) {
i__2 = *nrhs;
for (j = 1; j <= i__2; ++j) {
b[i__ + j * b_dim1] = 0.f;
/* L30: */
}
/* L40: */
}
/* B(1:N,1:NRHS) := Y' * B(1:N,1:NRHS) */
if (*rank < *n) {
i__1 = *rank;
for (i__ = 1; i__ <= i__1; ++i__) {
i__2 = *n - *rank + 1;
slatzm_("Left", &i__2, nrhs, &a[i__ + (*rank + 1) * a_dim1], lda,
&work[mn + i__], &b[i__ + b_dim1], &b[*rank + 1 + b_dim1],
ldb, &work[(mn << 1) + 1]);
/* L50: */
}
}
/* workspace NRHS */
/* B(1:N,1:NRHS) := P * B(1:N,1:NRHS) */
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
i__2 = *n;
for (i__ = 1; i__ <= i__2; ++i__) {
work[(mn << 1) + i__] = 1.f;
/* L60: */
}
i__2 = *n;
for (i__ = 1; i__ <= i__2; ++i__) {
if (work[(mn << 1) + i__] == 1.f) {
if (jpvt[i__] != i__) {
k = i__;
t1 = b[k + j * b_dim1];
t2 = b[jpvt[k] + j * b_dim1];
L70:
b[jpvt[k] + j * b_dim1] = t1;
work[(mn << 1) + k] = 0.f;
t1 = t2;
k = jpvt[k];
t2 = b[jpvt[k] + j * b_dim1];
if (jpvt[k] != i__) {
goto L70;
}
b[i__ + j * b_dim1] = t1;
work[(mn << 1) + k] = 0.f;
}
}
/* L80: */
}
/* L90: */
}
/* Undo scaling */
if (iascl == 1) {
slascl_("G", &c__0, &c__0, &anrm, &smlnum, n, nrhs, &b[b_offset], ldb,
info);
slascl_("U", &c__0, &c__0, &smlnum, &anrm, rank, rank, &a[a_offset],
lda, info);
} else if (iascl == 2) {
slascl_("G", &c__0, &c__0, &anrm, &bignum, n, nrhs, &b[b_offset], ldb,
info);
slascl_("U", &c__0, &c__0, &bignum, &anrm, rank, rank, &a[a_offset],
lda, info);
}
if (ibscl == 1) {
slascl_("G", &c__0, &c__0, &smlnum, &bnrm, n, nrhs, &b[b_offset], ldb,
info);
} else if (ibscl == 2) {
slascl_("G", &c__0, &c__0, &bignum, &bnrm, n, nrhs, &b[b_offset], ldb,
info);
}
L100:
return 0;
/* End of SGELSX */
} /* sgelsx_ */
/* Subroutine */ int sgerfsx_(char *trans, char *equed, integer *n, integer *
nrhs, real *a, integer *lda, real *af, integer *ldaf, integer *ipiv,
real *r__, real *c__, real *b, integer *ldb, real *x, integer *ldx,
real *rcond, real *berr, integer *n_err_bnds__, real *err_bnds_norm__,
real *err_bnds_comp__, integer *nparams, real *params, real *work,
integer *iwork, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, af_dim1, af_offset, b_dim1, b_offset, x_dim1,
x_offset, err_bnds_norm_dim1, err_bnds_norm_offset,
err_bnds_comp_dim1, err_bnds_comp_offset, i__1;
real r__1, r__2;
/* Builtin functions */
double sqrt(doublereal);
/* Local variables */
real illrcond_thresh__, unstable_thresh__, err_lbnd__;
integer ref_type__;
extern integer ilatrans_(char *);
integer j;
real rcond_tmp__;
integer prec_type__, trans_type__;
extern doublereal sla_gercond__(char *, integer *, real *, integer *,
real *, integer *, integer *, integer *, real *, integer *, real *
, integer *, ftnlen);
real cwise_wrong__;
extern /* Subroutine */ int sla_gerfsx_extended__(integer *, integer *,
integer *, integer *, real *, integer *, real *, integer *,
integer *, logical *, real *, real *, integer *, real *, integer *
, real *, integer *, real *, real *, real *, real *, real *, real
*, real *, integer *, real *, real *, logical *, integer *);
char norm[1];
logical ignore_cwise__;
extern logical lsame_(char *, char *);
real anorm;
extern doublereal slamch_(char *), slange_(char *, integer *,
integer *, real *, integer *, real *);
extern /* Subroutine */ int xerbla_(char *, integer *), sgecon_(
char *, integer *, real *, integer *, real *, real *, real *,
integer *, integer *);
logical colequ, notran, rowequ;
extern integer ilaprec_(char *);
integer ithresh, n_norms__;
real rthresh;
/* -- LAPACK routine (version 3.2.1) -- */
/* -- Contributed by James Demmel, Deaglan Halligan, Yozo Hida and -- */
/* -- Jason Riedy of Univ. of California Berkeley. -- */
/* -- April 2009 -- */
/* -- LAPACK is a software package provided by Univ. of Tennessee, -- */
/* -- Univ. of California Berkeley and NAG Ltd. -- */
/* .. */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* SGERFSX improves the computed solution to a system of linear */
/* equations and provides error bounds and backward error estimates */
/* for the solution. In addition to normwise error bound, the code */
/* provides maximum componentwise error bound if possible. See */
/* comments for ERR_BNDS_NORM and ERR_BNDS_COMP for details of the */
/* error bounds. */
/* The original system of linear equations may have been equilibrated */
/* before calling this routine, as described by arguments EQUED, R */
/* and C below. In this case, the solution and error bounds returned */
/* are for the original unequilibrated system. */
/* Arguments */
/* ========= */
/* Some optional parameters are bundled in the PARAMS array. These */
/* settings determine how refinement is performed, but often the */
/* defaults are acceptable. If the defaults are acceptable, users */
/* can pass NPARAMS = 0 which prevents the source code from accessing */
/* the PARAMS argument. */
/* TRANS (input) CHARACTER*1 */
/* Specifies the form of the system of equations: */
/* = 'N': A * X = B (No transpose) */
/* = 'T': A**T * X = B (Transpose) */
/* = 'C': A**H * X = B (Conjugate transpose = Transpose) */
/* EQUED (input) CHARACTER*1 */
/* Specifies the form of equilibration that was done to A */
/* before calling this routine. This is needed to compute */
/* the solution and error bounds correctly. */
/* = 'N': No equilibration */
/* = 'R': Row equilibration, i.e., A has been premultiplied by */
/* diag(R). */
/* = 'C': Column equilibration, i.e., A has been postmultiplied */
/* by diag(C). */
/* = 'B': Both row and column equilibration, i.e., A has been */
/* replaced by diag(R) * A * diag(C). */
/* The right hand side B has been changed accordingly. */
/* N (input) INTEGER */
/* The order of the matrix A. N >= 0. */
/* NRHS (input) INTEGER */
/* The number of right hand sides, i.e., the number of columns */
/* of the matrices B and X. NRHS >= 0. */
/* A (input) REAL array, dimension (LDA,N) */
/* The original N-by-N matrix A. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,N). */
/* AF (input) REAL array, dimension (LDAF,N) */
/* The factors L and U from the factorization A = P*L*U */
/* as computed by SGETRF. */
/* LDAF (input) INTEGER */
/* The leading dimension of the array AF. LDAF >= max(1,N). */
/* IPIV (input) INTEGER array, dimension (N) */
/* The pivot indices from SGETRF; for 1<=i<=N, row i of the */
/* matrix was interchanged with row IPIV(i). */
/* R (input or output) REAL array, dimension (N) */
/* The row scale factors for A. If EQUED = 'R' or 'B', A is */
/* multiplied on the left by diag(R); if EQUED = 'N' or 'C', R */
/* is not accessed. R is an input argument if FACT = 'F'; */
/* otherwise, R is an output argument. If FACT = 'F' and */
/* EQUED = 'R' or 'B', each element of R must be positive. */
/* If R is output, each element of R is a power of the radix. */
/* If R is input, each element of R should be a power of the radix */
/* to ensure a reliable solution and error estimates. Scaling by */
/* powers of the radix does not cause rounding errors unless the */
/* result underflows or overflows. Rounding errors during scaling */
/* lead to refining with a matrix that is not equivalent to the */
/* input matrix, producing error estimates that may not be */
/* reliable. */
/* C (input or output) REAL array, dimension (N) */
/* The column scale factors for A. If EQUED = 'C' or 'B', A is */
/* multiplied on the right by diag(C); if EQUED = 'N' or 'R', C */
/* is not accessed. C is an input argument if FACT = 'F'; */
/* otherwise, C is an output argument. If FACT = 'F' and */
/* EQUED = 'C' or 'B', each element of C must be positive. */
/* If C is output, each element of C is a power of the radix. */
/* If C is input, each element of C should be a power of the radix */
/* to ensure a reliable solution and error estimates. Scaling by */
/* powers of the radix does not cause rounding errors unless the */
/* result underflows or overflows. Rounding errors during scaling */
/* lead to refining with a matrix that is not equivalent to the */
/* input matrix, producing error estimates that may not be */
/* reliable. */
/* B (input) REAL array, dimension (LDB,NRHS) */
/* The right hand side matrix B. */
/* LDB (input) INTEGER */
/* The leading dimension of the array B. LDB >= max(1,N). */
/* X (input/output) REAL array, dimension (LDX,NRHS) */
/* On entry, the solution matrix X, as computed by SGETRS. */
/* On exit, the improved solution matrix X. */
/* LDX (input) INTEGER */
/* The leading dimension of the array X. LDX >= max(1,N). */
/* RCOND (output) REAL */
/* Reciprocal scaled condition number. This is an estimate of the */
/* reciprocal Skeel condition number of the matrix A after */
/* equilibration (if done). If this is less than the machine */
/* precision (in particular, if it is zero), the matrix is singular */
/* to working precision. Note that the error may still be small even */
/* if this number is very small and the matrix appears ill- */
/* conditioned. */
/* BERR (output) REAL array, dimension (NRHS) */
/* Componentwise relative backward error. This is the */
/* componentwise relative backward error of each solution vector X(j) */
/* (i.e., the smallest relative change in any element of A or B that */
/* makes X(j) an exact solution). */
/* N_ERR_BNDS (input) INTEGER */
/* Number of error bounds to return for each right hand side */
/* and each type (normwise or componentwise). See ERR_BNDS_NORM and */
/* ERR_BNDS_COMP below. */
/* ERR_BNDS_NORM (output) REAL array, dimension (NRHS, N_ERR_BNDS) */
/* For each right-hand side, this array contains information about */
/* various error bounds and condition numbers corresponding to the */
/* normwise relative error, which is defined as follows: */
/* Normwise relative error in the ith solution vector: */
/* max_j (abs(XTRUE(j,i) - X(j,i))) */
/* ------------------------------ */
/* max_j abs(X(j,i)) */
/* The array is indexed by the type of error information as described */
/* below. There currently are up to three pieces of information */
/* returned. */
/* The first index in ERR_BNDS_NORM(i,:) corresponds to the ith */
/* right-hand side. */
/* The second index in ERR_BNDS_NORM(:,err) contains the following */
/* three fields: */
/* err = 1 "Trust/don't trust" boolean. Trust the answer if the */
/* reciprocal condition number is less than the threshold */
/* sqrt(n) * slamch('Epsilon'). */
/* err = 2 "Guaranteed" error bound: The estimated forward error, */
/* almost certainly within a factor of 10 of the true error */
/* so long as the next entry is greater than the threshold */
/* sqrt(n) * slamch('Epsilon'). This error bound should only */
/* be trusted if the previous boolean is true. */
/* err = 3 Reciprocal condition number: Estimated normwise */
/* reciprocal condition number. Compared with the threshold */
/* sqrt(n) * slamch('Epsilon') to determine if the error */
/* estimate is "guaranteed". These reciprocal condition */
/* numbers are 1 / (norm(Z^{-1},inf) * norm(Z,inf)) for some */
/* appropriately scaled matrix Z. */
/* Let Z = S*A, where S scales each row by a power of the */
/* radix so all absolute row sums of Z are approximately 1. */
/* See Lapack Working Note 165 for further details and extra */
/* cautions. */
/* ERR_BNDS_COMP (output) REAL array, dimension (NRHS, N_ERR_BNDS) */
/* For each right-hand side, this array contains information about */
/* various error bounds and condition numbers corresponding to the */
/* componentwise relative error, which is defined as follows: */
/* Componentwise relative error in the ith solution vector: */
/* abs(XTRUE(j,i) - X(j,i)) */
/* max_j ---------------------- */
/* abs(X(j,i)) */
/* The array is indexed by the right-hand side i (on which the */
/* componentwise relative error depends), and the type of error */
/* information as described below. There currently are up to three */
/* pieces of information returned for each right-hand side. If */
/* componentwise accuracy is not requested (PARAMS(3) = 0.0), then */
/* ERR_BNDS_COMP is not accessed. If N_ERR_BNDS .LT. 3, then at most */
/* the first (:,N_ERR_BNDS) entries are returned. */
/* The first index in ERR_BNDS_COMP(i,:) corresponds to the ith */
/* right-hand side. */
/* The second index in ERR_BNDS_COMP(:,err) contains the following */
/* three fields: */
/* err = 1 "Trust/don't trust" boolean. Trust the answer if the */
/* reciprocal condition number is less than the threshold */
/* sqrt(n) * slamch('Epsilon'). */
/* err = 2 "Guaranteed" error bound: The estimated forward error, */
/* almost certainly within a factor of 10 of the true error */
/* so long as the next entry is greater than the threshold */
/* sqrt(n) * slamch('Epsilon'). This error bound should only */
/* be trusted if the previous boolean is true. */
/* err = 3 Reciprocal condition number: Estimated componentwise */
/* reciprocal condition number. Compared with the threshold */
/* sqrt(n) * slamch('Epsilon') to determine if the error */
/* estimate is "guaranteed". These reciprocal condition */
/* numbers are 1 / (norm(Z^{-1},inf) * norm(Z,inf)) for some */
/* appropriately scaled matrix Z. */
/* Let Z = S*(A*diag(x)), where x is the solution for the */
/* current right-hand side and S scales each row of */
/* A*diag(x) by a power of the radix so all absolute row */
/* sums of Z are approximately 1. */
/* See Lapack Working Note 165 for further details and extra */
/* cautions. */
/* NPARAMS (input) INTEGER */
/* Specifies the number of parameters set in PARAMS. If .LE. 0, the */
/* PARAMS array is never referenced and default values are used. */
/* PARAMS (input / output) REAL array, dimension NPARAMS */
/* Specifies algorithm parameters. If an entry is .LT. 0.0, then */
/* that entry will be filled with default value used for that */
/* parameter. Only positions up to NPARAMS are accessed; defaults */
/* are used for higher-numbered parameters. */
/* PARAMS(LA_LINRX_ITREF_I = 1) : Whether to perform iterative */
/* refinement or not. */
/* Default: 1.0 */
/* = 0.0 : No refinement is performed, and no error bounds are */
/* computed. */
/* = 1.0 : Use the double-precision refinement algorithm, */
/* possibly with doubled-single computations if the */
/* compilation environment does not support DOUBLE */
/* PRECISION. */
/* (other values are reserved for future use) */
/* PARAMS(LA_LINRX_ITHRESH_I = 2) : Maximum number of residual */
/* computations allowed for refinement. */
/* Default: 10 */
/* Aggressive: Set to 100 to permit convergence using approximate */
/* factorizations or factorizations other than LU. If */
/* the factorization uses a technique other than */
/* Gaussian elimination, the guarantees in */
/* err_bnds_norm and err_bnds_comp may no longer be */
/* trustworthy. */
/* PARAMS(LA_LINRX_CWISE_I = 3) : Flag determining if the code */
/* will attempt to find a solution with small componentwise */
/* relative error in the double-precision algorithm. Positive */
/* is true, 0.0 is false. */
/* Default: 1.0 (attempt componentwise convergence) */
/* WORK (workspace) REAL array, dimension (4*N) */
/* IWORK (workspace) INTEGER array, dimension (N) */
/* INFO (output) INTEGER */
/* = 0: Successful exit. The solution to every right-hand side is */
/* guaranteed. */
/* < 0: If INFO = -i, the i-th argument had an illegal value */
/* > 0 and <= N: U(INFO,INFO) is exactly zero. The factorization */
/* has been completed, but the factor U is exactly singular, so */
/* the solution and error bounds could not be computed. RCOND = 0 */
/* is returned. */
/* = N+J: The solution corresponding to the Jth right-hand side is */
/* not guaranteed. The solutions corresponding to other right- */
/* hand sides K with K > J may not be guaranteed as well, but */
/* only the first such right-hand side is reported. If a small */
/* componentwise error is not requested (PARAMS(3) = 0.0) then */
/* the Jth right-hand side is the first with a normwise error */
/* bound that is not guaranteed (the smallest J such */
/* that ERR_BNDS_NORM(J,1) = 0.0). By default (PARAMS(3) = 1.0) */
/* the Jth right-hand side is the first with either a normwise or */
/* componentwise error bound that is not guaranteed (the smallest */
/* J such that either ERR_BNDS_NORM(J,1) = 0.0 or */
/* ERR_BNDS_COMP(J,1) = 0.0). See the definition of */
/* ERR_BNDS_NORM(:,1) and ERR_BNDS_COMP(:,1). To get information */
/* about all of the right-hand sides check ERR_BNDS_NORM or */
/* ERR_BNDS_COMP. */
/* ================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Check the input parameters. */
/* Parameter adjustments */
err_bnds_comp_dim1 = *nrhs;
err_bnds_comp_offset = 1 + err_bnds_comp_dim1;
err_bnds_comp__ -= err_bnds_comp_offset;
err_bnds_norm_dim1 = *nrhs;
err_bnds_norm_offset = 1 + err_bnds_norm_dim1;
err_bnds_norm__ -= err_bnds_norm_offset;
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
af_dim1 = *ldaf;
af_offset = 1 + af_dim1;
af -= af_offset;
--ipiv;
--r__;
--c__;
b_dim1 = *ldb;
b_offset = 1 + b_dim1;
b -= b_offset;
x_dim1 = *ldx;
x_offset = 1 + x_dim1;
x -= x_offset;
--berr;
--params;
--work;
--iwork;
/* Function Body */
*info = 0;
trans_type__ = ilatrans_(trans);
ref_type__ = 1;
if (*nparams >= 1) {
if (params[1] < 0.f) {
params[1] = 1.f;
} else {
ref_type__ = params[1];
}
}
/* Set default parameters. */
illrcond_thresh__ = (real) (*n) * slamch_("Epsilon");
ithresh = 10;
rthresh = .5f;
unstable_thresh__ = .25f;
ignore_cwise__ = FALSE_;
if (*nparams >= 2) {
if (params[2] < 0.f) {
params[2] = (real) ithresh;
} else {
ithresh = (integer) params[2];
}
}
if (*nparams >= 3) {
if (params[3] < 0.f) {
if (ignore_cwise__) {
params[3] = 0.f;
} else {
params[3] = 1.f;
}
} else {
ignore_cwise__ = params[3] == 0.f;
}
}
if (ref_type__ == 0 || *n_err_bnds__ == 0) {
n_norms__ = 0;
} else if (ignore_cwise__) {
n_norms__ = 1;
} else {
n_norms__ = 2;
}
notran = lsame_(trans, "N");
rowequ = lsame_(equed, "R") || lsame_(equed, "B");
colequ = lsame_(equed, "C") || lsame_(equed, "B");
/* Test input parameters. */
if (trans_type__ == -1) {
*info = -1;
} else if (! rowequ && ! colequ && ! lsame_(equed, "N")) {
*info = -2;
} else if (*n < 0) {
*info = -3;
} else if (*nrhs < 0) {
*info = -4;
} else if (*lda < max(1,*n)) {
*info = -6;
} else if (*ldaf < max(1,*n)) {
*info = -8;
} else if (*ldb < max(1,*n)) {
*info = -13;
} else if (*ldx < max(1,*n)) {
*info = -15;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SGERFSX", &i__1);
return 0;
}
/* Quick return if possible. */
if (*n == 0 || *nrhs == 0) {
*rcond = 1.f;
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
berr[j] = 0.f;
if (*n_err_bnds__ >= 1) {
err_bnds_norm__[j + err_bnds_norm_dim1] = 1.f;
err_bnds_comp__[j + err_bnds_comp_dim1] = 1.f;
} else if (*n_err_bnds__ >= 2) {
err_bnds_norm__[j + (err_bnds_norm_dim1 << 1)] = 0.f;
err_bnds_comp__[j + (err_bnds_comp_dim1 << 1)] = 0.f;
} else if (*n_err_bnds__ >= 3) {
err_bnds_norm__[j + err_bnds_norm_dim1 * 3] = 1.f;
err_bnds_comp__[j + err_bnds_comp_dim1 * 3] = 1.f;
}
}
return 0;
}
/* Default to failure. */
*rcond = 0.f;
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
berr[j] = 1.f;
if (*n_err_bnds__ >= 1) {
err_bnds_norm__[j + err_bnds_norm_dim1] = 1.f;
err_bnds_comp__[j + err_bnds_comp_dim1] = 1.f;
} else if (*n_err_bnds__ >= 2) {
err_bnds_norm__[j + (err_bnds_norm_dim1 << 1)] = 1.f;
err_bnds_comp__[j + (err_bnds_comp_dim1 << 1)] = 1.f;
} else if (*n_err_bnds__ >= 3) {
err_bnds_norm__[j + err_bnds_norm_dim1 * 3] = 0.f;
err_bnds_comp__[j + err_bnds_comp_dim1 * 3] = 0.f;
}
}
/* Compute the norm of A and the reciprocal of the condition */
/* number of A. */
if (notran) {
*(unsigned char *)norm = 'I';
} else {
*(unsigned char *)norm = '1';
}
anorm = slange_(norm, n, n, &a[a_offset], lda, &work[1]);
sgecon_(norm, n, &af[af_offset], ldaf, &anorm, rcond, &work[1], &iwork[1],
info);
/* Perform refinement on each right-hand side */
if (ref_type__ != 0) {
prec_type__ = ilaprec_("D");
if (notran) {
sla_gerfsx_extended__(&prec_type__, &trans_type__, n, nrhs, &a[
a_offset], lda, &af[af_offset], ldaf, &ipiv[1], &colequ, &
c__[1], &b[b_offset], ldb, &x[x_offset], ldx, &berr[1], &
n_norms__, &err_bnds_norm__[err_bnds_norm_offset], &
err_bnds_comp__[err_bnds_comp_offset], &work[*n + 1], &
work[1], &work[(*n << 1) + 1], &work[1], rcond, &ithresh,
&rthresh, &unstable_thresh__, &ignore_cwise__, info);
} else {
sla_gerfsx_extended__(&prec_type__, &trans_type__, n, nrhs, &a[
a_offset], lda, &af[af_offset], ldaf, &ipiv[1], &rowequ, &
r__[1], &b[b_offset], ldb, &x[x_offset], ldx, &berr[1], &
n_norms__, &err_bnds_norm__[err_bnds_norm_offset], &
err_bnds_comp__[err_bnds_comp_offset], &work[*n + 1], &
work[1], &work[(*n << 1) + 1], &work[1], rcond, &ithresh,
&rthresh, &unstable_thresh__, &ignore_cwise__, info);
}
}
/* Computing MAX */
r__1 = 10.f, r__2 = sqrt((real) (*n));
err_lbnd__ = dmax(r__1,r__2) * slamch_("Epsilon");
if (*n_err_bnds__ >= 1 && n_norms__ >= 1) {
/* Compute scaled normwise condition number cond(A*C). */
if (colequ && notran) {
rcond_tmp__ = sla_gercond__(trans, n, &a[a_offset], lda, &af[
af_offset], ldaf, &ipiv[1], &c_n1, &c__[1], info, &work[1]
, &iwork[1], (ftnlen)1);
} else if (rowequ && ! notran) {
rcond_tmp__ = sla_gercond__(trans, n, &a[a_offset], lda, &af[
af_offset], ldaf, &ipiv[1], &c_n1, &r__[1], info, &work[1]
, &iwork[1], (ftnlen)1);
} else {
rcond_tmp__ = sla_gercond__(trans, n, &a[a_offset], lda, &af[
af_offset], ldaf, &ipiv[1], &c__0, &r__[1], info, &work[1]
, &iwork[1], (ftnlen)1);
}
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
/* Cap the error at 1.0. */
if (*n_err_bnds__ >= 2 && err_bnds_norm__[j + (err_bnds_norm_dim1
<< 1)] > 1.f) {
err_bnds_norm__[j + (err_bnds_norm_dim1 << 1)] = 1.f;
}
/* Threshold the error (see LAWN). */
if (rcond_tmp__ < illrcond_thresh__) {
err_bnds_norm__[j + (err_bnds_norm_dim1 << 1)] = 1.f;
err_bnds_norm__[j + err_bnds_norm_dim1] = 0.f;
if (*info <= *n) {
*info = *n + j;
}
} else if (err_bnds_norm__[j + (err_bnds_norm_dim1 << 1)] <
err_lbnd__) {
err_bnds_norm__[j + (err_bnds_norm_dim1 << 1)] = err_lbnd__;
err_bnds_norm__[j + err_bnds_norm_dim1] = 1.f;
}
/* Save the condition number. */
if (*n_err_bnds__ >= 3) {
err_bnds_norm__[j + err_bnds_norm_dim1 * 3] = rcond_tmp__;
}
}
}
if (*n_err_bnds__ >= 1 && n_norms__ >= 2) {
/* Compute componentwise condition number cond(A*diag(Y(:,J))) for */
/* each right-hand side using the current solution as an estimate of */
/* the true solution. If the componentwise error estimate is too */
/* large, then the solution is a lousy estimate of truth and the */
/* estimated RCOND may be too optimistic. To avoid misleading users, */
/* the inverse condition number is set to 0.0 when the estimated */
/* cwise error is at least CWISE_WRONG. */
cwise_wrong__ = sqrt(slamch_("Epsilon"));
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
if (err_bnds_comp__[j + (err_bnds_comp_dim1 << 1)] <
cwise_wrong__) {
rcond_tmp__ = sla_gercond__(trans, n, &a[a_offset], lda, &af[
af_offset], ldaf, &ipiv[1], &c__1, &x[j * x_dim1 + 1],
info, &work[1], &iwork[1], (ftnlen)1);
} else {
rcond_tmp__ = 0.f;
}
/* Cap the error at 1.0. */
if (*n_err_bnds__ >= 2 && err_bnds_comp__[j + (err_bnds_comp_dim1
<< 1)] > 1.f) {
err_bnds_comp__[j + (err_bnds_comp_dim1 << 1)] = 1.f;
}
/* Threshold the error (see LAWN). */
if (rcond_tmp__ < illrcond_thresh__) {
err_bnds_comp__[j + (err_bnds_comp_dim1 << 1)] = 1.f;
err_bnds_comp__[j + err_bnds_comp_dim1] = 0.f;
if (params[3] == 1.f && *info < *n + j) {
*info = *n + j;
}
} else if (err_bnds_comp__[j + (err_bnds_comp_dim1 << 1)] <
err_lbnd__) {
err_bnds_comp__[j + (err_bnds_comp_dim1 << 1)] = err_lbnd__;
err_bnds_comp__[j + err_bnds_comp_dim1] = 1.f;
}
/* Save the condition number. */
if (*n_err_bnds__ >= 3) {
err_bnds_comp__[j + err_bnds_comp_dim1 * 3] = rcond_tmp__;
}
}
}
return 0;
/* End of SGERFSX */
} /* sgerfsx_ */
doublereal sqrt11_(integer *m, integer *k, real *a, integer *lda, real *tau,
real *work, integer *lwork)
{
/* System generated locals */
integer a_dim1, a_offset, i__1;
real ret_val;
/* Local variables */
integer j, info;
real rdummy[1];
/* -- LAPACK routine (version 3.1) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* SQRT11 computes the test ratio */
/* || Q'*Q - I || / (eps * m) */
/* where the orthogonal matrix Q is represented as a product of */
/* elementary transformations. Each transformation has the form */
/* H(k) = I - tau(k) v(k) v(k)' */
/* where tau(k) is stored in TAU(k) and v(k) is an m-vector of the form */
/* [ 0 ... 0 1 x(k) ]', where x(k) is a vector of length m-k stored */
/* in A(k+1:m,k). */
/* Arguments */
/* ========= */
/* M (input) INTEGER */
/* The number of rows of the matrix A. */
/* K (input) INTEGER */
/* The number of columns of A whose subdiagonal entries */
/* contain information about orthogonal transformations. */
/* A (input) REAL array, dimension (LDA,K) */
/* The (possibly partial) output of a QR reduction routine. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. */
/* TAU (input) REAL array, dimension (K) */
/* The scaling factors tau for the elementary transformations as */
/* computed by the QR factorization routine. */
/* WORK (workspace) REAL array, dimension (LWORK) */
/* LWORK (input) INTEGER */
/* The length of the array WORK. LWORK >= M*M + M. */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Local Arrays .. */
/* .. */
/* .. Executable Statements .. */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
--tau;
--work;
/* Function Body */
ret_val = 0.f;
/* Test for sufficient workspace */
if (*lwork < *m * *m + *m) {
this_xerbla_("SQRT11", &c__7);
return ret_val;
}
/* Quick return if possible */
if (*m <= 0) {
return ret_val;
}
slaset_("Full", m, m, &c_b5, &c_b6, &work[1], m);
/* Form Q */
sorm2r_("Left", "No transpose", m, m, k, &a[a_offset], lda, &tau[1], &
work[1], m, &work[*m * *m + 1], &info);
/* Form Q'*Q */
sorm2r_("Left", "Transpose", m, m, k, &a[a_offset], lda, &tau[1], &work[1]
, m, &work[*m * *m + 1], &info);
i__1 = *m;
for (j = 1; j <= i__1; ++j) {
work[(j - 1) * *m + j] += -1.f;
/* L10: */
}
ret_val = slange_("One-norm", m, m, &work[1], m, rdummy) / ((
real) (*m) * slamch_("Epsilon"));
return ret_val;
/* End of SQRT11 */
} /* sqrt11_ */
/* Subroutine */ int sgelss_(integer *m, integer *n, integer *nrhs, real *a,
integer *lda, real *b, integer *ldb, real *s, real *rcond, integer *
rank, real *work, integer *lwork, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, i__1, i__2, i__3, i__4;
real r__1;
/* Local variables */
static real anrm, bnrm;
static integer itau;
static real vdum[1];
static integer i__, iascl, ibscl, chunk;
extern /* Subroutine */ int sgemm_(char *, char *, integer *, integer *,
integer *, real *, real *, integer *, real *, integer *, real *,
real *, integer *);
static real sfmin;
static integer minmn, maxmn;
extern /* Subroutine */ int sgemv_(char *, integer *, integer *, real *,
real *, integer *, real *, integer *, real *, real *, integer *);
static integer itaup, itauq;
extern /* Subroutine */ int srscl_(integer *, real *, real *, integer *);
static integer mnthr, iwork;
extern /* Subroutine */ int scopy_(integer *, real *, integer *, real *,
integer *);
static integer bl, ie, il;
extern /* Subroutine */ int slabad_(real *, real *);
static integer mm, bdspac;
extern /* Subroutine */ int sgebrd_(integer *, integer *, real *, integer
*, real *, real *, real *, real *, real *, integer *, integer *);
extern doublereal slamch_(char *), slange_(char *, integer *,
integer *, real *, integer *, real *);
extern /* Subroutine */ int xerbla_(char *, integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *, ftnlen, ftnlen);
static real bignum;
extern /* Subroutine */ int sgelqf_(integer *, integer *, real *, integer
*, real *, real *, integer *, integer *), slascl_(char *, integer
*, integer *, real *, real *, integer *, integer *, real *,
integer *, integer *), sgeqrf_(integer *, integer *, real
*, integer *, real *, real *, integer *, integer *), slacpy_(char
*, integer *, integer *, real *, integer *, real *, integer *), slaset_(char *, integer *, integer *, real *, real *,
real *, integer *), sbdsqr_(char *, integer *, integer *,
integer *, integer *, real *, real *, real *, integer *, real *,
integer *, real *, integer *, real *, integer *), sorgbr_(
char *, integer *, integer *, integer *, real *, integer *, real *
, real *, integer *, integer *);
static integer ldwork;
extern /* Subroutine */ int sormbr_(char *, char *, char *, integer *,
integer *, integer *, real *, integer *, real *, real *, integer *
, real *, integer *, integer *);
static integer minwrk, maxwrk;
static real smlnum;
extern /* Subroutine */ int sormlq_(char *, char *, integer *, integer *,
integer *, real *, integer *, real *, real *, integer *, real *,
integer *, integer *);
static logical lquery;
extern /* Subroutine */ int sormqr_(char *, char *, integer *, integer *,
integer *, real *, integer *, real *, real *, integer *, real *,
integer *, integer *);
static real eps, thr;
#define a_ref(a_1,a_2) a[(a_2)*a_dim1 + a_1]
#define b_ref(a_1,a_2) b[(a_2)*b_dim1 + a_1]
/* -- LAPACK driver routine (version 3.0) --
Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd.,
Courant Institute, Argonne National Lab, and Rice University
October 31, 1999
Purpose
=======
SGELSS computes the minimum norm solution to a real linear least
squares problem:
Minimize 2-norm(| b - A*x |).
using the singular value decomposition (SVD) of A. A is an M-by-N
matrix which may be rank-deficient.
