int ormqr(char side, char trans, int m, int n, int k, double *A, int lda, double *tau, double* C, int ldc) {
int info=0;
int lwork=-1;
double iwork;
dormqr_(&side, &trans, &m, &n, &k, A, &lda, tau, C, &ldc, &iwork, &lwork, &info);
lwork = (int)iwork;
double* work = new double[lwork];
dormqr_(&side, &trans, &m, &n, &k, A, &lda, tau, C, &ldc, work, &lwork, &info);
delete[] work;
return info;
}
void mtx_givens( Tmtx_ptr A, int k, double *rhs )
{
int i, pos, lwork, one=1, lda;
int info;
double a, b, c, s;
double *work, *tau;
char L = 'L', T = 'T';
lda = A->nrows;
lwork = 64*(k+1);
work = (double*)calloc( lwork, sizeof(double) );
tau = (double*)calloc( k+10, sizeof(double) );
//mtx_QR( A );
//dormqr_( &L, &T, &A->nrows, &one, &A->ncols, A->dat, &lda, A->tau, rhs, &lda, work, &lwork, &info );
//return;
/*
* eliminate upper block if it exists, use Householder reflections
*/
if( k )
{
int kplus1 = k+1;
pos = A->ncols-k;
// find QR decomp of block, storing in the block
dgeqrf_( &kplus1, &k, A->dat, &lda, tau, work, &lwork, &info );
// adjust the other columns in the first k+1 rows
dormqr_( &L, &T, &kplus1, &pos, &k, A->dat, &lda, tau, A->dat+lda*k, &lda, work, &lwork, &info );
// adjust the first k+1 elements of b
dormqr_( &L, &T, &kplus1, &one, &k, A->dat, &lda, tau, rhs, &lda, work, &lwork, &info );
}
/* eliminate sub-diagonal elements in remaining columns using Given's rotations */
for( i=k, pos=k*lda; i<A->ncols; i++, pos+=lda )
{
a = A->dat[pos+i];
b = A->dat[pos+i+1];
drotg( &a, &b, &c, &s );
drot( A->ncols-i, &A->dat[pos+i], lda, &A->dat[pos+i+1], lda, c, s );
drot( 1, rhs+i, lda, rhs+i+1, lda, c, s );
}
free( tau );
free( work );
}
void LapackQRDenseInternal::solve(double* x, int nrhs, bool transpose) {
// Properties of R
char uploR = 'U';
char diagR = 'N';
char sideR = 'L';
double alphaR = 1.;
char transR = transpose ? 'T' : 'N';
// Properties of Q
char transQ = transpose ? 'N' : 'T';
char sideQ = 'L';
int k = tau_.size(); // minimum of ncol_ and nrow_
int lwork = work_.size();
if (transpose) {
// Solve for transpose(R)
dtrsm_(&sideR, &uploR, &transR, &diagR, &ncol_, &nrhs, &alphaR,
getPtr(mat_), &ncol_, x, &ncol_);
// Multiply by Q
int info = 100;
dormqr_(&sideQ, &transQ, &ncol_, &nrhs, &k, getPtr(mat_), &ncol_, getPtr(tau_), x,
&ncol_, getPtr(work_), &lwork, &info);
if (info != 0) throw CasadiException("LapackQRDenseInternal::solve: dormqr_ failed "
"to solve the linear system");
} else {
// Multiply by transpose(Q)
int info = 100;
dormqr_(&sideQ, &transQ, &ncol_, &nrhs, &k, getPtr(mat_), &ncol_, getPtr(tau_), x,
&ncol_, getPtr(work_), &lwork, &info);
if (info != 0) throw CasadiException("LapackQRDenseInternal::solve: dormqr_ failed to "
"solve the linear system");
// Solve for R
dtrsm_(&sideR, &uploR, &transR, &diagR, &ncol_, &nrhs, &alphaR,
getPtr(mat_), &ncol_, x, &ncol_);
}
}
/* Multiply Q with a matrix using the output of geqrf */
void THLapack_(ormqr)(char side, char trans, int m, int n, int k, real *a, int lda, real *tau, real *c, int ldc, real *work, int lwork, int *info)
{
#ifdef USE_LAPACK
#if defined(TH_REAL_IS_DOUBLE)
dormqr_(&side, &trans, &m, &n, &k, a, &lda, tau, c, &ldc, work, &lwork, info);
#else
sormqr_(&side, &trans, &m, &n, &k, a, &lda, tau, c, &ldc, work, &lwork, info);
#endif
#else
THError("ormqr: Lapack library not found in compile time\n");
#endif
}
DLLEXPORT MKL_INT d_qr_solve_factored(MKL_INT m, MKL_INT n, MKL_INT bn, double r[], double b[], double tau[], double x[], double work[], MKL_INT len)
{
char side ='L';
char tran = 'T';
MKL_INT info = 0;
double* clone_b = new double[m*bn];
std::memcpy(clone_b, b, m*bn*sizeof(double));
dormqr_(&side, &tran, &m, &bn, &n, r, &m, tau, clone_b, &m, work, &len, &info);
cblas_dtrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit, n, bn, 1.0, r, m, clone_b, m);
for (MKL_INT i = 0; i < n; ++i)
{
for (MKL_INT j = 0; j < bn; ++j)
{
x[j * n + i] = clone_b[j * m + i];
}
}
delete[] clone_b;
return info;
}
DLLEXPORT int d_qr_solve(int m, int n, int bn, double r[], double b[], double x[], double work[], int len)
{
int info = 0;
double* clone_r = new double[m*n];
memcpy(clone_r, r, m*n*sizeof(double));
double* tau = new double[max(1, min(m,n))];
dgeqrf_(&m, &n, clone_r, &m, tau, work, &len, &info);
if (info != 0)
{
delete[] clone_r;
delete[] tau;
return info;
}
double* clone_b = new double[m*bn];
memcpy(clone_b, b, m*bn*sizeof(double));
char side ='L';
char tran = 'T';
dormqr_(&side, &tran, &m, &bn, &n, clone_r, &m, tau, clone_b, &m, work, &len, &info);
cblas_dtrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit, n, bn, 1.0, clone_r, m, clone_b, m);
for (int i = 0; i < n; ++i)
{
for (int j = 0; j < bn; ++j)
{
x[j * n + i] = clone_b[j * m + i];
}
}
delete[] clone_b;
delete[] tau;
delete[] clone_r;
return info;
}
/* Subroutine */ int dgelsd_(integer *m, integer *n, integer *nrhs,
doublereal *a, integer *lda, doublereal *b, integer *ldb, doublereal *
s, doublereal *rcond, integer *rank, doublereal *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 */
static doublereal anrm, bnrm;
static integer itau, nlvl, iascl, ibscl;
static doublereal sfmin;
static integer minmn, maxmn, itaup, itauq, mnthr, nwork;
extern /* Subroutine */ int dlabad_(doublereal *, doublereal *);
static integer ie, il;
extern /* Subroutine */ int dgebrd_(integer *, integer *, doublereal *,
integer *, doublereal *, doublereal *, doublereal *, doublereal *,
doublereal *, integer *, integer *);
extern doublereal dlamch_(char *);
static integer mm;
extern doublereal dlange_(char *, integer *, integer *, doublereal *,
integer *, doublereal *);
extern /* Subroutine */ int dgelqf_(integer *, integer *, doublereal *,
integer *, doublereal *, doublereal *, integer *, integer *),
dlalsd_(char *, integer *, integer *, integer *, doublereal *,
doublereal *, doublereal *, integer *, doublereal *, integer *,
doublereal *, integer *, integer *), dlascl_(char *,
integer *, integer *, doublereal *, doublereal *, integer *,
integer *, doublereal *, integer *, integer *), dgeqrf_(
integer *, integer *, doublereal *, integer *, doublereal *,
doublereal *, integer *, integer *), dlacpy_(char *, integer *,
integer *, doublereal *, integer *, doublereal *, integer *), dlaset_(char *, integer *, integer *, doublereal *,
doublereal *, doublereal *, integer *), xerbla_(char *,
integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *, ftnlen, ftnlen);
static doublereal bignum;
extern /* Subroutine */ int dormbr_(char *, char *, char *, integer *,
integer *, integer *, doublereal *, integer *, doublereal *,
doublereal *, integer *, doublereal *, integer *, integer *);
static integer wlalsd;
extern /* Subroutine */ int dormlq_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, doublereal *,
integer *, doublereal *, integer *, integer *);
static integer ldwork;
extern /* Subroutine */ int dormqr_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, doublereal *,
integer *, doublereal *, integer *, integer *);
static integer minwrk, maxwrk;
static doublereal smlnum;
static logical lquery;
static integer smlsiz;
static doublereal 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]
/* -- 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
=======
DGELSD 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) DOUBLE PRECISION 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) DOUBLE PRECISION 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) DOUBLE PRECISION 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) DOUBLE PRECISION
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) DOUBLE PRECISION array, dimension (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 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 (LIWORK)
LIWORK >= 3 * MINMN * NLVL + 11 * MINMN,
where MINMN = MIN( M,N ).
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
=====================================================================
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;
--iwork;
/* Function Body */
*info = 0;
minmn = min(*m,*n);
maxmn = max(*m,*n);
mnthr = ilaenv_(&c__6, "DGELSD", " ", 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;
}
smlsiz = ilaenv_(&c__9, "DGELSD", " ", &c__0, &c__0, &c__0, &c__0, (
ftnlen)6, (ftnlen)1);
/* 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;
minmn = max(1,minmn);
/* Computing MAX */
i__1 = (integer) (log((doublereal) minmn / (doublereal) (smlsiz + 1)) /
log(2.)) + 1;
nlvl = max(i__1,0);
if (*info == 0) {
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, "DGEQRF", " ", 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, "DORMQR", "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.
Computing MAX */
i__1 = maxwrk, i__2 = *n * 3 + (mm + *n) * ilaenv_(&c__1, "DGEBRD"
, " ", &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, "DORMBR",
"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, "DORMBR",
"PLN", n, nrhs, n, &c_n1, (ftnlen)6, (ftnlen)3);
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, "DGELQF", " ", 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, "DGEBRD", " ", 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, "DORMBR", "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, "DORMBR", "PLN", m, nrhs, m, &c_n1, (
ftnlen)6, (ftnlen)3);
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, "DORMLQ",
"LT", n, nrhs, m, &c_n1, (ftnlen)6, (ftnlen)2);
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, "DGEBRD", " ", m,
n, &c_n1, &c_n1, (ftnlen)6, (ftnlen)1);
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * 3 + *nrhs * ilaenv_(&c__1, "DORMBR"
, "QLT", m, nrhs, n, &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, "DORMBR",
"PLN", n, nrhs, m, &c_n1, (ftnlen)6, (ftnlen)3);
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] = (doublereal) maxwrk;
if (*lwork < minwrk && ! lquery) {
*info = -12;
}
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DGELSD", &i__1);
return 0;
} else if (lquery) {
goto L10;
}
/* Quick return if possible. */
if (*m == 0 || *n == 0) {
*rank = 0;
return 0;
}
/* Get machine parameters. */
eps = dlamch_("P");
sfmin = dlamch_("S");
smlnum = sfmin / eps;
bignum = 1. / smlnum;
dlabad_(&smlnum, &bignum);
/* Scale A if max entry outside range [SMLNUM,BIGNUM]. */
anrm = dlange_("M", m, n, &a[a_offset], lda, &work[1]);
iascl = 0;
if (anrm > 0. && anrm < smlnum) {
/* Scale matrix norm up to SMLNUM. */
dlascl_("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. */
dlascl_("G", &c__0, &c__0, &anrm, &bignum, m, n, &a[a_offset], lda,
info);
iascl = 2;
} else if (anrm == 0.) {
/* Matrix all zero. Return zero solution. */
i__1 = max(*m,*n);
dlaset_("F", &i__1, nrhs, &c_b82, &c_b82, &b[b_offset], ldb);
dlaset_("F", &minmn, &c__1, &c_b82, &c_b82, &s[1], &c__1);
*rank = 0;
goto L10;
}
/* Scale B if max entry outside range [SMLNUM,BIGNUM]. */
bnrm = dlange_("M", m, nrhs, &b[b_offset], ldb, &work[1]);
ibscl = 0;
if (bnrm > 0. && bnrm < smlnum) {
/* Scale matrix norm up to SMLNUM. */
dlascl_("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. */
dlascl_("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;
dlaset_("F", &i__1, nrhs, &c_b82, &c_b82, &b_ref(*m + 1, 1), 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;
dgeqrf_(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;
dormqr_("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;
dlaset_("L", &i__1, &i__2, &c_b82, &c_b82, &a_ref(2, 1), 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;
dgebrd_(&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;
dormbr_("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. */
dlalsd_("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;
dormbr_("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;
if (*n >= mnthr && *lwork >= (*m << 2) + *m * *m + max(i__1,i__2)) {
/* 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;
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;
dgelqf_(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. */
dlacpy_("L", m, m, &a[a_offset], lda, &work[il], &ldwork);
i__1 = *m - 1;
i__2 = *m - 1;
dlaset_("U", &i__1, &i__2, &c_b82, &c_b82, &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;
dgebrd_(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;
dormbr_("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. */
dlalsd_("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;
dormbr_("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;
dlaset_("F", &i__1, nrhs, &c_b82, &c_b82, &b_ref(*m + 1, 1), ldb);
nwork = itau + *m;
/* Multiply transpose(Q) by B.