Several right hand side vectors b and solution vectors x can be
handled in a single call; they are stored as the columns of the
M-by-NRHS right hand side matrix B and the N-by-NRHS solution matrix
X.
The effective rank of A is determined by treating as zero those
singular values which are less than RCOND times the largest singular
value.
Arguments
=========
M (input) INTEGER
The number of rows of the matrix A. M >= 0.
N (input) INTEGER
The number of columns of the matrix A. N >= 0.
NRHS (input) INTEGER
The number of right hand sides, i.e., the number of columns
of the matrices B and X. NRHS >= 0.
A (input/output) REAL array, dimension (LDA,N)
On entry, the M-by-N matrix A.
On exit, the first min(m,n) rows of A are overwritten with
its right singular vectors, stored rowwise.
LDA (input) INTEGER
The leading dimension of the array A. LDA >= max(1,M).
B (input/output) REAL array, dimension (LDB,NRHS)
On entry, the M-by-NRHS right hand side matrix B.
On exit, B is overwritten by the N-by-NRHS solution
matrix X. If m >= n and RANK = n, the residual
sum-of-squares for the solution in the i-th column is given
by the sum of squares of elements n+1:m in that column.
LDB (input) INTEGER
The leading dimension of the array B. LDB >= max(1,max(M,N)).
S (output) REAL array, dimension (min(M,N))
The singular values of A in decreasing order.
The condition number of A in the 2-norm = S(1)/S(min(m,n)).
RCOND (input) REAL
RCOND is used to determine the effective rank of A.
Singular values S(i) <= RCOND*S(1) are treated as zero.
If RCOND < 0, machine precision is used instead.
RANK (output) INTEGER
The effective rank of A, i.e., the number of singular values
which are greater than RCOND*S(1).
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. LWORK >= 1, and also:
LWORK >= 3*min(M,N) + max( 2*min(M,N), max(M,N), NRHS )
For good performance, LWORK should generally be larger.
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.
INFO (output) INTEGER
= 0: successful exit
< 0: if INFO = -i, the i-th argument had an illegal value.
> 0: the algorithm for computing the SVD failed to converge;
if INFO = i, i off-diagonal elements of an intermediate
bidiagonal form did not converge to zero.
=====================================================================
Test the input arguments
Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1 * 1;
a -= a_offset;
b_dim1 = *ldb;
b_offset = 1 + b_dim1 * 1;
b -= b_offset;
--s;
--work;
/* Function Body */
*info = 0;
minmn = min(*m,*n);
maxmn = max(*m,*n);
mnthr = ilaenv_(&c__6, "SGELSS", " ", m, n, nrhs, &c_n1, (ftnlen)6, (
ftnlen)1);
lquery = *lwork == -1;
if (*m < 0) {
*info = -1;
} else if (*n < 0) {
*info = -2;
} else if (*nrhs < 0) {
*info = -3;
} else if (*lda < max(1,*m)) {
*info = -5;
} else if (*ldb < max(1,maxmn)) {
*info = -7;
}
/* Compute workspace
(Note: Comments in the code beginning "Workspace:" describe the
minimal amount of workspace needed at that point in the code,
as well as the preferred amount for good performance.
NB refers to the optimal block size for the immediately
following subroutine, as returned by ILAENV.) */
minwrk = 1;
if (*info == 0 && (*lwork >= 1 || lquery)) {
maxwrk = 0;
mm = *m;
if (*m >= *n && *m >= mnthr) {
/* Path 1a - overdetermined, with many more rows than columns */
mm = *n;
/* Computing MAX */
i__1 = maxwrk, i__2 = *n + *n * ilaenv_(&c__1, "SGEQRF", " ", m,
n, &c_n1, &c_n1, (ftnlen)6, (ftnlen)1);
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk, i__2 = *n + *nrhs * ilaenv_(&c__1, "SORMQR", "LT",
m, nrhs, n, &c_n1, (ftnlen)6, (ftnlen)2);
maxwrk = max(i__1,i__2);
}
if (*m >= *n) {
/* Path 1 - overdetermined or exactly determined
Compute workspace needed for SBDSQR
Computing MAX */
i__1 = 1, i__2 = *n * 5;
bdspac = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk, i__2 = *n * 3 + (mm + *n) * ilaenv_(&c__1, "SGEBRD"
, " ", &mm, n, &c_n1, &c_n1, (ftnlen)6, (ftnlen)1);
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk, i__2 = *n * 3 + *nrhs * ilaenv_(&c__1, "SORMBR",
"QLT", &mm, nrhs, n, &c_n1, (ftnlen)6, (ftnlen)3);
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk, i__2 = *n * 3 + (*n - 1) * ilaenv_(&c__1, "SORGBR",
"P", n, n, n, &c_n1, (ftnlen)6, (ftnlen)1);
maxwrk = max(i__1,i__2);
maxwrk = max(maxwrk,bdspac);
/* Computing MAX */
i__1 = maxwrk, i__2 = *n * *nrhs;
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = *n * 3 + mm, i__2 = *n * 3 + *nrhs, i__1 = max(i__1,i__2);
minwrk = max(i__1,bdspac);
maxwrk = max(minwrk,maxwrk);
}
if (*n > *m) {
/* Compute workspace needed for SBDSQR
Computing MAX */
i__1 = 1, i__2 = *m * 5;
bdspac = max(i__1,i__2);
/* Computing MAX */
i__1 = *m * 3 + *nrhs, i__2 = *m * 3 + *n, i__1 = max(i__1,i__2);
minwrk = max(i__1,bdspac);
if (*n >= mnthr) {
/* Path 2a - underdetermined, with many more columns
than rows */
maxwrk = *m + *m * ilaenv_(&c__1, "SGELQF", " ", m, n, &c_n1,
&c_n1, (ftnlen)6, (ftnlen)1);
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * *m + (*m << 2) + (*m << 1) *
ilaenv_(&c__1, "SGEBRD", " ", m, m, &c_n1, &c_n1, (
ftnlen)6, (ftnlen)1);
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * *m + (*m << 2) + *nrhs * ilaenv_(&
c__1, "SORMBR", "QLT", m, nrhs, m, &c_n1, (ftnlen)6, (
ftnlen)3);
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * *m + (*m << 2) + (*m - 1) *
ilaenv_(&c__1, "SORGBR", "P", m, m, m, &c_n1, (ftnlen)
6, (ftnlen)1);
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * *m + *m + bdspac;
maxwrk = max(i__1,i__2);
if (*nrhs > 1) {
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * *m + *m + *m * *nrhs;
maxwrk = max(i__1,i__2);
} else {
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * *m + (*m << 1);
maxwrk = max(i__1,i__2);
}
/* Computing MAX */
i__1 = maxwrk, i__2 = *m + *nrhs * ilaenv_(&c__1, "SORMLQ",
"LT", n, nrhs, m, &c_n1, (ftnlen)6, (ftnlen)2);
maxwrk = max(i__1,i__2);
} else {
/* Path 2 - underdetermined */
maxwrk = *m * 3 + (*n + *m) * ilaenv_(&c__1, "SGEBRD", " ", m,
n, &c_n1, &c_n1, (ftnlen)6, (ftnlen)1);
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * 3 + *nrhs * ilaenv_(&c__1, "SORMBR"
, "QLT", m, nrhs, m, &c_n1, (ftnlen)6, (ftnlen)3);
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * 3 + *m * ilaenv_(&c__1, "SORGBR",
"P", m, n, m, &c_n1, (ftnlen)6, (ftnlen)1);
maxwrk = max(i__1,i__2);
maxwrk = max(maxwrk,bdspac);
/* Computing MAX */
i__1 = maxwrk, i__2 = *n * *nrhs;
maxwrk = max(i__1,i__2);
}
}
maxwrk = max(minwrk,maxwrk);
work[1] = (real) maxwrk;
}
minwrk = max(minwrk,1);
if (*lwork < minwrk && ! lquery) {
*info = -12;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SGELSS", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
if (*m == 0 || *n == 0) {
*rank = 0;
return 0;
}
/* Get machine parameters */
eps = slamch_("P");
sfmin = slamch_("S");
smlnum = sfmin / eps;
bignum = 1.f / smlnum;
slabad_(&smlnum, &bignum);
/* Scale A if max element outside range [SMLNUM,BIGNUM] */
anrm = slange_("M", m, n, &a[a_offset], lda, &work[1]);
iascl = 0;
if (anrm > 0.f && anrm < smlnum) {
/* Scale matrix norm up to SMLNUM */
slascl_("G", &c__0, &c__0, &anrm, &smlnum, m, n, &a[a_offset], lda,
info);
iascl = 1;
} else if (anrm > bignum) {
/* Scale matrix norm down to BIGNUM */
slascl_("G", &c__0, &c__0, &anrm, &bignum, m, n, &a[a_offset], lda,
info);
iascl = 2;
} else if (anrm == 0.f) {
/* Matrix all zero. Return zero solution. */
i__1 = max(*m,*n);
slaset_("F", &i__1, nrhs, &c_b74, &c_b74, &b[b_offset], ldb);
slaset_("F", &minmn, &c__1, &c_b74, &c_b74, &s[1], &c__1);
*rank = 0;
goto L70;
}
/* Scale B if max element outside range [SMLNUM,BIGNUM] */
bnrm = slange_("M", m, nrhs, &b[b_offset], ldb, &work[1]);
ibscl = 0;
if (bnrm > 0.f && bnrm < smlnum) {
/* Scale matrix norm up to SMLNUM */
slascl_("G", &c__0, &c__0, &bnrm, &smlnum, m, nrhs, &b[b_offset], ldb,
info);
ibscl = 1;
} else if (bnrm > bignum) {
/* Scale matrix norm down to BIGNUM */
slascl_("G", &c__0, &c__0, &bnrm, &bignum, m, nrhs, &b[b_offset], ldb,
info);
ibscl = 2;
}
/* Overdetermined case */
if (*m >= *n) {
/* Path 1 - overdetermined or exactly determined */
mm = *m;
if (*m >= mnthr) {
/* Path 1a - overdetermined, with many more rows than columns */
mm = *n;
itau = 1;
iwork = itau + *n;
/* Compute A=Q*R
(Workspace: need 2*N, prefer N+N*NB) */
i__1 = *lwork - iwork + 1;
sgeqrf_(m, n, &a[a_offset], lda, &work[itau], &work[iwork], &i__1,
info);
/* Multiply B by transpose(Q)
(Workspace: need N+NRHS, prefer N+NRHS*NB) */
i__1 = *lwork - iwork + 1;
sormqr_("L", "T", m, nrhs, n, &a[a_offset], lda, &work[itau], &b[
b_offset], ldb, &work[iwork], &i__1, info);
/* Zero out below R */
if (*n > 1) {
i__1 = *n - 1;
i__2 = *n - 1;
slaset_("L", &i__1, &i__2, &c_b74, &c_b74, &a_ref(2, 1), lda);
}
}
ie = 1;
itauq = ie + *n;
itaup = itauq + *n;
iwork = itaup + *n;
/* Bidiagonalize R in A
(Workspace: need 3*N+MM, prefer 3*N+(MM+N)*NB) */
i__1 = *lwork - iwork + 1;
sgebrd_(&mm, n, &a[a_offset], lda, &s[1], &work[ie], &work[itauq], &
work[itaup], &work[iwork], &i__1, info);
/* Multiply B by transpose of left bidiagonalizing vectors of R
(Workspace: need 3*N+NRHS, prefer 3*N+NRHS*NB) */
i__1 = *lwork - iwork + 1;
sormbr_("Q", "L", "T", &mm, nrhs, n, &a[a_offset], lda, &work[itauq],
&b[b_offset], ldb, &work[iwork], &i__1, info);
/* Generate right bidiagonalizing vectors of R in A
(Workspace: need 4*N-1, prefer 3*N+(N-1)*NB) */
i__1 = *lwork - iwork + 1;
sorgbr_("P", n, n, n, &a[a_offset], lda, &work[itaup], &work[iwork], &
i__1, info);
iwork = ie + *n;
/* Perform bidiagonal QR iteration
multiply B by transpose of left singular vectors
compute right singular vectors in A
(Workspace: need BDSPAC) */
sbdsqr_("U", n, n, &c__0, nrhs, &s[1], &work[ie], &a[a_offset], lda,
vdum, &c__1, &b[b_offset], ldb, &work[iwork], info)
;
if (*info != 0) {
goto L70;
}
/* Multiply B by reciprocals of singular values
Computing MAX */
r__1 = *rcond * s[1];
thr = dmax(r__1,sfmin);
if (*rcond < 0.f) {
/* Computing MAX */
r__1 = eps * s[1];
thr = dmax(r__1,sfmin);
}
*rank = 0;
i__1 = *n;
for (i__ = 1; i__ <= i__1; ++i__) {
if (s[i__] > thr) {
srscl_(nrhs, &s[i__], &b_ref(i__, 1), ldb);
++(*rank);
} else {
slaset_("F", &c__1, nrhs, &c_b74, &c_b74, &b_ref(i__, 1), ldb);
}
/* L10: */
}
/* Multiply B by right singular vectors
(Workspace: need N, prefer N*NRHS) */
if (*lwork >= *ldb * *nrhs && *nrhs > 1) {
sgemm_("T", "N", n, nrhs, n, &c_b108, &a[a_offset], lda, &b[
b_offset], ldb, &c_b74, &work[1], ldb);
slacpy_("G", n, nrhs, &work[1], ldb, &b[b_offset], ldb)
;
} else if (*nrhs > 1) {
chunk = *lwork / *n;
i__1 = *nrhs;
i__2 = chunk;
for (i__ = 1; i__2 < 0 ? i__ >= i__1 : i__ <= i__1; i__ += i__2) {
/* Computing MIN */
i__3 = *nrhs - i__ + 1;
bl = min(i__3,chunk);
sgemm_("T", "N", n, &bl, n, &c_b108, &a[a_offset], lda, &
b_ref(1, i__), ldb, &c_b74, &work[1], n);
slacpy_("G", n, &bl, &work[1], n, &b_ref(1, i__), ldb);
/* L20: */
}
} else {
sgemv_("T", n, n, &c_b108, &a[a_offset], lda, &b[b_offset], &c__1,
&c_b74, &work[1], &c__1);
scopy_(n, &work[1], &c__1, &b[b_offset], &c__1);
}
} else /* if(complicated condition) */ {
/* Computing MAX */
i__2 = *m, i__1 = (*m << 1) - 4, i__2 = max(i__2,i__1), i__2 = max(
i__2,*nrhs), i__1 = *n - *m * 3;
if (*n >= mnthr && *lwork >= (*m << 2) + *m * *m + max(i__2,i__1)) {
/* Path 2a - underdetermined, with many more columns than rows
and sufficient workspace for an efficient algorithm */
ldwork = *m;
/* Computing MAX
Computing MAX */
i__3 = *m, i__4 = (*m << 1) - 4, i__3 = max(i__3,i__4), i__3 =
max(i__3,*nrhs), i__4 = *n - *m * 3;
i__2 = (*m << 2) + *m * *lda + max(i__3,i__4), i__1 = *m * *lda +
*m + *m * *nrhs;
if (*lwork >= max(i__2,i__1)) {
ldwork = *lda;
}
itau = 1;
iwork = *m + 1;
/* Compute A=L*Q
(Workspace: need 2*M, prefer M+M*NB) */
i__2 = *lwork - iwork + 1;
sgelqf_(m, n, &a[a_offset], lda, &work[itau], &work[iwork], &i__2,
info);
il = iwork;
/* Copy L to WORK(IL), zeroing out above it */
slacpy_("L", m, m, &a[a_offset], lda, &work[il], &ldwork);
i__2 = *m - 1;
i__1 = *m - 1;
slaset_("U", &i__2, &i__1, &c_b74, &c_b74, &work[il + ldwork], &
ldwork);
ie = il + ldwork * *m;
itauq = ie + *m;
itaup = itauq + *m;
iwork = itaup + *m;
/* Bidiagonalize L in WORK(IL)
(Workspace: need M*M+5*M, prefer M*M+4*M+2*M*NB) */
i__2 = *lwork - iwork + 1;
sgebrd_(m, m, &work[il], &ldwork, &s[1], &work[ie], &work[itauq],
&work[itaup], &work[iwork], &i__2, info);
/* Multiply B by transpose of left bidiagonalizing vectors of L
(Workspace: need M*M+4*M+NRHS, prefer M*M+4*M+NRHS*NB) */
i__2 = *lwork - iwork + 1;
sormbr_("Q", "L", "T", m, nrhs, m, &work[il], &ldwork, &work[
itauq], &b[b_offset], ldb, &work[iwork], &i__2, info);
/* Generate right bidiagonalizing vectors of R in WORK(IL)
(Workspace: need M*M+5*M-1, prefer M*M+4*M+(M-1)*NB) */
i__2 = *lwork - iwork + 1;
sorgbr_("P", m, m, m, &work[il], &ldwork, &work[itaup], &work[
iwork], &i__2, info);
iwork = ie + *m;
/* Perform bidiagonal QR iteration,
computing right singular vectors of L in WORK(IL) and
multiplying B by transpose of left singular vectors
(Workspace: need M*M+M+BDSPAC) */
sbdsqr_("U", m, m, &c__0, nrhs, &s[1], &work[ie], &work[il], &
ldwork, &a[a_offset], lda, &b[b_offset], ldb, &work[iwork]
, info);
if (*info != 0) {
goto L70;
}
/* Multiply B by reciprocals of singular values
Computing MAX */
r__1 = *rcond * s[1];
thr = dmax(r__1,sfmin);
if (*rcond < 0.f) {
/* Computing MAX */
r__1 = eps * s[1];
thr = dmax(r__1,sfmin);
}
*rank = 0;
i__2 = *m;
for (i__ = 1; i__ <= i__2; ++i__) {
if (s[i__] > thr) {
srscl_(nrhs, &s[i__], &b_ref(i__, 1), ldb);
++(*rank);
} else {
slaset_("F", &c__1, nrhs, &c_b74, &c_b74, &b_ref(i__, 1),
ldb);
}
/* L30: */
}
iwork = ie;
/* Multiply B by right singular vectors of L in WORK(IL)
(Workspace: need M*M+2*M, prefer M*M+M+M*NRHS) */
if (*lwork >= *ldb * *nrhs + iwork - 1 && *nrhs > 1) {
sgemm_("T", "N", m, nrhs, m, &c_b108, &work[il], &ldwork, &b[
b_offset], ldb, &c_b74, &work[iwork], ldb);
slacpy_("G", m, nrhs, &work[iwork], ldb, &b[b_offset], ldb);
} else if (*nrhs > 1) {
chunk = (*lwork - iwork + 1) / *m;
i__2 = *nrhs;
i__1 = chunk;
for (i__ = 1; i__1 < 0 ? i__ >= i__2 : i__ <= i__2; i__ +=
i__1) {
/* Computing MIN */
i__3 = *nrhs - i__ + 1;
bl = min(i__3,chunk);
sgemm_("T", "N", m, &bl, m, &c_b108, &work[il], &ldwork, &
b_ref(1, i__), ldb, &c_b74, &work[iwork], n);
slacpy_("G", m, &bl, &work[iwork], n, &b_ref(1, i__), ldb);
/* L40: */
}
} else {
sgemv_("T", m, m, &c_b108, &work[il], &ldwork, &b_ref(1, 1), &
c__1, &c_b74, &work[iwork], &c__1);
scopy_(m, &work[iwork], &c__1, &b_ref(1, 1), &c__1);
}
/* Zero out below first M rows of B */
i__1 = *n - *m;
slaset_("F", &i__1, nrhs, &c_b74, &c_b74, &b_ref(*m + 1, 1), ldb);
iwork = itau + *m;
/* Multiply transpose(Q) by B
(Workspace: need M+NRHS, prefer M+NRHS*NB) */
i__1 = *lwork - iwork + 1;
sormlq_("L", "T", n, nrhs, m, &a[a_offset], lda, &work[itau], &b[
b_offset], ldb, &work[iwork], &i__1, info);
} else {
/* Path 2 - remaining underdetermined cases */
ie = 1;
itauq = ie + *m;
itaup = itauq + *m;
iwork = itaup + *m;
/* Bidiagonalize A
(Workspace: need 3*M+N, prefer 3*M+(M+N)*NB) */
i__1 = *lwork - iwork + 1;
sgebrd_(m, n, &a[a_offset], lda, &s[1], &work[ie], &work[itauq], &
work[itaup], &work[iwork], &i__1, info);
/* Multiply B by transpose of left bidiagonalizing vectors
(Workspace: need 3*M+NRHS, prefer 3*M+NRHS*NB) */
i__1 = *lwork - iwork + 1;
sormbr_("Q", "L", "T", m, nrhs, n, &a[a_offset], lda, &work[itauq]
, &b[b_offset], ldb, &work[iwork], &i__1, info);
/* Generate right bidiagonalizing vectors in A
(Workspace: need 4*M, prefer 3*M+M*NB) */
i__1 = *lwork - iwork + 1;
sorgbr_("P", m, n, m, &a[a_offset], lda, &work[itaup], &work[
iwork], &i__1, info);
iwork = ie + *m;
/* Perform bidiagonal QR iteration,
computing right singular vectors of A in A and
multiplying B by transpose of left singular vectors
(Workspace: need BDSPAC) */
sbdsqr_("L", m, n, &c__0, nrhs, &s[1], &work[ie], &a[a_offset],
lda, vdum, &c__1, &b[b_offset], ldb, &work[iwork], info);
if (*info != 0) {
goto L70;
}
/* Multiply B by reciprocals of singular values
Computing MAX */
r__1 = *rcond * s[1];
thr = dmax(r__1,sfmin);
if (*rcond < 0.f) {
/* Computing MAX */
r__1 = eps * s[1];
thr = dmax(r__1,sfmin);
}
*rank = 0;
i__1 = *m;
for (i__ = 1; i__ <= i__1; ++i__) {
if (s[i__] > thr) {
srscl_(nrhs, &s[i__], &b_ref(i__, 1), ldb);
++(*rank);
} else {
slaset_("F", &c__1, nrhs, &c_b74, &c_b74, &b_ref(i__, 1),
ldb);
}
/* L50: */
}
/* Multiply B by right singular vectors of A
(Workspace: need N, prefer N*NRHS) */
if (*lwork >= *ldb * *nrhs && *nrhs > 1) {
sgemm_("T", "N", n, nrhs, m, &c_b108, &a[a_offset], lda, &b[
b_offset], ldb, &c_b74, &work[1], ldb);
slacpy_("F", n, nrhs, &work[1], ldb, &b[b_offset], ldb);
} else if (*nrhs > 1) {
chunk = *lwork / *n;
i__1 = *nrhs;
i__2 = chunk;
for (i__ = 1; i__2 < 0 ? i__ >= i__1 : i__ <= i__1; i__ +=
i__2) {
/* Computing MIN */
i__3 = *nrhs - i__ + 1;
bl = min(i__3,chunk);
sgemm_("T", "N", n, &bl, m, &c_b108, &a[a_offset], lda, &
b_ref(1, i__), ldb, &c_b74, &work[1], n);
slacpy_("F", n, &bl, &work[1], n, &b_ref(1, i__), ldb);
/* L60: */
}
} else {
sgemv_("T", m, n, &c_b108, &a[a_offset], lda, &b[b_offset], &
c__1, &c_b74, &work[1], &c__1);
scopy_(n, &work[1], &c__1, &b[b_offset], &c__1);
}
}
}
/* Undo scaling */
if (iascl == 1) {
slascl_("G", &c__0, &c__0, &anrm, &smlnum, n, nrhs, &b[b_offset], ldb,
info);
slascl_("G", &c__0, &c__0, &smlnum, &anrm, &minmn, &c__1, &s[1], &
minmn, info);
} else if (iascl == 2) {
slascl_("G", &c__0, &c__0, &anrm, &bignum, n, nrhs, &b[b_offset], ldb,
info);
slascl_("G", &c__0, &c__0, &bignum, &anrm, &minmn, &c__1, &s[1], &
minmn, info);
}
if (ibscl == 1) {
slascl_("G", &c__0, &c__0, &smlnum, &bnrm, n, nrhs, &b[b_offset], ldb,
info);
} else if (ibscl == 2) {
slascl_("G", &c__0, &c__0, &bignum, &bnrm, n, nrhs, &b[b_offset], ldb,
info);
}
L70:
work[1] = (real) maxwrk;
return 0;
/* End of SGELSS */
} /* sgelss_ */
/* Subroutine */ int sgegv_(char *jobvl, char *jobvr, integer *n, real *a,
integer *lda, real *b, integer *ldb, real *alphar, real *alphai, real
*beta, real *vl, integer *ldvl, real *vr, integer *ldvr, real *work,
integer *lwork, 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
=======
This routine is deprecated and has been replaced by routine SGGEV.
SGEGV computes for a pair of n-by-n real nonsymmetric matrices A and
B, the generalized eigenvalues (alphar +/- alphai*i, beta), and
optionally, the left and/or right generalized eigenvectors (VL and
VR).
A generalized eigenvalue for a pair of matrices (A,B) is, roughly
speaking, a scalar w or a ratio alpha/beta = w, such that A - w*B
is singular. It is usually represented as the pair (alpha,beta),
as there is a reasonable interpretation for beta=0, and even for
both being zero. A good beginning reference is the book, "Matrix
Computations", by G. Golub & C. van Loan (Johns Hopkins U. Press)
A right generalized eigenvector corresponding to a generalized
eigenvalue w for a pair of matrices (A,B) is a vector r such
that (A - w B) r = 0 . A left generalized eigenvector is a vector
l such that l**H * (A - w B) = 0, where l**H is the
conjugate-transpose of l.
Note: this routine performs "full balancing" on A and B -- see
"Further Details", below.
Arguments
=========
JOBVL (input) CHARACTER*1
= 'N': do not compute the left generalized eigenvectors;
= 'V': compute the left generalized eigenvectors.
JOBVR (input) CHARACTER*1
= 'N': do not compute the right generalized eigenvectors;
= 'V': compute the right generalized eigenvectors.
N (input) INTEGER
The order of the matrices A, B, VL, and VR. N >= 0.
A (input/output) REAL array, dimension (LDA, N)
On entry, the first of the pair of matrices whose
generalized eigenvalues and (optionally) generalized
eigenvectors are to be computed.
On exit, the contents will have been destroyed. (For a
description of the contents of A on exit, see "Further
Details", below.)
LDA (input) INTEGER
The leading dimension of A. LDA >= max(1,N).
B (input/output) REAL array, dimension (LDB, N)
On entry, the second of the pair of matrices whose
generalized eigenvalues and (optionally) generalized
eigenvectors are to be computed.
On exit, the contents will have been destroyed. (For a
description of the contents of B on exit, see "Further
Details", below.)
LDB (input) INTEGER
The leading dimension of B. LDB >= max(1,N).
ALPHAR (output) REAL array, dimension (N)
ALPHAI (output) REAL array, dimension (N)
BETA (output) REAL array, dimension (N)
On exit, (ALPHAR(j) + ALPHAI(j)*i)/BETA(j), j=1,...,N, will
be the generalized eigenvalues. If ALPHAI(j) is zero, then
the j-th eigenvalue is real; if positive, then the j-th and
(j+1)-st eigenvalues are a complex conjugate pair, with
ALPHAI(j+1) negative.
Note: the quotients ALPHAR(j)/BETA(j) and ALPHAI(j)/BETA(j)
may easily over- or underflow, and BETA(j) may even be zero.
Thus, the user should avoid naively computing the ratio
alpha/beta. However, ALPHAR and ALPHAI will be always less
than and usually comparable with norm(A) in magnitude, and
BETA always less than and usually comparable with norm(B).
VL (output) REAL array, dimension (LDVL,N)
If JOBVL = 'V', the left generalized eigenvectors. (See
"Purpose", above.) Real eigenvectors take one column,
complex take two columns, the first for the real part and
the second for the imaginary part. Complex eigenvectors
correspond to an eigenvalue with positive imaginary part.
Each eigenvector will be scaled so the largest component
will have abs(real part) + abs(imag. part) = 1, *except*
that for eigenvalues with alpha=beta=0, a zero vector will
be returned as the corresponding eigenvector.
Not referenced if JOBVL = 'N'.
LDVL (input) INTEGER
The leading dimension of the matrix VL. LDVL >= 1, and
if JOBVL = 'V', LDVL >= N.
VR (output) REAL array, dimension (LDVR,N)
If JOBVR = 'V', the right generalized eigenvectors. (See
"Purpose", above.) Real eigenvectors take one column,
complex take two columns, the first for the real part and
the second for the imaginary part. Complex eigenvectors
correspond to an eigenvalue with positive imaginary part.
Each eigenvector will be scaled so the largest component
will have abs(real part) + abs(imag. part) = 1, *except*
that for eigenvalues with alpha=beta=0, a zero vector will
be returned as the corresponding eigenvector.
Not referenced if JOBVR = 'N'.
LDVR (input) INTEGER
The leading dimension of the matrix VR. LDVR >= 1, and
if JOBVR = 'V', LDVR >= N.
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. LWORK >= max(1,8*N).
For good performance, LWORK must generally be larger.
To compute the optimal value of LWORK, call ILAENV to get
blocksizes (for SGEQRF, SORMQR, and SORGQR.) Then compute:
NB -- MAX of the blocksizes for SGEQRF, SORMQR, and SORGQR;
The optimal LWORK is:
2*N + MAX( 6*N, N*(NB+1) ).
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.
INFO (output) INTEGER
= 0: successful exit
< 0: if INFO = -i, the i-th argument had an illegal value.
= 1,...,N:
The QZ iteration failed. No eigenvectors have been
calculated, but ALPHAR(j), ALPHAI(j), and BETA(j)
should be correct for j=INFO+1,...,N.
> N: errors that usually indicate LAPACK problems:
=N+1: error return from SGGBAL
=N+2: error return from SGEQRF
=N+3: error return from SORMQR
=N+4: error return from SORGQR
=N+5: error return from SGGHRD
=N+6: error return from SHGEQZ (other than failed
iteration)
=N+7: error return from STGEVC
=N+8: error return from SGGBAK (computing VL)
=N+9: error return from SGGBAK (computing VR)
=N+10: error return from SLASCL (various calls)
Further Details
===============
Balancing
---------
This driver calls SGGBAL to both permute and scale rows and columns
of A and B. The permutations PL and PR are chosen so that PL*A*PR
and PL*B*R will be upper triangular except for the diagonal blocks
A(i:j,i:j) and B(i:j,i:j), with i and j as close together as
possible. The diagonal scaling matrices DL and DR are chosen so
that the pair DL*PL*A*PR*DR, DL*PL*B*PR*DR have elements close to
one (except for the elements that start out zero.)
After the eigenvalues and eigenvectors of the balanced matrices
have been computed, SGGBAK transforms the eigenvectors back to what
they would have been (in perfect arithmetic) if they had not been
balanced.
Contents of A and B on Exit
-------- -- - --- - -- ----
If any eigenvectors are computed (either JOBVL='V' or JOBVR='V' or
both), then on exit the arrays A and B will contain the real Schur
form[*] of the "balanced" versions of A and B. If no eigenvectors
are computed, then only the diagonal blocks will be correct.
[*] See SHGEQZ, SGEGS, or read the book "Matrix Computations",
by Golub & van Loan, pub. by Johns Hopkins U. Press.