(Workspace: need M+NRHS, prefer M+NRHS*NB) */
i__1 = *lwork - nwork + 1;
dormlq_("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;
dgebrd_(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;
dormbr_("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. */
dlalsd_("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;
dormbr_("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) {
dlascl_("G", &c__0, &c__0, &anrm, &smlnum, n, nrhs, &b[b_offset], ldb,
info);
dlascl_("G", &c__0, &c__0, &smlnum, &anrm, &minmn, &c__1, &s[1], &
minmn, info);
} else if (iascl == 2) {
dlascl_("G", &c__0, &c__0, &anrm, &bignum, n, nrhs, &b[b_offset], ldb,
info);
dlascl_("G", &c__0, &c__0, &bignum, &anrm, &minmn, &c__1, &s[1], &
minmn, info);
}
if (ibscl == 1) {
dlascl_("G", &c__0, &c__0, &smlnum, &bnrm, n, nrhs, &b[b_offset], ldb,
info);
} else if (ibscl == 2) {
dlascl_("G", &c__0, &c__0, &bignum, &bnrm, n, nrhs, &b[b_offset], ldb,
info);
}
L10:
work[1] = (doublereal) maxwrk;
return 0;
/* End of DGELSD */
} /* dgelsd_ */
/* Subroutine */ int dggevx_(char *balanc, char *jobvl, char *jobvr, char *
sense, integer *n, doublereal *a, integer *lda, doublereal *b,
integer *ldb, doublereal *alphar, doublereal *alphai, doublereal *
beta, doublereal *vl, integer *ldvl, doublereal *vr, integer *ldvr,
integer *ilo, integer *ihi, doublereal *lscale, doublereal *rscale,
doublereal *abnrm, doublereal *bbnrm, doublereal *rconde, doublereal *
rcondv, doublereal *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;
doublereal d__1, d__2, d__3, d__4;
/* Local variables */
integer i__, j, m, jc, in, mm, jr;
doublereal eps;
logical ilv, pair;
doublereal anrm, bnrm;
integer ierr, itau;
doublereal temp;
logical ilvl, ilvr;
integer iwrk, iwrk1;
integer icols;
logical noscl;
integer irows;
logical ilascl, ilbscl;
logical ldumma[1];
char chtemp[1];
doublereal bignum;
integer ijobvl;
integer ijobvr;
logical wantsb;
doublereal anrmto;
logical wantse;
doublereal bnrmto;
integer minwrk, maxwrk;
logical wantsn;
doublereal smlnum;
logical lquery, wantsv;
/* -- LAPACK driver routine (version 3.2) -- */
/* November 2006 */
/* Purpose */
/* ======= */
/* DGGEVX computes for a pair of N-by-N real nonsymmetric matrices (A,B) */
/* the generalized eigenvalues, and optionally, the left and/or right */
/* generalized eigenvectors. */
/* Optionally also, it computes a balancing transformation to improve */
/* the conditioning of the eigenvalues and eigenvectors (ILO, IHI, */
/* LSCALE, RSCALE, ABNRM, and BBNRM), reciprocal condition numbers for */
/* the eigenvalues (RCONDE), and reciprocal condition numbers for the */
/* right eigenvectors (RCONDV). */
/* A generalized eigenvalue for a pair of matrices (A,B) is a scalar */
/* lambda or a ratio alpha/beta = lambda, such that A - lambda*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. */
/* The right eigenvector v(j) corresponding to the eigenvalue lambda(j) */
/* of (A,B) satisfies */
/* A * v(j) = lambda(j) * B * v(j) . */
/* The left eigenvector u(j) corresponding to the eigenvalue lambda(j) */
/* of (A,B) satisfies */
/* u(j)**H * A = lambda(j) * u(j)**H * B. */
/* where u(j)**H is the conjugate-transpose of u(j). */
/* Arguments */
/* ========= */
/* BALANC (input) CHARACTER*1 */
/* Specifies the balance option to be performed. */
/* = 'N': do not diagonally scale or permute; */
/* = 'P': permute only; */
/* = 'S': scale only; */
/* = 'B': both permute and scale. */
/* Computed reciprocal condition numbers will be for the */
/* matrices after permuting and/or balancing. Permuting does */
/* not change condition numbers (in exact arithmetic), but */
/* balancing does. */
/* 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. */
/* SENSE (input) CHARACTER*1 */
/* Determines which reciprocal condition numbers are computed. */
/* = 'N': none are computed; */
/* = 'E': computed for eigenvalues only; */
/* = 'V': computed for eigenvectors only; */
/* = 'B': computed for eigenvalues and eigenvectors. */
/* N (input) INTEGER */
/* The order of the matrices A, B, VL, and VR. N >= 0. */
/* A (input/output) DOUBLE PRECISION array, dimension (LDA, N) */
/* On entry, the matrix A in the pair (A,B). */
/* On exit, A has been overwritten. If JOBVL='V' or JOBVR='V' */
/* or both, then A contains the first part of the real Schur */
/* form of the "balanced" versions of the input A and B. */
/* LDA (input) INTEGER */
/* The leading dimension of A. LDA >= max(1,N). */
/* B (input/output) DOUBLE PRECISION array, dimension (LDB, N) */
/* On entry, the matrix B in the pair (A,B). */
/* On exit, B has been overwritten. If JOBVL='V' or JOBVR='V' */
/* or both, then B contains the second part of the real Schur */
/* form of the "balanced" versions of the input A and B. */
/* LDB (input) INTEGER */
/* The leading dimension of B. LDB >= max(1,N). */
/* ALPHAR (output) DOUBLE PRECISION array, dimension (N) */
/* ALPHAI (output) DOUBLE PRECISION array, dimension (N) */
/* BETA (output) DOUBLE PRECISION array, dimension (N) */
/* 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) DOUBLE PRECISION 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 the j-th eigenvalue is real, then */
/* u(j) = VL(:,j), the j-th column of VL. If the j-th and */
/* (j+1)-th 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). */
/* Each eigenvector will be scaled so the largest component have */
/* abs(real part) + abs(imag. part) = 1. */
/* 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) DOUBLE PRECISION 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 the j-th eigenvalue is real, then */
/* v(j) = VR(:,j), the j-th column of VR. If the j-th and */
/* (j+1)-th 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). */
/* Each eigenvector will be scaled so the largest component have */
/* abs(real part) + abs(imag. part) = 1. */
/* Not referenced if JOBVR = 'N'. */
/* LDVR (input) INTEGER */
/* The leading dimension of the matrix VR. LDVR >= 1, and */
/* if JOBVR = 'V', LDVR >= N. */
/* ILO (output) INTEGER */
/* IHI (output) INTEGER */
/* ILO and IHI are integer values such that on exit */
/* A(i,j) = 0 and B(i,j) = 0 if i > j and */
/* If BALANC = 'N' or 'S', ILO = 1 and IHI = N. */
/* LSCALE (output) DOUBLE PRECISION array, dimension (N) */
/* Details of the permutations and scaling factors applied */
/* to the left side of A and B. If PL(j) is the index of the */
/* row interchanged with row j, and DL(j) is the scaling */
/* factor applied to row j, then */
/* The order in which the interchanges are made is N to IHI+1, */
/* then 1 to ILO-1. */
/* RSCALE (output) DOUBLE PRECISION array, dimension (N) */
/* Details of the permutations and scaling factors applied */
/* to the right side of A and B. If PR(j) is the index of the */
/* column interchanged with column j, and DR(j) is the scaling */
/* factor applied to column j, then */
/* The order in which the interchanges are made is N to IHI+1, */
/* then 1 to ILO-1. */
/* ABNRM (output) DOUBLE PRECISION */
/* The one-norm of the balanced matrix A. */
/* BBNRM (output) DOUBLE PRECISION */
/* The one-norm of the balanced matrix B. */
/* RCONDE (output) DOUBLE PRECISION array, dimension (N) */
/* If SENSE = 'E' or 'B', the reciprocal condition numbers of */
/* the eigenvalues, stored in consecutive elements of the array. */
/* For a complex conjugate pair of eigenvalues two consecutive */
/* elements of RCONDE are set to the same value. Thus RCONDE(j), */
/* RCONDV(j), and the j-th columns of VL and VR all correspond */
/* to the j-th eigenpair. */
/* If SENSE = 'N or 'V', RCONDE is not referenced. */
/* RCONDV (output) DOUBLE PRECISION array, dimension (N) */
/* If SENSE = 'V' or 'B', the estimated reciprocal condition */
/* numbers of the eigenvectors, stored in consecutive elements */
/* of the array. For a complex eigenvector two consecutive */
/* elements of RCONDV are set to the same value. If the */
/* eigenvalues cannot be reordered to compute RCONDV(j), */
/* RCONDV(j) is set to 0; this can only occur when the true */
/* value would be very small anyway. */
/* If SENSE = 'N' or 'E', RCONDV is not referenced. */
/* WORK (workspace/output) DOUBLE PRECISION 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(1,2*N). */
/* If BALANC = 'S' or 'B', or JOBVL = 'V', or JOBVR = 'V', */
/* LWORK >= max(1,6*N). */
/* If SENSE = 'E' or 'B', LWORK >= max(1,10*N). */
/* If SENSE = 'V' or 'B', LWORK >= 2*N*N+8*N+16. */
/* 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 (N+6) */
/* If SENSE = 'E', IWORK is not referenced. */
/* BWORK (workspace) LOGICAL array, dimension (N) */
/* If SENSE = 'N', BWORK is not referenced. */
/* INFO (output) INTEGER */
/* = 0: successful exit */
/* < 0: if INFO = -i, the i-th argument had an illegal value. */
/* The QZ iteration failed. No eigenvectors have been */
/* calculated, but ALPHAR(j), ALPHAI(j), and BETA(j) */
/* > N: =N+1: other than QZ iteration failed in DHGEQZ. */
/* =N+2: error return from DTGEVC. */
/* Further Details */
/* =============== */
/* Balancing a matrix pair (A,B) includes, first, permuting rows and */
/* columns to isolate eigenvalues, second, applying diagonal similarity */
/* transformation to the rows and columns to make the rows and columns */
/* as close in norm as possible. 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 4.11.1.2 of LAPACK Users' Guide. */
/* An approximate error bound on the chordal distance between the i-th */
/* computed generalized eigenvalue w and the corresponding exact */
/* eigenvalue lambda is */
/* chord(w, lambda) <= EPS * norm(ABNRM, BBNRM) / RCONDE(I) */
/* An approximate error bound for the angle between the i-th computed */
/* eigenvector VL(i) or VR(i) is given by */
/* EPS * norm(ABNRM, BBNRM) / DIF(i). */
/* For further explanation of the reciprocal condition numbers RCONDE */
/* and RCONDV, see section 4.11 of LAPACK User's Guide. */
/* ===================================================================== */
/* 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 (! (lsame_(balanc, "N") || lsame_(balanc, "S") || lsame_(balanc, "P")
|| 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 || wantsb) {
minwrk = *n * 10;
}
if (wantsv || wantsb) {
/* Computing MAX */
i__1 = minwrk, i__2 = (*n << 1) * (*n + 4) + 16;
minwrk = max(i__1,i__2);
}
maxwrk = minwrk;
/* Computing MAX */
i__1 = maxwrk, i__2 = *n + *n * ilaenv_(&c__1, "DGEQRF", " ", n, &
c__1, n, &c__0);
maxwrk = max(i__1,i__2);
/* Computing MAX */
i__1 = maxwrk, i__2 = *n + *n * ilaenv_(&c__1, "DORMQR", " ", n, &
c__1, n, &c__0);
maxwrk = max(i__1,i__2);
if (ilvl) {
/* Computing MAX */
i__1 = maxwrk, i__2 = *n + *n * ilaenv_(&c__1, "DORGQR",
" ", n, &c__1, n, &c__0);
maxwrk = max(i__1,i__2);
}
}
work[1] = (doublereal) maxwrk;
if (*lwork < minwrk && ! lquery) {
*info = -26;
}
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DGGEVX", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
if (*n == 0) {
return 0;
}
/* Get machine constants */
eps = dlamch_("P");
smlnum = dlamch_("S");
bignum = 1. / smlnum;
dlabad_(&smlnum, &bignum);
smlnum = sqrt(smlnum) / eps;
bignum = 1. / smlnum;
/* Scale A if max element outside range [SMLNUM,BIGNUM] */
anrm = dlange_("M", n, n, &a[a_offset], lda, &work[1]);
ilascl = FALSE_;
if (anrm > 0. && anrm < smlnum) {
anrmto = smlnum;
ilascl = TRUE_;
} else if (anrm > bignum) {
anrmto = bignum;
ilascl = TRUE_;
}
if (ilascl) {
dlascl_("G", &c__0, &c__0, &anrm, &anrmto, n, n, &a[a_offset], lda, &
ierr);
}
/* Scale B if max element outside range [SMLNUM,BIGNUM] */
bnrm = dlange_("M", n, n, &b[b_offset], ldb, &work[1]);
ilbscl = FALSE_;
if (bnrm > 0. && bnrm < smlnum) {
bnrmto = smlnum;
ilbscl = TRUE_;
} else if (bnrm > bignum) {
bnrmto = bignum;
ilbscl = TRUE_;
}
if (ilbscl) {
dlascl_("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) */
dggbal_(balanc, n, &a[a_offset], lda, &b[b_offset], ldb, ilo, ihi, &
lscale[1], &rscale[1], &work[1], &ierr);
/* Compute ABNRM and BBNRM */
*abnrm = dlange_("1", n, n, &a[a_offset], lda, &work[1]);
if (ilascl) {
work[1] = *abnrm;
dlascl_("G", &c__0, &c__0, &anrmto, &anrm, &c__1, &c__1, &work[1], &
c__1, &ierr);
*abnrm = work[1];
}
*bbnrm = dlange_("1", n, n, &b[b_offset], ldb, &work[1]);
if (ilbscl) {
work[1] = *bbnrm;
dlascl_("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;
dgeqrf_(&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;
dormqr_("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) {
dlaset_("Full", n, n, &c_b59, &c_b60, &vl[vl_offset], ldvl)
;
if (irows > 1) {
i__1 = irows - 1;
i__2 = irows - 1;
dlacpy_("L", &i__1, &i__2, &b[*ilo + 1 + *ilo * b_dim1], ldb, &vl[
*ilo + 1 + *ilo * vl_dim1], ldvl);
}
i__1 = *lwork + 1 - iwrk;
dorgqr_(&irows, &irows, &irows, &vl[*ilo + *ilo * vl_dim1], ldvl, &
work[itau], &work[iwrk], &i__1, &ierr);
}
if (ilvr) {
dlaset_("Full", n, n, &c_b59, &c_b60, &vr[vr_offset], ldvr)
;
}
/* Reduce to generalized Hessenberg form */
/* (Workspace: none needed) */
if (ilv || ! wantsn) {
/* Eigenvectors requested -- work on whole matrix. */
dgghrd_(jobvl, jobvr, n, ilo, ihi, &a[a_offset], lda, &b[b_offset],
ldb, &vl[vl_offset], ldvl, &vr[vr_offset], ldvr, &ierr);
} else {
dgghrd_("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';
}
dhgeqz_(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: DTGEVC: need 6*N */
/* DTGSNA: 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';
}
dtgevc_(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 (DTGEVC) and estimate condition */
/* numbers (DTGSNA). 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.) {
pair = TRUE_;
mm = 2;
}
}
i__2 = *n;
for (j = 1; j <= i__2; ++j) {
bwork[j] = FALSE_;
}
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) {
dtgevc_("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;
dtgsna_(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) {
dggbak_(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.) {
goto L70;
}
temp = 0.;
if (alphai[jc] == 0.) {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
/* Computing MAX */
d__2 = temp, d__3 = (d__1 = vl[jr + jc * vl_dim1], abs(
d__1));
temp = max(d__2,d__3);
}
} else {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
/* Computing MAX */
d__3 = temp, d__4 = (d__1 = vl[jr + jc * vl_dim1], abs(
d__1)) + (d__2 = vl[jr + (jc + 1) * vl_dim1], abs(
d__2));
temp = max(d__3,d__4);
}
}
if (temp < smlnum) {
goto L70;
}
temp = 1. / temp;
if (alphai[jc] == 0.) {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
vl[jr + jc * vl_dim1] *= temp;
}
} else {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
vl[jr + jc * vl_dim1] *= temp;
vl[jr + (jc + 1) * vl_dim1] *= temp;
}
}
L70:
;
}
}
if (ilvr) {
dggbak_(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.) {
goto L120;
}
temp = 0.;
if (alphai[jc] == 0.) {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
/* Computing MAX */
d__2 = temp, d__3 = (d__1 = vr[jr + jc * vr_dim1], abs(
d__1));
temp = max(d__2,d__3);
}
} else {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
/* Computing MAX */
d__3 = temp, d__4 = (d__1 = vr[jr + jc * vr_dim1], abs(
d__1)) + (d__2 = vr[jr + (jc + 1) * vr_dim1], abs(
d__2));
temp = max(d__3,d__4);
}
}
if (temp < smlnum) {
goto L120;
}
temp = 1. / temp;
if (alphai[jc] == 0.) {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
vr[jr + jc * vr_dim1] *= temp;
}
} else {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
vr[jr + jc * vr_dim1] *= temp;
vr[jr + (jc + 1) * vr_dim1] *= temp;
}
}
L120:
;
}
}
/* Undo scaling if necessary */
if (ilascl) {
dlascl_("G", &c__0, &c__0, &anrmto, &anrm, n, &c__1, &alphar[1], n, &
ierr);
dlascl_("G", &c__0, &c__0, &anrmto, &anrm, n, &c__1, &alphai[1], n, &
ierr);
}
if (ilbscl) {
dlascl_("G", &c__0, &c__0, &bnrmto, &bnrm, n, &c__1, &beta[1], n, &
ierr);
}
L130:
work[1] = (doublereal) maxwrk;
return 0;
/* End of DGGEVX */
} /* dggevx_ */
/* Subroutine */ int dgegs_(char *jobvsl, char *jobvsr, integer *n,
doublereal *a, integer *lda, doublereal *b, integer *ldb, doublereal *
alphar, doublereal *alphai, doublereal *beta, doublereal *vsl,
integer *ldvsl, doublereal *vsr, integer *ldvsr, doublereal *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;
doublereal 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 dggbak_(char *, char *, integer *, integer *,
integer *, doublereal *, doublereal *, integer *, doublereal *,
integer *, integer *), dggbal_(char *, integer *,
doublereal *, integer *, doublereal *, integer *, integer *,
integer *, doublereal *, doublereal *, doublereal *, integer *);
extern doublereal dlamch_(char *), dlange_(char *, integer *,
integer *, doublereal *, integer *, doublereal *);
extern /* Subroutine */ int dgghrd_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, integer *,
doublereal *, integer *, doublereal *, integer *, integer *), dlascl_(char *, integer *, integer *, doublereal
*, doublereal *, integer *, integer *, doublereal *, integer *,
integer *);
logical ilascl, ilbscl;
extern /* Subroutine */ int dgeqrf_(integer *, integer *, doublereal *,
integer *, doublereal *, doublereal *, integer *, integer *),
dlacpy_(char *, integer *, integer *, doublereal *, integer *,
doublereal *, integer *);
doublereal safmin;
extern /* Subroutine */ int dlaset_(char *, integer *, integer *,
doublereal *, doublereal *, doublereal *, integer *),
xerbla_(char *, integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *);
doublereal bignum;
extern /* Subroutine */ int dhgeqz_(char *, char *, char *, integer *,
integer *, integer *, doublereal *, integer *, doublereal *,
integer *, doublereal *, doublereal *, doublereal *, doublereal *,
integer *, doublereal *, integer *, doublereal *, integer *,
integer *);
integer ijobvl, iright, ijobvr;
extern /* Subroutine */ int dorgqr_(integer *, integer *, integer *,
doublereal *, integer *, doublereal *, doublereal *, integer *,
integer *);
doublereal anrmto;
integer lwkmin;
doublereal bnrmto;
extern /* Subroutine */ int dormqr_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, doublereal *,
integer *, doublereal *, integer *, integer *);
doublereal smlnum;
integer lwkopt;
logical lquery;
/* -- 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 DGGES. */
/* DGEGS 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 */
/* DGEGV should be used instead. See DGEGV 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) DOUBLE PRECISION 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) DOUBLE PRECISION 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) DOUBLE PRECISION array, dimension (N) */
/* The real parts of each scalar alpha defining an eigenvalue */
/* of GNEP. */
/* ALPHAI (output) DOUBLE PRECISION 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) DOUBLE PRECISION 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) DOUBLE PRECISION 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) DOUBLE PRECISION 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) DOUBLE PRECISION 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 DGEQRF, DORMQR, and DORGQR.) Then compute: */
/* NB -- MAX of the blocksizes for DGEQRF, DORMQR, and DORGQR */
/* 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 DGGBAL */
/* =N+2: error return from DGEQRF */
/* =N+3: error return from DORMQR */
/* =N+4: error return from DORGQR */
/* =N+5: error return from DGGHRD */
/* =N+6: error return from DHGEQZ (other than failed */
/* iteration) */
/* =N+7: error return from DGGBAK (computing VSL) */
/* =N+8: error return from DGGBAK (computing VSR) */
/* =N+9: error return from DLASCL (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] = (doublereal) 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, "DGEQRF", " ", n, n, &c_n1, &c_n1);
nb2 = ilaenv_(&c__1, "DORMQR", " ", n, n, n, &c_n1);
nb3 = ilaenv_(&c__1, "DORGQR", " ", 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] = (doublereal) lopt;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DGEGS ", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
if (*n == 0) {
return 0;
}
/* Get machine constants */
eps = dlamch_("E") * dlamch_("B");
safmin = dlamch_("S");
smlnum = *n * safmin / eps;
bignum = 1. / smlnum;
/* Scale A if max element outside range [SMLNUM,BIGNUM] */
anrm = dlange_("M", n, n, &a[a_offset], lda, &work[1]);
ilascl = FALSE_;
if (anrm > 0. && anrm < smlnum) {
anrmto = smlnum;
ilascl = TRUE_;
} else if (anrm > bignum) {
anrmto = bignum;
ilascl = TRUE_;
}
if (ilascl) {
dlascl_("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 = dlange_("M", n, n, &b[b_offset], ldb, &work[1]);
ilbscl = FALSE_;
if (bnrm > 0. && bnrm < smlnum) {
bnrmto = smlnum;
ilbscl = TRUE_;
} else if (bnrm > bignum) {
bnrmto = bignum;
ilbscl = TRUE_;
}
if (ilbscl) {
dlascl_("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;
dggbal_("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;
dgeqrf_(&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;
dormqr_("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) {
dlaset_("Full", n, n, &c_b36, &c_b37, &vsl[vsl_offset], ldvsl);
i__1 = irows - 1;
i__2 = irows - 1;
dlacpy_("L", &i__1, &i__2, &b[ilo + 1 + ilo * b_dim1], ldb, &vsl[ilo
+ 1 + ilo * vsl_dim1], ldvsl);
i__1 = *lwork + 1 - iwork;
dorgqr_(&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) {
dlaset_("Full", n, n, &c_b36, &c_b37, &vsr[vsr_offset], ldvsr);
}
/* Reduce to generalized Hessenberg form */
dgghrd_(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;
dhgeqz_("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) {
dggbak_("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) {
dggbak_("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) {
dlascl_("H", &c_n1, &c_n1, &anrmto, &anrm, n, n, &a[a_offset], lda, &
iinfo);
if (iinfo != 0) {
*info = *n + 9;
return 0;
}
dlascl_("G", &c_n1, &c_n1, &anrmto, &anrm, n, &c__1, &alphar[1], n, &
iinfo);
if (iinfo != 0) {
*info = *n + 9;
return 0;
}
dlascl_("G", &c_n1, &c_n1, &anrmto, &anrm, n, &c__1, &alphai[1], n, &
iinfo);
if (iinfo != 0) {
*info = *n + 9;
return 0;
}
}
if (ilbscl) {
dlascl_("U", &c_n1, &c_n1, &bnrmto, &bnrm, n, n, &b[b_offset], ldb, &
iinfo);
if (iinfo != 0) {
*info = *n + 9;
return 0;
}
dlascl_("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] = (doublereal) lwkopt;
return 0;
/* End of DGEGS */
} /* dgegs_ */
void Stiefel::ObtainExtrHHR(Variable *x, Vector *intretax, Vector *result) const
{
if (!x->TempDataExist("HHR"))
{
const double *xM = x->ObtainReadData();
SharedSpace *HouseHolderResult = new SharedSpace(2, x->Getsize()[0], x->Getsize()[1]);
double *ptrHHR = HouseHolderResult->ObtainWriteEntireData();
SharedSpace *HHRTau = new SharedSpace(1, x->Getsize()[1]);
double *tau = HHRTau->ObtainWriteEntireData();
integer N = x->Getsize()[0], P = x->Getsize()[1], Length = N * P, inc = 1;
double one = 1, zero = 0;
dcopy_(&Length, const_cast<double *> (xM), &inc, ptrHHR, &inc);
integer *jpvt = new integer[P];
integer info;
integer lwork = -1;
double lworkopt;
for (integer i = 0; i < P; i++)
jpvt[i] = i + 1;
dgeqp3_(&N, &P, ptrHHR, &N, jpvt, tau, &lworkopt, &lwork, &info);
lwork = static_cast<integer> (lworkopt);
double *work = new double[lwork];
dgeqp3_(&N, &P, ptrHHR, &N, jpvt, tau, work, &lwork, &info);
x->AddToTempData("HHR", HouseHolderResult);
x->AddToTempData("HHRTau", HHRTau);
if (info < 0)
Rcpp::Rcout << "Error in qr decomposition!" << std::endl;
for (integer i = 0; i < P; i++)
{
if (jpvt[i] != (i + 1))
Rcpp::Rcout << "Error in qf retraction!" << std::endl;
}
delete[] jpvt;
delete[] work;
}
const double *xM = x->ObtainReadData();
const SharedSpace *HHR = x->ObtainReadTempData("HHR");
const SharedSpace *HHRTau = x->ObtainReadTempData("HHRTau");
const double *ptrHHR = HHR->ObtainReadData();
const double *ptrHHRTau = HHRTau->ObtainReadData();
const double *intretaxTV = intretax->ObtainReadData();
double *resultTV = result->ObtainWriteEntireData();
char *transn = const_cast<char *> ("n"), *sidel = const_cast<char *> ("l");
integer N = x->Getsize()[0], P = x->Getsize()[1], inc = 1, Length = N * P;
integer info;
double r2 = sqrt(2.0);
integer idx = 0;
for (integer i = 0; i < p; i++)
{
resultTV[i + i * n] = 0;
for (integer j = i + 1; j < p; j++)
{
resultTV[j + i * n] = intretaxTV[idx] / r2;
resultTV[i + j * n] = -resultTV[j + i * n];
idx++;
}
}
for (integer i = 0; i < p; i++)
{
for (integer j = p; j < n; j++)
{
resultTV[j + i * n] = intretaxTV[idx];
idx++;
}
}
double sign;
for (integer i = 0; i < p; i++)
{
sign = (ptrHHR[i + n * i] >= 0) ? 1 : -1;
dscal_(&P, &sign, resultTV + i, &N);
}
integer lwork = -1;
double lworkopt;
dormqr_(sidel, transn, &N, &P, &P, const_cast<double *> (ptrHHR), &N, const_cast<double *> (ptrHHRTau), resultTV, &N, &lworkopt, &lwork, &info);
lwork = static_cast<integer> (lworkopt);
double *work = new double[lwork];
dormqr_(sidel, transn, &N, &P, &P, const_cast<double *> (ptrHHR), &N, const_cast<double *> (ptrHHRTau), resultTV, &N, work, &lwork, &info);
delete[] work;
};
/* Subroutine */ int dormbr_(char *vect, char *side, char *trans, integer *m,
integer *n, integer *k, doublereal *a, integer *lda, doublereal *tau,
doublereal *c__, integer *ldc, doublereal *work, integer *lwork,
integer *info)
{
/* -- LAPACK routine (version 3.1) --
Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd..
November 2006
Purpose
=======
If VECT = 'Q', DORMBR overwrites the general real M-by-N matrix C
with
SIDE = 'L' SIDE = 'R'
TRANS = 'N': Q * C C * Q
TRANS = 'T': Q**T * C C * Q**T
If VECT = 'P', DORMBR overwrites the general real M-by-N matrix C
with
SIDE = 'L' SIDE = 'R'
TRANS = 'N': P * C C * P
TRANS = 'T': P**T * C C * P**T
Here Q and P**T are the orthogonal matrices determined by DGEBRD when
reducing a real matrix A to bidiagonal form: A = Q * B * P**T. Q and
P**T are defined as products of elementary reflectors H(i) and G(i)
respectively.
Let nq = m if SIDE = 'L' and nq = n if SIDE = 'R'. Thus nq is the
order of the orthogonal matrix Q or P**T that is applied.
If VECT = 'Q', A is assumed to have been an NQ-by-K matrix:
if nq >= k, Q = H(1) H(2) . . . H(k);
if nq < k, Q = H(1) H(2) . . . H(nq-1).
If VECT = 'P', A is assumed to have been a K-by-NQ matrix:
if k < nq, P = G(1) G(2) . . . G(k);
if k >= nq, P = G(1) G(2) . . . G(nq-1).
Arguments
=========
VECT (input) CHARACTER*1
= 'Q': apply Q or Q**T;
= 'P': apply P or P**T.
SIDE (input) CHARACTER*1
= 'L': apply Q, Q**T, P or P**T from the Left;
= 'R': apply Q, Q**T, P or P**T from the Right.
TRANS (input) CHARACTER*1
= 'N': No transpose, apply Q or P;
= 'T': Transpose, apply Q**T or P**T.
M (input) INTEGER
The number of rows of the matrix C. M >= 0.
N (input) INTEGER
The number of columns of the matrix C. N >= 0.
K (input) INTEGER
If VECT = 'Q', the number of columns in the original
matrix reduced by DGEBRD.
If VECT = 'P', the number of rows in the original
matrix reduced by DGEBRD.
K >= 0.
A (input) DOUBLE PRECISION array, dimension
(LDA,min(nq,K)) if VECT = 'Q'
(LDA,nq) if VECT = 'P'
The vectors which define the elementary reflectors H(i) and
G(i), whose products determine the matrices Q and P, as
returned by DGEBRD.
LDA (input) INTEGER
The leading dimension of the array A.
If VECT = 'Q', LDA >= max(1,nq);
if VECT = 'P', LDA >= max(1,min(nq,K)).
TAU (input) DOUBLE PRECISION array, dimension (min(nq,K))
TAU(i) must contain the scalar factor of the elementary
reflector H(i) or G(i) which determines Q or P, as returned
by DGEBRD in the array argument TAUQ or TAUP.
C (input/output) DOUBLE PRECISION array, dimension (LDC,N)
On entry, the M-by-N matrix C.
On exit, C is overwritten by Q*C or Q**T*C or C*Q**T or C*Q
or P*C or P**T*C or C*P or C*P**T.
LDC (input) INTEGER
The leading dimension of the array C. LDC >= max(1,M).
WORK (workspace/output) DOUBLE PRECISION 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 SIDE = 'L', LWORK >= max(1,N);
if SIDE = 'R', LWORK >= max(1,M).
For optimum performance LWORK >= N*NB if SIDE = 'L', and
LWORK >= M*NB if SIDE = 'R', where NB is the optimal
blocksize.