=====================================================================
Decode the input arguments
Parameter adjustments */
/* Table of constant values */
static integer c__1 = 1;
static integer c_n1 = -1;
static real c_b27 = 1.f;
static real c_b38 = 0.f;
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, vl_dim1, vl_offset, vr_dim1,
vr_offset, i__1, i__2;
real r__1, r__2, r__3, r__4;
/* Local variables */
static real absb, anrm, bnrm;
static integer itau;
static real temp;
static logical ilvl, ilvr;
static integer lopt;
static real anrm1, anrm2, bnrm1, bnrm2, absai, scale, absar, sbeta;
extern logical lsame_(char *, char *);
static integer ileft, iinfo, icols, iwork, irows, jc, nb, in, jr;
static real salfai;
extern /* Subroutine */ int sggbak_(char *, char *, integer *, integer *,
integer *, real *, real *, integer *, real *, integer *, integer *
), sggbal_(char *, integer *, real *, integer *,
real *, integer *, integer *, integer *, real *, real *, real *,
integer *);
static real salfar;
extern doublereal slamch_(char *), slange_(char *, integer *,
integer *, real *, integer *, real *);
static real safmin;
extern /* Subroutine */ int sgghrd_(char *, char *, integer *, integer *,
integer *, real *, integer *, real *, integer *, real *, integer *
, real *, integer *, integer *);
static real safmax;
static char chtemp[1];
static logical ldumma[1];
extern /* Subroutine */ int slascl_(char *, integer *, integer *, real *,
real *, integer *, integer *, real *, integer *, integer *), xerbla_(char *, integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *, ftnlen, ftnlen);
static integer ijobvl, iright;
static logical ilimit;
extern /* Subroutine */ int sgeqrf_(integer *, integer *, real *, integer
*, real *, real *, integer *, integer *);
static integer ijobvr;
extern /* Subroutine */ int slacpy_(char *, integer *, integer *, real *,
integer *, real *, integer *), slaset_(char *, integer *,
integer *, real *, real *, real *, integer *), stgevc_(
char *, char *, logical *, integer *, real *, integer *, real *,
integer *, real *, integer *, real *, integer *, integer *,
integer *, real *, integer *);
static real onepls;
static integer lwkmin, nb1, nb2, nb3;
extern /* Subroutine */ int shgeqz_(char *, char *, char *, integer *,
integer *, integer *, real *, integer *, real *, integer *, real *
, real *, real *, real *, integer *, real *, integer *, real *,
integer *, integer *), sorgqr_(integer *,
integer *, integer *, real *, integer *, real *, real *, integer *
, integer *);
static integer lwkopt;
static logical lquery;
extern /* Subroutine */ int sormqr_(char *, char *, integer *, integer *,
integer *, real *, integer *, real *, real *, integer *, real *,
integer *, integer *);
static integer ihi, ilo;
static real eps;
static logical ilv;
#define a_ref(a_1,a_2) a[(a_2)*a_dim1 + a_1]
#define b_ref(a_1,a_2) b[(a_2)*b_dim1 + a_1]
#define vl_ref(a_1,a_2) vl[(a_2)*vl_dim1 + a_1]
#define vr_ref(a_1,a_2) vr[(a_2)*vr_dim1 + a_1]
a_dim1 = *lda;
a_offset = 1 + a_dim1 * 1;
a -= a_offset;
b_dim1 = *ldb;
b_offset = 1 + b_dim1 * 1;
b -= b_offset;
--alphar;
--alphai;
--beta;
vl_dim1 = *ldvl;
vl_offset = 1 + vl_dim1 * 1;
vl -= vl_offset;
vr_dim1 = *ldvr;
vr_offset = 1 + vr_dim1 * 1;
vr -= vr_offset;
--work;
/* Function Body */
if (lsame_(jobvl, "N")) {
ijobvl = 1;
ilvl = FALSE_;
} else if (lsame_(jobvl, "V")) {
ijobvl = 2;
ilvl = TRUE_;
} else {
ijobvl = -1;
ilvl = FALSE_;
}
if (lsame_(jobvr, "N")) {
ijobvr = 1;
ilvr = FALSE_;
} else if (lsame_(jobvr, "V")) {
ijobvr = 2;
ilvr = TRUE_;
} else {
ijobvr = -1;
ilvr = FALSE_;
}
ilv = ilvl || ilvr;
/* Test the input arguments
Computing MAX */
i__1 = *n << 3;
lwkmin = max(i__1,1);
lwkopt = lwkmin;
work[1] = (real) lwkopt;
lquery = *lwork == -1;
*info = 0;
if (ijobvl <= 0) {
*info = -1;
} else if (ijobvr <= 0) {
*info = -2;
} else if (*n < 0) {
*info = -3;
} else if (*lda < max(1,*n)) {
*info = -5;
} else if (*ldb < max(1,*n)) {
*info = -7;
} else if (*ldvl < 1 || ilvl && *ldvl < *n) {
*info = -12;
} else if (*ldvr < 1 || ilvr && *ldvr < *n) {
*info = -14;
} else if (*lwork < lwkmin && ! lquery) {
*info = -16;
}
if (*info == 0) {
nb1 = ilaenv_(&c__1, "SGEQRF", " ", n, n, &c_n1, &c_n1, (ftnlen)6, (
ftnlen)1);
nb2 = ilaenv_(&c__1, "SORMQR", " ", n, n, n, &c_n1, (ftnlen)6, (
ftnlen)1);
nb3 = ilaenv_(&c__1, "SORGQR", " ", n, n, n, &c_n1, (ftnlen)6, (
ftnlen)1);
/* Computing MAX */
i__1 = max(nb1,nb2);
nb = max(i__1,nb3);
/* Computing MAX */
i__1 = *n * 6, i__2 = *n * (nb + 1);
lopt = (*n << 1) + max(i__1,i__2);
work[1] = (real) lopt;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SGEGV ", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
if (*n == 0) {
return 0;
}
/* Get machine constants */
eps = slamch_("E") * slamch_("B");
safmin = slamch_("S");
safmin += safmin;
safmax = 1.f / safmin;
onepls = eps * 4 + 1.f;
/* Scale A */
anrm = slange_("M", n, n, &a[a_offset], lda, &work[1]);
anrm1 = anrm;
anrm2 = 1.f;
if (anrm < 1.f) {
if (safmax * anrm < 1.f) {
anrm1 = safmin;
anrm2 = safmax * anrm;
}
}
if (anrm > 0.f) {
slascl_("G", &c_n1, &c_n1, &anrm, &c_b27, n, n, &a[a_offset], lda, &
iinfo);
if (iinfo != 0) {
*info = *n + 10;
return 0;
}
}
/* Scale B */
bnrm = slange_("M", n, n, &b[b_offset], ldb, &work[1]);
bnrm1 = bnrm;
bnrm2 = 1.f;
if (bnrm < 1.f) {
if (safmax * bnrm < 1.f) {
bnrm1 = safmin;
bnrm2 = safmax * bnrm;
}
}
if (bnrm > 0.f) {
slascl_("G", &c_n1, &c_n1, &bnrm, &c_b27, n, n, &b[b_offset], ldb, &
iinfo);
if (iinfo != 0) {
*info = *n + 10;
return 0;
}
}
/* Permute the matrix to make it more nearly triangular
Workspace layout: (8*N words -- "work" requires 6*N words)
left_permutation, right_permutation, work... */
ileft = 1;
iright = *n + 1;
iwork = iright + *n;
sggbal_("P", n, &a[a_offset], lda, &b[b_offset], ldb, &ilo, &ihi, &work[
ileft], &work[iright], &work[iwork], &iinfo);
if (iinfo != 0) {
*info = *n + 1;
goto L120;
}
/* Reduce B to triangular form, and initialize VL and/or VR
Workspace layout: ("work..." must have at least N words)
left_permutation, right_permutation, tau, work... */
irows = ihi + 1 - ilo;
if (ilv) {
icols = *n + 1 - ilo;
} else {
icols = irows;
}
itau = iwork;
iwork = itau + irows;
i__1 = *lwork + 1 - iwork;
sgeqrf_(&irows, &icols, &b_ref(ilo, ilo), ldb, &work[itau], &work[iwork],
&i__1, &iinfo);
if (iinfo >= 0) {
/* Computing MAX */
i__1 = lwkopt, i__2 = (integer) work[iwork] + iwork - 1;
lwkopt = max(i__1,i__2);
}
if (iinfo != 0) {
*info = *n + 2;
goto L120;
}
i__1 = *lwork + 1 - iwork;
sormqr_("L", "T", &irows, &icols, &irows, &b_ref(ilo, ilo), ldb, &work[
itau], &a_ref(ilo, ilo), lda, &work[iwork], &i__1, &iinfo);
if (iinfo >= 0) {
/* Computing MAX */
i__1 = lwkopt, i__2 = (integer) work[iwork] + iwork - 1;
lwkopt = max(i__1,i__2);
}
if (iinfo != 0) {
*info = *n + 3;
goto L120;
}
if (ilvl) {
slaset_("Full", n, n, &c_b38, &c_b27, &vl[vl_offset], ldvl)
;
i__1 = irows - 1;
i__2 = irows - 1;
slacpy_("L", &i__1, &i__2, &b_ref(ilo + 1, ilo), ldb, &vl_ref(ilo + 1,
ilo), ldvl);
i__1 = *lwork + 1 - iwork;
sorgqr_(&irows, &irows, &irows, &vl_ref(ilo, ilo), ldvl, &work[itau],
&work[iwork], &i__1, &iinfo);
if (iinfo >= 0) {
/* Computing MAX */
i__1 = lwkopt, i__2 = (integer) work[iwork] + iwork - 1;
lwkopt = max(i__1,i__2);
}
if (iinfo != 0) {
*info = *n + 4;
goto L120;
}
}
if (ilvr) {
slaset_("Full", n, n, &c_b38, &c_b27, &vr[vr_offset], ldvr)
;
}
/* Reduce to generalized Hessenberg form */
if (ilv) {
/* Eigenvectors requested -- work on whole matrix. */
sgghrd_(jobvl, jobvr, n, &ilo, &ihi, &a[a_offset], lda, &b[b_offset],
ldb, &vl[vl_offset], ldvl, &vr[vr_offset], ldvr, &iinfo);
} else {
sgghrd_("N", "N", &irows, &c__1, &irows, &a_ref(ilo, ilo), lda, &
b_ref(ilo, ilo), ldb, &vl[vl_offset], ldvl, &vr[vr_offset],
ldvr, &iinfo);
}
if (iinfo != 0) {
*info = *n + 5;
goto L120;
}
/* Perform QZ algorithm
Workspace layout: ("work..." must have at least 1 word)
left_permutation, right_permutation, work... */
iwork = itau;
if (ilv) {
*(unsigned char *)chtemp = 'S';
} else {
*(unsigned char *)chtemp = 'E';
}
i__1 = *lwork + 1 - iwork;
shgeqz_(chtemp, jobvl, jobvr, n, &ilo, &ihi, &a[a_offset], lda, &b[
b_offset], ldb, &alphar[1], &alphai[1], &beta[1], &vl[vl_offset],
ldvl, &vr[vr_offset], ldvr, &work[iwork], &i__1, &iinfo);
if (iinfo >= 0) {
/* Computing MAX */
i__1 = lwkopt, i__2 = (integer) work[iwork] + iwork - 1;
lwkopt = max(i__1,i__2);
}
if (iinfo != 0) {
if (iinfo > 0 && iinfo <= *n) {
*info = iinfo;
} else if (iinfo > *n && iinfo <= *n << 1) {
*info = iinfo - *n;
} else {
*info = *n + 6;
}
goto L120;
}
if (ilv) {
/* Compute Eigenvectors (STGEVC requires 6*N words of workspace) */
if (ilvl) {
if (ilvr) {
*(unsigned char *)chtemp = 'B';
} else {
*(unsigned char *)chtemp = 'L';
}
} else {
*(unsigned char *)chtemp = 'R';
}
stgevc_(chtemp, "B", ldumma, n, &a[a_offset], lda, &b[b_offset], ldb,
&vl[vl_offset], ldvl, &vr[vr_offset], ldvr, n, &in, &work[
iwork], &iinfo);
if (iinfo != 0) {
*info = *n + 7;
goto L120;
}
/* Undo balancing on VL and VR, rescale */
if (ilvl) {
sggbak_("P", "L", n, &ilo, &ihi, &work[ileft], &work[iright], n, &
vl[vl_offset], ldvl, &iinfo);
if (iinfo != 0) {
*info = *n + 8;
goto L120;
}
i__1 = *n;
for (jc = 1; jc <= i__1; ++jc) {
if (alphai[jc] < 0.f) {
goto L50;
}
temp = 0.f;
if (alphai[jc] == 0.f) {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
/* Computing MAX */
r__2 = temp, r__3 = (r__1 = vl_ref(jr, jc), dabs(r__1)
);
temp = dmax(r__2,r__3);
/* L10: */
}
} else {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
/* Computing MAX */
r__3 = temp, r__4 = (r__1 = vl_ref(jr, jc), dabs(r__1)
) + (r__2 = vl_ref(jr, jc + 1), dabs(r__2));
temp = dmax(r__3,r__4);
/* L20: */
}
}
if (temp < safmin) {
goto L50;
}
temp = 1.f / temp;
if (alphai[jc] == 0.f) {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
vl_ref(jr, jc) = vl_ref(jr, jc) * temp;
/* L30: */
}
} else {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
vl_ref(jr, jc) = vl_ref(jr, jc) * temp;
vl_ref(jr, jc + 1) = vl_ref(jr, jc + 1) * temp;
/* L40: */
}
}
L50:
;
}
}
if (ilvr) {
sggbak_("P", "R", n, &ilo, &ihi, &work[ileft], &work[iright], n, &
vr[vr_offset], ldvr, &iinfo);
if (iinfo != 0) {
*info = *n + 9;
goto L120;
}
i__1 = *n;
for (jc = 1; jc <= i__1; ++jc) {
if (alphai[jc] < 0.f) {
goto L100;
}
temp = 0.f;
if (alphai[jc] == 0.f) {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
/* Computing MAX */
r__2 = temp, r__3 = (r__1 = vr_ref(jr, jc), dabs(r__1)
);
temp = dmax(r__2,r__3);
/* L60: */
}
} else {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
/* Computing MAX */
r__3 = temp, r__4 = (r__1 = vr_ref(jr, jc), dabs(r__1)
) + (r__2 = vr_ref(jr, jc + 1), dabs(r__2));
temp = dmax(r__3,r__4);
/* L70: */
}
}
if (temp < safmin) {
goto L100;
}
temp = 1.f / temp;
if (alphai[jc] == 0.f) {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
vr_ref(jr, jc) = vr_ref(jr, jc) * temp;
/* L80: */
}
} else {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
vr_ref(jr, jc) = vr_ref(jr, jc) * temp;
vr_ref(jr, jc + 1) = vr_ref(jr, jc + 1) * temp;
/* L90: */
}
}
L100:
;
}
}
/* End of eigenvector calculation */
}
/* Undo scaling in alpha, beta
Note: this does not give the alpha and beta for the unscaled
problem.
Un-scaling is limited to avoid underflow in alpha and beta
if they are significant. */
i__1 = *n;
for (jc = 1; jc <= i__1; ++jc) {
absar = (r__1 = alphar[jc], dabs(r__1));
absai = (r__1 = alphai[jc], dabs(r__1));
absb = (r__1 = beta[jc], dabs(r__1));
salfar = anrm * alphar[jc];
salfai = anrm * alphai[jc];
sbeta = bnrm * beta[jc];
ilimit = FALSE_;
scale = 1.f;
/* Check for significant underflow in ALPHAI
Computing MAX */
r__1 = safmin, r__2 = eps * absar, r__1 = max(r__1,r__2), r__2 = eps *
absb;
if (dabs(salfai) < safmin && absai >= dmax(r__1,r__2)) {
ilimit = TRUE_;
/* Computing MAX */
r__1 = onepls * safmin, r__2 = anrm2 * absai;
scale = onepls * safmin / anrm1 / dmax(r__1,r__2);
} else if (salfai == 0.f) {
/* If insignificant underflow in ALPHAI, then make the
conjugate eigenvalue real. */
if (alphai[jc] < 0.f && jc > 1) {
alphai[jc - 1] = 0.f;
} else if (alphai[jc] > 0.f && jc < *n) {
alphai[jc + 1] = 0.f;
}
}
/* Check for significant underflow in ALPHAR
Computing MAX */
r__1 = safmin, r__2 = eps * absai, r__1 = max(r__1,r__2), r__2 = eps *
absb;
if (dabs(salfar) < safmin && absar >= dmax(r__1,r__2)) {
ilimit = TRUE_;
/* Computing MAX
Computing MAX */
r__3 = onepls * safmin, r__4 = anrm2 * absar;
r__1 = scale, r__2 = onepls * safmin / anrm1 / dmax(r__3,r__4);
scale = dmax(r__1,r__2);
}
/* Check for significant underflow in BETA
Computing MAX */
r__1 = safmin, r__2 = eps * absar, r__1 = max(r__1,r__2), r__2 = eps *
absai;
if (dabs(sbeta) < safmin && absb >= dmax(r__1,r__2)) {
ilimit = TRUE_;
/* Computing MAX
Computing MAX */
r__3 = onepls * safmin, r__4 = bnrm2 * absb;
r__1 = scale, r__2 = onepls * safmin / bnrm1 / dmax(r__3,r__4);
scale = dmax(r__1,r__2);
}
/* Check for possible overflow when limiting scaling */
if (ilimit) {
/* Computing MAX */
r__1 = dabs(salfar), r__2 = dabs(salfai), r__1 = max(r__1,r__2),
r__2 = dabs(sbeta);
temp = scale * safmin * dmax(r__1,r__2);
if (temp > 1.f) {
scale /= temp;
}
if (scale < 1.f) {
ilimit = FALSE_;
}
}
/* Recompute un-scaled ALPHAR, ALPHAI, BETA if necessary. */
if (ilimit) {
salfar = scale * alphar[jc] * anrm;
salfai = scale * alphai[jc] * anrm;
sbeta = scale * beta[jc] * bnrm;
}
alphar[jc] = salfar;
alphai[jc] = salfai;
beta[jc] = sbeta;
/* L110: */
}
L120:
work[1] = (real) lwkopt;
return 0;
/* End of SGEGV */
} /* sgegv_ */
/* Subroutine */ int sggesx_(char *jobvsl, char *jobvsr, char *sort, L_fp
selctg, char *sense, integer *n, real *a, integer *lda, real *b,
integer *ldb, integer *sdim, real *alphar, real *alphai, real *beta,
real *vsl, integer *ldvsl, real *vsr, integer *ldvsr, real *rconde,
real *rcondv, real *work, integer *lwork, integer *iwork, integer *
liwork, logical *bwork, 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
=======
SGGESX computes for a pair of N-by-N real nonsymmetric matrices
(A,B), the generalized eigenvalues, the real Schur form (S,T), and,
optionally, the left and/or right matrices of Schur vectors (VSL and
VSR). This gives the generalized Schur factorization
(A,B) = ( (VSL) S (VSR)**T, (VSL) T (VSR)**T )
Optionally, it also orders the eigenvalues so that a selected cluster
of eigenvalues appears in the leading diagonal blocks of the upper
quasi-triangular matrix S and the upper triangular matrix T; computes
a reciprocal condition number for the average of the selected
eigenvalues (RCONDE); and computes a reciprocal condition number for
the right and left deflating subspaces corresponding to the selected
eigenvalues (RCONDV). The leading columns of VSL and VSR then form
an orthonormal basis for the corresponding left and right eigenspaces
(deflating subspaces).
A generalized eigenvalue for a pair of matrices (A,B) is a scalar w
or a ratio alpha/beta = w, such that A - w*B is singular. It is
usually represented as the pair (alpha,beta), as there is a
reasonable interpretation for beta=0 or for both being zero.
A pair of matrices (S,T) is in generalized real Schur form if T is
upper triangular with non-negative diagonal and S is block upper
triangular with 1-by-1 and 2-by-2 blocks. 1-by-1 blocks correspond
to real generalized eigenvalues, while 2-by-2 blocks of S will be
"standardized" by making the corresponding elements of T have the
form:
[ a 0 ]
[ 0 b ]
and the pair of corresponding 2-by-2 blocks in S and T will have a
complex conjugate pair of generalized eigenvalues.
Arguments
=========
JOBVSL (input) CHARACTER*1
= 'N': do not compute the left Schur vectors;
= 'V': compute the left Schur vectors.
JOBVSR (input) CHARACTER*1
= 'N': do not compute the right Schur vectors;
= 'V': compute the right Schur vectors.
SORT (input) CHARACTER*1
Specifies whether or not to order the eigenvalues on the
diagonal of the generalized Schur form.
= 'N': Eigenvalues are not ordered;
= 'S': Eigenvalues are ordered (see SELCTG).
SELCTG (input) LOGICAL FUNCTION of three REAL arguments
SELCTG must be declared EXTERNAL in the calling subroutine.
If SORT = 'N', SELCTG is not referenced.
If SORT = 'S', SELCTG is used to select eigenvalues to sort
to the top left of the Schur form.
An eigenvalue (ALPHAR(j)+ALPHAI(j))/BETA(j) is selected if
SELCTG(ALPHAR(j),ALPHAI(j),BETA(j)) is true; i.e. if either
one of a complex conjugate pair of eigenvalues is selected,
then both complex eigenvalues are selected.
Note that a selected complex eigenvalue may no longer satisfy
SELCTG(ALPHAR(j),ALPHAI(j),BETA(j)) = .TRUE. after ordering,
since ordering may change the value of complex eigenvalues
(especially if the eigenvalue is ill-conditioned), in this
case INFO is set to N+3.
SENSE (input) CHARACTER
Determines which reciprocal condition numbers are computed.
= 'N' : None are computed;
= 'E' : Computed for average of selected eigenvalues only;
= 'V' : Computed for selected deflating subspaces only;
= 'B' : Computed for both.
If SENSE = 'E', 'V', or 'B', SORT must equal 'S'.
N (input) INTEGER
The order of the matrices A, B, VSL, and VSR. N >= 0.
A (input/output) REAL array, dimension (LDA, N)
On entry, the first of the pair of matrices.
On exit, A has been overwritten by its generalized Schur
form S.
LDA (input) INTEGER
The leading dimension of A. LDA >= max(1,N).
B (input/output) REAL array, dimension (LDB, N)
On entry, the second of the pair of matrices.
On exit, B has been overwritten by its generalized Schur
form T.
LDB (input) INTEGER
The leading dimension of B. LDB >= max(1,N).
SDIM (output) INTEGER
If SORT = 'N', SDIM = 0.
If SORT = 'S', SDIM = number of eigenvalues (after sorting)
for which SELCTG is true. (Complex conjugate pairs for which
SELCTG is true for either eigenvalue count as 2.)
ALPHAR (output) REAL array, dimension (N)
ALPHAI (output) REAL array, dimension (N)
BETA (output) REAL array, dimension (N)
On exit, (ALPHAR(j) + ALPHAI(j)*i)/BETA(j), j=1,...,N, will
be the generalized eigenvalues. ALPHAR(j) + ALPHAI(j)*i
and BETA(j),j=1,...,N are the diagonals of the complex Schur
form (S,T) that would result if the 2-by-2 diagonal blocks of
the real Schur form of (A,B) were further reduced to
triangular form using 2-by-2 complex unitary transformations.
If ALPHAI(j) is zero, then the j-th eigenvalue is real; if
positive, then the j-th and (j+1)-st eigenvalues are a
complex conjugate pair, with ALPHAI(j+1) negative.
Note: the quotients ALPHAR(j)/BETA(j) and ALPHAI(j)/BETA(j)
may easily over- or underflow, and BETA(j) may even be zero.
Thus, the user should avoid naively computing the ratio.
However, ALPHAR and ALPHAI will be always less than and
usually comparable with norm(A) in magnitude, and BETA always
less than and usually comparable with norm(B).
VSL (output) REAL array, dimension (LDVSL,N)
If JOBVSL = 'V', VSL will contain the left Schur vectors.
Not referenced if JOBVSL = 'N'.
LDVSL (input) INTEGER
The leading dimension of the matrix VSL. LDVSL >=1, and
if JOBVSL = 'V', LDVSL >= N.
VSR (output) REAL array, dimension (LDVSR,N)
If JOBVSR = 'V', VSR will contain the right Schur vectors.
Not referenced if JOBVSR = 'N'.
LDVSR (input) INTEGER
The leading dimension of the matrix VSR. LDVSR >= 1, and
if JOBVSR = 'V', LDVSR >= N.
RCONDE (output) REAL array, dimension ( 2 )
If SENSE = 'E' or 'B', RCONDE(1) and RCONDE(2) contain the
reciprocal condition numbers for the average of the selected
eigenvalues.
Not referenced if SENSE = 'N' or 'V'.
RCONDV (output) REAL array, dimension ( 2 )
If SENSE = 'V' or 'B', RCONDV(1) and RCONDV(2) contain the
reciprocal condition numbers for the selected deflating
subspaces.
Not referenced if SENSE = 'N' or 'E'.
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. LWORK >= 8*(N+1)+16.
If SENSE = 'E', 'V', or 'B',
LWORK >= MAX( 8*(N+1)+16, 2*SDIM*(N-SDIM) ).
IWORK (workspace) INTEGER array, dimension (LIWORK)
Not referenced if SENSE = 'N'.
LIWORK (input) INTEGER
The dimension of the array WORK. LIWORK >= N+6.
BWORK (workspace) LOGICAL array, dimension (N)
Not referenced if SORT = 'N'.
INFO (output) INTEGER
= 0: successful exit
< 0: if INFO = -i, the i-th argument had an illegal value.
= 1,...,N:
The QZ iteration failed. (A,B) are not in Schur
form, but ALPHAR(j), ALPHAI(j), and BETA(j) should
be correct for j=INFO+1,...,N.
> N: =N+1: other than QZ iteration failed in SHGEQZ
=N+2: after reordering, roundoff changed values of
some complex eigenvalues so that leading
eigenvalues in the Generalized Schur form no
longer satisfy SELCTG=.TRUE. This could also
be caused due to scaling.
=N+3: reordering failed in STGSEN.
Further details
===============
An approximate (asymptotic) bound on the average absolute error of
the selected eigenvalues is
EPS * norm((A, B)) / RCONDE( 1 ).
An approximate (asymptotic) bound on the maximum angular error in
the computed deflating subspaces is
EPS * norm((A, B)) / RCONDV( 2 ).
See LAPACK User's Guide, section 4.11 for more information.
=====================================================================
Decode the input arguments
Parameter adjustments */
/* Table of constant values */
static integer c__1 = 1;
static integer c__0 = 0;
static integer c_n1 = -1;
static real c_b37 = 0.f;
static real c_b38 = 1.f;
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, vsl_dim1, vsl_offset,
vsr_dim1, vsr_offset, i__1, i__2;
real r__1;
/* Builtin functions */
double sqrt(doublereal);
/* Local variables */
static integer ijob;
static real anrm, bnrm;
static integer ierr, itau, iwrk, i__;
extern logical lsame_(char *, char *);
static integer ileft, icols;
static logical cursl, ilvsl, ilvsr;
static integer irows;
static logical lst2sl;
extern /* Subroutine */ int slabad_(real *, real *);
static integer ip;
static real pl;
extern /* Subroutine */ int sggbak_(char *, char *, integer *, integer *,
integer *, real *, real *, integer *, real *, integer *, integer *
), sggbal_(char *, integer *, real *, integer *,
real *, integer *, integer *, integer *, real *, real *, real *,
integer *);
static real pr;
static logical ilascl, ilbscl;
extern doublereal slamch_(char *), slange_(char *, integer *,
integer *, real *, integer *, real *);
static real safmin;
extern /* Subroutine */ int sgghrd_(char *, char *, integer *, integer *,
integer *, real *, integer *, real *, integer *, real *, integer *
, real *, integer *, integer *);
static real safmax;
extern /* Subroutine */ int xerbla_(char *, integer *);
static real bignum;
extern /* Subroutine */ int slascl_(char *, integer *, integer *, real *,
real *, integer *, integer *, real *, integer *, integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *, ftnlen, ftnlen);
static integer ijobvl, iright;
extern /* Subroutine */ int sgeqrf_(integer *, integer *, real *, integer
*, real *, real *, integer *, integer *);
static integer ijobvr;
extern /* Subroutine */ int slacpy_(char *, integer *, integer *, real *,
integer *, real *, integer *);
static logical wantsb, wantse, lastsl;
static integer liwmin;
static real anrmto, bnrmto;
static integer minwrk, maxwrk;
static logical wantsn;
static real smlnum;
extern /* Subroutine */ int shgeqz_(char *, char *, char *, integer *,
integer *, integer *, real *, integer *, real *, integer *, real *
, real *, real *, real *, integer *, real *, integer *, real *,
integer *, integer *), slaset_(char *,
integer *, integer *, real *, real *, real *, integer *),
sorgqr_(integer *, integer *, integer *, real *, integer *, real *
, real *, integer *, integer *), stgsen_(integer *, logical *,
logical *, logical *, integer *, real *, integer *, real *,
integer *, real *, real *, real *, real *, integer *, real *,
integer *, integer *, real *, real *, real *, real *, integer *,
integer *, integer *, integer *);
static logical wantst, wantsv;
extern /* Subroutine */ int sormqr_(char *, char *, integer *, integer *,
integer *, real *, integer *, real *, real *, integer *, real *,
integer *, integer *);
static real dif[2];
static integer ihi, ilo;
static real eps;
#define a_ref(a_1,a_2) a[(a_2)*a_dim1 + a_1]
#define b_ref(a_1,a_2) b[(a_2)*b_dim1 + a_1]
#define vsl_ref(a_1,a_2) vsl[(a_2)*vsl_dim1 + a_1]
a_dim1 = *lda;
a_offset = 1 + a_dim1 * 1;
a -= a_offset;
b_dim1 = *ldb;
b_offset = 1 + b_dim1 * 1;
b -= b_offset;
--alphar;
--alphai;
--beta;
vsl_dim1 = *ldvsl;
vsl_offset = 1 + vsl_dim1 * 1;
vsl -= vsl_offset;
vsr_dim1 = *ldvsr;
vsr_offset = 1 + vsr_dim1 * 1;
vsr -= vsr_offset;
--rconde;
--rcondv;
--work;
--iwork;
--bwork;
/* Function Body */
if (lsame_(jobvsl, "N")) {
ijobvl = 1;
ilvsl = FALSE_;
} else if (lsame_(jobvsl, "V")) {
ijobvl = 2;
ilvsl = TRUE_;
} else {
ijobvl = -1;
ilvsl = FALSE_;
}
if (lsame_(jobvsr, "N")) {
ijobvr = 1;
ilvsr = FALSE_;
} else if (lsame_(jobvsr, "V")) {
ijobvr = 2;
ilvsr = TRUE_;
} else {
ijobvr = -1;
ilvsr = FALSE_;
}
wantst = lsame_(sort, "S");
wantsn = lsame_(sense, "N");
wantse = lsame_(sense, "E");
wantsv = lsame_(sense, "V");
wantsb = lsame_(sense, "B");
if (wantsn) {
ijob = 0;
iwork[1] = 1;
} else if (wantse) {
ijob = 1;
} else if (wantsv) {
ijob = 2;
} else if (wantsb) {
ijob = 4;
}
/* Test the input arguments */
*info = 0;
if (ijobvl <= 0) {
*info = -1;
} else if (ijobvr <= 0) {
*info = -2;
} else if (! wantst && ! lsame_(sort, "N")) {
*info = -3;
} else if (! (wantsn || wantse || wantsv || wantsb) || ! wantst && !
wantsn) {
*info = -5;
} else if (*n < 0) {
*info = -6;
} else if (*lda < max(1,*n)) {
*info = -8;
} else if (*ldb < max(1,*n)) {
*info = -10;
} else if (*ldvsl < 1 || ilvsl && *ldvsl < *n) {
*info = -16;
} else if (*ldvsr < 1 || ilvsr && *ldvsr < *n) {
*info = -18;
}
/* Compute workspace
(Note: Comments in the code beginning "Workspace:" describe the
minimal amount of workspace needed at that point in the code,
as well as the preferred amount for good performance.