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
=====================================================================
Test the input arguments
Parameter adjustments */
/* Table of constant values */
static const integer c__1 = 1;
static const integer c_n1 = -1;
static const integer c__2 = 2;
/* System generated locals */
address a__1[2];
integer a_dim1, a_offset, c_dim1, c_offset, i__1, i__2, i__3[2];
char ch__1[2];
/* Builtin functions
Subroutine */ int s_cat(const char *, char **, const integer *, const integer *, const ftnlen);
/* Local variables */
_THREAD_STATIC_ integer i1, i2, nb, mi, ni, nq, nw;
_THREAD_STATIC_ logical left;
extern logical lsame_(const char *, const char *);
_THREAD_STATIC_ integer iinfo;
extern /* Subroutine */ int xerbla_(const char *, const integer *);
extern integer ilaenv_(const integer *, const char *, const char *, const integer *, const integer *,
const integer *, const integer *, const ftnlen, const ftnlen);
extern /* Subroutine */ int dormlq_(const char *, const char *, const integer *, const integer *,
const integer *, const doublereal *, const integer *, const doublereal *, const doublereal *,
const integer *, const doublereal *, const integer *, const integer *);
_THREAD_STATIC_ logical notran;
extern /* Subroutine */ int dormqr_(const char *, const char *, const integer *, const integer *,
const integer *, const doublereal *, const integer *, const doublereal *, const doublereal *,
const integer *, const doublereal *, const integer *, const integer *);
_THREAD_STATIC_ logical applyq;
_THREAD_STATIC_ char transt[1];
_THREAD_STATIC_ integer lwkopt;
_THREAD_STATIC_ logical lquery;
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
--tau;
c_dim1 = *ldc;
c_offset = 1 + c_dim1;
c__ -= c_offset;
--work;
/* Function Body */
*info = 0;
applyq = lsame_(vect, "Q");
left = lsame_(side, "L");
notran = lsame_(trans, "N");
lquery = *lwork == -1;
/* NQ is the order of Q or P and NW is the minimum dimension of WORK */
if (left) {
nq = *m;
nw = *n;
} else {
nq = *n;
nw = *m;
}
if (! applyq && ! lsame_(vect, "P")) {
*info = -1;
} else if (! left && ! lsame_(side, "R")) {
*info = -2;
} else if (! notran && ! lsame_(trans, "T")) {
*info = -3;
} else if (*m < 0) {
*info = -4;
} else if (*n < 0) {
*info = -5;
} else if (*k < 0) {
*info = -6;
} else /* if(complicated condition) */ {
/* Computing MAX */
i__1 = 1, i__2 = min(nq,*k);
if (applyq && *lda < max(1,nq) || ! applyq && *lda < max(i__1,i__2)) {
*info = -8;
} else if (*ldc < max(1,*m)) {
*info = -11;
} else if (*lwork < max(1,nw) && ! lquery) {
*info = -13;
}
}
if (*info == 0) {
if (applyq) {
if (left) {
/* Writing concatenation */
i__3[0] = 1, a__1[0] = side;
i__3[1] = 1, a__1[1] = trans;
s_cat(ch__1, a__1, i__3, &c__2, (ftnlen)2);
i__1 = *m - 1;
i__2 = *m - 1;
nb = ilaenv_(&c__1, "DORMQR", ch__1, &i__1, n, &i__2, &c_n1, (
ftnlen)6, (ftnlen)2);
} else {
/* Writing concatenation */
i__3[0] = 1, a__1[0] = side;
i__3[1] = 1, a__1[1] = trans;
s_cat(ch__1, a__1, i__3, &c__2, (ftnlen)2);
i__1 = *n - 1;
i__2 = *n - 1;
nb = ilaenv_(&c__1, "DORMQR", ch__1, m, &i__1, &i__2, &c_n1, (
ftnlen)6, (ftnlen)2);
}
} else {
if (left) {
/* Writing concatenation */
i__3[0] = 1, a__1[0] = side;
i__3[1] = 1, a__1[1] = trans;
s_cat(ch__1, a__1, i__3, &c__2, (ftnlen)2);
i__1 = *m - 1;
i__2 = *m - 1;
nb = ilaenv_(&c__1, "DORMLQ", ch__1, &i__1, n, &i__2, &c_n1, (
ftnlen)6, (ftnlen)2);
} else {
/* Writing concatenation */
i__3[0] = 1, a__1[0] = side;
i__3[1] = 1, a__1[1] = trans;
s_cat(ch__1, a__1, i__3, &c__2, (ftnlen)2);
i__1 = *n - 1;
i__2 = *n - 1;
nb = ilaenv_(&c__1, "DORMLQ", ch__1, m, &i__1, &i__2, &c_n1, (
ftnlen)6, (ftnlen)2);
}
}
lwkopt = max(1,nw) * nb;
work[1] = (doublereal) lwkopt;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DORMBR", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
work[1] = 1.;
if (*m == 0 || *n == 0) {
return 0;
}
if (applyq) {
/* Apply Q */
if (nq >= *k) {
/* Q was determined by a call to DGEBRD with nq >= k */
dormqr_(side, trans, m, n, k, &a[a_offset], lda, &tau[1], &c__[
c_offset], ldc, &work[1], lwork, &iinfo);
} else if (nq > 1) {
/* Q was determined by a call to DGEBRD with nq < k */
if (left) {
mi = *m - 1;
ni = *n;
i1 = 2;
i2 = 1;
} else {
mi = *m;
ni = *n - 1;
i1 = 1;
i2 = 2;
}
i__1 = nq - 1;
dormqr_(side, trans, &mi, &ni, &i__1, &a[a_dim1 + 2], lda, &tau[1]
, &c__[i1 + i2 * c_dim1], ldc, &work[1], lwork, &iinfo);
}
} else {
/* Apply P */
if (notran) {
*(unsigned char *)transt = 'T';
} else {
*(unsigned char *)transt = 'N';
}
if (nq > *k) {
/* P was determined by a call to DGEBRD with nq > k */
dormlq_(side, transt, m, n, k, &a[a_offset], lda, &tau[1], &c__[
c_offset], ldc, &work[1], lwork, &iinfo);
} else if (nq > 1) {
/* P was determined by a call to DGEBRD with nq <= k */
if (left) {
mi = *m - 1;
ni = *n;
i1 = 2;
i2 = 1;
} else {
mi = *m;
ni = *n - 1;
i1 = 1;
i2 = 2;
}
i__1 = nq - 1;
dormlq_(side, transt, &mi, &ni, &i__1, &a[(a_dim1 << 1) + 1], lda,
&tau[1], &c__[i1 + i2 * c_dim1], ldc, &work[1], lwork, &
iinfo);
}
}
work[1] = (doublereal) lwkopt;
return 0;
/* End of DORMBR */
} /* dormbr_ */
/* Subroutine */ int derrqr_(char *path, integer *nunit)
{
/* Builtin functions */
integer s_wsle(cilist *), e_wsle(void);
/* Subroutine */ int s_copy(char *, char *, ftnlen, ftnlen);
/* Local variables */
doublereal a[4] /* was [2][2] */, b[2];
integer i__, j;
doublereal w[2], x[2], af[4] /* was [2][2] */;
integer info;
extern /* Subroutine */ int dgeqr2_(integer *, integer *, doublereal *,
integer *, doublereal *, doublereal *, integer *), dorg2r_(
integer *, integer *, integer *, doublereal *, integer *,
doublereal *, doublereal *, integer *), dorm2r_(char *, char *,
integer *, integer *, integer *, doublereal *, integer *,
doublereal *, doublereal *, integer *, doublereal *, integer *), alaesm_(char *, logical *, integer *),
dgeqrf_(integer *, integer *, doublereal *, integer *, doublereal
*, doublereal *, integer *, integer *), chkxer_(char *, integer *,
integer *, logical *, logical *), dgeqrs_(integer *,
integer *, integer *, doublereal *, integer *, doublereal *,
doublereal *, integer *, doublereal *, integer *, integer *),
dorgqr_(integer *, integer *, integer *, doublereal *, integer *,
doublereal *, doublereal *, integer *, integer *), dormqr_(char *,
char *, integer *, integer *, integer *, doublereal *, integer *,
doublereal *, doublereal *, integer *, doublereal *, integer *,
integer *);
/* Fortran I/O blocks */
static cilist io___1 = { 0, 0, 0, 0, 0 };
/* -- LAPACK test routine (version 3.1) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* DERRQR tests the error exits for the DOUBLE PRECISION routines */
/* that use the QR decomposition of a general matrix. */
/* Arguments */
/* ========= */
/* PATH (input) CHARACTER*3 */
/* The LAPACK path name for the routines to be tested. */
/* NUNIT (input) INTEGER */
/* The unit number for output. */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. Local Arrays .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Scalars in Common .. */
/* .. */
/* .. Common blocks .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
infoc_1.nout = *nunit;
io___1.ciunit = infoc_1.nout;
s_wsle(&io___1);
e_wsle();
/* Set the variables to innocuous values. */
for (j = 1; j <= 2; ++j) {
for (i__ = 1; i__ <= 2; ++i__) {
a[i__ + (j << 1) - 3] = 1. / (doublereal) (i__ + j);
af[i__ + (j << 1) - 3] = 1. / (doublereal) (i__ + j);
/* L10: */
}
b[j - 1] = 0.;
w[j - 1] = 0.;
x[j - 1] = 0.;
/* L20: */
}
infoc_1.ok = TRUE_;
/* Error exits for QR factorization */
/* DGEQRF */
s_copy(srnamc_1.srnamt, "DGEQRF", (ftnlen)6, (ftnlen)6);
infoc_1.infot = 1;
dgeqrf_(&c_n1, &c__0, a, &c__1, b, w, &c__1, &info);
chkxer_("DGEQRF", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 2;
dgeqrf_(&c__0, &c_n1, a, &c__1, b, w, &c__1, &info);
chkxer_("DGEQRF", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 4;
dgeqrf_(&c__2, &c__1, a, &c__1, b, w, &c__1, &info);
chkxer_("DGEQRF", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 7;
dgeqrf_(&c__1, &c__2, a, &c__1, b, w, &c__1, &info);
chkxer_("DGEQRF", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
/* DGEQR2 */
s_copy(srnamc_1.srnamt, "DGEQR2", (ftnlen)6, (ftnlen)6);
infoc_1.infot = 1;
dgeqr2_(&c_n1, &c__0, a, &c__1, b, w, &info);
chkxer_("DGEQR2", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 2;
dgeqr2_(&c__0, &c_n1, a, &c__1, b, w, &info);
chkxer_("DGEQR2", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 4;
dgeqr2_(&c__2, &c__1, a, &c__1, b, w, &info);
chkxer_("DGEQR2", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
/* DGEQRS */
s_copy(srnamc_1.srnamt, "DGEQRS", (ftnlen)6, (ftnlen)6);
infoc_1.infot = 1;
dgeqrs_(&c_n1, &c__0, &c__0, a, &c__1, x, b, &c__1, w, &c__1, &info);
chkxer_("DGEQRS", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 2;
dgeqrs_(&c__0, &c_n1, &c__0, a, &c__1, x, b, &c__1, w, &c__1, &info);
chkxer_("DGEQRS", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 2;
dgeqrs_(&c__1, &c__2, &c__0, a, &c__2, x, b, &c__2, w, &c__1, &info);
chkxer_("DGEQRS", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 3;
dgeqrs_(&c__0, &c__0, &c_n1, a, &c__1, x, b, &c__1, w, &c__1, &info);
chkxer_("DGEQRS", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 5;
dgeqrs_(&c__2, &c__1, &c__0, a, &c__1, x, b, &c__2, w, &c__1, &info);
chkxer_("DGEQRS", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 8;
dgeqrs_(&c__2, &c__1, &c__0, a, &c__2, x, b, &c__1, w, &c__1, &info);
chkxer_("DGEQRS", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 10;
dgeqrs_(&c__1, &c__1, &c__2, a, &c__1, x, b, &c__1, w, &c__1, &info);
chkxer_("DGEQRS", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
/* DORGQR */
s_copy(srnamc_1.srnamt, "DORGQR", (ftnlen)6, (ftnlen)6);
infoc_1.infot = 1;
dorgqr_(&c_n1, &c__0, &c__0, a, &c__1, x, w, &c__1, &info);
chkxer_("DORGQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 2;
dorgqr_(&c__0, &c_n1, &c__0, a, &c__1, x, w, &c__1, &info);
chkxer_("DORGQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 2;
dorgqr_(&c__1, &c__2, &c__0, a, &c__1, x, w, &c__2, &info);
chkxer_("DORGQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 3;
dorgqr_(&c__0, &c__0, &c_n1, a, &c__1, x, w, &c__1, &info);
chkxer_("DORGQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 3;
dorgqr_(&c__1, &c__1, &c__2, a, &c__1, x, w, &c__1, &info);
chkxer_("DORGQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 5;
dorgqr_(&c__2, &c__2, &c__0, a, &c__1, x, w, &c__2, &info);
chkxer_("DORGQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 8;
dorgqr_(&c__2, &c__2, &c__0, a, &c__2, x, w, &c__1, &info);
chkxer_("DORGQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
/* DORG2R */
s_copy(srnamc_1.srnamt, "DORG2R", (ftnlen)6, (ftnlen)6);
infoc_1.infot = 1;
dorg2r_(&c_n1, &c__0, &c__0, a, &c__1, x, w, &info);
chkxer_("DORG2R", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 2;
dorg2r_(&c__0, &c_n1, &c__0, a, &c__1, x, w, &info);
chkxer_("DORG2R", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 2;
dorg2r_(&c__1, &c__2, &c__0, a, &c__1, x, w, &info);
chkxer_("DORG2R", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 3;
dorg2r_(&c__0, &c__0, &c_n1, a, &c__1, x, w, &info);
chkxer_("DORG2R", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 3;
dorg2r_(&c__2, &c__1, &c__2, a, &c__2, x, w, &info);
chkxer_("DORG2R", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 5;
dorg2r_(&c__2, &c__1, &c__0, a, &c__1, x, w, &info);
chkxer_("DORG2R", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
/* DORMQR */
s_copy(srnamc_1.srnamt, "DORMQR", (ftnlen)6, (ftnlen)6);
infoc_1.infot = 1;
dormqr_("/", "N", &c__0, &c__0, &c__0, a, &c__1, x, af, &c__1, w, &c__1, &
info);
chkxer_("DORMQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 2;
dormqr_("L", "/", &c__0, &c__0, &c__0, a, &c__1, x, af, &c__1, w, &c__1, &
info);
chkxer_("DORMQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 3;
dormqr_("L", "N", &c_n1, &c__0, &c__0, a, &c__1, x, af, &c__1, w, &c__1, &
info);
chkxer_("DORMQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 4;
dormqr_("L", "N", &c__0, &c_n1, &c__0, a, &c__1, x, af, &c__1, w, &c__1, &
info);
chkxer_("DORMQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 5;
dormqr_("L", "N", &c__0, &c__0, &c_n1, a, &c__1, x, af, &c__1, w, &c__1, &
info);
chkxer_("DORMQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 5;
dormqr_("L", "N", &c__0, &c__1, &c__1, a, &c__1, x, af, &c__1, w, &c__1, &
info);
chkxer_("DORMQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 5;
dormqr_("R", "N", &c__1, &c__0, &c__1, a, &c__1, x, af, &c__1, w, &c__1, &
info);
chkxer_("DORMQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 7;
dormqr_("L", "N", &c__2, &c__1, &c__0, a, &c__1, x, af, &c__2, w, &c__1, &
info);
chkxer_("DORMQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 7;
dormqr_("R", "N", &c__1, &c__2, &c__0, a, &c__1, x, af, &c__1, w, &c__1, &
info);
chkxer_("DORMQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 10;
dormqr_("L", "N", &c__2, &c__1, &c__0, a, &c__2, x, af, &c__1, w, &c__1, &
info);
chkxer_("DORMQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 12;
dormqr_("L", "N", &c__1, &c__2, &c__0, a, &c__1, x, af, &c__1, w, &c__1, &
info);
chkxer_("DORMQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 12;
dormqr_("R", "N", &c__2, &c__1, &c__0, a, &c__1, x, af, &c__2, w, &c__1, &
info);
chkxer_("DORMQR", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
/* DORM2R */
s_copy(srnamc_1.srnamt, "DORM2R", (ftnlen)6, (ftnlen)6);
infoc_1.infot = 1;
dorm2r_("/", "N", &c__0, &c__0, &c__0, a, &c__1, x, af, &c__1, w, &info);
chkxer_("DORM2R", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 2;
dorm2r_("L", "/", &c__0, &c__0, &c__0, a, &c__1, x, af, &c__1, w, &info);
chkxer_("DORM2R", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 3;
dorm2r_("L", "N", &c_n1, &c__0, &c__0, a, &c__1, x, af, &c__1, w, &info);
chkxer_("DORM2R", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 4;
dorm2r_("L", "N", &c__0, &c_n1, &c__0, a, &c__1, x, af, &c__1, w, &info);
chkxer_("DORM2R", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 5;
dorm2r_("L", "N", &c__0, &c__0, &c_n1, a, &c__1, x, af, &c__1, w, &info);
chkxer_("DORM2R", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 5;
dorm2r_("L", "N", &c__0, &c__1, &c__1, a, &c__1, x, af, &c__1, w, &info);
chkxer_("DORM2R", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 5;
dorm2r_("R", "N", &c__1, &c__0, &c__1, a, &c__1, x, af, &c__1, w, &info);
chkxer_("DORM2R", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 7;
dorm2r_("L", "N", &c__2, &c__1, &c__0, a, &c__1, x, af, &c__2, w, &info);
chkxer_("DORM2R", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 7;
dorm2r_("R", "N", &c__1, &c__2, &c__0, a, &c__1, x, af, &c__1, w, &info);
chkxer_("DORM2R", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
infoc_1.infot = 10;
dorm2r_("L", "N", &c__2, &c__1, &c__0, a, &c__2, x, af, &c__1, w, &info);
chkxer_("DORM2R", &infoc_1.infot, &infoc_1.nout, &infoc_1.lerr, &
infoc_1.ok);
/* Print a summary line. */
alaesm_(path, &infoc_1.ok, &infoc_1.nout);
return 0;
/* End of DERRQR */
} /* derrqr_ */
/* Subroutine */ int dgelss_(integer *m, integer *n, integer *nrhs,
doublereal *a, integer *lda, doublereal *b, integer *ldb, doublereal *
s, doublereal *rcond, integer *rank, doublereal *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;
doublereal d__1;
/* Local variables */
static doublereal anrm, bnrm;
static integer itau;
static doublereal vdum[1];
static integer i__;
extern /* Subroutine */ int dgemm_(char *, char *, integer *, integer *,
integer *, doublereal *, doublereal *, integer *, doublereal *,
integer *, doublereal *, doublereal *, integer *);
static integer iascl, ibscl;
extern /* Subroutine */ int dgemv_(char *, integer *, integer *,
doublereal *, doublereal *, integer *, doublereal *, integer *,
doublereal *, doublereal *, integer *), drscl_(integer *,
doublereal *, doublereal *, integer *);
static integer chunk;
static doublereal sfmin;
static integer minmn;
extern /* Subroutine */ int dcopy_(integer *, doublereal *, integer *,
doublereal *, integer *);
static integer maxmn, itaup, itauq, mnthr, iwork;
extern /* Subroutine */ int dlabad_(doublereal *, doublereal *);
static integer bl, ie, il;
extern /* Subroutine */ int dgebrd_(integer *, integer *, doublereal *,
integer *, doublereal *, doublereal *, doublereal *, doublereal *,
doublereal *, integer *, integer *);
extern doublereal dlamch_(char *);
static integer mm;
extern doublereal dlange_(char *, integer *, integer *, doublereal *,
integer *, doublereal *);
static integer bdspac;
extern /* Subroutine */ int dgelqf_(integer *, integer *, doublereal *,
integer *, doublereal *, doublereal *, integer *, integer *),
dlascl_(char *, integer *, integer *, doublereal *, doublereal *,
integer *, integer *, doublereal *, integer *, integer *),
dgeqrf_(integer *, integer *, doublereal *, integer *,
doublereal *, doublereal *, integer *, integer *), dlacpy_(char *,
integer *, integer *, doublereal *, integer *, doublereal *,
integer *), dlaset_(char *, integer *, integer *,
doublereal *, doublereal *, doublereal *, integer *),
xerbla_(char *, integer *), dbdsqr_(char *, integer *,
integer *, integer *, integer *, doublereal *, doublereal *,
doublereal *, integer *, doublereal *, integer *, doublereal *,
integer *, doublereal *, integer *), dorgbr_(char *,
integer *, integer *, integer *, doublereal *, integer *,
doublereal *, doublereal *, integer *, integer *);
static doublereal bignum;
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *, ftnlen, ftnlen);
extern /* Subroutine */ int dormbr_(char *, char *, char *, integer *,
integer *, integer *, doublereal *, integer *, doublereal *,
doublereal *, integer *, doublereal *, integer *, integer *), dormlq_(char *, char *, integer *,
integer *, integer *, doublereal *, integer *, doublereal *,
doublereal *, integer *, doublereal *, integer *, integer *);
static integer ldwork;
extern /* Subroutine */ int dormqr_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, doublereal *,
integer *, doublereal *, integer *, integer *);
static integer minwrk, maxwrk;
static doublereal smlnum;
static logical lquery;
static doublereal 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
=======
DGELSS 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) DOUBLE PRECISION 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) DOUBLE PRECISION 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) DOUBLE PRECISION 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) DOUBLE PRECISION
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) DOUBLE PRECISION 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, "DGELSS", " ", 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, "DGEQRF", " ", 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, "DORMQR", "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 DBDSQR
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, "DGEBRD"
, " ", &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, "DORMBR",
"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, "DORGBR",
"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 DBDSQR
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, "DGELQF", " ", 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, "DGEBRD", " ", 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, "DORMBR", "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, "DORGBR", "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, "DORMLQ",
"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, "DGEBRD", " ", m,
n, &c_n1, &c_n1, (ftnlen)6, (ftnlen)1);
/* Computing MAX */
i__1 = maxwrk, i__2 = *m * 3 + *nrhs * ilaenv_(&c__1, "DORMBR"
, "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, "DORGBR",
"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] = (doublereal) maxwrk;
}
minwrk = max(minwrk,1);
if (*lwork < minwrk && ! lquery) {
*info = -12;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DGELSS", &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 = dlamch_("P");
sfmin = dlamch_("S");
smlnum = sfmin / eps;
bignum = 1. / smlnum;
dlabad_(&smlnum, &bignum);
/* Scale A if max element outside range [SMLNUM,BIGNUM] */
anrm = dlange_("M", m, n, &a[a_offset], lda, &work[1]);
iascl = 0;
if (anrm > 0. && anrm < smlnum) {
/* Scale matrix norm up to SMLNUM */
dlascl_("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 */
dlascl_("G", &c__0, &c__0, &anrm, &bignum, m, n, &a[a_offset], lda,
info);
iascl = 2;
} else if (anrm == 0.) {
/* Matrix all zero. Return zero solution. */
i__1 = max(*m,*n);
dlaset_("F", &i__1, nrhs, &c_b74, &c_b74, &b[b_offset], ldb);
dlaset_("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 = dlange_("M", m, nrhs, &b[b_offset], ldb, &work[1]);
ibscl = 0;
if (bnrm > 0. && bnrm < smlnum) {
/* Scale matrix norm up to SMLNUM */
dlascl_("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 */
dlascl_("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;
dgeqrf_(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;
dormqr_("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;
dlaset_("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;
dgebrd_(&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;
dormbr_("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;
dorgbr_("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) */
dbdsqr_("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 */
d__1 = *rcond * s[1];
thr = max(d__1,sfmin);
if (*rcond < 0.) {
/* Computing MAX */
d__1 = eps * s[1];
thr = max(d__1,sfmin);
}
*rank = 0;
i__1 = *n;
for (i__ = 1; i__ <= i__1; ++i__) {
if (s[i__] > thr) {
drscl_(nrhs, &s[i__], &b_ref(i__, 1), ldb);
++(*rank);
} else {
dlaset_("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) {
dgemm_("T", "N", n, nrhs, n, &c_b108, &a[a_offset], lda, &b[
b_offset], ldb, &c_b74, &work[1], ldb);
dlacpy_("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);
dgemm_("T", "N", n, &bl, n, &c_b108, &a[a_offset], lda, &
b_ref(1, i__), ldb, &c_b74, &work[1], n);
dlacpy_("G", n, &bl, &work[1], n, &b_ref(1, i__), ldb);
/* L20: */
}
} else {
dgemv_("T", n, n, &c_b108, &a[a_offset], lda, &b[b_offset], &c__1,
&c_b74, &work[1], &c__1);
dcopy_(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;
dgelqf_(m, n, &a[a_offset], lda, &work[itau], &work[iwork], &i__2,
info);
il = iwork;
/* Copy L to WORK(IL), zeroing out above it */
dlacpy_("L", m, m, &a[a_offset], lda, &work[il], &ldwork);
i__2 = *m - 1;
i__1 = *m - 1;
dlaset_("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;
dgebrd_(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;
dormbr_("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;
dorgbr_("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) */
dbdsqr_("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 */
d__1 = *rcond * s[1];
thr = max(d__1,sfmin);
if (*rcond < 0.) {
/* Computing MAX */
d__1 = eps * s[1];
thr = max(d__1,sfmin);
}
*rank = 0;
i__2 = *m;
for (i__ = 1; i__ <= i__2; ++i__) {
if (s[i__] > thr) {
drscl_(nrhs, &s[i__], &b_ref(i__, 1), ldb);
++(*rank);
} else {
dlaset_("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) {
dgemm_("T", "N", m, nrhs, m, &c_b108, &work[il], &ldwork, &b[
b_offset], ldb, &c_b74, &work[iwork], ldb);
dlacpy_("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);
dgemm_("T", "N", m, &bl, m, &c_b108, &work[il], &ldwork, &
b_ref(1, i__), ldb, &c_b74, &work[iwork], n);
dlacpy_("G", m, &bl, &work[iwork], n, &b_ref(1, i__), ldb);
/* L40: */
}
} else {
dgemv_("T", m, m, &c_b108, &work[il], &ldwork, &b_ref(1, 1), &
c__1, &c_b74, &work[iwork], &c__1);
dcopy_(m, &work[iwork], &c__1, &b_ref(1, 1), &c__1);
}
/* Zero out below first M rows of B */
i__1 = *n - *m;
dlaset_("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;
dormlq_("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;
dgebrd_(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;
dormbr_("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;
dorgbr_("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) */
dbdsqr_("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 */
d__1 = *rcond * s[1];
thr = max(d__1,sfmin);
if (*rcond < 0.) {
/* Computing MAX */
d__1 = eps * s[1];
thr = max(d__1,sfmin);
}
*rank = 0;
i__1 = *m;
for (i__ = 1; i__ <= i__1; ++i__) {
if (s[i__] > thr) {
drscl_(nrhs, &s[i__], &b_ref(i__, 1), ldb);
++(*rank);
} else {
dlaset_("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) {
dgemm_("T", "N", n, nrhs, m, &c_b108, &a[a_offset], lda, &b[
b_offset], ldb, &c_b74, &work[1], ldb);
dlacpy_("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);
dgemm_("T", "N", n, &bl, m, &c_b108, &a[a_offset], lda, &
b_ref(1, i__), ldb, &c_b74, &work[1], n);
dlacpy_("F", n, &bl, &work[1], n, &b_ref(1, i__), ldb);
/* L60: */
}
} else {
dgemv_("T", m, n, &c_b108, &a[a_offset], lda, &b[b_offset], &
c__1, &c_b74, &work[1], &c__1);
dcopy_(n, &work[1], &c__1, &b[b_offset], &c__1);
}
}
}
/* Undo scaling */
if (iascl == 1) {
dlascl_("G", &c__0, &c__0, &anrm, &smlnum, n, nrhs, &b[b_offset], ldb,
info);
dlascl_("G", &c__0, &c__0, &smlnum, &anrm, &minmn, &c__1, &s[1], &
minmn, info);
} else if (iascl == 2) {
dlascl_("G", &c__0, &c__0, &anrm, &bignum, n, nrhs, &b[b_offset], ldb,
info);
dlascl_("G", &c__0, &c__0, &bignum, &anrm, &minmn, &c__1, &s[1], &
minmn, info);
}
if (ibscl == 1) {
dlascl_("G", &c__0, &c__0, &smlnum, &bnrm, n, nrhs, &b[b_offset], ldb,
info);
} else if (ibscl == 2) {
dlascl_("G", &c__0, &c__0, &bignum, &bnrm, n, nrhs, &b[b_offset], ldb,
info);
}
L70:
work[1] = (doublereal) maxwrk;
return 0;
/* End of DGELSS */
} /* dgelss_ */
/* Subroutine */ int dormhr_(char *side, char *trans, integer *m, integer *n,
integer *ilo, integer *ihi, doublereal *a, integer *lda, doublereal *
tau, doublereal *c__, integer *ldc, doublereal *work, integer *lwork,
integer *info)
{
/* System generated locals */
address a__1[2];
integer a_dim1, a_offset, c_dim1, c_offset, i__1[2], i__2;
char ch__1[2];
/* Builtin functions */
/* Subroutine */ int s_cat(char *, char **, integer *, integer *, ftnlen);
/* Local variables */
integer i1, i2, nb, mi, nh, ni, nq, nw;
logical left;
extern logical lsame_(char *, char *);
integer iinfo;
extern /* Subroutine */ int xerbla_(char *, integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *);
extern /* Subroutine */ int dormqr_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, doublereal *,
integer *, doublereal *, integer *, integer *);
integer lwkopt;
logical lquery;
/* -- LAPACK routine (version 3.1) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* DORMHR overwrites the general real M-by-N matrix C with */
/* SIDE = 'L' SIDE = 'R' */
/* TRANS = 'N': Q * C C * Q */
/* TRANS = 'T': Q**T * C C * Q**T */
/* where Q is a real orthogonal matrix of order nq, with nq = m if */
/* SIDE = 'L' and nq = n if SIDE = 'R'. Q is defined as the product of */
/* IHI-ILO elementary reflectors, as returned by DGEHRD: */
/* Q = H(ilo) H(ilo+1) . . . H(ihi-1). */
/* Arguments */
/* ========= */
/* SIDE (input) CHARACTER*1 */
/* = 'L': apply Q or Q**T from the Left; */
/* = 'R': apply Q or Q**T from the Right. */
/* TRANS (input) CHARACTER*1 */
/* = 'N': No transpose, apply Q; */
/* = 'T': Transpose, apply Q**T. */
/* M (input) INTEGER */
/* The number of rows of the matrix C. M >= 0. */
/* N (input) INTEGER */
/* The number of columns of the matrix C. N >= 0. */
/* ILO (input) INTEGER */
/* IHI (input) INTEGER */
/* ILO and IHI must have the same values as in the previous call */
/* of DGEHRD. Q is equal to the unit matrix except in the */
/* submatrix Q(ilo+1:ihi,ilo+1:ihi). */
/* If SIDE = 'L', then 1 <= ILO <= IHI <= M, if M > 0, and */
/* ILO = 1 and IHI = 0, if M = 0; */
/* if SIDE = 'R', then 1 <= ILO <= IHI <= N, if N > 0, and */
/* ILO = 1 and IHI = 0, if N = 0. */
/* A (input) DOUBLE PRECISION array, dimension */
/* (LDA,M) if SIDE = 'L' */
/* (LDA,N) if SIDE = 'R' */
/* The vectors which define the elementary reflectors, as */
/* returned by DGEHRD. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. */
/* LDA >= max(1,M) if SIDE = 'L'; LDA >= max(1,N) if SIDE = 'R'. */
/* TAU (input) DOUBLE PRECISION array, dimension */
/* (M-1) if SIDE = 'L' */
/* (N-1) if SIDE = 'R' */
/* TAU(i) must contain the scalar factor of the elementary */
/* reflector H(i), as returned by DGEHRD. */
/* C (input/output) DOUBLE PRECISION array, dimension (LDC,N) */
/* On entry, the M-by-N matrix C. */
/* On exit, C is overwritten by Q*C or Q**T*C or C*Q**T or C*Q. */
/* LDC (input) INTEGER */
/* The leading dimension of the array C. LDC >= max(1,M). */
/* WORK (workspace/output) DOUBLE PRECISION 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 SIDE = 'L', LWORK >= max(1,N); */
/* if SIDE = 'R', LWORK >= max(1,M). */
/* For optimum performance LWORK >= N*NB if SIDE = 'L', and */
/* LWORK >= M*NB if SIDE = 'R', where NB is the optimal */
/* blocksize. */
/* 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 */
/* ===================================================================== */
/* .. Local Scalars .. */
/* .. */
/* .. 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;
--tau;
c_dim1 = *ldc;
c_offset = 1 + c_dim1;
c__ -= c_offset;
--work;
/* Function Body */
*info = 0;
nh = *ihi - *ilo;
left = lsame_(side, "L");
lquery = *lwork == -1;
/* NQ is the order of Q and NW is the minimum dimension of WORK */
if (left) {
nq = *m;
nw = *n;
} else {
nq = *n;
nw = *m;
}
if (! left && ! lsame_(side, "R")) {
*info = -1;
} else if (! lsame_(trans, "N") && ! lsame_(trans,
"T")) {
*info = -2;
} else if (*m < 0) {
*info = -3;
} else if (*n < 0) {
*info = -4;
} else if (*ilo < 1 || *ilo > max(1,nq)) {
*info = -5;
} else if (*ihi < min(*ilo,nq) || *ihi > nq) {
*info = -6;
} else if (*lda < max(1,nq)) {
*info = -8;
} else if (*ldc < max(1,*m)) {
*info = -11;
} else if (*lwork < max(1,nw) && ! lquery) {
*info = -13;
}
if (*info == 0) {
if (left) {