NB refers to the optimal block size for the immediately
following subroutine, as returned by ILAENV.) */
minwrk = 1;
if (*info == 0 && *lwork >= 1) {
minwrk = (*n + 1 << 3) + 16;
maxwrk = (*n + 1) * 7 + *n * ilaenv_(&c__1, "SGEQRF", " ", n, &c__1,
n, &c__0, (ftnlen)6, (ftnlen)1) + 16;
if (ilvsl) {
/* Computing MAX */
i__1 = maxwrk, i__2 = (*n + 1 << 3) + *n * ilaenv_(&c__1, "SORGQR"
, " ", n, &c__1, n, &c_n1, (ftnlen)6, (ftnlen)1) + 16;
maxwrk = max(i__1,i__2);
}
work[1] = (real) maxwrk;
}
if (! wantsn) {
liwmin = 1;
} else {
liwmin = *n + 6;
}
iwork[1] = liwmin;
if (*info == 0 && *lwork < minwrk) {
*info = -22;
} else if (*info == 0 && ijob >= 1) {
if (*liwork < liwmin) {
*info = -24;
}
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SGGESX", &i__1);
return 0;
}
/* Quick return if possible */
if (*n == 0) {
*sdim = 0;
return 0;
}
/* Get machine constants */
eps = slamch_("P");
safmin = slamch_("S");
safmax = 1.f / safmin;
slabad_(&safmin, &safmax);
smlnum = sqrt(safmin) / eps;
bignum = 1.f / smlnum;
/* Scale A if max element outside range [SMLNUM,BIGNUM] */
anrm = slange_("M", n, n, &a[a_offset], lda, &work[1]);
ilascl = FALSE_;
if (anrm > 0.f && anrm < smlnum) {
anrmto = smlnum;
ilascl = TRUE_;
} else if (anrm > bignum) {
anrmto = bignum;
ilascl = TRUE_;
}
if (ilascl) {
slascl_("G", &c__0, &c__0, &anrm, &anrmto, n, n, &a[a_offset], lda, &
ierr);
}
/* Scale B if max element outside range [SMLNUM,BIGNUM] */
bnrm = slange_("M", n, n, &b[b_offset], ldb, &work[1]);
ilbscl = FALSE_;
if (bnrm > 0.f && bnrm < smlnum) {
bnrmto = smlnum;
ilbscl = TRUE_;
} else if (bnrm > bignum) {
bnrmto = bignum;
ilbscl = TRUE_;
}
if (ilbscl) {
slascl_("G", &c__0, &c__0, &bnrm, &bnrmto, n, n, &b[b_offset], ldb, &
ierr);
}
/* Permute the matrix to make it more nearly triangular
(Workspace: need 6*N + 2*N for permutation parameters) */
ileft = 1;
iright = *n + 1;
iwrk = iright + *n;
sggbal_("P", n, &a[a_offset], lda, &b[b_offset], ldb, &ilo, &ihi, &work[
ileft], &work[iright], &work[iwrk], &ierr);
/* Reduce B to triangular form (QR decomposition of B)
(Workspace: need N, prefer N*NB) */
irows = ihi + 1 - ilo;
icols = *n + 1 - ilo;
itau = iwrk;
iwrk = itau + irows;
i__1 = *lwork + 1 - iwrk;
sgeqrf_(&irows, &icols, &b_ref(ilo, ilo), ldb, &work[itau], &work[iwrk], &
i__1, &ierr);
/* Apply the orthogonal transformation to matrix A
(Workspace: need N, prefer N*NB) */
i__1 = *lwork + 1 - iwrk;
sormqr_("L", "T", &irows, &icols, &irows, &b_ref(ilo, ilo), ldb, &work[
itau], &a_ref(ilo, ilo), lda, &work[iwrk], &i__1, &ierr);
/* Initialize VSL
(Workspace: need N, prefer N*NB) */
if (ilvsl) {
slaset_("Full", n, n, &c_b37, &c_b38, &vsl[vsl_offset], ldvsl);
i__1 = irows - 1;
i__2 = irows - 1;
slacpy_("L", &i__1, &i__2, &b_ref(ilo + 1, ilo), ldb, &vsl_ref(ilo +
1, ilo), ldvsl);
i__1 = *lwork + 1 - iwrk;
sorgqr_(&irows, &irows, &irows, &vsl_ref(ilo, ilo), ldvsl, &work[itau]
, &work[iwrk], &i__1, &ierr);
}
/* Initialize VSR */
if (ilvsr) {
slaset_("Full", n, n, &c_b37, &c_b38, &vsr[vsr_offset], ldvsr);
}
/* Reduce to generalized Hessenberg form
(Workspace: none needed) */
sgghrd_(jobvsl, jobvsr, n, &ilo, &ihi, &a[a_offset], lda, &b[b_offset],
ldb, &vsl[vsl_offset], ldvsl, &vsr[vsr_offset], ldvsr, &ierr);
*sdim = 0;
/* Perform QZ algorithm, computing Schur vectors if desired
(Workspace: need N) */
iwrk = itau;
i__1 = *lwork + 1 - iwrk;
shgeqz_("S", jobvsl, jobvsr, n, &ilo, &ihi, &a[a_offset], lda, &b[
b_offset], ldb, &alphar[1], &alphai[1], &beta[1], &vsl[vsl_offset]
, ldvsl, &vsr[vsr_offset], ldvsr, &work[iwrk], &i__1, &ierr);
if (ierr != 0) {
if (ierr > 0 && ierr <= *n) {
*info = ierr;
} else if (ierr > *n && ierr <= *n << 1) {
*info = ierr - *n;
} else {
*info = *n + 1;
}
goto L50;
}
/* Sort eigenvalues ALPHA/BETA and compute the reciprocal of
condition number(s)
(Workspace: If IJOB >= 1, need MAX( 8*(N+1), 2*SDIM*(N-SDIM) )
otherwise, need 8*(N+1) ) */
if (wantst) {
/* Undo scaling on eigenvalues before SELCTGing */
if (ilascl) {
slascl_("G", &c__0, &c__0, &anrmto, &anrm, n, &c__1, &alphar[1],
n, &ierr);
slascl_("G", &c__0, &c__0, &anrmto, &anrm, n, &c__1, &alphai[1],
n, &ierr);
}
if (ilbscl) {
slascl_("G", &c__0, &c__0, &bnrmto, &bnrm, n, &c__1, &beta[1], n,
&ierr);
}
/* Select eigenvalues */
i__1 = *n;
for (i__ = 1; i__ <= i__1; ++i__) {
bwork[i__] = (*selctg)(&alphar[i__], &alphai[i__], &beta[i__]);
/* L10: */
}
/* Reorder eigenvalues, transform Generalized Schur vectors, and
compute reciprocal condition numbers */
i__1 = *lwork - iwrk + 1;
stgsen_(&ijob, &ilvsl, &ilvsr, &bwork[1], n, &a[a_offset], lda, &b[
b_offset], ldb, &alphar[1], &alphai[1], &beta[1], &vsl[
vsl_offset], ldvsl, &vsr[vsr_offset], ldvsr, sdim, &pl, &pr,
dif, &work[iwrk], &i__1, &iwork[1], liwork, &ierr);
if (ijob >= 1) {
/* Computing MAX */
i__1 = maxwrk, i__2 = (*sdim << 1) * (*n - *sdim);
maxwrk = max(i__1,i__2);
}
if (ierr == -22) {
/* not enough real workspace */
*info = -22;
} else {
rconde[1] = pl;
rconde[2] = pr;
rcondv[1] = dif[0];
rcondv[2] = dif[1];
if (ierr == 1) {
*info = *n + 3;
}
}
}
/* Apply permutation to VSL and VSR
(Workspace: none needed) */
if (ilvsl) {
sggbak_("P", "L", n, &ilo, &ihi, &work[ileft], &work[iright], n, &vsl[
vsl_offset], ldvsl, &ierr);
}
if (ilvsr) {
sggbak_("P", "R", n, &ilo, &ihi, &work[ileft], &work[iright], n, &vsr[
vsr_offset], ldvsr, &ierr);
}
/* Check if unscaling would cause over/underflow, if so, rescale
(ALPHAR(I),ALPHAI(I),BETA(I)) so BETA(I) is on the order of
B(I,I) and ALPHAR(I) and ALPHAI(I) are on the order of A(I,I) */
if (ilascl) {
i__1 = *n;
for (i__ = 1; i__ <= i__1; ++i__) {
if (alphai[i__] != 0.f) {
if (alphar[i__] / safmax > anrmto / anrm || safmin / alphar[
i__] > anrm / anrmto) {
work[1] = (r__1 = a_ref(i__, i__) / alphar[i__], dabs(
r__1));
beta[i__] *= work[1];
alphar[i__] *= work[1];
alphai[i__] *= work[1];
} else if (alphai[i__] / safmax > anrmto / anrm || safmin /
alphai[i__] > anrm / anrmto) {
work[1] = (r__1 = a_ref(i__, i__ + 1) / alphai[i__], dabs(
r__1));
beta[i__] *= work[1];
alphar[i__] *= work[1];
alphai[i__] *= work[1];
}
}
/* L20: */
}
}
if (ilbscl) {
i__1 = *n;
for (i__ = 1; i__ <= i__1; ++i__) {
if (alphai[i__] != 0.f) {
if (beta[i__] / safmax > bnrmto / bnrm || safmin / beta[i__]
> bnrm / bnrmto) {
work[1] = (r__1 = b_ref(i__, i__) / beta[i__], dabs(r__1))
;
beta[i__] *= work[1];
alphar[i__] *= work[1];
alphai[i__] *= work[1];
}
}
/* L25: */
}
}
/* Undo scaling */
if (ilascl) {
slascl_("H", &c__0, &c__0, &anrmto, &anrm, n, n, &a[a_offset], lda, &
ierr);
slascl_("G", &c__0, &c__0, &anrmto, &anrm, n, &c__1, &alphar[1], n, &
ierr);
slascl_("G", &c__0, &c__0, &anrmto, &anrm, n, &c__1, &alphai[1], n, &
ierr);
}
if (ilbscl) {
slascl_("U", &c__0, &c__0, &bnrmto, &bnrm, n, n, &b[b_offset], ldb, &
ierr);
slascl_("G", &c__0, &c__0, &bnrmto, &bnrm, n, &c__1, &beta[1], n, &
ierr);
}
/* L30: */
if (wantst) {
/* Check if reordering is correct */
lastsl = TRUE_;
lst2sl = TRUE_;
*sdim = 0;
ip = 0;
i__1 = *n;
for (i__ = 1; i__ <= i__1; ++i__) {
cursl = (*selctg)(&alphar[i__], &alphai[i__], &beta[i__]);
if (alphai[i__] == 0.f) {
if (cursl) {
++(*sdim);
}
ip = 0;
if (cursl && ! lastsl) {
*info = *n + 2;
}
} else {
if (ip == 1) {
/* Last eigenvalue of conjugate pair */
cursl = cursl || lastsl;
lastsl = cursl;
if (cursl) {
*sdim += 2;
}
ip = -1;
if (cursl && ! lst2sl) {
*info = *n + 2;
}
} else {
/* First eigenvalue of conjugate pair */
ip = 1;
}
}
lst2sl = lastsl;
lastsl = cursl;
/* L40: */
}
}
L50:
work[1] = (real) maxwrk;
iwork[1] = liwmin;
return 0;
/* End of SGGESX */
} /* sggesx_ */
/* Subroutine */ int sgels_(char *trans, integer *m, integer *n, integer *
nrhs, real *a, integer *lda, real *b, integer *ldb, real *work,
integer *lwork, integer *info, ftnlen trans_len)
{
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, i__1, i__2;
/* Local variables */
static integer i__, j, nb, mn;
static real anrm, bnrm;
static integer brow;
static logical tpsd;
static integer iascl, ibscl;
extern logical lsame_(char *, char *, ftnlen, ftnlen);
static integer wsize;
static real rwork[1];
extern /* Subroutine */ int strsm_(char *, char *, char *, char *,
integer *, integer *, real *, real *, integer *, real *, integer *
, ftnlen, ftnlen, ftnlen, ftnlen), slabad_(real *, real *);
extern doublereal slamch_(char *, ftnlen), slange_(char *, integer *,
integer *, real *, integer *, real *, ftnlen);
extern /* Subroutine */ int xerbla_(char *, integer *, ftnlen);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *, ftnlen, ftnlen);
static integer scllen;
static real bignum;
extern /* Subroutine */ int sgelqf_(integer *, integer *, real *, integer
*, real *, real *, integer *, integer *), slascl_(char *, integer
*, integer *, real *, real *, integer *, integer *, real *,
integer *, integer *, ftnlen), sgeqrf_(integer *, integer *, real
*, integer *, real *, real *, integer *, integer *), slaset_(char
*, integer *, integer *, real *, real *, real *, integer *,
ftnlen);
static real smlnum;
extern /* Subroutine */ int sormlq_(char *, char *, integer *, integer *,
integer *, real *, integer *, real *, real *, integer *, real *,
integer *, integer *, ftnlen, ftnlen);
static logical lquery;
extern /* Subroutine */ int sormqr_(char *, char *, integer *, integer *,
integer *, real *, integer *, real *, real *, integer *, real *,
integer *, integer *, ftnlen, ftnlen);
/* -- 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 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* SGELS solves overdetermined or underdetermined real linear systems */
/* involving an M-by-N matrix A, or its transpose, using a QR or LQ */
/* factorization of A. It is assumed that A has full rank. */
/* The following options are provided: */
/* 1. If TRANS = 'N' and m >= n: find the least squares solution of */
/* an overdetermined system, i.e., solve the least squares problem */
/* minimize || B - A*X ||. */
/* 2. If TRANS = 'N' and m < n: find the minimum norm solution of */
/* an underdetermined system A * X = B. */
/* 3. If TRANS = 'T' and m >= n: find the minimum norm solution of */
/* an undetermined system A**T * X = B. */
/* 4. If TRANS = 'T' and m < n: find the least squares solution of */
/* an overdetermined system, i.e., solve the least squares problem */
/* minimize || B - A**T * X ||. */
/* Several right hand side vectors b and solution vectors x can be */
/* handled in a single call; they are stored as the columns of the */
/* M-by-NRHS right hand side matrix B and the N-by-NRHS solution */
/* matrix X. */
/* Arguments */
/* ========= */
/* TRANS (input) CHARACTER */
/* = 'N': the linear system involves A; */
/* = 'T': the linear system involves A**T. */
/* M (input) INTEGER */
/* The number of rows of the matrix A. M >= 0. */
/* N (input) INTEGER */
/* The number of columns of the matrix A. N >= 0. */
/* NRHS (input) INTEGER */
/* The number of right hand sides, i.e., the number of */
/* columns of the matrices B and X. NRHS >=0. */
/* A (input/output) REAL array, dimension (LDA,N) */
/* On entry, the M-by-N matrix A. */
/* On exit, */
/* if M >= N, A is overwritten by details of its QR */
/* factorization as returned by SGEQRF; */
/* if M < N, A is overwritten by details of its LQ */
/* factorization as returned by SGELQF. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,M). */
/* B (input/output) REAL array, dimension (LDB,NRHS) */
/* On entry, the matrix B of right hand side vectors, stored */
/* columnwise; B is M-by-NRHS if TRANS = 'N', or N-by-NRHS */
/* if TRANS = 'T'. */
/* On exit, B is overwritten by the solution vectors, stored */
/* columnwise: */
/* if TRANS = 'N' and m >= n, rows 1 to n of B contain the least */
/* squares solution vectors; the residual sum of squares for the */
/* solution in each column is given by the sum of squares of */
/* elements N+1 to M in that column; */
/* if TRANS = 'N' and m < n, rows 1 to N of B contain the */
/* minimum norm solution vectors; */
/* if TRANS = 'T' and m >= n, rows 1 to M of B contain the */
/* minimum norm solution vectors; */
/* if TRANS = 'T' and m < n, rows 1 to M of B contain the */
/* least squares solution vectors; the residual sum of squares */
/* for the solution in each column is given by the sum of */
/* squares of elements M+1 to N in that column. */
/* LDB (input) INTEGER */
/* The leading dimension of the array B. LDB >= MAX(1,M,N). */
/* 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. */
/* LWORK >= max( 1, MN + max( MN, NRHS ) ). */
/* For optimal performance, */
/* LWORK >= max( 1, MN + max( MN, NRHS )*NB ). */
/* where MN = min(M,N) and NB is the optimum block size. */
/* 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. */
/* INFO (output) INTEGER */
/* = 0: successful exit */
/* < 0: if INFO = -i, the i-th argument had an illegal value */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. Local Arrays .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Test the input arguments. */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
b_dim1 = *ldb;
b_offset = 1 + b_dim1;
b -= b_offset;
--work;
/* Function Body */
*info = 0;
mn = min(*m,*n);
lquery = *lwork == -1;
if (! (lsame_(trans, "N", (ftnlen)1, (ftnlen)1) || lsame_(trans, "T", (
ftnlen)1, (ftnlen)1))) {
*info = -1;
} else if (*m < 0) {
*info = -2;
} else if (*n < 0) {
*info = -3;
} else if (*nrhs < 0) {
*info = -4;
} else if (*lda < max(1,*m)) {
*info = -6;
} else { /* if(complicated condition) */
/* Computing MAX */
i__1 = max(1,*m);
if (*ldb < max(i__1,*n)) {
*info = -8;
} else { /* if(complicated condition) */
/* Computing MAX */
i__1 = 1, i__2 = mn + max(mn,*nrhs);
if (*lwork < max(i__1,i__2) && ! lquery) {
*info = -10;
}
}
}
/* Figure out optimal block size */
if (*info == 0 || *info == -10) {
tpsd = TRUE_;
if (lsame_(trans, "N", (ftnlen)1, (ftnlen)1)) {
tpsd = FALSE_;
}
if (*m >= *n) {
nb = ilaenv_(&c__1, "SGEQRF", " ", m, n, &c_n1, &c_n1, (ftnlen)6,
(ftnlen)1);
if (tpsd) {
/* Computing MAX */
i__1 = nb, i__2 = ilaenv_(&c__1, "SORMQR", "LN", m, nrhs, n, &
c_n1, (ftnlen)6, (ftnlen)2);
nb = max(i__1,i__2);
} else {
/* Computing MAX */
i__1 = nb, i__2 = ilaenv_(&c__1, "SORMQR", "LT", m, nrhs, n, &
c_n1, (ftnlen)6, (ftnlen)2);
nb = max(i__1,i__2);
}
} else {
nb = ilaenv_(&c__1, "SGELQF", " ", m, n, &c_n1, &c_n1, (ftnlen)6,
(ftnlen)1);
if (tpsd) {
/* Computing MAX */
i__1 = nb, i__2 = ilaenv_(&c__1, "SORMLQ", "LT", n, nrhs, m, &
c_n1, (ftnlen)6, (ftnlen)2);
nb = max(i__1,i__2);
} else {
/* Computing MAX */
i__1 = nb, i__2 = ilaenv_(&c__1, "SORMLQ", "LN", n, nrhs, m, &
c_n1, (ftnlen)6, (ftnlen)2);
nb = max(i__1,i__2);
}
}
/* Computing MAX */
i__1 = 1, i__2 = mn + max(mn,*nrhs) * nb;
wsize = max(i__1,i__2);
work[1] = (real) wsize;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SGELS ", &i__1, (ftnlen)6);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
/* Computing MIN */
i__1 = min(*m,*n);
if (min(i__1,*nrhs) == 0) {
i__1 = max(*m,*n);
slaset_("Full", &i__1, nrhs, &c_b33, &c_b33, &b[b_offset], ldb, (
ftnlen)4);
return 0;
}
/* Get machine parameters */
smlnum = slamch_("S", (ftnlen)1) / slamch_("P", (ftnlen)1);
bignum = 1.f / smlnum;
slabad_(&smlnum, &bignum);
/* Scale A, B if max element outside range [SMLNUM,BIGNUM] */
anrm = slange_("M", m, n, &a[a_offset], lda, rwork, (ftnlen)1);
iascl = 0;
if (anrm > 0.f && anrm < smlnum) {
/* Scale matrix norm up to SMLNUM */
slascl_("G", &c__0, &c__0, &anrm, &smlnum, m, n, &a[a_offset], lda,
info, (ftnlen)1);
iascl = 1;
} else if (anrm > bignum) {
/* Scale matrix norm down to BIGNUM */
slascl_("G", &c__0, &c__0, &anrm, &bignum, m, n, &a[a_offset], lda,
info, (ftnlen)1);
iascl = 2;
} else if (anrm == 0.f) {
/* Matrix all zero. Return zero solution. */
i__1 = max(*m,*n);
slaset_("F", &i__1, nrhs, &c_b33, &c_b33, &b[b_offset], ldb, (ftnlen)
1);
goto L50;
}
brow = *m;
if (tpsd) {
brow = *n;
}
bnrm = slange_("M", &brow, nrhs, &b[b_offset], ldb, rwork, (ftnlen)1);
ibscl = 0;
if (bnrm > 0.f && bnrm < smlnum) {
/* Scale matrix norm up to SMLNUM */
slascl_("G", &c__0, &c__0, &bnrm, &smlnum, &brow, nrhs, &b[b_offset],
ldb, info, (ftnlen)1);
ibscl = 1;
} else if (bnrm > bignum) {
/* Scale matrix norm down to BIGNUM */
slascl_("G", &c__0, &c__0, &bnrm, &bignum, &brow, nrhs, &b[b_offset],
ldb, info, (ftnlen)1);
ibscl = 2;
}
if (*m >= *n) {
/* compute QR factorization of A */
i__1 = *lwork - mn;
sgeqrf_(m, n, &a[a_offset], lda, &work[1], &work[mn + 1], &i__1, info)
;
/* workspace at least N, optimally N*NB */
if (! tpsd) {
/* Least-Squares Problem min || A * X - B || */
/* B(1:M,1:NRHS) := Q' * B(1:M,1:NRHS) */
i__1 = *lwork - mn;
sormqr_("Left", "Transpose", m, nrhs, n, &a[a_offset], lda, &work[
1], &b[b_offset], ldb, &work[mn + 1], &i__1, info, (
ftnlen)4, (ftnlen)9);
/* workspace at least NRHS, optimally NRHS*NB */
/* B(1:N,1:NRHS) := inv(R) * B(1:N,1:NRHS) */
strsm_("Left", "Upper", "No transpose", "Non-unit", n, nrhs, &
c_b61, &a[a_offset], lda, &b[b_offset], ldb, (ftnlen)4, (
ftnlen)5, (ftnlen)12, (ftnlen)8);
scllen = *n;
} else {
/* Overdetermined system of equations A' * X = B */
/* B(1:N,1:NRHS) := inv(R') * B(1:N,1:NRHS) */
strsm_("Left", "Upper", "Transpose", "Non-unit", n, nrhs, &c_b61,
&a[a_offset], lda, &b[b_offset], ldb, (ftnlen)4, (ftnlen)
5, (ftnlen)9, (ftnlen)8);
/* B(N+1:M,1:NRHS) = ZERO */
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
i__2 = *m;
for (i__ = *n + 1; i__ <= i__2; ++i__) {
b[i__ + j * b_dim1] = 0.f;
/* L10: */
}
/* L20: */
}
/* B(1:M,1:NRHS) := Q(1:N,:) * B(1:N,1:NRHS) */
i__1 = *lwork - mn;
sormqr_("Left", "No transpose", m, nrhs, n, &a[a_offset], lda, &
work[1], &b[b_offset], ldb, &work[mn + 1], &i__1, info, (
ftnlen)4, (ftnlen)12);
/* workspace at least NRHS, optimally NRHS*NB */
scllen = *m;
}
} else {
/* Compute LQ factorization of A */
i__1 = *lwork - mn;
sgelqf_(m, n, &a[a_offset], lda, &work[1], &work[mn + 1], &i__1, info)
;
/* workspace at least M, optimally M*NB. */
if (! tpsd) {
/* underdetermined system of equations A * X = B */
/* B(1:M,1:NRHS) := inv(L) * B(1:M,1:NRHS) */
strsm_("Left", "Lower", "No transpose", "Non-unit", m, nrhs, &
c_b61, &a[a_offset], lda, &b[b_offset], ldb, (ftnlen)4, (
ftnlen)5, (ftnlen)12, (ftnlen)8);
/* B(M+1:N,1:NRHS) = 0 */
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
i__2 = *n;
for (i__ = *m + 1; i__ <= i__2; ++i__) {
b[i__ + j * b_dim1] = 0.f;
/* L30: */
}
/* L40: */
}
/* B(1:N,1:NRHS) := Q(1:N,:)' * B(1:M,1:NRHS) */
i__1 = *lwork - mn;
sormlq_("Left", "Transpose", n, nrhs, m, &a[a_offset], lda, &work[
1], &b[b_offset], ldb, &work[mn + 1], &i__1, info, (
ftnlen)4, (ftnlen)9);
/* workspace at least NRHS, optimally NRHS*NB */
scllen = *n;
} else {
/* overdetermined system min || A' * X - B || */
/* B(1:N,1:NRHS) := Q * B(1:N,1:NRHS) */
i__1 = *lwork - mn;
sormlq_("Left", "No transpose", n, nrhs, m, &a[a_offset], lda, &
work[1], &b[b_offset], ldb, &work[mn + 1], &i__1, info, (
ftnlen)4, (ftnlen)12);
/* workspace at least NRHS, optimally NRHS*NB */
/* B(1:M,1:NRHS) := inv(L') * B(1:M,1:NRHS) */
strsm_("Left", "Lower", "Transpose", "Non-unit", m, nrhs, &c_b61,
&a[a_offset], lda, &b[b_offset], ldb, (ftnlen)4, (ftnlen)
5, (ftnlen)9, (ftnlen)8);
scllen = *m;
}
}
/* Undo scaling */
if (iascl == 1) {
slascl_("G", &c__0, &c__0, &anrm, &smlnum, &scllen, nrhs, &b[b_offset]
, ldb, info, (ftnlen)1);
} else if (iascl == 2) {
slascl_("G", &c__0, &c__0, &anrm, &bignum, &scllen, nrhs, &b[b_offset]
, ldb, info, (ftnlen)1);
}
if (ibscl == 1) {
slascl_("G", &c__0, &c__0, &smlnum, &bnrm, &scllen, nrhs, &b[b_offset]
, ldb, info, (ftnlen)1);
} else if (ibscl == 2) {
slascl_("G", &c__0, &c__0, &bignum, &bnrm, &scllen, nrhs, &b[b_offset]
, ldb, info, (ftnlen)1);
}
L50:
work[1] = (real) wsize;
return 0;
/* End of SGELS */
} /* sgels_ */
/* Subroutine */ int sgesvx_(char *fact, char *trans, integer *n, integer *
nrhs, real *a, integer *lda, real *af, integer *ldaf, integer *ipiv,
char *equed, real *r__, real *c__, real *b, integer *ldb, real *x,
integer *ldx, real *rcond, real *ferr, real *berr, real *work,
integer *iwork, 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
=======
SGESVX uses the LU factorization to compute the solution to a real
system of linear equations
A * X = B,
where A is an N-by-N matrix and X and B are N-by-NRHS matrices.
Error bounds on the solution and a condition estimate are also
provided.
Description
===========
The following steps are performed:
1. If FACT = 'E', real scaling factors are computed to equilibrate
the system:
TRANS = 'N': diag(R)*A*diag(C) *inv(diag(C))*X = diag(R)*B
TRANS = 'T': (diag(R)*A*diag(C))**T *inv(diag(R))*X = diag(C)*B
TRANS = 'C': (diag(R)*A*diag(C))**H *inv(diag(R))*X = diag(C)*B
Whether or not the system will be equilibrated depends on the
scaling of the matrix A, but if equilibration is used, A is
overwritten by diag(R)*A*diag(C) and B by diag(R)*B (if TRANS='N')
or diag(C)*B (if TRANS = 'T' or 'C').
2. If FACT = 'N' or 'E', the LU decomposition is used to factor the
matrix A (after equilibration if FACT = 'E') as
A = P * L * U,
where P is a permutation matrix, L is a unit lower triangular
matrix, and U is upper triangular.
3. If some U(i,i)=0, so that U is exactly singular, then the routine
returns with INFO = i. Otherwise, the factored form of A is used
to estimate the condition number of the matrix A. If the
reciprocal of the condition number is less than machine precision,
INFO = N+1 is returned as a warning, but the routine still goes on
to solve for X and compute error bounds as described below.
4. The system of equations is solved for X using the factored form
of A.
5. Iterative refinement is applied to improve the computed solution
matrix and calculate error bounds and backward error estimates
for it.
6. If equilibration was used, the matrix X is premultiplied by
diag(C) (if TRANS = 'N') or diag(R) (if TRANS = 'T' or 'C') so
that it solves the original system before equilibration.
Arguments
=========
FACT (input) CHARACTER*1
Specifies whether or not the factored form of the matrix A is
supplied on entry, and if not, whether the matrix A should be
equilibrated before it is factored.
= 'F': On entry, AF and IPIV contain the factored form of A.
If EQUED is not 'N', the matrix A has been
equilibrated with scaling factors given by R and C.
A, AF, and IPIV are not modified.
= 'N': The matrix A will be copied to AF and factored.
= 'E': The matrix A will be equilibrated if necessary, then
copied to AF and factored.
TRANS (input) CHARACTER*1
Specifies the form of the system of equations:
= 'N': A * X = B (No transpose)
= 'T': A**T * X = B (Transpose)
= 'C': A**H * X = B (Transpose)
N (input) INTEGER
The number of linear equations, i.e., the order of the
matrix A. N >= 0.
NRHS (input) INTEGER
The number of right hand sides, i.e., the number of columns
of the matrices B and X. NRHS >= 0.
A (input/output) REAL array, dimension (LDA,N)
On entry, the N-by-N matrix A. If FACT = 'F' and EQUED is
not 'N', then A must have been equilibrated by the scaling
factors in R and/or C. A is not modified if FACT = 'F' or
'N', or if FACT = 'E' and EQUED = 'N' on exit.
On exit, if EQUED .ne. 'N', A is scaled as follows:
EQUED = 'R': A := diag(R) * A
EQUED = 'C': A := A * diag(C)
EQUED = 'B': A := diag(R) * A * diag(C).
LDA (input) INTEGER
The leading dimension of the array A. LDA >= max(1,N).
AF (input or output) REAL array, dimension (LDAF,N)
If FACT = 'F', then AF is an input argument and on entry
contains the factors L and U from the factorization
A = P*L*U as computed by SGETRF. If EQUED .ne. 'N', then
AF is the factored form of the equilibrated matrix A.
If FACT = 'N', then AF is an output argument and on exit
returns the factors L and U from the factorization A = P*L*U
of the original matrix A.
If FACT = 'E', then AF is an output argument and on exit
returns the factors L and U from the factorization A = P*L*U
of the equilibrated matrix A (see the description of A for
the form of the equilibrated matrix).
LDAF (input) INTEGER
The leading dimension of the array AF. LDAF >= max(1,N).
IPIV (input or output) INTEGER array, dimension (N)
If FACT = 'F', then IPIV is an input argument and on entry
contains the pivot indices from the factorization A = P*L*U
as computed by SGETRF; row i of the matrix was interchanged
with row IPIV(i).
If FACT = 'N', then IPIV is an output argument and on exit
contains the pivot indices from the factorization A = P*L*U
of the original matrix A.
If FACT = 'E', then IPIV is an output argument and on exit
contains the pivot indices from the factorization A = P*L*U
of the equilibrated matrix A.
EQUED (input or output) CHARACTER*1
Specifies the form of equilibration that was done.
= 'N': No equilibration (always true if FACT = 'N').
= 'R': Row equilibration, i.e., A has been premultiplied by
diag(R).
= 'C': Column equilibration, i.e., A has been postmultiplied
by diag(C).
= 'B': Both row and column equilibration, i.e., A has been
replaced by diag(R) * A * diag(C).
EQUED is an input argument if FACT = 'F'; otherwise, it is an
output argument.
R (input or output) REAL array, dimension (N)
The row scale factors for A. If EQUED = 'R' or 'B', A is
multiplied on the left by diag(R); if EQUED = 'N' or 'C', R
is not accessed. R is an input argument if FACT = 'F';
otherwise, R is an output argument. If FACT = 'F' and
EQUED = 'R' or 'B', each element of R must be positive.
C (input or output) REAL array, dimension (N)
The column scale factors for A. If EQUED = 'C' or 'B', A is
multiplied on the right by diag(C); if EQUED = 'N' or 'R', C
is not accessed. C is an input argument if FACT = 'F';
otherwise, C is an output argument. If FACT = 'F' and
EQUED = 'C' or 'B', each element of C must be positive.
B (input/output) REAL array, dimension (LDB,NRHS)
On entry, the N-by-NRHS right hand side matrix B.
On exit,
if EQUED = 'N', B is not modified;
if TRANS = 'N' and EQUED = 'R' or 'B', B is overwritten by
diag(R)*B;
if TRANS = 'T' or 'C' and EQUED = 'C' or 'B', B is
overwritten by diag(C)*B.
LDB (input) INTEGER
The leading dimension of the array B. LDB >= max(1,N).
X (output) REAL array, dimension (LDX,NRHS)
If INFO = 0 or INFO = N+1, the N-by-NRHS solution matrix X
to the original system of equations. Note that A and B are
modified on exit if EQUED .ne. 'N', and the solution to the
equilibrated system is inv(diag(C))*X if TRANS = 'N' and
EQUED = 'C' or 'B', or inv(diag(R))*X if TRANS = 'T' or 'C'
and EQUED = 'R' or 'B'.