/* Writing concatenation */
i__1[0] = 1, a__1[0] = side;
i__1[1] = 1, a__1[1] = trans;
s_cat(ch__1, a__1, i__1, &c__2, (ftnlen)2);
nb = ilaenv_(&c__1, "DORMQR", ch__1, &nh, n, &nh, &c_n1);
} else {
/* Writing concatenation */
i__1[0] = 1, a__1[0] = side;
i__1[1] = 1, a__1[1] = trans;
s_cat(ch__1, a__1, i__1, &c__2, (ftnlen)2);
nb = ilaenv_(&c__1, "DORMQR", ch__1, m, &nh, &nh, &c_n1);
}
lwkopt = max(1,nw) * nb;
work[1] = (doublereal) lwkopt;
}
if (*info != 0) {
i__2 = -(*info);
xerbla_("DORMHR", &i__2);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
if (*m == 0 || *n == 0 || nh == 0) {
work[1] = 1.;
return 0;
}
if (left) {
mi = nh;
ni = *n;
i1 = *ilo + 1;
i2 = 1;
} else {
mi = *m;
ni = nh;
i1 = 1;
i2 = *ilo + 1;
}
dormqr_(side, trans, &mi, &ni, &nh, &a[*ilo + 1 + *ilo * a_dim1], lda, &
tau[*ilo], &c__[i1 + i2 * c_dim1], ldc, &work[1], lwork, &iinfo);
work[1] = (doublereal) lwkopt;
return 0;
/* End of DORMHR */
} /* dormhr_ */
/* Subroutine */ int dggglm_(integer *n, integer *m, integer *p, doublereal *
a, integer *lda, doublereal *b, integer *ldb, doublereal *d__,
doublereal *x, doublereal *y, doublereal *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;
/* Local variables */
integer i__, nb, np, nb1, nb2, nb3, nb4, lopt;
extern /* Subroutine */ int dgemv_(char *, integer *, integer *,
doublereal *, doublereal *, integer *, doublereal *, integer *,
doublereal *, doublereal *, integer *), dcopy_(integer *,
doublereal *, integer *, doublereal *, integer *), dggqrf_(
integer *, integer *, integer *, doublereal *, integer *,
doublereal *, doublereal *, integer *, doublereal *, doublereal *,
integer *, integer *), xerbla_(char *, integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *);
integer lwkmin;
extern /* Subroutine */ int dormqr_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, doublereal *,
integer *, doublereal *, integer *, integer *),
dormrq_(char *, char *, integer *, integer *, integer *,
doublereal *, integer *, doublereal *, doublereal *, integer *,
doublereal *, integer *, integer *);
integer lwkopt;
logical lquery;
extern /* Subroutine */ int dtrtrs_(char *, char *, char *, integer *,
integer *, doublereal *, integer *, doublereal *, 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 */
/* ======= */
/* DGGGLM solves a general Gauss-Markov linear model (GLM) problem: */
/* minimize || y ||_2 subject to d = A*x + B*y */
/* x */
/* where A is an N-by-M matrix, B is an N-by-P matrix, and d is a */
/* given N-vector. It is assumed that M <= N <= M+P, and */
/* rank(A) = M and rank( A B ) = N. */
/* Under these assumptions, the constrained equation is always */
/* consistent, and there is a unique solution x and a minimal 2-norm */
/* solution y, which is obtained using a generalized QR factorization */
/* of the matrices (A, B) given by */
/* A = Q*(R), B = Q*T*Z. */
/* (0) */
/* In particular, if matrix B is square nonsingular, then the problem */
/* GLM is equivalent to the following weighted linear least squares */
/* problem */
/* minimize || inv(B)*(d-A*x) ||_2 */
/* x */
/* where inv(B) denotes the inverse of B. */
/* Arguments */
/* ========= */
/* N (input) INTEGER */
/* The number of rows of the matrices A and B. N >= 0. */
/* M (input) INTEGER */
/* The number of columns of the matrix A. 0 <= M <= N. */
/* P (input) INTEGER */
/* The number of columns of the matrix B. P >= N-M. */
/* A (input/output) DOUBLE PRECISION array, dimension (LDA,M) */
/* On entry, the N-by-M matrix A. */
/* On exit, the upper triangular part of the array A contains */
/* the M-by-M upper triangular matrix R. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,N). */
/* B (input/output) DOUBLE PRECISION array, dimension (LDB,P) */
/* On entry, the N-by-P matrix B. */
/* On exit, if N <= P, the upper triangle of the subarray */
/* B(1:N,P-N+1:P) contains the N-by-N upper triangular matrix T; */
/* if N > P, the elements on and above the (N-P)th subdiagonal */
/* contain the N-by-P upper trapezoidal matrix T. */
/* LDB (input) INTEGER */
/* The leading dimension of the array B. LDB >= max(1,N). */
/* D (input/output) DOUBLE PRECISION array, dimension (N) */
/* On entry, D is the left hand side of the GLM equation. */
/* On exit, D is destroyed. */
/* X (output) DOUBLE PRECISION array, dimension (M) */
/* Y (output) DOUBLE PRECISION array, dimension (P) */
/* On exit, X and Y are the solutions of the GLM problem. */
/* WORK (workspace/output) DOUBLE PRECISION 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,N+M+P). */
/* For optimum performance, LWORK >= M+min(N,P)+max(N,P)*NB, */
/* where NB is an upper bound for the optimal blocksizes for */
/* DGEQRF, SGERQF, DORMQR and SORMRQ. */
/* 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: the upper triangular factor R associated with A in the */
/* generalized QR factorization of the pair (A, B) is */
/* singular, so that rank(A) < M; the least squares */
/* solution could not be computed. */
/* = 2: the bottom (N-M) by (N-M) part of the upper trapezoidal */
/* factor T associated with B in the generalized QR */
/* factorization of the pair (A, B) is singular, so that */
/* rank( A B ) < N; the least squares solution could not */
/* be computed. */
/* =================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Test the input parameters */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
b_dim1 = *ldb;
b_offset = 1 + b_dim1;
b -= b_offset;
--d__;
--x;
--y;
--work;
/* Function Body */
*info = 0;
np = min(*n,*p);
lquery = *lwork == -1;
if (*n < 0) {
*info = -1;
} else if (*m < 0 || *m > *n) {
*info = -2;
} else if (*p < 0 || *p < *n - *m) {
*info = -3;
} else if (*lda < max(1,*n)) {
*info = -5;
} else if (*ldb < max(1,*n)) {
*info = -7;
}
/* Calculate workspace */
if (*info == 0) {
if (*n == 0) {
lwkmin = 1;
lwkopt = 1;
} else {
nb1 = ilaenv_(&c__1, "DGEQRF", " ", n, m, &c_n1, &c_n1);
nb2 = ilaenv_(&c__1, "DGERQF", " ", n, m, &c_n1, &c_n1);
nb3 = ilaenv_(&c__1, "DORMQR", " ", n, m, p, &c_n1);
nb4 = ilaenv_(&c__1, "DORMRQ", " ", n, m, p, &c_n1);
/* Computing MAX */
i__1 = max(nb1,nb2), i__1 = max(i__1,nb3);
nb = max(i__1,nb4);
lwkmin = *m + *n + *p;
lwkopt = *m + np + max(*n,*p) * nb;
}
work[1] = (doublereal) lwkopt;
if (*lwork < lwkmin && ! lquery) {
*info = -12;
}
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DGGGLM", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
if (*n == 0) {
return 0;
}
/* Compute the GQR factorization of matrices A and B: */
/* Q'*A = ( R11 ) M, Q'*B*Z' = ( T11 T12 ) M */
/* ( 0 ) N-M ( 0 T22 ) N-M */
/* M M+P-N N-M */
/* where R11 and T22 are upper triangular, and Q and Z are */
/* orthogonal. */
i__1 = *lwork - *m - np;
dggqrf_(n, m, p, &a[a_offset], lda, &work[1], &b[b_offset], ldb, &work[*m
+ 1], &work[*m + np + 1], &i__1, info);
lopt = (integer) work[*m + np + 1];
/* Update left-hand-side vector d = Q'*d = ( d1 ) M */
/* ( d2 ) N-M */
i__1 = max(1,*n);
i__2 = *lwork - *m - np;
dormqr_("Left", "Transpose", n, &c__1, m, &a[a_offset], lda, &work[1], &
d__[1], &i__1, &work[*m + np + 1], &i__2, info);
/* Computing MAX */
i__1 = lopt, i__2 = (integer) work[*m + np + 1];
lopt = max(i__1,i__2);
/* Solve T22*y2 = d2 for y2 */
if (*n > *m) {
i__1 = *n - *m;
i__2 = *n - *m;
dtrtrs_("Upper", "No transpose", "Non unit", &i__1, &c__1, &b[*m + 1
+ (*m + *p - *n + 1) * b_dim1], ldb, &d__[*m + 1], &i__2,
info);
if (*info > 0) {
*info = 1;
return 0;
}
i__1 = *n - *m;
dcopy_(&i__1, &d__[*m + 1], &c__1, &y[*m + *p - *n + 1], &c__1);
}
/* Set y1 = 0 */
i__1 = *m + *p - *n;
for (i__ = 1; i__ <= i__1; ++i__) {
y[i__] = 0.;
/* L10: */
}
/* Update d1 = d1 - T12*y2 */
i__1 = *n - *m;
dgemv_("No transpose", m, &i__1, &c_b32, &b[(*m + *p - *n + 1) * b_dim1 +
1], ldb, &y[*m + *p - *n + 1], &c__1, &c_b34, &d__[1], &c__1);
/* Solve triangular system: R11*x = d1 */
if (*m > 0) {
dtrtrs_("Upper", "No Transpose", "Non unit", m, &c__1, &a[a_offset],
lda, &d__[1], m, info);
if (*info > 0) {
*info = 2;
return 0;
}
/* Copy D to X */
dcopy_(m, &d__[1], &c__1, &x[1], &c__1);
}
/* Backward transformation y = Z'*y */
/* Computing MAX */
i__1 = 1, i__2 = *n - *p + 1;
i__3 = max(1,*p);
i__4 = *lwork - *m - np;
dormrq_("Left", "Transpose", p, &c__1, &np, &b[max(i__1, i__2)+ b_dim1],
ldb, &work[*m + 1], &y[1], &i__3, &work[*m + np + 1], &i__4, info);
/* Computing MAX */
i__1 = lopt, i__2 = (integer) work[*m + np + 1];
work[1] = (doublereal) (*m + np + max(i__1,i__2));
return 0;
/* End of DGGGLM */
} /* dggglm_ */
/* Subroutine */ int dqrt03_(integer *m, integer *n, integer *k, doublereal *
af, doublereal *c__, doublereal *cc, doublereal *q, integer *lda,
doublereal *tau, doublereal *work, integer *lwork, doublereal *rwork,
doublereal *result)
{
/* Initialized data */
static integer iseed[4] = { 1988,1989,1990,1991 };
/* System generated locals */
integer af_dim1, af_offset, c_dim1, c_offset, cc_dim1, cc_offset, q_dim1,
q_offset, i__1;
/* Builtin functions */
/* Subroutine */ int s_copy(char *, char *, ftnlen, ftnlen);
/* Local variables */
integer j, mc, nc;
doublereal eps;
char side[1];
integer info;
extern /* Subroutine */ int dgemm_(char *, char *, integer *, integer *,
integer *, doublereal *, doublereal *, integer *, doublereal *,
integer *, doublereal *, doublereal *, integer *);
integer iside;
extern logical lsame_(char *, char *);
doublereal resid, cnorm;
char trans[1];
extern doublereal dlamch_(char *), dlange_(char *, integer *,
integer *, doublereal *, integer *, doublereal *);
extern /* Subroutine */ int dlacpy_(char *, integer *, integer *,
doublereal *, integer *, doublereal *, integer *),
dlaset_(char *, integer *, integer *, doublereal *, doublereal *,
doublereal *, integer *), dlarnv_(integer *, integer *,
integer *, doublereal *), dorgqr_(integer *, integer *, integer *,
doublereal *, integer *, doublereal *, doublereal *, integer *,
integer *);
integer itrans;
extern /* Subroutine */ int dormqr_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, doublereal *,
integer *, doublereal *, integer *, integer *);
/* -- LAPACK test routine (version 3.1) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* DQRT03 tests DORMQR, which computes Q*C, Q'*C, C*Q or C*Q'. */
/* DQRT03 compares the results of a call to DORMQR with the results of */
/* forming Q explicitly by a call to DORGQR and then performing matrix */
/* multiplication by a call to DGEMM. */
/* Arguments */
/* ========= */
/* M (input) INTEGER */
/* The order of the orthogonal matrix Q. M >= 0. */
/* N (input) INTEGER */
/* The number of rows or columns of the matrix C; C is m-by-n if */
/* Q is applied from the left, or n-by-m if Q is applied from */
/* the right. N >= 0. */
/* K (input) INTEGER */
/* The number of elementary reflectors whose product defines the */
/* orthogonal matrix Q. M >= K >= 0. */
/* AF (input) DOUBLE PRECISION array, dimension (LDA,N) */
/* Details of the QR factorization of an m-by-n matrix, as */
/* returnedby DGEQRF. See SGEQRF for further details. */
/* C (workspace) DOUBLE PRECISION array, dimension (LDA,N) */
/* CC (workspace) DOUBLE PRECISION array, dimension (LDA,N) */
/* Q (workspace) DOUBLE PRECISION array, dimension (LDA,M) */
/* LDA (input) INTEGER */
/* The leading dimension of the arrays AF, C, CC, and Q. */
/* TAU (input) DOUBLE PRECISION array, dimension (min(M,N)) */
/* The scalar factors of the elementary reflectors corresponding */
/* to the QR factorization in AF. */
/* WORK (workspace) DOUBLE PRECISION array, dimension (LWORK) */
/* LWORK (input) INTEGER */
/* The length of WORK. LWORK must be at least M, and should be */
/* M*NB, where NB is the blocksize for this environment. */
/* RWORK (workspace) DOUBLE PRECISION array, dimension (M) */
/* RESULT (output) DOUBLE PRECISION array, dimension (4) */
/* The test ratios compare two techniques for multiplying a */
/* random matrix C by an m-by-m orthogonal matrix Q. */
/* RESULT(1) = norm( Q*C - Q*C ) / ( M * norm(C) * EPS ) */
/* RESULT(2) = norm( C*Q - C*Q ) / ( M * norm(C) * EPS ) */
/* RESULT(3) = norm( Q'*C - Q'*C )/ ( M * norm(C) * EPS ) */
/* RESULT(4) = norm( C*Q' - C*Q' )/ ( M * norm(C) * EPS ) */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Local Arrays .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Scalars in Common .. */
/* .. */
/* .. Common blocks .. */
/* .. */
/* .. Data statements .. */
/* Parameter adjustments */
q_dim1 = *lda;
q_offset = 1 + q_dim1;
q -= q_offset;
cc_dim1 = *lda;
cc_offset = 1 + cc_dim1;
cc -= cc_offset;
c_dim1 = *lda;
c_offset = 1 + c_dim1;
c__ -= c_offset;
af_dim1 = *lda;
af_offset = 1 + af_dim1;
af -= af_offset;
--tau;
--work;
--rwork;
--result;
/* Function Body */
/* .. */
/* .. Executable Statements .. */
eps = dlamch_("Epsilon");
/* Copy the first k columns of the factorization to the array Q */
dlaset_("Full", m, m, &c_b4, &c_b4, &q[q_offset], lda);
i__1 = *m - 1;
dlacpy_("Lower", &i__1, k, &af[af_dim1 + 2], lda, &q[q_dim1 + 2], lda);
/* Generate the m-by-m matrix Q */
s_copy(srnamc_1.srnamt, "DORGQR", (ftnlen)6, (ftnlen)6);
dorgqr_(m, m, k, &q[q_offset], lda, &tau[1], &work[1], lwork, &info);
for (iside = 1; iside <= 2; ++iside) {
if (iside == 1) {
*(unsigned char *)side = 'L';
mc = *m;
nc = *n;
} else {
*(unsigned char *)side = 'R';
mc = *n;
nc = *m;
}
/* Generate MC by NC matrix C */
i__1 = nc;
for (j = 1; j <= i__1; ++j) {
dlarnv_(&c__2, iseed, &mc, &c__[j * c_dim1 + 1]);
/* L10: */
}
cnorm = dlange_("1", &mc, &nc, &c__[c_offset], lda, &rwork[1]);
if (cnorm == 0.) {
cnorm = 1.;
}
for (itrans = 1; itrans <= 2; ++itrans) {
if (itrans == 1) {
*(unsigned char *)trans = 'N';
} else {
*(unsigned char *)trans = 'T';
}
/* Copy C */
dlacpy_("Full", &mc, &nc, &c__[c_offset], lda, &cc[cc_offset],
lda);
/* Apply Q or Q' to C */
s_copy(srnamc_1.srnamt, "DORMQR", (ftnlen)6, (ftnlen)6);
dormqr_(side, trans, &mc, &nc, k, &af[af_offset], lda, &tau[1], &
cc[cc_offset], lda, &work[1], lwork, &info);
/* Form explicit product and subtract */
if (lsame_(side, "L")) {
dgemm_(trans, "No transpose", &mc, &nc, &mc, &c_b21, &q[
q_offset], lda, &c__[c_offset], lda, &c_b22, &cc[
cc_offset], lda);
} else {
dgemm_("No transpose", trans, &mc, &nc, &nc, &c_b21, &c__[
c_offset], lda, &q[q_offset], lda, &c_b22, &cc[
cc_offset], lda);
}
/* Compute error in the difference */
resid = dlange_("1", &mc, &nc, &cc[cc_offset], lda, &rwork[1]);
result[(iside - 1 << 1) + itrans] = resid / ((doublereal) max(1,*
m) * cnorm * eps);
/* L20: */
}
/* L30: */
}
return 0;
/* End of DQRT03 */
} /* dqrt03_ */
/* Subroutine */ int dgglse_(integer *m, integer *n, integer *p, doublereal *
a, integer *lda, doublereal *b, integer *ldb, doublereal *c__,
doublereal *d__, doublereal *x, doublereal *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
=======
DGGLSE solves the linear equality-constrained least squares (LSE)
problem:
minimize || c - A*x ||_2 subject to B*x = d
where A is an M-by-N matrix, B is a P-by-N matrix, c is a given
M-vector, and d is a given P-vector. It is assumed that
P <= N <= M+P, and
rank(B) = P and rank( ( A ) ) = N.
( ( B ) )
These conditions ensure that the LSE problem has a unique solution,
which is obtained using a GRQ factorization of the matrices B and A.
Arguments
=========
M (input) INTEGER
The number of rows of the matrix A. M >= 0.
N (input) INTEGER
The number of columns of the matrices A and B. N >= 0.
P (input) INTEGER
The number of rows of the matrix B. 0 <= P <= N <= M+P.
A (input/output) DOUBLE PRECISION array, dimension (LDA,N)
On entry, the M-by-N matrix A.
On exit, A is destroyed.
LDA (input) INTEGER
The leading dimension of the array A. LDA >= max(1,M).
B (input/output) DOUBLE PRECISION array, dimension (LDB,N)
On entry, the P-by-N matrix B.
On exit, B is destroyed.
LDB (input) INTEGER
The leading dimension of the array B. LDB >= max(1,P).
C (input/output) DOUBLE PRECISION array, dimension (M)
On entry, C contains the right hand side vector for the
least squares part of the LSE problem.
On exit, the residual sum of squares for the solution
is given by the sum of squares of elements N-P+1 to M of
vector C.
D (input/output) DOUBLE PRECISION array, dimension (P)
On entry, D contains the right hand side vector for the
constrained equation.
On exit, D is destroyed.
X (output) DOUBLE PRECISION array, dimension (N)
On exit, X is the solution of the LSE problem.
WORK (workspace/output) DOUBLE PRECISION 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,M+N+P).
For optimum performance LWORK >= P+min(M,N)+max(M,N)*NB,
where NB is an upper bound for the optimal blocksizes for
DGEQRF, SGERQF, DORMQR and SORMRQ.
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.
=====================================================================
Test the input parameters
Parameter adjustments */
/* Table of constant values */
static integer c__1 = 1;
static integer c_n1 = -1;
static doublereal c_b29 = -1.;
static doublereal c_b31 = 1.;
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, i__1, i__2;
/* Local variables */
static integer lopt;
extern /* Subroutine */ int dgemv_(char *, integer *, integer *,
doublereal *, doublereal *, integer *, doublereal *, integer *,
doublereal *, doublereal *, integer *), dcopy_(integer *,
doublereal *, integer *, doublereal *, integer *), daxpy_(integer
*, doublereal *, doublereal *, integer *, doublereal *, integer *)
, dtrmv_(char *, char *, char *, integer *, doublereal *, integer
*, doublereal *, integer *), dtrsv_(char *
, char *, char *, integer *, doublereal *, integer *, doublereal *
, integer *);
static integer nb, mn, nr;
extern /* Subroutine */ int dggrqf_(integer *, integer *, integer *,
doublereal *, integer *, doublereal *, doublereal *, integer *,
doublereal *, doublereal *, integer *, integer *), xerbla_(char *,
integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *, ftnlen, ftnlen);
static integer nb1, nb2, nb3, nb4;
extern /* Subroutine */ int dormqr_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, doublereal *,
integer *, doublereal *, integer *, integer *),
dormrq_(char *, char *, integer *, integer *, integer *,
doublereal *, integer *, doublereal *, doublereal *, integer *,
doublereal *, integer *, integer *);
static integer lwkopt;
static logical lquery;
#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]
a_dim1 = *lda;
a_offset = 1 + a_dim1 * 1;
a -= a_offset;
b_dim1 = *ldb;
b_offset = 1 + b_dim1 * 1;
b -= b_offset;
--c__;
--d__;
--x;
--work;
/* Function Body */
*info = 0;
mn = min(*m,*n);
nb1 = ilaenv_(&c__1, "DGEQRF", " ", m, n, &c_n1, &c_n1, (ftnlen)6, (
ftnlen)1);
nb2 = ilaenv_(&c__1, "DGERQF", " ", m, n, &c_n1, &c_n1, (ftnlen)6, (
ftnlen)1);
nb3 = ilaenv_(&c__1, "DORMQR", " ", m, n, p, &c_n1, (ftnlen)6, (ftnlen)1);
nb4 = ilaenv_(&c__1, "DORMRQ", " ", m, n, p, &c_n1, (ftnlen)6, (ftnlen)1);
/* Computing MAX */
i__1 = max(nb1,nb2), i__1 = max(i__1,nb3);
nb = max(i__1,nb4);
lwkopt = *p + mn + max(*m,*n) * nb;
work[1] = (doublereal) lwkopt;
lquery = *lwork == -1;
if (*m < 0) {
*info = -1;
} else if (*n < 0) {
*info = -2;
} else if (*p < 0 || *p > *n || *p < *n - *m) {
*info = -3;
} else if (*lda < max(1,*m)) {
*info = -5;
} else if (*ldb < max(1,*p)) {
*info = -7;
} else /* if(complicated condition) */ {
/* Computing MAX */
i__1 = 1, i__2 = *m + *n + *p;
if (*lwork < max(i__1,i__2) && ! lquery) {
*info = -12;
}
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DGGLSE", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
if (*n == 0) {
return 0;
}
/* Compute the GRQ factorization of matrices B and A:
B*Q' = ( 0 T12 ) P Z'*A*Q' = ( R11 R12 ) N-P
N-P P ( 0 R22 ) M+P-N
N-P P
where T12 and R11 are upper triangular, and Q and Z are
orthogonal. */
i__1 = *lwork - *p - mn;
dggrqf_(p, m, n, &b[b_offset], ldb, &work[1], &a[a_offset], lda, &work[*p
+ 1], &work[*p + mn + 1], &i__1, info);
lopt = (integer) work[*p + mn + 1];
/* Update c = Z'*c = ( c1 ) N-P
( c2 ) M+P-N */
i__1 = max(1,*m);
i__2 = *lwork - *p - mn;
dormqr_("Left", "Transpose", m, &c__1, &mn, &a[a_offset], lda, &work[*p +
1], &c__[1], &i__1, &work[*p + mn + 1], &i__2, info);
/* Computing MAX */
i__1 = lopt, i__2 = (integer) work[*p + mn + 1];
lopt = max(i__1,i__2);
/* Solve T12*x2 = d for x2 */
dtrsv_("Upper", "No transpose", "Non unit", p, &b_ref(1, *n - *p + 1),
ldb, &d__[1], &c__1);
/* Update c1 */
i__1 = *n - *p;
dgemv_("No transpose", &i__1, p, &c_b29, &a_ref(1, *n - *p + 1), lda, &
d__[1], &c__1, &c_b31, &c__[1], &c__1);
/* Sovle R11*x1 = c1 for x1 */
i__1 = *n - *p;
dtrsv_("Upper", "No transpose", "Non unit", &i__1, &a[a_offset], lda, &
c__[1], &c__1);
/* Put the solutions in X */
i__1 = *n - *p;
dcopy_(&i__1, &c__[1], &c__1, &x[1], &c__1);
dcopy_(p, &d__[1], &c__1, &x[*n - *p + 1], &c__1);
/* Compute the residual vector: */
if (*m < *n) {
nr = *m + *p - *n;
i__1 = *n - *m;
dgemv_("No transpose", &nr, &i__1, &c_b29, &a_ref(*n - *p + 1, *m + 1)
, lda, &d__[nr + 1], &c__1, &c_b31, &c__[*n - *p + 1], &c__1);
} else {
nr = *p;
}
dtrmv_("Upper", "No transpose", "Non unit", &nr, &a_ref(*n - *p + 1, *n -
*p + 1), lda, &d__[1], &c__1);
daxpy_(&nr, &c_b29, &d__[1], &c__1, &c__[*n - *p + 1], &c__1);
/* Backward transformation x = Q'*x */
i__1 = *lwork - *p - mn;
dormrq_("Left", "Transpose", n, &c__1, p, &b[b_offset], ldb, &work[1], &x[
1], n, &work[*p + mn + 1], &i__1, info);
/* Computing MAX */
i__1 = lopt, i__2 = (integer) work[*p + mn + 1];
work[1] = (doublereal) (*p + mn + max(i__1,i__2));
return 0;
/* End of DGGLSE */
} /* dgglse_ */
/* Subroutine */ int dggqrf_(integer *n, integer *m, integer *p, doublereal *
a, integer *lda, doublereal *taua, doublereal *b, integer *ldb,
doublereal *taub, doublereal *work, integer *lwork, integer *info)
{
/* -- LAPACK 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
=======
DGGQRF computes a generalized QR factorization of an N-by-M matrix A
and an N-by-P matrix B:
A = Q*R, B = Q*T*Z,
where Q is an N-by-N orthogonal matrix, Z is a P-by-P orthogonal
matrix, and R and T assume one of the forms:
if N >= M, R = ( R11 ) M , or if N < M, R = ( R11 R12 ) N,
( 0 ) N-M N M-N
M
where R11 is upper triangular, and
if N <= P, T = ( 0 T12 ) N, or if N > P, T = ( T11 ) N-P,
P-N N ( T21 ) P
P
where T12 or T21 is upper triangular.
In particular, if B is square and nonsingular, the GQR factorization
of A and B implicitly gives the QR factorization of inv(B)*A:
inv(B)*A = Z'*(inv(T)*R)
where inv(B) denotes the inverse of the matrix B, and Z' denotes the
transpose of the matrix Z.
Arguments
=========
N (input) INTEGER
The number of rows of the matrices A and B. N >= 0.
M (input) INTEGER
The number of columns of the matrix A. M >= 0.
P (input) INTEGER
The number of columns of the matrix B. P >= 0.
A (input/output) DOUBLE PRECISION array, dimension (LDA,M)
On entry, the N-by-M matrix A.
On exit, the elements on and above the diagonal of the array
contain the min(N,M)-by-M upper trapezoidal matrix R (R is
upper triangular if N >= M); the elements below the diagonal,
with the array TAUA, represent the orthogonal matrix Q as a
product of min(N,M) elementary reflectors (see Further
Details).
LDA (input) INTEGER
The leading dimension of the array A. LDA >= max(1,N).
TAUA (output) DOUBLE PRECISION array, dimension (min(N,M))
The scalar factors of the elementary reflectors which
represent the orthogonal matrix Q (see Further Details).
B (input/output) DOUBLE PRECISION array, dimension (LDB,P)
On entry, the N-by-P matrix B.
On exit, if N <= P, the upper triangle of the subarray
B(1:N,P-N+1:P) contains the N-by-N upper triangular matrix T;
if N > P, the elements on and above the (N-P)-th subdiagonal
contain the N-by-P upper trapezoidal matrix T; the remaining
elements, with the array TAUB, represent the orthogonal
matrix Z as a product of elementary reflectors (see Further
Details).
LDB (input) INTEGER
The leading dimension of the array B. LDB >= max(1,N).
TAUB (output) DOUBLE PRECISION array, dimension (min(N,P))
The scalar factors of the elementary reflectors which
represent the orthogonal matrix Z (see Further Details).
WORK (workspace/output) DOUBLE PRECISION 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,N,M,P).
For optimum performance LWORK >= max(N,M,P)*max(NB1,NB2,NB3),
where NB1 is the optimal blocksize for the QR factorization
of an N-by-M matrix, NB2 is the optimal blocksize for the
RQ factorization of an N-by-P matrix, and NB3 is the optimal
blocksize for a call of DORMQR.
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.
Further Details
===============
The matrix Q is represented as a product of elementary reflectors
Q = H(1) H(2) . . . H(k), where k = min(n,m).
Each H(i) has the form
H(i) = I - taua * v * v'
where taua is a real scalar, and v is a real vector with
v(1:i-1) = 0 and v(i) = 1; v(i+1:n) is stored on exit in A(i+1:n,i),
and taua in TAUA(i).