LDX (input) INTEGER
The leading dimension of the array X. LDX >= max(1,N).
RCOND (output) REAL
The estimate of the reciprocal condition number of the matrix
A after equilibration (if done). If RCOND is less than the
machine precision (in particular, if RCOND = 0), the matrix
is singular to working precision. This condition is
indicated by a return code of INFO > 0.
FERR (output) REAL array, dimension (NRHS)
The estimated forward error bound for each solution vector
X(j) (the j-th column of the solution matrix X).
If XTRUE is the true solution corresponding to X(j), FERR(j)
is an estimated upper bound for the magnitude of the largest
element in (X(j) - XTRUE) divided by the magnitude of the
largest element in X(j). The estimate is as reliable as
the estimate for RCOND, and is almost always a slight
overestimate of the true error.
BERR (output) REAL array, dimension (NRHS)
The componentwise relative backward error of each solution
vector X(j) (i.e., the smallest relative change in
any element of A or B that makes X(j) an exact solution).
WORK (workspace/output) REAL array, dimension (4*N)
On exit, WORK(1) contains the reciprocal pivot growth
factor norm(A)/norm(U). The "max absolute element" norm is
used. If WORK(1) is much less than 1, then the stability
of the LU factorization of the (equilibrated) matrix A
could be poor. This also means that the solution X, condition
estimator RCOND, and forward error bound FERR could be
unreliable. If factorization fails with 0<INFO<=N, then
WORK(1) contains the reciprocal pivot growth factor for the
leading INFO columns of A.
IWORK (workspace) INTEGER array, dimension (N)
INFO (output) INTEGER
= 0: successful exit
< 0: if INFO = -i, the i-th argument had an illegal value
> 0: if INFO = i, and i is
<= N: U(i,i) is exactly zero. The factorization has
been completed, but the factor U is exactly
singular, so the solution and error bounds
could not be computed. RCOND = 0 is returned.
= N+1: U is nonsingular, but RCOND is less than machine
precision, meaning that the matrix is singular
to working precision. Nevertheless, the
solution and error bounds are computed because
there are a number of situations where the
computed solution can be more accurate than the
value of RCOND would suggest.
=====================================================================
Parameter adjustments */
/* System generated locals */
integer a_dim1, a_offset, af_dim1, af_offset, b_dim1, b_offset, x_dim1,
x_offset, i__1, i__2;
real r__1, r__2;
/* Local variables */
static real amax;
static char norm[1];
static integer i__, j;
extern logical lsame_(char *, char *);
static real rcmin, rcmax, anorm;
static logical equil;
static real colcnd;
extern doublereal slamch_(char *), slange_(char *, integer *,
integer *, real *, integer *, real *);
static logical nofact;
extern /* Subroutine */ int slaqge_(integer *, integer *, real *, integer
*, real *, real *, real *, real *, real *, char *),
xerbla_(char *, integer *), sgecon_(char *, integer *,
real *, integer *, real *, real *, real *, integer *, integer *);
static real bignum;
static integer infequ;
static logical colequ;
extern /* Subroutine */ int sgeequ_(integer *, integer *, real *, integer
*, real *, real *, real *, real *, real *, integer *), sgerfs_(
char *, integer *, integer *, real *, integer *, real *, integer *
, integer *, real *, integer *, real *, integer *, real *, real *,
real *, integer *, integer *), sgetrf_(integer *,
integer *, real *, integer *, integer *, integer *);
static real rowcnd;
extern /* Subroutine */ int slacpy_(char *, integer *, integer *, real *,
integer *, real *, integer *);
static logical notran;
extern doublereal slantr_(char *, char *, char *, integer *, integer *,
real *, integer *, real *);
extern /* Subroutine */ int sgetrs_(char *, integer *, integer *, real *,
integer *, integer *, real *, integer *, integer *);
static real smlnum;
static logical rowequ;
static real rpvgrw;
#define b_ref(a_1,a_2) b[(a_2)*b_dim1 + a_1]
#define x_ref(a_1,a_2) x[(a_2)*x_dim1 + a_1]
a_dim1 = *lda;
a_offset = 1 + a_dim1 * 1;
a -= a_offset;
af_dim1 = *ldaf;
af_offset = 1 + af_dim1 * 1;
af -= af_offset;
--ipiv;
--r__;
--c__;
b_dim1 = *ldb;
b_offset = 1 + b_dim1 * 1;
b -= b_offset;
x_dim1 = *ldx;
x_offset = 1 + x_dim1 * 1;
x -= x_offset;
--ferr;
--berr;
--work;
--iwork;
/* Function Body */
*info = 0;
nofact = lsame_(fact, "N");
equil = lsame_(fact, "E");
notran = lsame_(trans, "N");
if (nofact || equil) {
*(unsigned char *)equed = 'N';
rowequ = FALSE_;
colequ = FALSE_;
} else {
rowequ = lsame_(equed, "R") || lsame_(equed,
"B");
colequ = lsame_(equed, "C") || lsame_(equed,
"B");
smlnum = slamch_("Safe minimum");
bignum = 1.f / smlnum;
}
/* Test the input parameters. */
if (! nofact && ! equil && ! lsame_(fact, "F")) {
*info = -1;
} else if (! notran && ! lsame_(trans, "T") && !
lsame_(trans, "C")) {
*info = -2;
} else if (*n < 0) {
*info = -3;
} else if (*nrhs < 0) {
*info = -4;
} else if (*lda < max(1,*n)) {
*info = -6;
} else if (*ldaf < max(1,*n)) {
*info = -8;
} else if (lsame_(fact, "F") && ! (rowequ || colequ
|| lsame_(equed, "N"))) {
*info = -10;
} else {
if (rowequ) {
rcmin = bignum;
rcmax = 0.f;
i__1 = *n;
for (j = 1; j <= i__1; ++j) {
/* Computing MIN */
r__1 = rcmin, r__2 = r__[j];
rcmin = dmin(r__1,r__2);
/* Computing MAX */
r__1 = rcmax, r__2 = r__[j];
rcmax = dmax(r__1,r__2);
/* L10: */
}
if (rcmin <= 0.f) {
*info = -11;
} else if (*n > 0) {
rowcnd = dmax(rcmin,smlnum) / dmin(rcmax,bignum);
} else {
rowcnd = 1.f;
}
}
if (colequ && *info == 0) {
rcmin = bignum;
rcmax = 0.f;
i__1 = *n;
for (j = 1; j <= i__1; ++j) {
/* Computing MIN */
r__1 = rcmin, r__2 = c__[j];
rcmin = dmin(r__1,r__2);
/* Computing MAX */
r__1 = rcmax, r__2 = c__[j];
rcmax = dmax(r__1,r__2);
/* L20: */
}
if (rcmin <= 0.f) {
*info = -12;
} else if (*n > 0) {
colcnd = dmax(rcmin,smlnum) / dmin(rcmax,bignum);
} else {
colcnd = 1.f;
}
}
if (*info == 0) {
if (*ldb < max(1,*n)) {
*info = -14;
} else if (*ldx < max(1,*n)) {
*info = -16;
}
}
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SGESVX", &i__1);
return 0;
}
if (equil) {
/* Compute row and column scalings to equilibrate the matrix A. */
sgeequ_(n, n, &a[a_offset], lda, &r__[1], &c__[1], &rowcnd, &colcnd, &
amax, &infequ);
if (infequ == 0) {
/* Equilibrate the matrix. */
slaqge_(n, n, &a[a_offset], lda, &r__[1], &c__[1], &rowcnd, &
colcnd, &amax, equed);
rowequ = lsame_(equed, "R") || lsame_(equed,
"B");
colequ = lsame_(equed, "C") || lsame_(equed,
"B");
}
}
/* Scale the right hand side. */
if (notran) {
if (rowequ) {
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
i__2 = *n;
for (i__ = 1; i__ <= i__2; ++i__) {
b_ref(i__, j) = r__[i__] * b_ref(i__, j);
/* L30: */
}
/* L40: */
}
}
} else if (colequ) {
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
i__2 = *n;
for (i__ = 1; i__ <= i__2; ++i__) {
b_ref(i__, j) = c__[i__] * b_ref(i__, j);
/* L50: */
}
/* L60: */
}
}
if (nofact || equil) {
/* Compute the LU factorization of A. */
slacpy_("Full", n, n, &a[a_offset], lda, &af[af_offset], ldaf);
sgetrf_(n, n, &af[af_offset], ldaf, &ipiv[1], info);
/* Return if INFO is non-zero. */
if (*info != 0) {
if (*info > 0) {
/* Compute the reciprocal pivot growth factor of the
leading rank-deficient INFO columns of A. */
rpvgrw = slantr_("M", "U", "N", info, info, &af[af_offset],
ldaf, &work[1]);
if (rpvgrw == 0.f) {
rpvgrw = 1.f;
} else {
rpvgrw = slange_("M", n, info, &a[a_offset], lda, &work[1]
) / rpvgrw;
}
work[1] = rpvgrw;
*rcond = 0.f;
}
return 0;
}
}
/* Compute the norm of the matrix A and the
reciprocal pivot growth factor RPVGRW. */
if (notran) {
*(unsigned char *)norm = '1';
} else {
*(unsigned char *)norm = 'I';
}
anorm = slange_(norm, n, n, &a[a_offset], lda, &work[1]);
rpvgrw = slantr_("M", "U", "N", n, n, &af[af_offset], ldaf, &work[1]);
if (rpvgrw == 0.f) {
rpvgrw = 1.f;
} else {
rpvgrw = slange_("M", n, n, &a[a_offset], lda, &work[1]) /
rpvgrw;
}
/* Compute the reciprocal of the condition number of A. */
sgecon_(norm, n, &af[af_offset], ldaf, &anorm, rcond, &work[1], &iwork[1],
info);
/* Set INFO = N+1 if the matrix is singular to working precision. */
if (*rcond < slamch_("Epsilon")) {
*info = *n + 1;
}
/* Compute the solution matrix X. */
slacpy_("Full", n, nrhs, &b[b_offset], ldb, &x[x_offset], ldx);
sgetrs_(trans, n, nrhs, &af[af_offset], ldaf, &ipiv[1], &x[x_offset], ldx,
info);
/* Use iterative refinement to improve the computed solution and
compute error bounds and backward error estimates for it. */
sgerfs_(trans, n, nrhs, &a[a_offset], lda, &af[af_offset], ldaf, &ipiv[1],
&b[b_offset], ldb, &x[x_offset], ldx, &ferr[1], &berr[1], &work[
1], &iwork[1], info);
/* Transform the solution matrix X to a solution of the original
system. */
if (notran) {
if (colequ) {
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
i__2 = *n;
for (i__ = 1; i__ <= i__2; ++i__) {
x_ref(i__, j) = c__[i__] * x_ref(i__, j);
/* L70: */
}
/* L80: */
}
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
ferr[j] /= colcnd;
/* L90: */
}
}
} else if (rowequ) {
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
i__2 = *n;
for (i__ = 1; i__ <= i__2; ++i__) {
x_ref(i__, j) = r__[i__] * x_ref(i__, j);
/* L100: */
}
/* L110: */
}
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
ferr[j] /= rowcnd;
/* L120: */
}
}
work[1] = rpvgrw;
return 0;
/* End of SGESVX */
} /* sgesvx_ */
/* Subroutine */ int sgegs_(char *jobvsl, char *jobvsr, integer *n, real *a,
integer *lda, real *b, integer *ldb, real *alphar, real *alphai, real
*beta, real *vsl, integer *ldvsl, real *vsr, integer *ldvsr, real *
work, integer *lwork, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, vsl_dim1, vsl_offset,
vsr_dim1, vsr_offset, i__1, i__2;
/* Local variables */
integer nb, nb1, nb2, nb3, ihi, ilo;
real eps, anrm, bnrm;
integer itau, lopt;
extern logical lsame_(char *, char *);
integer ileft, iinfo, icols;
logical ilvsl;
integer iwork;
logical ilvsr;
integer irows;
extern /* Subroutine */ int sggbak_(char *, char *, integer *, integer *,
integer *, real *, real *, integer *, real *, integer *, integer *
), sggbal_(char *, integer *, real *, integer *,
real *, integer *, integer *, integer *, real *, real *, real *,
integer *);
logical ilascl, ilbscl;
extern doublereal slamch_(char *), slange_(char *, integer *,
integer *, real *, integer *, real *);
real safmin;
extern /* Subroutine */ int sgghrd_(char *, char *, integer *, integer *,
integer *, real *, integer *, real *, integer *, real *, integer *
, real *, integer *, integer *), xerbla_(char *,
integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *);
real bignum;
extern /* Subroutine */ int slascl_(char *, integer *, integer *, real *,
real *, integer *, integer *, real *, integer *, integer *);
integer ijobvl, iright;
extern /* Subroutine */ int sgeqrf_(integer *, integer *, real *, integer
*, real *, real *, integer *, integer *);
integer ijobvr;
extern /* Subroutine */ int slacpy_(char *, integer *, integer *, real *,
integer *, real *, integer *), slaset_(char *, integer *,
integer *, real *, real *, real *, integer *);
real anrmto;
integer lwkmin;
real bnrmto;
extern /* Subroutine */ int shgeqz_(char *, char *, char *, integer *,
integer *, integer *, real *, integer *, real *, integer *, real *
, real *, real *, real *, integer *, real *, integer *, real *,
integer *, integer *);
real smlnum;
extern /* Subroutine */ int sorgqr_(integer *, integer *, integer *, real
*, integer *, real *, real *, integer *, integer *);
integer lwkopt;
logical lquery;
extern /* Subroutine */ int sormqr_(char *, char *, integer *, integer *,
integer *, real *, integer *, real *, real *, integer *, real *,
integer *, integer *);
/* -- LAPACK driver routine (version 3.2) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* This routine is deprecated and has been replaced by routine SGGES. */
/* SGEGS computes the eigenvalues, real Schur form, and, optionally, */
/* left and or/right Schur vectors of a real matrix pair (A,B). */
/* Given two square matrices A and B, the generalized real Schur */
/* factorization has the form */
/* A = Q*S*Z**T, B = Q*T*Z**T */
/* where Q and Z are orthogonal matrices, T is upper triangular, and S */
/* is an upper quasi-triangular matrix with 1-by-1 and 2-by-2 diagonal */
/* blocks, the 2-by-2 blocks corresponding to complex conjugate pairs */
/* of eigenvalues of (A,B). The columns of Q are the left Schur vectors */
/* and the columns of Z are the right Schur vectors. */
/* If only the eigenvalues of (A,B) are needed, the driver routine */
/* SGEGV should be used instead. See SGEGV for a description of the */
/* eigenvalues of the generalized nonsymmetric eigenvalue problem */
/* (GNEP). */
/* Arguments */
/* ========= */
/* JOBVSL (input) CHARACTER*1 */
/* = 'N': do not compute the left Schur vectors; */
/* = 'V': compute the left Schur vectors (returned in VSL). */
/* JOBVSR (input) CHARACTER*1 */
/* = 'N': do not compute the right Schur vectors; */
/* = 'V': compute the right Schur vectors (returned in VSR). */
/* N (input) INTEGER */
/* The order of the matrices A, B, VSL, and VSR. N >= 0. */
/* A (input/output) REAL array, dimension (LDA, N) */
/* On entry, the matrix A. */
/* On exit, the upper quasi-triangular matrix S from the */
/* generalized real Schur factorization. */
/* LDA (input) INTEGER */
/* The leading dimension of A. LDA >= max(1,N). */
/* B (input/output) REAL array, dimension (LDB, N) */
/* On entry, the matrix B. */
/* On exit, the upper triangular matrix T from the generalized */
/* real Schur factorization. */
/* LDB (input) INTEGER */
/* The leading dimension of B. LDB >= max(1,N). */
/* ALPHAR (output) REAL array, dimension (N) */
/* The real parts of each scalar alpha defining an eigenvalue */
/* of GNEP. */
/* ALPHAI (output) REAL array, dimension (N) */
/* The imaginary parts of each scalar alpha defining an */
/* eigenvalue of GNEP. If ALPHAI(j) is zero, then the j-th */
/* eigenvalue is real; if positive, then the j-th and (j+1)-st */
/* eigenvalues are a complex conjugate pair, with */
/* ALPHAI(j+1) = -ALPHAI(j). */
/* BETA (output) REAL array, dimension (N) */
/* The scalars beta that define the eigenvalues of GNEP. */
/* Together, the quantities alpha = (ALPHAR(j),ALPHAI(j)) and */
/* beta = BETA(j) represent the j-th eigenvalue of the matrix */
/* pair (A,B), in one of the forms lambda = alpha/beta or */
/* mu = beta/alpha. Since either lambda or mu may overflow, */
/* they should not, in general, be computed. */
/* VSL (output) REAL array, dimension (LDVSL,N) */
/* If JOBVSL = 'V', the matrix of left Schur vectors Q. */
/* Not referenced if JOBVSL = 'N'. */
/* LDVSL (input) INTEGER */
/* The leading dimension of the matrix VSL. LDVSL >=1, and */
/* if JOBVSL = 'V', LDVSL >= N. */
/* VSR (output) REAL array, dimension (LDVSR,N) */
/* If JOBVSR = 'V', the matrix of right Schur vectors Z. */
/* Not referenced if JOBVSR = 'N'. */
/* LDVSR (input) INTEGER */
/* The leading dimension of the matrix VSR. LDVSR >= 1, and */
/* if JOBVSR = 'V', LDVSR >= 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 dimension of the array WORK. LWORK >= max(1,4*N). */
/* For good performance, LWORK must generally be larger. */
/* To compute the optimal value of LWORK, call ILAENV to get */
/* blocksizes (for SGEQRF, SORMQR, and SORGQR.) Then compute: */
/* NB -- MAX of the blocksizes for SGEQRF, SORMQR, and SORGQR */
/* The optimal LWORK is 2*N + N*(NB+1). */
/* 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. */
/* INFO (output) INTEGER */
/* = 0: successful exit */
/* < 0: if INFO = -i, the i-th argument had an illegal value. */
/* = 1,...,N: */
/* The QZ iteration failed. (A,B) are not in Schur */
/* form, but ALPHAR(j), ALPHAI(j), and BETA(j) should */
/* be correct for j=INFO+1,...,N. */
/* > N: errors that usually indicate LAPACK problems: */
/* =N+1: error return from SGGBAL */
/* =N+2: error return from SGEQRF */
/* =N+3: error return from SORMQR */
/* =N+4: error return from SORGQR */
/* =N+5: error return from SGGHRD */
/* =N+6: error return from SHGEQZ (other than failed */
/* iteration) */
/* =N+7: error return from SGGBAK (computing VSL) */
/* =N+8: error return from SGGBAK (computing VSR) */
/* =N+9: error return from SLASCL (various places) */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Decode the input arguments */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
b_dim1 = *ldb;
b_offset = 1 + b_dim1;
b -= b_offset;
--alphar;
--alphai;
--beta;
vsl_dim1 = *ldvsl;
vsl_offset = 1 + vsl_dim1;
vsl -= vsl_offset;
vsr_dim1 = *ldvsr;
vsr_offset = 1 + vsr_dim1;
vsr -= vsr_offset;
--work;
/* Function Body */
if (lsame_(jobvsl, "N")) {
ijobvl = 1;
ilvsl = FALSE_;
} else if (lsame_(jobvsl, "V")) {
ijobvl = 2;
ilvsl = TRUE_;
} else {
ijobvl = -1;
ilvsl = FALSE_;
}
if (lsame_(jobvsr, "N")) {
ijobvr = 1;
ilvsr = FALSE_;
} else if (lsame_(jobvsr, "V")) {
ijobvr = 2;
ilvsr = TRUE_;
} else {
ijobvr = -1;
ilvsr = FALSE_;
}
/* Test the input arguments */
/* Computing MAX */
i__1 = *n << 2;
lwkmin = max(i__1,1);
lwkopt = lwkmin;
work[1] = (real) lwkopt;
lquery = *lwork == -1;
*info = 0;
if (ijobvl <= 0) {
*info = -1;
} else if (ijobvr <= 0) {
*info = -2;
} else if (*n < 0) {
*info = -3;
} else if (*lda < max(1,*n)) {
*info = -5;
} else if (*ldb < max(1,*n)) {
*info = -7;
} else if (*ldvsl < 1 || ilvsl && *ldvsl < *n) {
*info = -12;
} else if (*ldvsr < 1 || ilvsr && *ldvsr < *n) {
*info = -14;
} else if (*lwork < lwkmin && ! lquery) {
*info = -16;
}
if (*info == 0) {
nb1 = ilaenv_(&c__1, "SGEQRF", " ", n, n, &c_n1, &c_n1);
nb2 = ilaenv_(&c__1, "SORMQR", " ", n, n, n, &c_n1);
nb3 = ilaenv_(&c__1, "SORGQR", " ", n, n, n, &c_n1);
/* Computing MAX */
i__1 = max(nb1,nb2);
nb = max(i__1,nb3);
lopt = (*n << 1) + *n * (nb + 1);
work[1] = (real) lopt;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SGEGS ", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
if (*n == 0) {
return 0;
}
/* Get machine constants */
eps = slamch_("E") * slamch_("B");
safmin = slamch_("S");
smlnum = *n * safmin / eps;
bignum = 1.f / smlnum;
/* Scale A if max element outside range [SMLNUM,BIGNUM] */
anrm = slange_("M", n, n, &a[a_offset], lda, &work[1]);
ilascl = FALSE_;
if (anrm > 0.f && anrm < smlnum) {
anrmto = smlnum;
ilascl = TRUE_;
} else if (anrm > bignum) {
anrmto = bignum;
ilascl = TRUE_;
}
if (ilascl) {
slascl_("G", &c_n1, &c_n1, &anrm, &anrmto, n, n, &a[a_offset], lda, &
iinfo);
if (iinfo != 0) {
*info = *n + 9;
return 0;
}
}
/* Scale B if max element outside range [SMLNUM,BIGNUM] */
bnrm = slange_("M", n, n, &b[b_offset], ldb, &work[1]);
ilbscl = FALSE_;
if (bnrm > 0.f && bnrm < smlnum) {
bnrmto = smlnum;
ilbscl = TRUE_;
} else if (bnrm > bignum) {
bnrmto = bignum;
ilbscl = TRUE_;
}
if (ilbscl) {
slascl_("G", &c_n1, &c_n1, &bnrm, &bnrmto, n, n, &b[b_offset], ldb, &
iinfo);
if (iinfo != 0) {
*info = *n + 9;
return 0;
}
}
/* Permute the matrix to make it more nearly triangular */
/* Workspace layout: (2*N words -- "work..." not actually used) */
/* left_permutation, right_permutation, work... */
ileft = 1;
iright = *n + 1;
iwork = iright + *n;
sggbal_("P", n, &a[a_offset], lda, &b[b_offset], ldb, &ilo, &ihi, &work[
ileft], &work[iright], &work[iwork], &iinfo);
if (iinfo != 0) {
*info = *n + 1;
goto L10;
}
/* Reduce B to triangular form, and initialize VSL and/or VSR */
/* Workspace layout: ("work..." must have at least N words) */
/* left_permutation, right_permutation, tau, work... */
irows = ihi + 1 - ilo;
icols = *n + 1 - ilo;
itau = iwork;
iwork = itau + irows;
i__1 = *lwork + 1 - iwork;
sgeqrf_(&irows, &icols, &b[ilo + ilo * b_dim1], ldb, &work[itau], &work[
iwork], &i__1, &iinfo);
if (iinfo >= 0) {
/* Computing MAX */
i__1 = lwkopt, i__2 = (integer) work[iwork] + iwork - 1;
lwkopt = max(i__1,i__2);
}
if (iinfo != 0) {
*info = *n + 2;
goto L10;
}
i__1 = *lwork + 1 - iwork;
sormqr_("L", "T", &irows, &icols, &irows, &b[ilo + ilo * b_dim1], ldb, &
work[itau], &a[ilo + ilo * a_dim1], lda, &work[iwork], &i__1, &
iinfo);
if (iinfo >= 0) {
/* Computing MAX */
i__1 = lwkopt, i__2 = (integer) work[iwork] + iwork - 1;
lwkopt = max(i__1,i__2);
}
if (iinfo != 0) {
*info = *n + 3;
goto L10;
}
if (ilvsl) {
slaset_("Full", n, n, &c_b36, &c_b37, &vsl[vsl_offset], ldvsl);
i__1 = irows - 1;
i__2 = irows - 1;
slacpy_("L", &i__1, &i__2, &b[ilo + 1 + ilo * b_dim1], ldb, &vsl[ilo
+ 1 + ilo * vsl_dim1], ldvsl);
i__1 = *lwork + 1 - iwork;
sorgqr_(&irows, &irows, &irows, &vsl[ilo + ilo * vsl_dim1], ldvsl, &
work[itau], &work[iwork], &i__1, &iinfo);
if (iinfo >= 0) {
/* Computing MAX */
i__1 = lwkopt, i__2 = (integer) work[iwork] + iwork - 1;
lwkopt = max(i__1,i__2);
}
if (iinfo != 0) {
*info = *n + 4;
goto L10;
}
}
if (ilvsr) {
slaset_("Full", n, n, &c_b36, &c_b37, &vsr[vsr_offset], ldvsr);
}
/* Reduce to generalized Hessenberg form */
sgghrd_(jobvsl, jobvsr, n, &ilo, &ihi, &a[a_offset], lda, &b[b_offset],
ldb, &vsl[vsl_offset], ldvsl, &vsr[vsr_offset], ldvsr, &iinfo);
if (iinfo != 0) {
*info = *n + 5;
goto L10;
}
/* Perform QZ algorithm, computing Schur vectors if desired */
/* Workspace layout: ("work..." must have at least 1 word) */
/* left_permutation, right_permutation, work... */
iwork = itau;
i__1 = *lwork + 1 - iwork;
shgeqz_("S", jobvsl, jobvsr, n, &ilo, &ihi, &a[a_offset], lda, &b[
b_offset], ldb, &alphar[1], &alphai[1], &beta[1], &vsl[vsl_offset]
, ldvsl, &vsr[vsr_offset], ldvsr, &work[iwork], &i__1, &iinfo);
if (iinfo >= 0) {
/* Computing MAX */
i__1 = lwkopt, i__2 = (integer) work[iwork] + iwork - 1;
lwkopt = max(i__1,i__2);
}
if (iinfo != 0) {
if (iinfo > 0 && iinfo <= *n) {
*info = iinfo;
} else if (iinfo > *n && iinfo <= *n << 1) {
*info = iinfo - *n;
} else {
*info = *n + 6;
}
goto L10;
}
/* Apply permutation to VSL and VSR */
if (ilvsl) {
sggbak_("P", "L", n, &ilo, &ihi, &work[ileft], &work[iright], n, &vsl[
vsl_offset], ldvsl, &iinfo);
if (iinfo != 0) {
*info = *n + 7;
goto L10;
}
}
if (ilvsr) {
sggbak_("P", "R", n, &ilo, &ihi, &work[ileft], &work[iright], n, &vsr[
vsr_offset], ldvsr, &iinfo);
if (iinfo != 0) {
*info = *n + 8;
goto L10;
}
}
/* Undo scaling */
if (ilascl) {
slascl_("H", &c_n1, &c_n1, &anrmto, &anrm, n, n, &a[a_offset], lda, &
iinfo);
if (iinfo != 0) {
*info = *n + 9;
return 0;
}
slascl_("G", &c_n1, &c_n1, &anrmto, &anrm, n, &c__1, &alphar[1], n, &
iinfo);
if (iinfo != 0) {
*info = *n + 9;
return 0;
}
slascl_("G", &c_n1, &c_n1, &anrmto, &anrm, n, &c__1, &alphai[1], n, &
iinfo);
if (iinfo != 0) {
*info = *n + 9;
return 0;
}
}
if (ilbscl) {
slascl_("U", &c_n1, &c_n1, &bnrmto, &bnrm, n, n, &b[b_offset], ldb, &
iinfo);
if (iinfo != 0) {
*info = *n + 9;
return 0;
}
slascl_("G", &c_n1, &c_n1, &bnrmto, &bnrm, n, &c__1, &beta[1], n, &
iinfo);
if (iinfo != 0) {
*info = *n + 9;
return 0;
}
}
L10:
work[1] = (real) lwkopt;
return 0;
/* End of SGEGS */
} /* sgegs_ */
/* Subroutine */ int sqlt02_(integer *m, integer *n, integer *k, real *a,
real *af, real *q, real *l, integer *lda, real *tau, real *work,
integer *lwork, real *rwork, real *result)
{
/* System generated locals */
integer a_dim1, a_offset, af_dim1, af_offset, l_dim1, l_offset, q_dim1,
q_offset, i__1, i__2;
/* Builtin functions */
/* Subroutine */ int s_copy(char *, char *, ftnlen, ftnlen);
/* Local variables */
real eps;
integer info;
real resid;
extern /* Subroutine */ int sgemm_(char *, char *, integer *, integer *,
integer *, real *, real *, integer *, real *, integer *, real *,
real *, integer *);
real anorm;
extern /* Subroutine */ int ssyrk_(char *, char *, integer *, integer *,
real *, real *, integer *, real *, real *, integer *);
extern doublereal slamch_(char *), slange_(char *, integer *,
integer *, real *, integer *, real *);
extern /* Subroutine */ int slacpy_(char *, integer *, integer *, real *,
integer *, real *, integer *), slaset_(char *, integer *,
integer *, real *, real *, real *, integer *), sorgql_(
integer *, integer *, integer *, real *, integer *, real *, real *
, integer *, integer *);
extern doublereal slansy_(char *, char *, integer *, real *, integer *,
real *);
/* -- LAPACK test routine (version 3.1) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* SQLT02 tests SORGQL, which generates an m-by-n matrix Q with */
/* orthonornmal columns that is defined as the product of k elementary */
/* reflectors. */
/* Given the QL factorization of an m-by-n matrix A, SQLT02 generates */
/* the orthogonal matrix Q defined by the factorization of the last k */
/* columns of A; it compares L(m-n+1:m,n-k+1:n) with */
/* Q(1:m,m-n+1:m)'*A(1:m,n-k+1:n), and checks that the columns of Q are */
/* orthonormal. */
/* Arguments */
/* ========= */
/* M (input) INTEGER */
/* The number of rows of the matrix Q to be generated. M >= 0. */
/* N (input) INTEGER */
/* The number of columns of the matrix Q to be generated. */
/* M >= N >= 0. */
/* K (input) INTEGER */
/* The number of elementary reflectors whose product defines the */
/* matrix Q. N >= K >= 0. */
/* A (input) REAL array, dimension (LDA,N) */
/* The m-by-n matrix A which was factorized by SQLT01. */
/* AF (input) REAL array, dimension (LDA,N) */
/* Details of the QL factorization of A, as returned by SGEQLF. */
/* See SGEQLF for further details. */
/* Q (workspace) REAL array, dimension (LDA,N) */
/* L (workspace) REAL array, dimension (LDA,N) */
/* LDA (input) INTEGER */
/* The leading dimension of the arrays A, AF, Q and L. LDA >= M. */
/* TAU (input) REAL array, dimension (N) */
/* The scalar factors of the elementary reflectors corresponding */
/* to the QL factorization in AF. */
/* WORK (workspace) REAL array, dimension (LWORK) */
/* LWORK (input) INTEGER */
/* The dimension of the array WORK. */
/* RWORK (workspace) REAL array, dimension (M) */
/* RESULT (output) REAL array, dimension (2) */
/* The test ratios: */
/* RESULT(1) = norm( L - Q'*A ) / ( M * norm(A) * EPS ) */
/* RESULT(2) = norm( I - Q'*Q ) / ( M * EPS ) */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Scalars in Common .. */
/* .. */
/* .. Common blocks .. */
/* .. */
/* .. Executable Statements .. */
/* Quick return if possible */
/* Parameter adjustments */
l_dim1 = *lda;
l_offset = 1 + l_dim1;
l -= l_offset;
q_dim1 = *lda;
q_offset = 1 + q_dim1;
q -= q_offset;
af_dim1 = *lda;
af_offset = 1 + af_dim1;
af -= af_offset;
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
--tau;
--work;
--rwork;
--result;
/* Function Body */
if (*m == 0 || *n == 0 || *k == 0) {
result[1] = 0.f;
result[2] = 0.f;
return 0;
}
eps = slamch_("Epsilon");
/* Copy the last k columns of the factorization to the array Q */
slaset_("Full", m, n, &c_b4, &c_b4, &q[q_offset], lda);
if (*k < *m) {
i__1 = *m - *k;
slacpy_("Full", &i__1, k, &af[(*n - *k + 1) * af_dim1 + 1], lda, &q[(*
n - *k + 1) * q_dim1 + 1], lda);
}
if (*k > 1) {
i__1 = *k - 1;
i__2 = *k - 1;
slacpy_("Upper", &i__1, &i__2, &af[*m - *k + 1 + (*n - *k + 2) *
af_dim1], lda, &q[*m - *k + 1 + (*n - *k + 2) * q_dim1], lda);
}
/* Generate the last n columns of the matrix Q */
s_copy(srnamc_1.srnamt, "SORGQL", (ftnlen)6, (ftnlen)6);
sorgql_(m, n, k, &q[q_offset], lda, &tau[*n - *k + 1], &work[1], lwork, &
info);
/* Copy L(m-n+1:m,n-k+1:n) */
slaset_("Full", n, k, &c_b10, &c_b10, &l[*m - *n + 1 + (*n - *k + 1) *
l_dim1], lda);
slacpy_("Lower", k, k, &af[*m - *k + 1 + (*n - *k + 1) * af_dim1], lda, &
l[*m - *k + 1 + (*n - *k + 1) * l_dim1], lda);
/* Compute L(m-n+1:m,n-k+1:n) - Q(1:m,m-n+1:m)' * A(1:m,n-k+1:n) */
sgemm_("Transpose", "No transpose", n, k, m, &c_b15, &q[q_offset], lda, &
a[(*n - *k + 1) * a_dim1 + 1], lda, &c_b16, &l[*m - *n + 1 + (*n
- *k + 1) * l_dim1], lda);
/* Compute norm( L - Q'*A ) / ( M * norm(A) * EPS ) . */
anorm = slange_("1", m, k, &a[(*n - *k + 1) * a_dim1 + 1], lda, &rwork[1]);
resid = slange_("1", n, k, &l[*m - *n + 1 + (*n - *k + 1) * l_dim1], lda,
&rwork[1]);
if (anorm > 0.f) {
result[1] = resid / (real) max(1,*m) / anorm / eps;
} else {
result[1] = 0.f;
}
/* Compute I - Q'*Q */
slaset_("Full", n, n, &c_b10, &c_b16, &l[l_offset], lda);
ssyrk_("Upper", "Transpose", n, m, &c_b15, &q[q_offset], lda, &c_b16, &l[
l_offset], lda);
/* Compute norm( I - Q'*Q ) / ( M * EPS ) . */
resid = slansy_("1", "Upper", n, &l[l_offset], lda, &rwork[1]);
result[2] = resid / (real) max(1,*m) / eps;
return 0;
/* End of SQLT02 */
} /* sqlt02_ */
/* Subroutine */ int sstt22_(integer *n, integer *m, integer *kband, real *ad,
real *ae, real *sd, real *se, real *u, integer *ldu, real *work,
integer *ldwork, real *result)
{
/* System generated locals */
integer u_dim1, u_offset, work_dim1, work_offset, i__1, i__2, i__3;
real r__1, r__2, r__3, r__4, r__5;
/* Local variables */
integer i__, j, k;
real ulp, aukj, unfl;
extern /* Subroutine */ int sgemm_(char *, char *, integer *, integer *,
integer *, real *, real *, integer *, real *, integer *, real *,
real *, integer *);
real anorm, wnorm;
extern doublereal slamch_(char *), slange_(char *, integer *,
integer *, real *, integer *, real *), slansy_(char *,
char *, integer *, real *, integer *, real *);
/* -- LAPACK test routine (version 3.1) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* SSTT22 checks a set of M eigenvalues and eigenvectors, */
/* A U = U S */
/* where A is symmetric tridiagonal, the columns of U are orthogonal, */
/* and S is diagonal (if KBAND=0) or symmetric tridiagonal (if KBAND=1). */
/* Two tests are performed: */
/* RESULT(1) = | U' A U - S | / ( |A| m ulp ) */
/* RESULT(2) = | I - U'U | / ( m ulp ) */
/* Arguments */
/* ========= */
/* N (input) INTEGER */
/* The size of the matrix. If it is zero, SSTT22 does nothing. */
/* It must be at least zero. */
/* M (input) INTEGER */
/* The number of eigenpairs to check. If it is zero, SSTT22 */
/* does nothing. It must be at least zero. */
/* KBAND (input) INTEGER */
/* The bandwidth of the matrix S. It may only be zero or one. */
/* If zero, then S is diagonal, and SE is not referenced. If */
/* one, then S is symmetric tri-diagonal. */
/* AD (input) REAL array, dimension (N) */
/* The diagonal of the original (unfactored) matrix A. A is */
/* assumed to be symmetric tridiagonal. */
/* AE (input) REAL array, dimension (N) */
/* The off-diagonal of the original (unfactored) matrix A. A */
/* is assumed to be symmetric tridiagonal. AE(1) is ignored, */
/* AE(2) is the (1,2) and (2,1) element, etc. */
/* SD (input) REAL array, dimension (N) */
/* The diagonal of the (symmetric tri-) diagonal matrix S. */
/* SE (input) REAL array, dimension (N) */
/* The off-diagonal of the (symmetric tri-) diagonal matrix S. */
/* Not referenced if KBSND=0. If KBAND=1, then AE(1) is */
/* ignored, SE(2) is the (1,2) and (2,1) element, etc. */
/* U (input) REAL array, dimension (LDU, N) */
/* The orthogonal matrix in the decomposition. */
/* LDU (input) INTEGER */
/* The leading dimension of U. LDU must be at least N. */
/* WORK (workspace) REAL array, dimension (LDWORK, M+1) */
/* LDWORK (input) INTEGER */
/* The leading dimension of WORK. LDWORK must be at least */
/* max(1,M). */
/* RESULT (output) REAL array, dimension (2) */
/* The values computed by the two tests described above. The */
/* values are currently limited to 1/ulp, to avoid overflow. */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Parameter adjustments */
--ad;
--ae;
--sd;
--se;
u_dim1 = *ldu;
u_offset = 1 + u_dim1;
u -= u_offset;
work_dim1 = *ldwork;
work_offset = 1 + work_dim1;
work -= work_offset;
--result;
/* Function Body */
result[1] = 0.f;
result[2] = 0.f;
if (*n <= 0 || *m <= 0) {
return 0;
}
unfl = slamch_("Safe minimum");
ulp = slamch_("Epsilon");
/* Do Test 1 */
/* Compute the 1-norm of A. */
if (*n > 1) {
anorm = dabs(ad[1]) + dabs(ae[1]);
i__1 = *n - 1;
for (j = 2; j <= i__1; ++j) {
/* Computing MAX */
r__4 = anorm, r__5 = (r__1 = ad[j], dabs(r__1)) + (r__2 = ae[j],
dabs(r__2)) + (r__3 = ae[j - 1], dabs(r__3));
anorm = dmax(r__4,r__5);
/* L10: */
}
/* Computing MAX */
r__3 = anorm, r__4 = (r__1 = ad[*n], dabs(r__1)) + (r__2 = ae[*n - 1],
dabs(r__2));
anorm = dmax(r__3,r__4);
} else {
anorm = dabs(ad[1]);
}
anorm = dmax(anorm,unfl);
/* Norm of U'AU - S */
i__1 = *m;
for (i__ = 1; i__ <= i__1; ++i__) {
i__2 = *m;
for (j = 1; j <= i__2; ++j) {
work[i__ + j * work_dim1] = 0.f;
i__3 = *n;
for (k = 1; k <= i__3; ++k) {
aukj = ad[k] * u[k + j * u_dim1];
if (k != *n) {
aukj += ae[k] * u[k + 1 + j * u_dim1];
}
if (k != 1) {
aukj += ae[k - 1] * u[k - 1 + j * u_dim1];
}
work[i__ + j * work_dim1] += u[k + i__ * u_dim1] * aukj;
/* L20: */
}
/* L30: */
}
work[i__ + i__ * work_dim1] -= sd[i__];
if (*kband == 1) {
if (i__ != 1) {
work[i__ + (i__ - 1) * work_dim1] -= se[i__ - 1];
}
if (i__ != *n) {
work[i__ + (i__ + 1) * work_dim1] -= se[i__];
}
}
/* L40: */
}
wnorm = slansy_("1", "L", m, &work[work_offset], m, &work[(*m + 1) *
work_dim1 + 1]);
if (anorm > wnorm) {
result[1] = wnorm / anorm / (*m * ulp);
} else {
if (anorm < 1.f) {
/* Computing MIN */
r__1 = wnorm, r__2 = *m * anorm;
result[1] = dmin(r__1,r__2) / anorm / (*m * ulp);
} else {
/* Computing MIN */
r__1 = wnorm / anorm, r__2 = (real) (*m);
result[1] = dmin(r__1,r__2) / (*m * ulp);
}
}
/* Do Test 2 */
/* Compute U'U - I */
sgemm_("T", "N", m, m, n, &c_b12, &u[u_offset], ldu, &u[u_offset], ldu, &
c_b13, &work[work_offset], m);
i__1 = *m;
for (j = 1; j <= i__1; ++j) {
work[j + j * work_dim1] += -1.f;
/* L50: */
}
/* Computing MIN */
r__1 = (real) (*m), r__2 = slange_("1", m, m, &work[work_offset], m, &
work[(*m + 1) * work_dim1 + 1]);
result[2] = dmin(r__1,r__2) / (*m * ulp);
return 0;
/* End of SSTT22 */
} /* sstt22_ */
/* Subroutine */ int sget52_(logical *left, integer *n, real *a, integer *lda,
real *b, integer *ldb, real *e, integer *lde, real *alphar, real *
alphai, real *beta, real *work, real *result)
{
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, e_dim1, e_offset, i__1, i__2;
real r__1, r__2, r__3, r__4;
/* Local variables */
integer j;
real ulp;
integer jvec;
real temp1, acoef, scale, abmax, salfi, sbeta, salfr, anorm, bnorm, enorm;
extern /* Subroutine */ int sgemv_(char *, integer *, integer *, real *,
real *, integer *, real *, integer *, real *, real *, integer *);
char trans[1];
real bcoefi, bcoefr, alfmax;
extern doublereal slamch_(char *), slange_(char *, integer *,
integer *, real *, integer *, real *);
real safmin;
char normab[1];
real safmax, betmax, enrmer;
logical ilcplx;
real errnrm;
/* -- LAPACK test routine (version 3.1) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* SGET52 does an eigenvector check for the generalized eigenvalue */
/* problem. */
/* The basic test for right eigenvectors is: */
/* | b(j) A E(j) - a(j) B E(j) | */
/* RESULT(1) = max ------------------------------- */
/* j n ulp max( |b(j) A|, |a(j) B| ) */
/* using the 1-norm. Here, a(j)/b(j) = w is the j-th generalized */
/* eigenvalue of A - w B, or, equivalently, b(j)/a(j) = m is the j-th */
/* generalized eigenvalue of m A - B. */
/* For real eigenvalues, the test is straightforward. For complex */
/* eigenvalues, E(j) and a(j) are complex, represented by */
/* Er(j) + i*Ei(j) and ar(j) + i*ai(j), resp., so the test for that */
/* eigenvector becomes */
/* max( |Wr|, |Wi| ) */
/* -------------------------------------------- */
/* n ulp max( |b(j) A|, (|ar(j)|+|ai(j)|) |B| ) */
/* where */
/* Wr = b(j) A Er(j) - ar(j) B Er(j) + ai(j) B Ei(j) */
/* Wi = b(j) A Ei(j) - ai(j) B Er(j) - ar(j) B Ei(j) */
/* T T _ */
/* For left eigenvectors, A , B , a, and b are used. */
/* SGET52 also tests the normalization of E. Each eigenvector is */
/* supposed to be normalized so that the maximum "absolute value" */
/* of its elements is 1, where in this case, "absolute value" */
/* of a complex value x is |Re(x)| + |Im(x)| ; let us call this */
/* maximum "absolute value" norm of a vector v M(v). */
/* if a(j)=b(j)=0, then the eigenvector is set to be the jth coordinate */
/* vector. The normalization test is: */
/* RESULT(2) = max | M(v(j)) - 1 | / ( n ulp ) */
/* eigenvectors v(j) */
/* Arguments */
/* ========= */
/* LEFT (input) LOGICAL */
/* =.TRUE.: The eigenvectors in the columns of E are assumed */
/* to be *left* eigenvectors. */
/* =.FALSE.: The eigenvectors in the columns of E are assumed */
/* to be *right* eigenvectors. */
/* N (input) INTEGER */
/* The size of the matrices. If it is zero, SGET52 does */
/* nothing. It must be at least zero. */
/* A (input) REAL array, dimension (LDA, N) */
/* The matrix A. */
/* LDA (input) INTEGER */
/* The leading dimension of A. It must be at least 1 */
/* and at least N. */
/* B (input) REAL array, dimension (LDB, N) */
/* The matrix B. */
/* LDB (input) INTEGER */
/* The leading dimension of B. It must be at least 1 */
/* and at least N. */
/* E (input) REAL array, dimension (LDE, N) */
/* The matrix of eigenvectors. It must be O( 1 ). Complex */
/* eigenvalues and eigenvectors always come in pairs, the */
/* eigenvalue and its conjugate being stored in adjacent */
/* elements of ALPHAR, ALPHAI, and BETA. Thus, if a(j)/b(j) */
/* and a(j+1)/b(j+1) are a complex conjugate pair of */
/* generalized eigenvalues, then E(,j) contains the real part */
/* of the eigenvector and E(,j+1) contains the imaginary part. */
/* Note that whether E(,j) is a real eigenvector or part of a */
/* complex one is specified by whether ALPHAI(j) is zero or not. */
/* LDE (input) INTEGER */
/* The leading dimension of E. It must be at least 1 and at */
/* least N. */
/* ALPHAR (input) REAL array, dimension (N) */
/* The real parts of the values a(j) as described above, which, */
/* along with b(j), define the generalized eigenvalues. */
/* Complex eigenvalues always come in complex conjugate pairs */
/* a(j)/b(j) and a(j+1)/b(j+1), which are stored in adjacent */
/* elements in ALPHAR, ALPHAI, and BETA. Thus, if the j-th */
/* and (j+1)-st eigenvalues form a pair, ALPHAR(j+1)/BETA(j+1) */
/* is assumed to be equal to ALPHAR(j)/BETA(j). */
/* ALPHAI (input) REAL array, dimension (N) */
/* The imaginary parts of the values a(j) as described above, */
/* which, along with b(j), define the generalized eigenvalues. */
/* If ALPHAI(j)=0, then the eigenvalue is real, otherwise it */
/* is part of a complex conjugate pair. Complex eigenvalues */
/* always come in complex conjugate pairs a(j)/b(j) and */
/* a(j+1)/b(j+1), which are stored in adjacent elements in */
/* ALPHAR, ALPHAI, and BETA. Thus, if the j-th and (j+1)-st */
/* eigenvalues form a pair, ALPHAI(j+1)/BETA(j+1) is assumed to */
/* be equal to -ALPHAI(j)/BETA(j). Also, nonzero values in */
/* ALPHAI are assumed to always come in adjacent pairs. */
/* BETA (input) REAL array, dimension (N) */
/* The values b(j) as described above, which, along with a(j), */
/* define the generalized eigenvalues. */
/* WORK (workspace) REAL array, dimension (N**2+N) */
/* RESULT (output) REAL array, dimension (2) */
/* The values computed by the test described above. If A E or */
/* B E is likely to overflow, then RESULT(1:2) is set to */
/* 10 / ulp. */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
b_dim1 = *ldb;
b_offset = 1 + b_dim1;
b -= b_offset;
e_dim1 = *lde;
e_offset = 1 + e_dim1;
e -= e_offset;
--alphar;
--alphai;
--beta;
--work;
--result;
/* Function Body */
result[1] = 0.f;
result[2] = 0.f;
if (*n <= 0) {
return 0;
}
safmin = slamch_("Safe minimum");
safmax = 1.f / safmin;
ulp = slamch_("Epsilon") * slamch_("Base");
if (*left) {
*(unsigned char *)trans = 'T';
*(unsigned char *)normab = 'I';
} else {
*(unsigned char *)trans = 'N';
*(unsigned char *)normab = 'O';
}
/* Norm of A, B, and E: */
/* Computing MAX */
r__1 = slange_(normab, n, n, &a[a_offset], lda, &work[1]);
anorm = dmax(r__1,safmin);
/* Computing MAX */
r__1 = slange_(normab, n, n, &b[b_offset], ldb, &work[1]);
bnorm = dmax(r__1,safmin);
/* Computing MAX */
r__1 = slange_("O", n, n, &e[e_offset], lde, &work[1]);
enorm = dmax(r__1,ulp);
alfmax = safmax / dmax(1.f,bnorm);
betmax = safmax / dmax(1.f,anorm);
/* Compute error matrix. */
/* Column i = ( b(i) A - a(i) B ) E(i) / max( |a(i) B| |b(i) A| ) */
ilcplx = FALSE_;
i__1 = *n;
for (jvec = 1; jvec <= i__1; ++jvec) {
if (ilcplx) {
/* 2nd Eigenvalue/-vector of pair -- do nothing */
ilcplx = FALSE_;
} else {
salfr = alphar[jvec];
salfi = alphai[jvec];
sbeta = beta[jvec];
if (salfi == 0.f) {
/* Real eigenvalue and -vector */
/* Computing MAX */
r__1 = dabs(salfr), r__2 = dabs(sbeta);
abmax = dmax(r__1,r__2);
if (dabs(salfr) > alfmax || dabs(sbeta) > betmax || abmax <
1.f) {
scale = 1.f / dmax(abmax,safmin);
salfr = scale * salfr;
sbeta = scale * sbeta;
}
/* Computing MAX */
r__1 = dabs(salfr) * bnorm, r__2 = dabs(sbeta) * anorm, r__1 =
max(r__1,r__2);
scale = 1.f / dmax(r__1,safmin);
acoef = scale * sbeta;
bcoefr = scale * salfr;
sgemv_(trans, n, n, &acoef, &a[a_offset], lda, &e[jvec *
e_dim1 + 1], &c__1, &c_b12, &work[*n * (jvec - 1) + 1]
, &c__1);
r__1 = -bcoefr;
sgemv_(trans, n, n, &r__1, &b[b_offset], lda, &e[jvec *
e_dim1 + 1], &c__1, &c_b15, &work[*n * (jvec - 1) + 1]
, &c__1);
} else {
/* Complex conjugate pair */
ilcplx = TRUE_;
if (jvec == *n) {
result[1] = 10.f / ulp;
return 0;
}
/* Computing MAX */
r__1 = dabs(salfr) + dabs(salfi), r__2 = dabs(sbeta);
abmax = dmax(r__1,r__2);
if (dabs(salfr) + dabs(salfi) > alfmax || dabs(sbeta) >
betmax || abmax < 1.f) {
scale = 1.f / dmax(abmax,safmin);
salfr = scale * salfr;
salfi = scale * salfi;
sbeta = scale * sbeta;
}
/* Computing MAX */
r__1 = (dabs(salfr) + dabs(salfi)) * bnorm, r__2 = dabs(sbeta)
* anorm, r__1 = max(r__1,r__2);
scale = 1.f / dmax(r__1,safmin);
acoef = scale * sbeta;
bcoefr = scale * salfr;
bcoefi = scale * salfi;
if (*left) {
bcoefi = -bcoefi;
}
sgemv_(trans, n, n, &acoef, &a[a_offset], lda, &e[jvec *
e_dim1 + 1], &c__1, &c_b12, &work[*n * (jvec - 1) + 1]
, &c__1);
r__1 = -bcoefr;
sgemv_(trans, n, n, &r__1, &b[b_offset], lda, &e[jvec *
e_dim1 + 1], &c__1, &c_b15, &work[*n * (jvec - 1) + 1]
, &c__1);
sgemv_(trans, n, n, &bcoefi, &b[b_offset], lda, &e[(jvec + 1)
* e_dim1 + 1], &c__1, &c_b15, &work[*n * (jvec - 1) +
1], &c__1);
sgemv_(trans, n, n, &acoef, &a[a_offset], lda, &e[(jvec + 1) *
e_dim1 + 1], &c__1, &c_b12, &work[*n * jvec + 1], &
c__1);
r__1 = -bcoefi;
sgemv_(trans, n, n, &r__1, &b[b_offset], lda, &e[jvec *
e_dim1 + 1], &c__1, &c_b15, &work[*n * jvec + 1], &
c__1);
r__1 = -bcoefr;
sgemv_(trans, n, n, &r__1, &b[b_offset], lda, &e[(jvec + 1) *
e_dim1 + 1], &c__1, &c_b15, &work[*n * jvec + 1], &
c__1);
}
}
/* L10: */
}
/* Computing 2nd power */
i__1 = *n;
errnrm = slange_("One", n, n, &work[1], n, &work[i__1 * i__1 + 1]) / enorm;
/* Compute RESULT(1) */
result[1] = errnrm / ulp;
/* Normalization of E: */
enrmer = 0.f;
ilcplx = FALSE_;
i__1 = *n;
for (jvec = 1; jvec <= i__1; ++jvec) {
if (ilcplx) {
ilcplx = FALSE_;
} else {
temp1 = 0.f;
if (alphai[jvec] == 0.f) {
i__2 = *n;
for (j = 1; j <= i__2; ++j) {
/* Computing MAX */
r__2 = temp1, r__3 = (r__1 = e[j + jvec * e_dim1], dabs(
r__1));
temp1 = dmax(r__2,r__3);
/* L20: */
}
/* Computing MAX */
r__1 = enrmer, r__2 = temp1 - 1.f;
enrmer = dmax(r__1,r__2);
} else {
ilcplx = TRUE_;
i__2 = *n;
for (j = 1; j <= i__2; ++j) {
/* Computing MAX */
r__3 = temp1, r__4 = (r__1 = e[j + jvec * e_dim1], dabs(
r__1)) + (r__2 = e[j + (jvec + 1) * e_dim1], dabs(
r__2));
temp1 = dmax(r__3,r__4);
/* L30: */
}
/* Computing MAX */
r__1 = enrmer, r__2 = temp1 - 1.f;
enrmer = dmax(r__1,r__2);
}
}
/* L40: */
}
/* Compute RESULT(2) : the normalization error in E. */
result[2] = enrmer / ((real) (*n) * ulp);
return 0;
/* End of SGET52 */
} /* sget52_ */
/* Subroutine */
int sggevx_(char *balanc, char *jobvl, char *jobvr, char * sense, integer *n, real *a, integer *lda, real *b, integer *ldb, real *alphar, real *alphai, real *beta, real *vl, integer *ldvl, real *vr, integer *ldvr, integer *ilo, integer *ihi, real *lscale, real *rscale, real *abnrm, real *bbnrm, real *rconde, real *rcondv, real *work, integer *lwork, integer *iwork, logical *bwork, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, vl_dim1, vl_offset, vr_dim1, vr_offset, i__1, i__2;
real r__1, r__2, r__3, r__4;
/* Builtin functions */
double sqrt(doublereal);
/* Local variables */
integer i__, j, m, jc, in, mm, jr;
real eps;
logical ilv, pair;
real anrm, bnrm;
integer ierr, itau;
real temp;
logical ilvl, ilvr;
integer iwrk, iwrk1;
extern logical lsame_(char *, char *);
integer icols;
logical noscl;
integer irows;
extern /* Subroutine */
int slabad_(real *, real *), sggbak_(char *, char *, integer *, integer *, integer *, real *, real *, integer *, real *, integer *, integer *), sggbal_(char *, integer *, real *, integer *, real *, integer *, integer *, integer *, real *, real *, real *, integer *);
logical ilascl, ilbscl;
extern real slamch_(char *);
extern /* Subroutine */
int xerbla_(char *, integer *), sgghrd_( char *, char *, integer *, integer *, integer *, real *, integer * , real *, integer *, real *, integer *, real *, integer *, integer *);
logical ldumma[1];
char chtemp[1];
real bignum;
extern /* Subroutine */
int slascl_(char *, integer *, integer *, real *, real *, integer *, integer *, real *, integer *, integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *, integer *, integer *);
extern real slange_(char *, integer *, integer *, real *, integer *, real *);
integer ijobvl;
extern /* Subroutine */
int sgeqrf_(integer *, integer *, real *, integer *, real *, real *, integer *, integer *);
integer ijobvr;
extern /* Subroutine */
int slacpy_(char *, integer *, integer *, real *, integer *, real *, integer *);
logical wantsb;
extern /* Subroutine */
int slaset_(char *, integer *, integer *, real *, real *, real *, integer *);
real anrmto;
logical wantse;
real bnrmto;
extern /* Subroutine */
int shgeqz_(char *, char *, char *, integer *, integer *, integer *, real *, integer *, real *, integer *, real * , real *, real *, real *, integer *, real *, integer *, real *, integer *, integer *), stgevc_(char *, char *, logical *, integer *, real *, integer *, real *, integer * , real *, integer *, real *, integer *, integer *, integer *, real *, integer *), stgsna_(char *, char *, logical *, integer *, real *, integer *, real *, integer *, real * , integer *, real *, integer *, real *, real *, integer *, integer *, real *, integer *, integer *, integer *);
integer minwrk, maxwrk;
logical wantsn;
real smlnum;
extern /* Subroutine */
int sorgqr_(integer *, integer *, integer *, real *, integer *, real *, real *, integer *, integer *);
logical lquery, wantsv;
extern /* Subroutine */
int sormqr_(char *, char *, integer *, integer *, integer *, real *, integer *, real *, real *, integer *, real *, integer *, integer *);
/* -- LAPACK driver routine (version 3.4.1) -- */
/* -- LAPACK is a software package provided by Univ. of Tennessee, -- */
/* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- */
/* April 2012 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. Local Arrays .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Decode the input arguments */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
b_dim1 = *ldb;
b_offset = 1 + b_dim1;
b -= b_offset;
--alphar;
--alphai;
--beta;
vl_dim1 = *ldvl;
vl_offset = 1 + vl_dim1;
vl -= vl_offset;
vr_dim1 = *ldvr;
vr_offset = 1 + vr_dim1;
vr -= vr_offset;
--lscale;
--rscale;
--rconde;
--rcondv;
--work;
--iwork;
--bwork;
/* Function Body */
if (lsame_(jobvl, "N"))
{
ijobvl = 1;
ilvl = FALSE_;
}
else if (lsame_(jobvl, "V"))
{
ijobvl = 2;
ilvl = TRUE_;
}
else
{
ijobvl = -1;
ilvl = FALSE_;
}
if (lsame_(jobvr, "N"))
{
ijobvr = 1;
ilvr = FALSE_;
}
else if (lsame_(jobvr, "V"))
{
ijobvr = 2;
ilvr = TRUE_;
}
else
{
ijobvr = -1;
ilvr = FALSE_;
}
ilv = ilvl || ilvr;
noscl = lsame_(balanc, "N") || lsame_(balanc, "P");
wantsn = lsame_(sense, "N");
wantse = lsame_(sense, "E");
wantsv = lsame_(sense, "V");
wantsb = lsame_(sense, "B");
/* Test the input arguments */
*info = 0;
lquery = *lwork == -1;
if (! (noscl || lsame_(balanc, "S") || lsame_( balanc, "B")))
{
*info = -1;
}
else if (ijobvl <= 0)
{
*info = -2;
}
else if (ijobvr <= 0)
{
*info = -3;
}
else if (! (wantsn || wantse || wantsb || wantsv))
{
*info = -4;
}
else if (*n < 0)
{
*info = -5;
}
else if (*lda < max(1,*n))
{
*info = -7;
}
else if (*ldb < max(1,*n))
{
*info = -9;
}
else if (*ldvl < 1 || ilvl && *ldvl < *n)
{
*info = -14;
}
else if (*ldvr < 1 || ilvr && *ldvr < *n)
{
*info = -16;
}
/* Compute workspace */
/* (Note: Comments in the code beginning "Workspace:" describe the */
/* minimal amount of workspace needed at that point in the code, */
/* as well as the preferred amount for good performance. */
/* NB refers to the optimal block size for the immediately */
/* following subroutine, as returned by ILAENV. The workspace is */
/* computed assuming ILO = 1 and IHI = N, the worst case.) */
if (*info == 0)
{
if (*n == 0)
{
minwrk = 1;
maxwrk = 1;
}
else
{
if (noscl && ! ilv)
{
minwrk = *n << 1;
}
else
{
minwrk = *n * 6;
}
if (wantse)
{
minwrk = *n * 10;
}
else if (wantsv || wantsb)
{
minwrk = (*n << 1) * (*n + 4) + 16;
}
maxwrk = minwrk;
/* Computing MAX */
i__1 = maxwrk;
i__2 = *n + *n * ilaenv_(&c__1, "SGEQRF", " ", n, & c__1, n, &c__0); // , expr subst
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk;
i__2 = *n + *n * ilaenv_(&c__1, "SORMQR", " ", n, & c__1, n, &c__0); // , expr subst
maxwrk = max(i__1,i__2);
if (ilvl)
{
/* Computing MAX */
i__1 = maxwrk;
i__2 = *n + *n * ilaenv_(&c__1, "SORGQR", " ", n, &c__1, n, &c__0); // , expr subst
maxwrk = max(i__1,i__2);
}
}
work[1] = (real) maxwrk;
if (*lwork < minwrk && ! lquery)
{
*info = -26;
}
}
if (*info != 0)
{
i__1 = -(*info);
xerbla_("SGGEVX", &i__1);
return 0;
}
else if (lquery)
{
return 0;
}
/* Quick return if possible */
if (*n == 0)
{
return 0;
}
/* Get machine constants */
eps = slamch_("P");
smlnum = slamch_("S");
bignum = 1.f / smlnum;
slabad_(&smlnum, &bignum);
smlnum = sqrt(smlnum) / eps;
bignum = 1.f / smlnum;
/* Scale A if max element outside range [SMLNUM,BIGNUM] */
anrm = slange_("M", n, n, &a[a_offset], lda, &work[1]);
ilascl = FALSE_;
if (anrm > 0.f && anrm < smlnum)
{
anrmto = smlnum;
ilascl = TRUE_;
}
else if (anrm > bignum)
{
anrmto = bignum;
ilascl = TRUE_;
}
if (ilascl)
{
slascl_("G", &c__0, &c__0, &anrm, &anrmto, n, n, &a[a_offset], lda, & ierr);
}
/* Scale B if max element outside range [SMLNUM,BIGNUM] */
bnrm = slange_("M", n, n, &b[b_offset], ldb, &work[1]);
ilbscl = FALSE_;
if (bnrm > 0.f && bnrm < smlnum)
{
bnrmto = smlnum;
ilbscl = TRUE_;
}
else if (bnrm > bignum)
{
bnrmto = bignum;
ilbscl = TRUE_;
}
if (ilbscl)
{
slascl_("G", &c__0, &c__0, &bnrm, &bnrmto, n, n, &b[b_offset], ldb, & ierr);
}
/* Permute and/or balance the matrix pair (A,B) */
/* (Workspace: need 6*N if BALANC = 'S' or 'B', 1 otherwise) */
sggbal_(balanc, n, &a[a_offset], lda, &b[b_offset], ldb, ilo, ihi, & lscale[1], &rscale[1], &work[1], &ierr);
/* Compute ABNRM and BBNRM */
*abnrm = slange_("1", n, n, &a[a_offset], lda, &work[1]);
if (ilascl)
{
work[1] = *abnrm;
slascl_("G", &c__0, &c__0, &anrmto, &anrm, &c__1, &c__1, &work[1], & c__1, &ierr);
*abnrm = work[1];
}
*bbnrm = slange_("1", n, n, &b[b_offset], ldb, &work[1]);
if (ilbscl)
{
work[1] = *bbnrm;
slascl_("G", &c__0, &c__0, &bnrmto, &bnrm, &c__1, &c__1, &work[1], & c__1, &ierr);
*bbnrm = work[1];
}
/* Reduce B to triangular form (QR decomposition of B) */
/* (Workspace: need N, prefer N*NB ) */
irows = *ihi + 1 - *ilo;
if (ilv || ! wantsn)
{
icols = *n + 1 - *ilo;
}
else
{
icols = irows;
}
itau = 1;
iwrk = itau + irows;
i__1 = *lwork + 1 - iwrk;
sgeqrf_(&irows, &icols, &b[*ilo + *ilo * b_dim1], ldb, &work[itau], &work[ iwrk], &i__1, &ierr);
/* Apply the orthogonal transformation to A */
/* (Workspace: need N, prefer N*NB) */
i__1 = *lwork + 1 - iwrk;
sormqr_("L", "T", &irows, &icols, &irows, &b[*ilo + *ilo * b_dim1], ldb, & work[itau], &a[*ilo + *ilo * a_dim1], lda, &work[iwrk], &i__1, & ierr);
/* Initialize VL and/or VR */
/* (Workspace: need N, prefer N*NB) */
if (ilvl)
{
slaset_("Full", n, n, &c_b57, &c_b58, &vl[vl_offset], ldvl) ;
if (irows > 1)
{
i__1 = irows - 1;
i__2 = irows - 1;
slacpy_("L", &i__1, &i__2, &b[*ilo + 1 + *ilo * b_dim1], ldb, &vl[ *ilo + 1 + *ilo * vl_dim1], ldvl);
}
i__1 = *lwork + 1 - iwrk;
sorgqr_(&irows, &irows, &irows, &vl[*ilo + *ilo * vl_dim1], ldvl, & work[itau], &work[iwrk], &i__1, &ierr);
}
if (ilvr)
{
slaset_("Full", n, n, &c_b57, &c_b58, &vr[vr_offset], ldvr) ;
}
/* Reduce to generalized Hessenberg form */
/* (Workspace: none needed) */
if (ilv || ! wantsn)
{
/* Eigenvectors requested -- work on whole matrix. */
sgghrd_(jobvl, jobvr, n, ilo, ihi, &a[a_offset], lda, &b[b_offset], ldb, &vl[vl_offset], ldvl, &vr[vr_offset], ldvr, &ierr);
}
else
{
sgghrd_("N", "N", &irows, &c__1, &irows, &a[*ilo + *ilo * a_dim1], lda, &b[*ilo + *ilo * b_dim1], ldb, &vl[vl_offset], ldvl, &vr[ vr_offset], ldvr, &ierr);
}
/* Perform QZ algorithm (Compute eigenvalues, and optionally, the */
/* Schur forms and Schur vectors) */
/* (Workspace: need N) */
if (ilv || ! wantsn)
{
*(unsigned char *)chtemp = 'S';
}
else
{
*(unsigned char *)chtemp = 'E';
}
shgeqz_(chtemp, jobvl, jobvr, n, ilo, ihi, &a[a_offset], lda, &b[b_offset] , ldb, &alphar[1], &alphai[1], &beta[1], &vl[vl_offset], ldvl, & vr[vr_offset], ldvr, &work[1], lwork, &ierr);
if (ierr != 0)
{
if (ierr > 0 && ierr <= *n)
{
*info = ierr;
}
else if (ierr > *n && ierr <= *n << 1)
{
*info = ierr - *n;
}
else
{
*info = *n + 1;
}
goto L130;
}
/* Compute Eigenvectors and estimate condition numbers if desired */
/* (Workspace: STGEVC: need 6*N */
/* STGSNA: need 2*N*(N+2)+16 if SENSE = 'V' or 'B', */
/* need N otherwise ) */
if (ilv || ! wantsn)
{
if (ilv)
{
if (ilvl)
{
if (ilvr)
{
*(unsigned char *)chtemp = 'B';
}
else
{
*(unsigned char *)chtemp = 'L';
}
}
else
{
*(unsigned char *)chtemp = 'R';
}
stgevc_(chtemp, "B", ldumma, n, &a[a_offset], lda, &b[b_offset], ldb, &vl[vl_offset], ldvl, &vr[vr_offset], ldvr, n, &in, & work[1], &ierr);
if (ierr != 0)
{
*info = *n + 2;
goto L130;
}
}
if (! wantsn)
{
/* compute eigenvectors (STGEVC) and estimate condition */
/* numbers (STGSNA). Note that the definition of the condition */
/* number is not invariant under transformation (u,v) to */
/* (Q*u, Z*v), where (u,v) are eigenvectors of the generalized */
/* Schur form (S,T), Q and Z are orthogonal matrices. In order */
/* to avoid using extra 2*N*N workspace, we have to recalculate */
/* eigenvectors and estimate one condition numbers at a time. */
pair = FALSE_;
i__1 = *n;
for (i__ = 1;
i__ <= i__1;
++i__)
{
if (pair)
{
pair = FALSE_;
goto L20;
}
mm = 1;
if (i__ < *n)
{
if (a[i__ + 1 + i__ * a_dim1] != 0.f)
{
pair = TRUE_;
mm = 2;
}
}
i__2 = *n;
for (j = 1;
j <= i__2;
++j)
{
bwork[j] = FALSE_;
/* L10: */
}
if (mm == 1)
{
bwork[i__] = TRUE_;
}
else if (mm == 2)
{
bwork[i__] = TRUE_;
bwork[i__ + 1] = TRUE_;
}
iwrk = mm * *n + 1;
iwrk1 = iwrk + mm * *n;
/* Compute a pair of left and right eigenvectors. */
/* (compute workspace: need up to 4*N + 6*N) */
if (wantse || wantsb)
{
stgevc_("B", "S", &bwork[1], n, &a[a_offset], lda, &b[ b_offset], ldb, &work[1], n, &work[iwrk], n, &mm, &m, &work[iwrk1], &ierr);
if (ierr != 0)
{
*info = *n + 2;
goto L130;
}
}
i__2 = *lwork - iwrk1 + 1;
stgsna_(sense, "S", &bwork[1], n, &a[a_offset], lda, &b[ b_offset], ldb, &work[1], n, &work[iwrk], n, &rconde[ i__], &rcondv[i__], &mm, &m, &work[iwrk1], &i__2, & iwork[1], &ierr);
L20:
;
}
}
}
/* Undo balancing on VL and VR and normalization */
/* (Workspace: none needed) */
if (ilvl)
{
sggbak_(balanc, "L", n, ilo, ihi, &lscale[1], &rscale[1], n, &vl[ vl_offset], ldvl, &ierr);
i__1 = *n;
for (jc = 1;
jc <= i__1;
++jc)
{
if (alphai[jc] < 0.f)
{
goto L70;
}
temp = 0.f;
if (alphai[jc] == 0.f)
{
i__2 = *n;
for (jr = 1;
jr <= i__2;
++jr)
{
/* Computing MAX */
r__2 = temp;
r__3 = (r__1 = vl[jr + jc * vl_dim1], abs( r__1)); // , expr subst
temp = max(r__2,r__3);
/* L30: */
}
}
else
{
i__2 = *n;
for (jr = 1;
jr <= i__2;
++jr)
{
/* Computing MAX */
r__3 = temp;
r__4 = (r__1 = vl[jr + jc * vl_dim1], abs( r__1)) + (r__2 = vl[jr + (jc + 1) * vl_dim1], abs( r__2)); // , expr subst
temp = max(r__3,r__4);
/* L40: */
}
}
if (temp < smlnum)
{
goto L70;
}
temp = 1.f / temp;
if (alphai[jc] == 0.f)
{
i__2 = *n;
for (jr = 1;
jr <= i__2;
++jr)
{
vl[jr + jc * vl_dim1] *= temp;
/* L50: */
}
}
else
{
i__2 = *n;
for (jr = 1;
jr <= i__2;
++jr)
{
vl[jr + jc * vl_dim1] *= temp;
vl[jr + (jc + 1) * vl_dim1] *= temp;
/* L60: */
}
}
L70:
;
}
}
if (ilvr)
{
sggbak_(balanc, "R", n, ilo, ihi, &lscale[1], &rscale[1], n, &vr[ vr_offset], ldvr, &ierr);
i__1 = *n;
for (jc = 1;
jc <= i__1;
++jc)
{
if (alphai[jc] < 0.f)
{
goto L120;
}
temp = 0.f;
if (alphai[jc] == 0.f)
{
i__2 = *n;
for (jr = 1;
jr <= i__2;
++jr)
{
/* Computing MAX */
r__2 = temp;
r__3 = (r__1 = vr[jr + jc * vr_dim1], abs( r__1)); // , expr subst
temp = max(r__2,r__3);
/* L80: */
}
}
else
{
i__2 = *n;
for (jr = 1;
jr <= i__2;
++jr)
{
/* Computing MAX */
r__3 = temp;
r__4 = (r__1 = vr[jr + jc * vr_dim1], abs( r__1)) + (r__2 = vr[jr + (jc + 1) * vr_dim1], abs( r__2)); // , expr subst
temp = max(r__3,r__4);
/* L90: */
}
}
if (temp < smlnum)
{
goto L120;
}
temp = 1.f / temp;
if (alphai[jc] == 0.f)
{
i__2 = *n;
for (jr = 1;
jr <= i__2;
++jr)
{
vr[jr + jc * vr_dim1] *= temp;
/* L100: */
}
}
else
{
i__2 = *n;
for (jr = 1;
jr <= i__2;
++jr)
{
vr[jr + jc * vr_dim1] *= temp;
vr[jr + (jc + 1) * vr_dim1] *= temp;
/* L110: */
}
}
L120:
;
}
}
/* Undo scaling if necessary */
L130:
if (ilascl)
{
slascl_("G", &c__0, &c__0, &anrmto, &anrm, n, &c__1, &alphar[1], n, & ierr);
slascl_("G", &c__0, &c__0, &anrmto, &anrm, n, &c__1, &alphai[1], n, & ierr);
}
if (ilbscl)
{
slascl_("G", &c__0, &c__0, &bnrmto, &bnrm, n, &c__1, &beta[1], n, & ierr);
}
work[1] = (real) maxwrk;
return 0;
/* End of SGGEVX */
}
/* Subroutine */ int sgeevx_(char *balanc, char *jobvl, char *jobvr, char *
sense, integer *n, real *a, integer *lda, real *wr, real *wi, real *
vl, integer *ldvl, real *vr, integer *ldvr, integer *ilo, integer *
ihi, real *scale, real *abnrm, real *rconde, real *rcondv, real *work,
integer *lwork, integer *iwork, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, vl_dim1, vl_offset, vr_dim1, vr_offset, i__1,
i__2, i__3;
real r__1, r__2;
/* Builtin functions */
double sqrt(doublereal);
/* Local variables */
integer i__, k;
real r__, cs, sn;
char job[1];
real scl, dum[1], eps;
char side[1];
real anrm;
integer ierr, itau, iwrk, nout;
extern /* Subroutine */ int srot_(integer *, real *, integer *, real *,
integer *, real *, real *);
extern doublereal snrm2_(integer *, real *, integer *);
integer icond;
extern logical lsame_(char *, char *);
extern /* Subroutine */ int sscal_(integer *, real *, real *, integer *);
extern doublereal slapy2_(real *, real *);
extern /* Subroutine */ int slabad_(real *, real *);
logical scalea;
real cscale;
extern /* Subroutine */ int sgebak_(char *, char *, integer *, integer *,
integer *, real *, integer *, real *, integer *, integer *), sgebal_(char *, integer *, real *, integer *,
integer *, integer *, real *, integer *);
extern doublereal slamch_(char *), slange_(char *, integer *,
integer *, real *, integer *, real *);
extern /* Subroutine */ int sgehrd_(integer *, integer *, integer *, real
*, integer *, real *, real *, integer *, integer *), xerbla_(char
*, integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *);
logical select[1];
real bignum;
extern /* Subroutine */ int slascl_(char *, integer *, integer *, real *,
real *, integer *, integer *, real *, integer *, integer *);
extern integer isamax_(integer *, real *, integer *);
extern /* Subroutine */ int slacpy_(char *, integer *, integer *, real *,
integer *, real *, integer *), slartg_(real *, real *,
real *, real *, real *), sorghr_(integer *, integer *, integer *,
real *, integer *, real *, real *, integer *, integer *), shseqr_(
char *, char *, integer *, integer *, integer *, real *, integer *
, real *, real *, real *, integer *, real *, integer *, integer *), strevc_(char *, char *, logical *, integer *,
real *, integer *, real *, integer *, real *, integer *, integer *
, integer *, real *, integer *);
integer minwrk, maxwrk;
extern /* Subroutine */ int strsna_(char *, char *, logical *, integer *,
real *, integer *, real *, integer *, real *, integer *, real *,
real *, integer *, integer *, real *, integer *, integer *,
integer *);
logical wantvl, wntsnb;
integer hswork;
logical wntsne;
real smlnum;
logical lquery, wantvr, wntsnn, wntsnv;
/* -- LAPACK driver routine (version 3.2) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* SGEEVX computes for an N-by-N real nonsymmetric matrix A, the */
/* eigenvalues and, optionally, the left and/or right eigenvectors. */
/* Optionally also, it computes a balancing transformation to improve */
/* the conditioning of the eigenvalues and eigenvectors (ILO, IHI, */
/* SCALE, and ABNRM), reciprocal condition numbers for the eigenvalues */
/* (RCONDE), and reciprocal condition numbers for the right */
/* eigenvectors (RCONDV). */
/* The right eigenvector v(j) of A satisfies */
/* A * v(j) = lambda(j) * v(j) */
/* where lambda(j) is its eigenvalue. */
/* The left eigenvector u(j) of A satisfies */
/* u(j)**H * A = lambda(j) * u(j)**H */
/* where u(j)**H denotes the conjugate transpose of u(j). */
/* The computed eigenvectors are normalized to have Euclidean norm */
/* equal to 1 and largest component real. */
/* Balancing a matrix means permuting the rows and columns to make it */
/* more nearly upper triangular, and applying a diagonal similarity */
/* transformation D * A * D**(-1), where D is a diagonal matrix, to */
/* make its rows and columns closer in norm and the condition numbers */
/* of its eigenvalues and eigenvectors smaller. The computed */
/* reciprocal condition numbers correspond to the balanced matrix. */
/* Permuting rows and columns will not change the condition numbers */
/* (in exact arithmetic) but diagonal scaling will. For further */
/* explanation of balancing, see section 4.10.2 of the LAPACK */
/* Users' Guide. */
/* Arguments */
/* ========= */
/* BALANC (input) CHARACTER*1 */
/* Indicates how the input matrix should be diagonally scaled */
/* and/or permuted to improve the conditioning of its */
/* eigenvalues. */
/* = 'N': Do not diagonally scale or permute; */
/* = 'P': Perform permutations to make the matrix more nearly */
/* upper triangular. Do not diagonally scale; */
/* = 'S': Diagonally scale the matrix, i.e. replace A by */
/* D*A*D**(-1), where D is a diagonal matrix chosen */
/* to make the rows and columns of A more equal in */
/* norm. Do not permute; */
/* = 'B': Both diagonally scale and permute A. */
/* Computed reciprocal condition numbers will be for the matrix */
/* after balancing and/or permuting. Permuting does not change */
/* condition numbers (in exact arithmetic), but balancing does. */
/* JOBVL (input) CHARACTER*1 */
/* = 'N': left eigenvectors of A are not computed; */
/* = 'V': left eigenvectors of A are computed. */
/* If SENSE = 'E' or 'B', JOBVL must = 'V'. */
/* JOBVR (input) CHARACTER*1 */
/* = 'N': right eigenvectors of A are not computed; */
/* = 'V': right eigenvectors of A are computed. */
/* If SENSE = 'E' or 'B', JOBVR must = 'V'. */
/* SENSE (input) CHARACTER*1 */
/* Determines which reciprocal condition numbers are computed. */
/* = 'N': None are computed; */
/* = 'E': Computed for eigenvalues only; */
/* = 'V': Computed for right eigenvectors only; */
/* = 'B': Computed for eigenvalues and right eigenvectors. */
/* If SENSE = 'E' or 'B', both left and right eigenvectors */
/* must also be computed (JOBVL = 'V' and JOBVR = 'V'). */
/* N (input) INTEGER */
/* The order of the matrix A. N >= 0. */
/* A (input/output) REAL array, dimension (LDA,N) */
/* On entry, the N-by-N matrix A. */
/* On exit, A has been overwritten. If JOBVL = 'V' or */
/* JOBVR = 'V', A contains the real Schur form of the balanced */
/* version of the input matrix A. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,N). */
/* WR (output) REAL array, dimension (N) */
/* WI (output) REAL array, dimension (N) */
/* WR and WI contain the real and imaginary parts, */
/* respectively, of the computed eigenvalues. Complex */
/* conjugate pairs of eigenvalues will appear consecutively */
/* with the eigenvalue having the positive imaginary part */
/* first. */
/* VL (output) REAL array, dimension (LDVL,N) */
/* If JOBVL = 'V', the left eigenvectors u(j) are stored one */
/* after another in the columns of VL, in the same order */
/* as their eigenvalues. */
/* If JOBVL = 'N', VL is not referenced. */
/* If the j-th eigenvalue is real, then u(j) = VL(:,j), */
/* the j-th column of VL. */
/* If the j-th and (j+1)-st eigenvalues form a complex */
/* conjugate pair, then u(j) = VL(:,j) + i*VL(:,j+1) and */
/* u(j+1) = VL(:,j) - i*VL(:,j+1). */
/* LDVL (input) INTEGER */
/* The leading dimension of the array VL. LDVL >= 1; if */
/* JOBVL = 'V', LDVL >= N. */
/* VR (output) REAL array, dimension (LDVR,N) */
/* If JOBVR = 'V', the right eigenvectors v(j) are stored one */
/* after another in the columns of VR, in the same order */
/* as their eigenvalues. */
/* If JOBVR = 'N', VR is not referenced. */
/* If the j-th eigenvalue is real, then v(j) = VR(:,j), */
/* the j-th column of VR. */
/* If the j-th and (j+1)-st eigenvalues form a complex */
/* conjugate pair, then v(j) = VR(:,j) + i*VR(:,j+1) and */
/* v(j+1) = VR(:,j) - i*VR(:,j+1). */
/* LDVR (input) INTEGER */
/* The leading dimension of the array VR. LDVR >= 1, and if */
/* JOBVR = 'V', LDVR >= N. */
/* ILO (output) INTEGER */
/* IHI (output) INTEGER */
/* ILO and IHI are integer values determined when A was */
/* balanced. The balanced A(i,j) = 0 if I > J and */
/* J = 1,...,ILO-1 or I = IHI+1,...,N. */
/* SCALE (output) REAL array, dimension (N) */
/* Details of the permutations and scaling factors applied */
/* when balancing A. If P(j) is the index of the row and column */
/* interchanged with row and column j, and D(j) is the scaling */
/* factor applied to row and column j, then */
/* SCALE(J) = P(J), for J = 1,...,ILO-1 */
/* = D(J), for J = ILO,...,IHI */
/* = P(J) for J = IHI+1,...,N. */
/* The order in which the interchanges are made is N to IHI+1, */
/* then 1 to ILO-1. */
/* ABNRM (output) REAL */
/* The one-norm of the balanced matrix (the maximum */
/* of the sum of absolute values of elements of any column). */
/* RCONDE (output) REAL array, dimension (N) */
/* RCONDE(j) is the reciprocal condition number of the j-th */
/* eigenvalue. */
/* RCONDV (output) REAL array, dimension (N) */
/* RCONDV(j) is the reciprocal condition number of the j-th */
/* right eigenvector. */
/* 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. If SENSE = 'N' or 'E', */
/* LWORK >= max(1,2*N), and if JOBVL = 'V' or JOBVR = 'V', */
/* LWORK >= 3*N. If SENSE = 'V' or 'B', LWORK >= N*(N+6). */
/* For good performance, LWORK must generally be larger. */
/* 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 (2*N-2) */
/* If SENSE = 'N' or 'E', not referenced. */
/* INFO (output) INTEGER */
/* = 0: successful exit */
/* < 0: if INFO = -i, the i-th argument had an illegal value. */
/* > 0: if INFO = i, the QR algorithm failed to compute all the */
/* eigenvalues, and no eigenvectors or condition numbers */
/* have been computed; elements 1:ILO-1 and i+1:N of WR */
/* and WI contain eigenvalues which have converged. */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. Local Arrays .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Test the input arguments */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
--wr;
--wi;
vl_dim1 = *ldvl;
vl_offset = 1 + vl_dim1;
vl -= vl_offset;
vr_dim1 = *ldvr;
vr_offset = 1 + vr_dim1;
vr -= vr_offset;
--scale;
--rconde;
--rcondv;
--work;
--iwork;
/* Function Body */
*info = 0;
lquery = *lwork == -1;
wantvl = lsame_(jobvl, "V");
wantvr = lsame_(jobvr, "V");
wntsnn = lsame_(sense, "N");
wntsne = lsame_(sense, "E");
wntsnv = lsame_(sense, "V");
wntsnb = lsame_(sense, "B");
if (! (lsame_(balanc, "N") || lsame_(balanc, "S") || lsame_(balanc, "P")
|| lsame_(balanc, "B"))) {
*info = -1;
} else if (! wantvl && ! lsame_(jobvl, "N")) {
*info = -2;
} else if (! wantvr && ! lsame_(jobvr, "N")) {
*info = -3;
} else if (! (wntsnn || wntsne || wntsnb || wntsnv) || (wntsne || wntsnb)
&& ! (wantvl && wantvr)) {
*info = -4;
} else if (*n < 0) {
*info = -5;
} else if (*lda < max(1,*n)) {
*info = -7;
} else if (*ldvl < 1 || wantvl && *ldvl < *n) {
*info = -11;
} else if (*ldvr < 1 || wantvr && *ldvr < *n) {
*info = -13;
}
/* Compute workspace */
/* (Note: Comments in the code beginning "Workspace:" describe the */
/* minimal amount of workspace needed at that point in the code, */
/* as well as the preferred amount for good performance. */
/* NB refers to the optimal block size for the immediately */
/* following subroutine, as returned by ILAENV. */
/* HSWORK refers to the workspace preferred by SHSEQR, as */
/* calculated below. HSWORK is computed assuming ILO=1 and IHI=N, */
/* the worst case.) */
if (*info == 0) {
if (*n == 0) {
minwrk = 1;
maxwrk = 1;
} else {
maxwrk = *n + *n * ilaenv_(&c__1, "SGEHRD", " ", n, &c__1, n, &
c__0);
if (wantvl) {
shseqr_("S", "V", n, &c__1, n, &a[a_offset], lda, &wr[1], &wi[
1], &vl[vl_offset], ldvl, &work[1], &c_n1, info);
} else if (wantvr) {
shseqr_("S", "V", n, &c__1, n, &a[a_offset], lda, &wr[1], &wi[
1], &vr[vr_offset], ldvr, &work[1], &c_n1, info);
} else {
if (wntsnn) {
shseqr_("E", "N", n, &c__1, n, &a[a_offset], lda, &wr[1],
&wi[1], &vr[vr_offset], ldvr, &work[1], &c_n1,
info);
} else {
shseqr_("S", "N", n, &c__1, n, &a[a_offset], lda, &wr[1],
&wi[1], &vr[vr_offset], ldvr, &work[1], &c_n1,
info);
}
}
hswork = work[1];
if (! wantvl && ! wantvr) {
minwrk = *n << 1;
if (! wntsnn) {
/* Computing MAX */
i__1 = minwrk, i__2 = *n * *n + *n * 6;
minwrk = max(i__1,i__2);
}
maxwrk = max(maxwrk,hswork);
if (! wntsnn) {
/* Computing MAX */
i__1 = maxwrk, i__2 = *n * *n + *n * 6;
maxwrk = max(i__1,i__2);
}
} else {
minwrk = *n * 3;
if (! wntsnn && ! wntsne) {
/* Computing MAX */
i__1 = minwrk, i__2 = *n * *n + *n * 6;
minwrk = max(i__1,i__2);
}
maxwrk = max(maxwrk,hswork);
/* Computing MAX */
i__1 = maxwrk, i__2 = *n + (*n - 1) * ilaenv_(&c__1, "SORGHR",
" ", n, &c__1, n, &c_n1);
maxwrk = max(i__1,i__2);
if (! wntsnn && ! wntsne) {
/* Computing MAX */
i__1 = maxwrk, i__2 = *n * *n + *n * 6;
maxwrk = max(i__1,i__2);
}
/* Computing MAX */
i__1 = maxwrk, i__2 = *n * 3;
maxwrk = max(i__1,i__2);
}
maxwrk = max(maxwrk,minwrk);
}
work[1] = (real) maxwrk;
if (*lwork < minwrk && ! lquery) {
*info = -21;
}
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SGEEVX", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
if (*n == 0) {
return 0;
}
/* Get machine constants */
eps = slamch_("P");
smlnum = slamch_("S");
bignum = 1.f / smlnum;
slabad_(&smlnum, &bignum);
smlnum = sqrt(smlnum) / eps;
bignum = 1.f / smlnum;
/* Scale A if max element outside range [SMLNUM,BIGNUM] */
icond = 0;
anrm = slange_("M", n, n, &a[a_offset], lda, dum);
scalea = FALSE_;
if (anrm > 0.f && anrm < smlnum) {
scalea = TRUE_;
cscale = smlnum;
} else if (anrm > bignum) {
scalea = TRUE_;
cscale = bignum;
}
if (scalea) {
slascl_("G", &c__0, &c__0, &anrm, &cscale, n, n, &a[a_offset], lda, &
ierr);
}
/* Balance the matrix and compute ABNRM */
sgebal_(balanc, n, &a[a_offset], lda, ilo, ihi, &scale[1], &ierr);
*abnrm = slange_("1", n, n, &a[a_offset], lda, dum);
if (scalea) {
dum[0] = *abnrm;
slascl_("G", &c__0, &c__0, &cscale, &anrm, &c__1, &c__1, dum, &c__1, &
ierr);
*abnrm = dum[0];
}
/* Reduce to upper Hessenberg form */
/* (Workspace: need 2*N, prefer N+N*NB) */
itau = 1;
iwrk = itau + *n;
i__1 = *lwork - iwrk + 1;
sgehrd_(n, ilo, ihi, &a[a_offset], lda, &work[itau], &work[iwrk], &i__1, &
ierr);
if (wantvl) {
/* Want left eigenvectors */
/* Copy Householder vectors to VL */
*(unsigned char *)side = 'L';
slacpy_("L", n, n, &a[a_offset], lda, &vl[vl_offset], ldvl)
;
/* Generate orthogonal matrix in VL */
/* (Workspace: need 2*N-1, prefer N+(N-1)*NB) */
i__1 = *lwork - iwrk + 1;
sorghr_(n, ilo, ihi, &vl[vl_offset], ldvl, &work[itau], &work[iwrk], &
i__1, &ierr);
/* Perform QR iteration, accumulating Schur vectors in VL */
/* (Workspace: need 1, prefer HSWORK (see comments) ) */
iwrk = itau;
i__1 = *lwork - iwrk + 1;
shseqr_("S", "V", n, ilo, ihi, &a[a_offset], lda, &wr[1], &wi[1], &vl[
vl_offset], ldvl, &work[iwrk], &i__1, info);
if (wantvr) {
/* Want left and right eigenvectors */
/* Copy Schur vectors to VR */
*(unsigned char *)side = 'B';
slacpy_("F", n, n, &vl[vl_offset], ldvl, &vr[vr_offset], ldvr);
}
} else if (wantvr) {
/* Want right eigenvectors */
/* Copy Householder vectors to VR */
*(unsigned char *)side = 'R';
slacpy_("L", n, n, &a[a_offset], lda, &vr[vr_offset], ldvr)
;
/* Generate orthogonal matrix in VR */
/* (Workspace: need 2*N-1, prefer N+(N-1)*NB) */
i__1 = *lwork - iwrk + 1;
sorghr_(n, ilo, ihi, &vr[vr_offset], ldvr, &work[itau], &work[iwrk], &
i__1, &ierr);
/* Perform QR iteration, accumulating Schur vectors in VR */
/* (Workspace: need 1, prefer HSWORK (see comments) ) */
iwrk = itau;
i__1 = *lwork - iwrk + 1;
shseqr_("S", "V", n, ilo, ihi, &a[a_offset], lda, &wr[1], &wi[1], &vr[
vr_offset], ldvr, &work[iwrk], &i__1, info);
} else {
/* Compute eigenvalues only */
/* If condition numbers desired, compute Schur form */
if (wntsnn) {
*(unsigned char *)job = 'E';
} else {
*(unsigned char *)job = 'S';
}
/* (Workspace: need 1, prefer HSWORK (see comments) ) */
iwrk = itau;
i__1 = *lwork - iwrk + 1;
shseqr_(job, "N", n, ilo, ihi, &a[a_offset], lda, &wr[1], &wi[1], &vr[
vr_offset], ldvr, &work[iwrk], &i__1, info);
}
/* If INFO > 0 from SHSEQR, then quit */
if (*info > 0) {
goto L50;
}
if (wantvl || wantvr) {
/* Compute left and/or right eigenvectors */
/* (Workspace: need 3*N) */
strevc_(side, "B", select, n, &a[a_offset], lda, &vl[vl_offset], ldvl,
&vr[vr_offset], ldvr, n, &nout, &work[iwrk], &ierr);
}
/* Compute condition numbers if desired */
/* (Workspace: need N*N+6*N unless SENSE = 'E') */
if (! wntsnn) {
strsna_(sense, "A", select, n, &a[a_offset], lda, &vl[vl_offset],
ldvl, &vr[vr_offset], ldvr, &rconde[1], &rcondv[1], n, &nout,
&work[iwrk], n, &iwork[1], &icond);
}
if (wantvl) {
/* Undo balancing of left eigenvectors */
sgebak_(balanc, "L", n, ilo, ihi, &scale[1], n, &vl[vl_offset], ldvl,
&ierr);
/* Normalize left eigenvectors and make largest component real */
i__1 = *n;
for (i__ = 1; i__ <= i__1; ++i__) {
if (wi[i__] == 0.f) {
scl = 1.f / snrm2_(n, &vl[i__ * vl_dim1 + 1], &c__1);
sscal_(n, &scl, &vl[i__ * vl_dim1 + 1], &c__1);
} else if (wi[i__] > 0.f) {
r__1 = snrm2_(n, &vl[i__ * vl_dim1 + 1], &c__1);
r__2 = snrm2_(n, &vl[(i__ + 1) * vl_dim1 + 1], &c__1);
scl = 1.f / slapy2_(&r__1, &r__2);
sscal_(n, &scl, &vl[i__ * vl_dim1 + 1], &c__1);
sscal_(n, &scl, &vl[(i__ + 1) * vl_dim1 + 1], &c__1);
i__2 = *n;
for (k = 1; k <= i__2; ++k) {
/* Computing 2nd power */
r__1 = vl[k + i__ * vl_dim1];
/* Computing 2nd power */
r__2 = vl[k + (i__ + 1) * vl_dim1];
work[k] = r__1 * r__1 + r__2 * r__2;
/* L10: */
}
k = isamax_(n, &work[1], &c__1);
slartg_(&vl[k + i__ * vl_dim1], &vl[k + (i__ + 1) * vl_dim1],
&cs, &sn, &r__);
srot_(n, &vl[i__ * vl_dim1 + 1], &c__1, &vl[(i__ + 1) *
vl_dim1 + 1], &c__1, &cs, &sn);
vl[k + (i__ + 1) * vl_dim1] = 0.f;
}
/* L20: */
}
}
if (wantvr) {
/* Undo balancing of right eigenvectors */
sgebak_(balanc, "R", n, ilo, ihi, &scale[1], n, &vr[vr_offset], ldvr,
&ierr);
/* Normalize right eigenvectors and make largest component real */
i__1 = *n;
for (i__ = 1; i__ <= i__1; ++i__) {
if (wi[i__] == 0.f) {
scl = 1.f / snrm2_(n, &vr[i__ * vr_dim1 + 1], &c__1);
sscal_(n, &scl, &vr[i__ * vr_dim1 + 1], &c__1);
} else if (wi[i__] > 0.f) {
r__1 = snrm2_(n, &vr[i__ * vr_dim1 + 1], &c__1);
r__2 = snrm2_(n, &vr[(i__ + 1) * vr_dim1 + 1], &c__1);
scl = 1.f / slapy2_(&r__1, &r__2);
sscal_(n, &scl, &vr[i__ * vr_dim1 + 1], &c__1);
sscal_(n, &scl, &vr[(i__ + 1) * vr_dim1 + 1], &c__1);
i__2 = *n;
for (k = 1; k <= i__2; ++k) {
/* Computing 2nd power */
r__1 = vr[k + i__ * vr_dim1];
/* Computing 2nd power */
r__2 = vr[k + (i__ + 1) * vr_dim1];
work[k] = r__1 * r__1 + r__2 * r__2;
/* L30: */
}
k = isamax_(n, &work[1], &c__1);
slartg_(&vr[k + i__ * vr_dim1], &vr[k + (i__ + 1) * vr_dim1],
&cs, &sn, &r__);
srot_(n, &vr[i__ * vr_dim1 + 1], &c__1, &vr[(i__ + 1) *
vr_dim1 + 1], &c__1, &cs, &sn);
vr[k + (i__ + 1) * vr_dim1] = 0.f;
}
/* L40: */
}
}
/* Undo scaling if necessary */
L50:
if (scalea) {
i__1 = *n - *info;
/* Computing MAX */
i__3 = *n - *info;
i__2 = max(i__3,1);
slascl_("G", &c__0, &c__0, &cscale, &anrm, &i__1, &c__1, &wr[*info +
1], &i__2, &ierr);
i__1 = *n - *info;
/* Computing MAX */
i__3 = *n - *info;
i__2 = max(i__3,1);
slascl_("G", &c__0, &c__0, &cscale, &anrm, &i__1, &c__1, &wi[*info +
1], &i__2, &ierr);
if (*info == 0) {
if ((wntsnv || wntsnb) && icond == 0) {
slascl_("G", &c__0, &c__0, &cscale, &anrm, n, &c__1, &rcondv[
1], n, &ierr);
}
} else {
i__1 = *ilo - 1;
slascl_("G", &c__0, &c__0, &cscale, &anrm, &i__1, &c__1, &wr[1],
n, &ierr);
i__1 = *ilo - 1;
slascl_("G", &c__0, &c__0, &cscale, &anrm, &i__1, &c__1, &wi[1],
n, &ierr);
}
}
work[1] = (real) maxwrk;
return 0;
/* End of SGEEVX */
} /* sgeevx_ */
/* Subroutine */ int schkgl_(integer *nin, integer *nout)
{
/* Format strings */
static char fmt_9999[] = "(1x,\002.. test output of SGGBAL .. \002)";
static char fmt_9998[] = "(1x,\002value of largest test error "
" = \002,e12.3)";
static char fmt_9997[] = "(1x,\002example number where info is not zero "
" = \002,i4)";
static char fmt_9996[] = "(1x,\002example number where ILO or IHI wrong "
" = \002,i4)";
static char fmt_9995[] = "(1x,\002example number having largest error "
" = \002,i4)";
static char fmt_9994[] = "(1x,\002number of examples where info is not 0"
" = \002,i4)";
static char fmt_9993[] = "(1x,\002total number of examples tested "
" = \002,i4)";
/* System generated locals */
integer i__1, i__2;
real r__1, r__2, r__3;
/* Builtin functions */
integer s_rsle(cilist *), do_lio(integer *, integer *, char *, ftnlen),
e_rsle(void), s_wsfe(cilist *), e_wsfe(void), do_fio(integer *,
char *, ftnlen);
/* Local variables */
real a[400] /* was [20][20] */, b[400] /* was [20][20] */;
integer i__, j, n;
real ain[400] /* was [20][20] */, bin[400] /* was [20][20] */;
integer ihi, ilo;
real eps;
integer knt, info, lmax[5];
real rmax, vmax, work[120];
integer ihiin, ninfo, iloin;
real anorm, bnorm;
extern /* Subroutine */ int sggbal_(char *, integer *, real *, integer *,
real *, integer *, integer *, integer *, real *, real *, real *,
integer *);
real lscale[20];
extern doublereal slamch_(char *);
real rscale[20];
extern doublereal slange_(char *, integer *, integer *, real *, integer *,
real *);
real lsclin[20], rsclin[20];
/* Fortran I/O blocks */
static cilist io___6 = { 0, 0, 0, 0, 0 };
static cilist io___9 = { 0, 0, 0, 0, 0 };
static cilist io___12 = { 0, 0, 0, 0, 0 };
static cilist io___14 = { 0, 0, 0, 0, 0 };
static cilist io___17 = { 0, 0, 0, 0, 0 };
static cilist io___19 = { 0, 0, 0, 0, 0 };
static cilist io___21 = { 0, 0, 0, 0, 0 };
static cilist io___23 = { 0, 0, 0, 0, 0 };
static cilist io___34 = { 0, 0, 0, fmt_9999, 0 };
static cilist io___35 = { 0, 0, 0, fmt_9998, 0 };
static cilist io___36 = { 0, 0, 0, fmt_9997, 0 };
static cilist io___37 = { 0, 0, 0, fmt_9996, 0 };
static cilist io___38 = { 0, 0, 0, fmt_9995, 0 };
static cilist io___39 = { 0, 0, 0, fmt_9994, 0 };
static cilist io___40 = { 0, 0, 0, fmt_9993, 0 };
/* -- LAPACK test routine (version 3.1) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* SCHKGL tests SGGBAL, a routine for balancing a matrix pair (A, B). */
/* Arguments */
/* ========= */
/* NIN (input) INTEGER */
/* The logical unit number for input. NIN > 0. */
/* NOUT (input) INTEGER */
/* The logical unit number for output. NOUT > 0. */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. Local Arrays .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
lmax[0] = 0;
lmax[1] = 0;
lmax[2] = 0;
ninfo = 0;
knt = 0;
rmax = 0.f;
eps = slamch_("Precision");
L10:
io___6.ciunit = *nin;
s_rsle(&io___6);
do_lio(&c__3, &c__1, (char *)&n, (ftnlen)sizeof(integer));
e_rsle();
if (n == 0) {
goto L90;
}
i__1 = n;
for (i__ = 1; i__ <= i__1; ++i__) {
io___9.ciunit = *nin;
s_rsle(&io___9);
i__2 = n;
for (j = 1; j <= i__2; ++j) {
do_lio(&c__4, &c__1, (char *)&a[i__ + j * 20 - 21], (ftnlen)
sizeof(real));
}
e_rsle();
/* L20: */
}
i__1 = n;
for (i__ = 1; i__ <= i__1; ++i__) {
io___12.ciunit = *nin;
s_rsle(&io___12);
i__2 = n;
for (j = 1; j <= i__2; ++j) {
do_lio(&c__4, &c__1, (char *)&b[i__ + j * 20 - 21], (ftnlen)
sizeof(real));
}
e_rsle();
/* L30: */
}
io___14.ciunit = *nin;
s_rsle(&io___14);
do_lio(&c__3, &c__1, (char *)&iloin, (ftnlen)sizeof(integer));
do_lio(&c__3, &c__1, (char *)&ihiin, (ftnlen)sizeof(integer));
e_rsle();
i__1 = n;
for (i__ = 1; i__ <= i__1; ++i__) {
io___17.ciunit = *nin;
s_rsle(&io___17);
i__2 = n;
for (j = 1; j <= i__2; ++j) {
do_lio(&c__4, &c__1, (char *)&ain[i__ + j * 20 - 21], (ftnlen)
sizeof(real));
}
e_rsle();
/* L40: */
}
i__1 = n;
for (i__ = 1; i__ <= i__1; ++i__) {
io___19.ciunit = *nin;
s_rsle(&io___19);
i__2 = n;
for (j = 1; j <= i__2; ++j) {
do_lio(&c__4, &c__1, (char *)&bin[i__ + j * 20 - 21], (ftnlen)
sizeof(real));
}
e_rsle();
/* L50: */
}
io___21.ciunit = *nin;
s_rsle(&io___21);
i__1 = n;
for (i__ = 1; i__ <= i__1; ++i__) {
do_lio(&c__4, &c__1, (char *)&lsclin[i__ - 1], (ftnlen)sizeof(real));
}
e_rsle();
io___23.ciunit = *nin;
s_rsle(&io___23);
i__1 = n;
for (i__ = 1; i__ <= i__1; ++i__) {
do_lio(&c__4, &c__1, (char *)&rsclin[i__ - 1], (ftnlen)sizeof(real));
}
e_rsle();
anorm = slange_("M", &n, &n, a, &c__20, work);
bnorm = slange_("M", &n, &n, b, &c__20, work);
++knt;
sggbal_("B", &n, a, &c__20, b, &c__20, &ilo, &ihi, lscale, rscale, work, &
info);
if (info != 0) {
++ninfo;
lmax[0] = knt;
}
if (ilo != iloin || ihi != ihiin) {
++ninfo;
lmax[1] = knt;
}
vmax = 0.f;
i__1 = n;
for (i__ = 1; i__ <= i__1; ++i__) {
i__2 = n;
for (j = 1; j <= i__2; ++j) {
/* Computing MAX */
r__2 = vmax, r__3 = (r__1 = a[i__ + j * 20 - 21] - ain[i__ + j *
20 - 21], dabs(r__1));
vmax = dmax(r__2,r__3);
/* Computing MAX */
r__2 = vmax, r__3 = (r__1 = b[i__ + j * 20 - 21] - bin[i__ + j *
20 - 21], dabs(r__1));
vmax = dmax(r__2,r__3);
/* L60: */
}
/* L70: */
}
i__1 = n;
for (i__ = 1; i__ <= i__1; ++i__) {
/* Computing MAX */
r__2 = vmax, r__3 = (r__1 = lscale[i__ - 1] - lsclin[i__ - 1], dabs(
r__1));
vmax = dmax(r__2,r__3);
/* Computing MAX */
r__2 = vmax, r__3 = (r__1 = rscale[i__ - 1] - rsclin[i__ - 1], dabs(
r__1));
vmax = dmax(r__2,r__3);
/* L80: */
}
vmax /= eps * dmax(anorm,bnorm);
if (vmax > rmax) {
lmax[2] = knt;
rmax = vmax;
}
goto L10;
L90:
io___34.ciunit = *nout;
s_wsfe(&io___34);
e_wsfe();
io___35.ciunit = *nout;
s_wsfe(&io___35);
do_fio(&c__1, (char *)&rmax, (ftnlen)sizeof(real));
e_wsfe();
io___36.ciunit = *nout;
s_wsfe(&io___36);
do_fio(&c__1, (char *)&lmax[0], (ftnlen)sizeof(integer));
e_wsfe();
io___37.ciunit = *nout;
s_wsfe(&io___37);
do_fio(&c__1, (char *)&lmax[1], (ftnlen)sizeof(integer));
e_wsfe();
io___38.ciunit = *nout;
s_wsfe(&io___38);
do_fio(&c__1, (char *)&lmax[2], (ftnlen)sizeof(integer));
e_wsfe();
io___39.ciunit = *nout;
s_wsfe(&io___39);
do_fio(&c__1, (char *)&ninfo, (ftnlen)sizeof(integer));
e_wsfe();
io___40.ciunit = *nout;
s_wsfe(&io___40);
do_fio(&c__1, (char *)&knt, (ftnlen)sizeof(integer));
e_wsfe();
return 0;
/* End of SCHKGL */
} /* schkgl_ */
/* Subroutine */ int sgelsd_(integer *m, integer *n, integer *nrhs, real *a,
integer *lda, real *b, integer *ldb, real *s, real *rcond, integer *
rank, real *work, integer *lwork, integer *iwork, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, i__1, i__2, i__3, i__4;
/* Builtin functions */
double log(doublereal);
/* Local variables */
integer ie, il, mm;
real eps, anrm, bnrm;
integer itau, nlvl, iascl, ibscl;
real sfmin;
integer minmn, maxmn, itaup, itauq, mnthr, nwork;
extern /* Subroutine */ int slabad_(real *, real *), sgebrd_(integer *,
integer *, real *, integer *, real *, real *, real *, real *,
real *, integer *, integer *);
extern doublereal slamch_(char *), slange_(char *, integer *,
integer *, real *, integer *, real *);
extern /* Subroutine */ int xerbla_(char *, integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *);
real bignum;
extern /* Subroutine */ int sgelqf_(integer *, integer *, real *, integer
*, real *, real *, integer *, integer *), slalsd_(char *, integer
*, integer *, integer *, real *, real *, real *, integer *, real *
, integer *, real *, integer *, integer *), slascl_(char *
, integer *, integer *, real *, real *, integer *, integer *,
real *, integer *, integer *);
integer wlalsd;
extern /* Subroutine */ int sgeqrf_(integer *, integer *, real *, integer
*, real *, real *, integer *, integer *), slacpy_(char *, integer
*, integer *, real *, integer *, real *, integer *),
slaset_(char *, integer *, integer *, real *, real *, real *,
integer *);
integer ldwork;
extern /* Subroutine */ int sormbr_(char *, char *, char *, integer *,
integer *, integer *, real *, integer *, real *, real *, integer *
, real *, integer *, integer *);
integer liwork, minwrk, maxwrk;
real smlnum;
extern /* Subroutine */ int sormlq_(char *, char *, integer *, integer *,
integer *, real *, integer *, real *, real *, integer *, real *,
integer *, integer *);
logical lquery;
integer smlsiz;
extern /* Subroutine */ int sormqr_(char *, char *, integer *, integer *,
integer *, real *, integer *, real *, real *, integer *, 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 */
/* ======= */
/* SGELSD computes the minimum-norm solution to a real linear least */
/* squares problem: */
/* minimize 2-norm(| b - A*x |) */
/* using the singular value decomposition (SVD) of A. A is an M-by-N */
/* matrix which may be rank-deficient. */
/* Several right hand side vectors b and solution vectors x can be */
/* handled in a single call; they are stored as the columns of the */
/* M-by-NRHS right hand side matrix B and the N-by-NRHS solution */
/* matrix X. */
/* The problem is solved in three steps: */
/* (1) Reduce the coefficient matrix A to bidiagonal form with */
/* Householder transformations, reducing the original problem */
/* into a "bidiagonal least squares problem" (BLS) */
/* (2) Solve the BLS using a divide and conquer approach. */
/* (3) Apply back all the Householder tranformations to solve */
/* the original least squares problem. */
/* The effective rank of A is determined by treating as zero those */
/* singular values which are less than RCOND times the largest singular */
/* value. */
/* 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. */
/* Arguments */
/* ========= */
/* M (input) INTEGER */
/* The number of rows of A. M >= 0. */
/* N (input) INTEGER */
/* The number of columns of A. N >= 0. */
/* NRHS (input) INTEGER */
/* The number of right hand sides, i.e., the number of columns */
/* of the matrices B and X. NRHS >= 0. */
/* A (input) REAL array, dimension (LDA,N) */
/* On entry, the M-by-N matrix A. */
/* On exit, A has been destroyed. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,M). */
/* B (input/output) REAL array, dimension (LDB,NRHS) */
/* On entry, the M-by-NRHS right hand side matrix B. */
/* On exit, B is overwritten by the N-by-NRHS solution */
/* matrix X. If m >= n and RANK = n, the residual */
/* sum-of-squares for the solution in the i-th column is given */
/* by the sum of squares of elements n+1:m in that column. */
/* LDB (input) INTEGER */
/* The leading dimension of the array B. LDB >= max(1,max(M,N)). */
/* S (output) REAL array, dimension (min(M,N)) */
/* The singular values of A in decreasing order. */
/* The condition number of A in the 2-norm = S(1)/S(min(m,n)). */
/* RCOND (input) REAL */
/* RCOND is used to determine the effective rank of A. */
/* Singular values S(i) <= RCOND*S(1) are treated as zero. */
/* If RCOND < 0, machine precision is used instead. */
/* RANK (output) INTEGER */
/* The effective rank of A, i.e., the number of singular values */
/* which are greater than RCOND*S(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 must be at least 1. */
/* The exact minimum amount of workspace needed depends on M, */
/* N and NRHS. As long as LWORK is at least */
/* 12*N + 2*N*SMLSIZ + 8*N*NLVL + N*NRHS + (SMLSIZ+1)**2, */
/* if M is greater than or equal to N or */
/* 12*M + 2*M*SMLSIZ + 8*M*NLVL + M*NRHS + (SMLSIZ+1)**2, */
/* if M is less than N, the code will execute correctly. */
/* SMLSIZ is returned by ILAENV and is equal to the maximum */
/* size of the subproblems at the bottom of the computation */
/* tree (usually about 25), and */
/* NLVL = MAX( 0, INT( LOG_2( MIN( M,N )/(SMLSIZ+1) ) ) + 1 ) */
/* For good performance, LWORK should generally be larger. */
/* If LWORK = -1, then a workspace query is assumed; the routine */
/* only calculates the optimal size of the array WORK and the */
/* minimum size of the array IWORK, and returns these values as */
/* the first entries of the WORK and IWORK arrays, and no error */
/* message related to LWORK is issued by XERBLA. */
/* IWORK (workspace) INTEGER array, dimension (MAX(1,LIWORK)) */
/* LIWORK >= max(1, 3*MINMN*NLVL + 11*MINMN), */
/* where MINMN = MIN( M,N ). */
/* On exit, if INFO = 0, IWORK(1) returns the minimum LIWORK. */
/* INFO (output) INTEGER */
/* = 0: successful exit */
/* < 0: if INFO = -i, the i-th argument had an illegal value. */
/* > 0: the algorithm for computing the SVD failed to converge; */
/* if INFO = i, i off-diagonal elements of an intermediate */
/* bidiagonal form did not converge to zero. */
/* Further Details */
/* =============== */
/* Based on contributions by */
/* Ming Gu and Ren-Cang Li, Computer Science Division, University of */
/* California at Berkeley, USA */
/* Osni Marques, LBNL/NERSC, USA */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Test the input arguments. */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
b_dim1 = *ldb;
b_offset = 1 + b_dim1;
b -= b_offset;
--s;
--work;
--iwork;
/* Function Body */
*info = 0;
minmn = min(*m,*n);
maxmn = max(*m,*n);
lquery = *lwork == -1;
if (*m < 0) {
*info = -1;
} else if (*n < 0) {
*info = -2;
} else if (*nrhs < 0) {
*info = -3;
} else if (*lda < max(1,*m)) {
*info = -5;
} else if (*ldb < max(1,maxmn)) {
*info = -7;
}
/* Compute workspace. */
/* (Note: Comments in the code beginning "Workspace:" describe the */
/* minimal amount of workspace needed at that point in the code, */
/* as well as the preferred amount for good performance. */
/* NB refers to the optimal block size for the immediately */
/* following subroutine, as returned by ILAENV.) */
if (*info == 0) {
minwrk = 1;
maxwrk = 1;
liwork = 1;
if (minmn > 0) {
smlsiz = ilaenv_(&c__9, "SGELSD", " ", &c__0, &c__0, &c__0, &c__0);
mnthr = ilaenv_(&c__6, "SGELSD", " ", m, n, nrhs, &c_n1);
/* Computing MAX */
i__1 = (integer) (log((real) minmn / (real) (smlsiz + 1)) / log(
2.f)) + 1;
nlvl = max(i__1,0);
liwork = minmn * 3 * nlvl + minmn * 11;
mm = *m;
if (*m >= *n && *m >= mnthr) {
/* Path 1a - overdetermined, with many more rows than */
/* columns. */
mm = *n;
/* Computing MAX */
i__1 = maxwrk, i__2 = *n + *n * ilaenv_(&c__1, "SGEQRF",
" ", m, n, &c_n1, &c_n1);
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk, i__2 = *n + *nrhs * ilaenv_(&c__1, "SORMQR",
"LT", m, nrhs, n, &c_n1);
maxwrk = max(i__1,i__2);
}
if (*m >= *n) {
/* Path 1 - overdetermined or exactly determined. */
/* Computing MAX */
i__1 = maxwrk, i__2 = *n * 3 + (mm + *n) * ilaenv_(&c__1,
"SGEBRD", " ", &mm, n, &c_n1, &c_n1);
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk, i__2 = *n * 3 + *nrhs * ilaenv_(&c__1, "SORMBR"
, "QLT", &mm, nrhs, n, &c_n1);
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk, i__2 = *n * 3 + (*n - 1) * ilaenv_(&c__1,
"SORMBR", "PLN", n, nrhs, n, &c_n1);
maxwrk = max(i__1,i__2);
/* Computing 2nd power */
i__1 = smlsiz + 1;
wlalsd = *n * 9 + (*n << 1) * smlsiz + (*n << 3) * nlvl + *n *
*nrhs + i__1 * i__1;
/* Computing MAX */
i__1 = maxwrk, i__2 = *n * 3 + wlalsd;
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = *n * 3 + mm, i__2 = *n * 3 + *nrhs, i__1 = max(i__1,
i__2), i__2 = *n * 3 + wlalsd;
minwrk = max(i__1,i__2);
}
if (*n > *m) {
/* Computing 2nd power */
i__1 = smlsiz + 1;
wlalsd = *m * 9 + (*m << 1) * smlsiz + (*m << 3) * nlvl + *m *
*nrhs + i__1 * i__1;
if (*n >= mnthr) {
/* Path 2a - underdetermined, with many more columns */
/* than rows. */
maxwrk = *m + *m * ilaenv_(&c__1, "SGELQF", " ", m, n, &
c_n1, &c_n1);
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * *m + (*m << 2) + (*m << 1) *
ilaenv_(&c__1, "SGEBRD", " ", m, m, &c_n1, &c_n1);
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * *m + (*m << 2) + *nrhs *
ilaenv_(&c__1, "SORMBR", "QLT", m, nrhs, m, &c_n1);
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * *m + (*m << 2) + (*m - 1) *
ilaenv_(&c__1, "SORMBR", "PLN", m, nrhs, m, &c_n1);
maxwrk = max(i__1,i__2);
if (*nrhs > 1) {
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * *m + *m + *m * *nrhs;
maxwrk = max(i__1,i__2);
} else {
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * *m + (*m << 1);
maxwrk = max(i__1,i__2);
}
/* Computing MAX */
i__1 = maxwrk, i__2 = *m + *nrhs * ilaenv_(&c__1, "SORMLQ"
, "LT", n, nrhs, m, &c_n1);
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * *m + (*m << 2) + wlalsd;
maxwrk = max(i__1,i__2);
} else {
/* Path 2 - remaining underdetermined cases. */
maxwrk = *m * 3 + (*n + *m) * ilaenv_(&c__1, "SGEBRD",
" ", m, n, &c_n1, &c_n1);
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * 3 + *nrhs * ilaenv_(&c__1,
"SORMBR", "QLT", m, nrhs, n, &c_n1);
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * 3 + *m * ilaenv_(&c__1, "SORM"
"BR", "PLN", n, nrhs, m, &c_n1);
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * 3 + wlalsd;
maxwrk = max(i__1,i__2);
}
/* Computing MAX */
i__1 = *m * 3 + *nrhs, i__2 = *m * 3 + *m, i__1 = max(i__1,
i__2), i__2 = *m * 3 + wlalsd;
minwrk = max(i__1,i__2);
}
}
minwrk = min(minwrk,maxwrk);
work[1] = (real) maxwrk;
iwork[1] = liwork;
if (*lwork < minwrk && ! lquery) {
*info = -12;
}
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SGELSD", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible. */
if (*m == 0 || *n == 0) {
*rank = 0;
return 0;
}
/* Get machine parameters. */
eps = slamch_("P");
sfmin = slamch_("S");
smlnum = sfmin / eps;
bignum = 1.f / smlnum;
slabad_(&smlnum, &bignum);
/* Scale A if max entry outside range [SMLNUM,BIGNUM]. */
anrm = slange_("M", m, n, &a[a_offset], lda, &work[1]);
iascl = 0;
if (anrm > 0.f && anrm < smlnum) {
/* Scale matrix norm up to SMLNUM. */
slascl_("G", &c__0, &c__0, &anrm, &smlnum, m, n, &a[a_offset], lda,
info);
iascl = 1;
} else if (anrm > bignum) {
/* Scale matrix norm down to BIGNUM. */
slascl_("G", &c__0, &c__0, &anrm, &bignum, m, n, &a[a_offset], lda,
info);
iascl = 2;
} else if (anrm == 0.f) {
/* Matrix all zero. Return zero solution. */
i__1 = max(*m,*n);
slaset_("F", &i__1, nrhs, &c_b81, &c_b81, &b[b_offset], ldb);
slaset_("F", &minmn, &c__1, &c_b81, &c_b81, &s[1], &c__1);
*rank = 0;
goto L10;
}
/* Scale B if max entry outside range [SMLNUM,BIGNUM]. */
bnrm = slange_("M", m, nrhs, &b[b_offset], ldb, &work[1]);
ibscl = 0;
if (bnrm > 0.f && bnrm < smlnum) {
/* Scale matrix norm up to SMLNUM. */
slascl_("G", &c__0, &c__0, &bnrm, &smlnum, m, nrhs, &b[b_offset], ldb,
info);
ibscl = 1;
} else if (bnrm > bignum) {
/* Scale matrix norm down to BIGNUM. */
slascl_("G", &c__0, &c__0, &bnrm, &bignum, m, nrhs, &b[b_offset], ldb,
info);
ibscl = 2;
}
/* If M < N make sure certain entries of B are zero. */
if (*m < *n) {
i__1 = *n - *m;
slaset_("F", &i__1, nrhs, &c_b81, &c_b81, &b[*m + 1 + b_dim1], ldb);
}
/* Overdetermined case. */
if (*m >= *n) {
/* Path 1 - overdetermined or exactly determined. */
mm = *m;
if (*m >= mnthr) {
/* Path 1a - overdetermined, with many more rows than columns. */
mm = *n;
itau = 1;
nwork = itau + *n;
/* Compute A=Q*R. */
/* (Workspace: need 2*N, prefer N+N*NB) */
i__1 = *lwork - nwork + 1;
sgeqrf_(m, n, &a[a_offset], lda, &work[itau], &work[nwork], &i__1,
info);
/* Multiply B by transpose(Q). */
/* (Workspace: need N+NRHS, prefer N+NRHS*NB) */
i__1 = *lwork - nwork + 1;
sormqr_("L", "T", m, nrhs, n, &a[a_offset], lda, &work[itau], &b[
b_offset], ldb, &work[nwork], &i__1, info);
/* Zero out below R. */
if (*n > 1) {
i__1 = *n - 1;
i__2 = *n - 1;
slaset_("L", &i__1, &i__2, &c_b81, &c_b81, &a[a_dim1 + 2],
lda);
}
}
ie = 1;
itauq = ie + *n;
itaup = itauq + *n;
nwork = itaup + *n;
/* Bidiagonalize R in A. */
/* (Workspace: need 3*N+MM, prefer 3*N+(MM+N)*NB) */
i__1 = *lwork - nwork + 1;
sgebrd_(&mm, n, &a[a_offset], lda, &s[1], &work[ie], &work[itauq], &
work[itaup], &work[nwork], &i__1, info);
/* Multiply B by transpose of left bidiagonalizing vectors of R. */
/* (Workspace: need 3*N+NRHS, prefer 3*N+NRHS*NB) */
i__1 = *lwork - nwork + 1;
sormbr_("Q", "L", "T", &mm, nrhs, n, &a[a_offset], lda, &work[itauq],
&b[b_offset], ldb, &work[nwork], &i__1, info);
/* Solve the bidiagonal least squares problem. */
slalsd_("U", &smlsiz, n, nrhs, &s[1], &work[ie], &b[b_offset], ldb,
rcond, rank, &work[nwork], &iwork[1], info);
if (*info != 0) {
goto L10;
}
/* Multiply B by right bidiagonalizing vectors of R. */
i__1 = *lwork - nwork + 1;
sormbr_("P", "L", "N", n, nrhs, n, &a[a_offset], lda, &work[itaup], &
b[b_offset], ldb, &work[nwork], &i__1, info);
} else /* if(complicated condition) */ {
/* Computing MAX */
i__1 = *m, i__2 = (*m << 1) - 4, i__1 = max(i__1,i__2), i__1 = max(
i__1,*nrhs), i__2 = *n - *m * 3, i__1 = max(i__1,i__2);
if (*n >= mnthr && *lwork >= (*m << 2) + *m * *m + max(i__1,wlalsd)) {
/* Path 2a - underdetermined, with many more columns than rows */
/* and sufficient workspace for an efficient algorithm. */
ldwork = *m;
/* Computing MAX */
/* Computing MAX */
i__3 = *m, i__4 = (*m << 1) - 4, i__3 = max(i__3,i__4), i__3 =
max(i__3,*nrhs), i__4 = *n - *m * 3;
i__1 = (*m << 2) + *m * *lda + max(i__3,i__4), i__2 = *m * *lda +
*m + *m * *nrhs, i__1 = max(i__1,i__2), i__2 = (*m << 2)
+ *m * *lda + wlalsd;
if (*lwork >= max(i__1,i__2)) {
ldwork = *lda;
}
itau = 1;
nwork = *m + 1;
/* Compute A=L*Q. */
/* (Workspace: need 2*M, prefer M+M*NB) */
i__1 = *lwork - nwork + 1;
sgelqf_(m, n, &a[a_offset], lda, &work[itau], &work[nwork], &i__1,
info);
il = nwork;
/* Copy L to WORK(IL), zeroing out above its diagonal. */
slacpy_("L", m, m, &a[a_offset], lda, &work[il], &ldwork);
i__1 = *m - 1;
i__2 = *m - 1;
slaset_("U", &i__1, &i__2, &c_b81, &c_b81, &work[il + ldwork], &
ldwork);
ie = il + ldwork * *m;
itauq = ie + *m;
itaup = itauq + *m;
nwork = itaup + *m;
/* Bidiagonalize L in WORK(IL). */
/* (Workspace: need M*M+5*M, prefer M*M+4*M+2*M*NB) */
i__1 = *lwork - nwork + 1;
sgebrd_(m, m, &work[il], &ldwork, &s[1], &work[ie], &work[itauq],
&work[itaup], &work[nwork], &i__1, info);
/* Multiply B by transpose of left bidiagonalizing vectors of L. */
/* (Workspace: need M*M+4*M+NRHS, prefer M*M+4*M+NRHS*NB) */
i__1 = *lwork - nwork + 1;
sormbr_("Q", "L", "T", m, nrhs, m, &work[il], &ldwork, &work[
itauq], &b[b_offset], ldb, &work[nwork], &i__1, info);
/* Solve the bidiagonal least squares problem. */
slalsd_("U", &smlsiz, m, nrhs, &s[1], &work[ie], &b[b_offset],
ldb, rcond, rank, &work[nwork], &iwork[1], info);
if (*info != 0) {
goto L10;
}
/* Multiply B by right bidiagonalizing vectors of L. */
i__1 = *lwork - nwork + 1;
sormbr_("P", "L", "N", m, nrhs, m, &work[il], &ldwork, &work[
itaup], &b[b_offset], ldb, &work[nwork], &i__1, info);
/* Zero out below first M rows of B. */
i__1 = *n - *m;
slaset_("F", &i__1, nrhs, &c_b81, &c_b81, &b[*m + 1 + b_dim1],
ldb);
nwork = itau + *m;
/* Multiply transpose(Q) by B. */
/* (Workspace: need M+NRHS, prefer M+NRHS*NB) */
i__1 = *lwork - nwork + 1;
sormlq_("L", "T", n, nrhs, m, &a[a_offset], lda, &work[itau], &b[
b_offset], ldb, &work[nwork], &i__1, info);
} else {
/* Path 2 - remaining underdetermined cases. */
ie = 1;
itauq = ie + *m;
itaup = itauq + *m;
nwork = itaup + *m;
/* Bidiagonalize A. */
/* (Workspace: need 3*M+N, prefer 3*M+(M+N)*NB) */
i__1 = *lwork - nwork + 1;
sgebrd_(m, n, &a[a_offset], lda, &s[1], &work[ie], &work[itauq], &
work[itaup], &work[nwork], &i__1, info);
/* Multiply B by transpose of left bidiagonalizing vectors. */
/* (Workspace: need 3*M+NRHS, prefer 3*M+NRHS*NB) */
i__1 = *lwork - nwork + 1;
sormbr_("Q", "L", "T", m, nrhs, n, &a[a_offset], lda, &work[itauq]
, &b[b_offset], ldb, &work[nwork], &i__1, info);
/* Solve the bidiagonal least squares problem. */
slalsd_("L", &smlsiz, m, nrhs, &s[1], &work[ie], &b[b_offset],
ldb, rcond, rank, &work[nwork], &iwork[1], info);
if (*info != 0) {
goto L10;
}
/* Multiply B by right bidiagonalizing vectors of A. */
i__1 = *lwork - nwork + 1;
sormbr_("P", "L", "N", n, nrhs, m, &a[a_offset], lda, &work[itaup]
, &b[b_offset], ldb, &work[nwork], &i__1, info);
}
}
/* Undo scaling. */
if (iascl == 1) {
slascl_("G", &c__0, &c__0, &anrm, &smlnum, n, nrhs, &b[b_offset], ldb,
info);
slascl_("G", &c__0, &c__0, &smlnum, &anrm, &minmn, &c__1, &s[1], &
minmn, info);
} else if (iascl == 2) {
slascl_("G", &c__0, &c__0, &anrm, &bignum, n, nrhs, &b[b_offset], ldb,
info);
slascl_("G", &c__0, &c__0, &bignum, &anrm, &minmn, &c__1, &s[1], &
minmn, info);
}
if (ibscl == 1) {
slascl_("G", &c__0, &c__0, &smlnum, &bnrm, n, nrhs, &b[b_offset], ldb,
info);
} else if (ibscl == 2) {
slascl_("G", &c__0, &c__0, &bignum, &bnrm, n, nrhs, &b[b_offset], ldb,
info);
}
L10:
work[1] = (real) maxwrk;
iwork[1] = liwork;
return 0;
/* End of SGELSD */
} /* sgelsd_ */