To form Q explicitly, use LAPACK subroutine DORGQR.
To use Q to update another matrix, use LAPACK subroutine DORMQR.
The matrix Z is represented as a product of elementary reflectors
Z = H(1) H(2) . . . H(k), where k = min(n,p).
Each H(i) has the form
H(i) = I - taub * v * v'
where taub is a real scalar, and v is a real vector with
v(p-k+i+1:p) = 0 and v(p-k+i) = 1; v(1:p-k+i-1) is stored on exit in
B(n-k+i,1:p-k+i-1), and taub in TAUB(i).
To form Z explicitly, use LAPACK subroutine DORGRQ.
To use Z to update another matrix, use LAPACK subroutine DORMRQ.
=====================================================================
Test the input parameters
Parameter adjustments */
/* Table of constant values */
static integer c__1 = 1;
static integer c_n1 = -1;
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, i__1, i__2;
/* Local variables */
static integer lopt, nb;
extern /* Subroutine */ int dgeqrf_(integer *, integer *, doublereal *,
integer *, doublereal *, doublereal *, integer *, integer *),
dgerqf_(integer *, integer *, doublereal *, integer *, doublereal
*, doublereal *, integer *, integer *), xerbla_(char *, integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *, ftnlen, ftnlen);
static integer nb1, nb2, nb3;
extern /* Subroutine */ int dormqr_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, doublereal *,
integer *, doublereal *, integer *, integer *);
static integer lwkopt;
static logical lquery;
a_dim1 = *lda;
a_offset = 1 + a_dim1 * 1;
a -= a_offset;
--taua;
b_dim1 = *ldb;
b_offset = 1 + b_dim1 * 1;
b -= b_offset;
--taub;
--work;
/* Function Body */
*info = 0;
nb1 = ilaenv_(&c__1, "DGEQRF", " ", n, m, &c_n1, &c_n1, (ftnlen)6, (
ftnlen)1);
nb2 = ilaenv_(&c__1, "DGERQF", " ", n, p, &c_n1, &c_n1, (ftnlen)6, (
ftnlen)1);
nb3 = ilaenv_(&c__1, "DORMQR", " ", n, m, p, &c_n1, (ftnlen)6, (ftnlen)1);
/* Computing MAX */
i__1 = max(nb1,nb2);
nb = max(i__1,nb3);
/* Computing MAX */
i__1 = max(*n,*m);
lwkopt = max(i__1,*p) * nb;
work[1] = (doublereal) lwkopt;
lquery = *lwork == -1;
if (*n < 0) {
*info = -1;
} else if (*m < 0) {
*info = -2;
} else if (*p < 0) {
*info = -3;
} else if (*lda < max(1,*n)) {
*info = -5;
} else if (*ldb < max(1,*n)) {
*info = -8;
} else /* if(complicated condition) */ {
/* Computing MAX */
i__1 = max(1,*n), i__1 = max(i__1,*m);
if (*lwork < max(i__1,*p) && ! lquery) {
*info = -11;
}
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DGGQRF", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* QR factorization of N-by-M matrix A: A = Q*R */
dgeqrf_(n, m, &a[a_offset], lda, &taua[1], &work[1], lwork, info);
lopt = (integer) work[1];
/* Update B := Q'*B. */
i__1 = min(*n,*m);
dormqr_("Left", "Transpose", n, p, &i__1, &a[a_offset], lda, &taua[1], &b[
b_offset], ldb, &work[1], lwork, info);
/* Computing MAX */
i__1 = lopt, i__2 = (integer) work[1];
lopt = max(i__1,i__2);
/* RQ factorization of N-by-P matrix B: B = T*Z. */
dgerqf_(n, p, &b[b_offset], ldb, &taub[1], &work[1], lwork, info);
/* Computing MAX */
i__1 = lopt, i__2 = (integer) work[1];
work[1] = (doublereal) max(i__1,i__2);
return 0;
/* End of DGGQRF */
} /* dggqrf_ */
vpMatrix vpMatrix::inverseByQRLapack() const{
int rowNum_ = (int)this->getRows();
int colNum_ = (int)this->getCols();
int lda = (int)rowNum_; //lda is the number of rows because we don't use a submatrix
int dimTau = std::min(rowNum_,colNum_);
int dimWork = -1;
double *tau = new double[dimTau];
double *work = new double[1];
int info;
vpMatrix C;
vpMatrix A = *this;
try{
//1) Extract householder reflections (useful to compute Q) and R
dgeqrf_(
&rowNum_, //The number of rows of the matrix A. M >= 0.
&colNum_, //The number of columns of the matrix A. N >= 0.
A.data, /*On entry, the M-by-N matrix A.
On exit, the elements on and above the diagonal of the array
contain the min(M,N)-by-N upper trapezoidal matrix R (R is
upper triangular if m >= n); the elements below the diagonal,
with the array TAU, represent the orthogonal matrix Q as a
product of min(m,n) elementary reflectors.
*/
&lda, //The leading dimension of the array A. LDA >= max(1,M).
tau, /*Dimension (min(M,N))
The scalar factors of the elementary reflectors
*/
work, //Internal working array. dimension (MAX(1,LWORK))
&dimWork, //The dimension of the array WORK. LWORK >= max(1,N).
&info //status
);
if(info != 0){
std::cout << "dgeqrf_:Preparation:" << -info << "th element had an illegal value" << std::endl;
throw vpMatrixException::badValue;
}
dimWork = allocate_work(&work);
dgeqrf_(
&rowNum_, //The number of rows of the matrix A. M >= 0.
&colNum_, //The number of columns of the matrix A. N >= 0.
A.data, /*On entry, the M-by-N matrix A.
On exit, the elements on and above the diagonal of the array
contain the min(M,N)-by-N upper trapezoidal matrix R (R is
upper triangular if m >= n); the elements below the diagonal,
with the array TAU, represent the orthogonal matrix Q as a
product of min(m,n) elementary reflectors.
*/
&lda, //The leading dimension of the array A. LDA >= max(1,M).
tau, /*Dimension (min(M,N))
The scalar factors of the elementary reflectors
*/
work, //Internal working array. dimension (MAX(1,LWORK))
&dimWork, //The dimension of the array WORK. LWORK >= max(1,N).
&info //status
);
if(info != 0){
std::cout << "dgeqrf_:" << -info << "th element had an illegal value" << std::endl;
throw vpMatrixException::badValue;
}
//A now contains the R matrix in its upper triangular (in lapack convention)
C = A;
//2) Invert R
dtrtri_((char*)"U",(char*)"N",&dimTau,C.data,&lda,&info);
if(info!=0){
if(info < 0)
std::cout << "dtrtri_:"<< -info << "th element had an illegal value" << std::endl;
else if(info > 0){
std::cout << "dtrtri_:R("<< info << "," <<info << ")"<< " is exactly zero. The triangular matrix is singular and its inverse can not be computed." << std::endl;
std::cout << "R=" << std::endl << C << std::endl;
}
throw vpMatrixException::badValue;
}
//3) Zero-fill R^-1
//the matrix is upper triangular for lapack but lower triangular for visp
//we fill it with zeros above the diagonal (where we don't need the values)
for(unsigned int i=0;i<C.getRows();i++)
for(unsigned int j=0;j<C.getRows();j++)
if(j>i) C[i][j] = 0.;
dimWork = -1;
int ldc = lda;
//4) Transpose Q and left-multiply it by R^-1
//get R^-1*tQ
//C contains R^-1
//A contains Q
dormqr_((char*)"R", (char*)"T", &rowNum_, &colNum_, &dimTau, A.data, &lda, tau, C.data, &ldc, work, &dimWork, &info);
if(info != 0){
std::cout << "dormqr_:Preparation"<< -info << "th element had an illegal value" << std::endl;
throw vpMatrixException::badValue;
}
dimWork = allocate_work(&work);
dormqr_((char*)"R", (char*)"T", &rowNum_, &colNum_, &dimTau, A.data, &lda, tau, C.data, &ldc, work, &dimWork, &info);
if(info != 0){
std::cout << "dormqr_:"<< -info << "th element had an illegal value" << std::endl;
throw vpMatrixException::badValue;
}
delete[] tau;
delete[] work;
}catch(vpMatrixException&){
delete[] tau;
delete[] work;
throw;
}
return C;
}
/* Subroutine */ int dggglm_(integer *n, integer *m, integer *p, doublereal *
a, integer *lda, doublereal *b, integer *ldb, doublereal *d__,
doublereal *x, doublereal *y, doublereal *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
=======
DGGGLM solves a general Gauss-Markov linear model (GLM) problem:
minimize || y ||_2 subject to d = A*x + B*y
x
where A is an N-by-M matrix, B is an N-by-P matrix, and d is a
given N-vector. It is assumed that M <= N <= M+P, and
rank(A) = M and rank( A B ) = N.
Under these assumptions, the constrained equation is always
consistent, and there is a unique solution x and a minimal 2-norm
solution y, which is obtained using a generalized QR factorization
of A and B.
In particular, if matrix B is square nonsingular, then the problem
GLM is equivalent to the following weighted linear least squares
problem
minimize || inv(B)*(d-A*x) ||_2
x
where inv(B) denotes the inverse of B.
Arguments
=========
N (input) INTEGER
The number of rows of the matrices A and B. N >= 0.
M (input) INTEGER
The number of columns of the matrix A. 0 <= M <= N.
P (input) INTEGER
The number of columns of the matrix B. P >= N-M.
A (input/output) DOUBLE PRECISION array, dimension (LDA,M)
On entry, the N-by-M matrix A.
On exit, A is destroyed.
LDA (input) INTEGER
The leading dimension of the array A. LDA >= max(1,N).
B (input/output) DOUBLE PRECISION array, dimension (LDB,P)
On entry, the N-by-P matrix B.
On exit, B is destroyed.
LDB (input) INTEGER
The leading dimension of the array B. LDB >= max(1,N).
D (input/output) DOUBLE PRECISION array, dimension (N)
On entry, D is the left hand side of the GLM equation.
On exit, D is destroyed.
X (output) DOUBLE PRECISION array, dimension (M)
Y (output) DOUBLE PRECISION array, dimension (P)
On exit, X and Y are the solutions of the GLM problem.
WORK (workspace/output) DOUBLE PRECISION 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,N+M+P).
For optimum performance, LWORK >= M+min(N,P)+max(N,P)*NB,
where NB is an upper bound for the optimal blocksizes for
DGEQRF, SGERQF, DORMQR and SORMRQ.
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.
===================================================================
Test the input parameters
Parameter adjustments */
/* Table of constant values */
static integer c__1 = 1;
static integer c_n1 = -1;
static doublereal c_b32 = -1.;
static doublereal c_b34 = 1.;
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, i__1, i__2, i__3, i__4;
/* Local variables */
static integer lopt, i__;
extern /* Subroutine */ int dgemv_(char *, integer *, integer *,
doublereal *, doublereal *, integer *, doublereal *, integer *,
doublereal *, doublereal *, integer *), dcopy_(integer *,
doublereal *, integer *, doublereal *, integer *), dtrsv_(char *,
char *, char *, integer *, doublereal *, integer *, doublereal *,
integer *);
static integer nb, np;
extern /* Subroutine */ int dggqrf_(integer *, integer *, integer *,
doublereal *, integer *, doublereal *, doublereal *, integer *,
doublereal *, doublereal *, integer *, integer *), xerbla_(char *,
integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *, ftnlen, ftnlen);
static integer nb1, nb2, nb3, nb4;
extern /* Subroutine */ int dormqr_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, doublereal *,
integer *, doublereal *, integer *, integer *),
dormrq_(char *, char *, integer *, integer *, integer *,
doublereal *, integer *, doublereal *, doublereal *, integer *,
doublereal *, integer *, integer *);
static integer lwkopt;
static logical lquery;
#define b_ref(a_1,a_2) b[(a_2)*b_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;
--d__;
--x;
--y;
--work;
/* Function Body */
*info = 0;
np = min(*n,*p);
nb1 = ilaenv_(&c__1, "DGEQRF", " ", n, m, &c_n1, &c_n1, (ftnlen)6, (
ftnlen)1);
nb2 = ilaenv_(&c__1, "DGERQF", " ", n, m, &c_n1, &c_n1, (ftnlen)6, (
ftnlen)1);
nb3 = ilaenv_(&c__1, "DORMQR", " ", n, m, p, &c_n1, (ftnlen)6, (ftnlen)1);
nb4 = ilaenv_(&c__1, "DORMRQ", " ", n, m, p, &c_n1, (ftnlen)6, (ftnlen)1);
/* Computing MAX */
i__1 = max(nb1,nb2), i__1 = max(i__1,nb3);
nb = max(i__1,nb4);
lwkopt = *m + np + max(*n,*p) * nb;
work[1] = (doublereal) lwkopt;
lquery = *lwork == -1;
if (*n < 0) {
*info = -1;
} else if (*m < 0 || *m > *n) {
*info = -2;
} else if (*p < 0 || *p < *n - *m) {
*info = -3;
} else if (*lda < max(1,*n)) {
*info = -5;
} else if (*ldb < max(1,*n)) {
*info = -7;
} else /* if(complicated condition) */ {
/* Computing MAX */
i__1 = 1, i__2 = *n + *m + *p;
if (*lwork < max(i__1,i__2) && ! lquery) {
*info = -12;
}
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DGGGLM", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
if (*n == 0) {
return 0;
}
/* Compute the GQR factorization of matrices A and B:
Q'*A = ( R11 ) M, Q'*B*Z' = ( T11 T12 ) M
( 0 ) N-M ( 0 T22 ) N-M
M M+P-N N-M
where R11 and T22 are upper triangular, and Q and Z are
orthogonal. */
i__1 = *lwork - *m - np;
dggqrf_(n, m, p, &a[a_offset], lda, &work[1], &b[b_offset], ldb, &work[*m
+ 1], &work[*m + np + 1], &i__1, info);
lopt = (integer) work[*m + np + 1];
/* Update left-hand-side vector d = Q'*d = ( d1 ) M
( d2 ) N-M */
i__1 = max(1,*n);
i__2 = *lwork - *m - np;
dormqr_("Left", "Transpose", n, &c__1, m, &a[a_offset], lda, &work[1], &
d__[1], &i__1, &work[*m + np + 1], &i__2, info);
/* Computing MAX */
i__1 = lopt, i__2 = (integer) work[*m + np + 1];
lopt = max(i__1,i__2);
/* Solve T22*y2 = d2 for y2 */
i__1 = *n - *m;
dtrsv_("Upper", "No transpose", "Non unit", &i__1, &b_ref(*m + 1, *m + *p
- *n + 1), ldb, &d__[*m + 1], &c__1);
i__1 = *n - *m;
dcopy_(&i__1, &d__[*m + 1], &c__1, &y[*m + *p - *n + 1], &c__1);
/* Set y1 = 0 */
i__1 = *m + *p - *n;
for (i__ = 1; i__ <= i__1; ++i__) {
y[i__] = 0.;
/* L10: */
}
/* Update d1 = d1 - T12*y2 */
i__1 = *n - *m;
dgemv_("No transpose", m, &i__1, &c_b32, &b_ref(1, *m + *p - *n + 1), ldb,
&y[*m + *p - *n + 1], &c__1, &c_b34, &d__[1], &c__1);
/* Solve triangular system: R11*x = d1 */
dtrsv_("Upper", "No Transpose", "Non unit", m, &a[a_offset], lda, &d__[1],
&c__1);
/* Copy D to X */
dcopy_(m, &d__[1], &c__1, &x[1], &c__1);
/* Backward transformation y = Z'*y
Computing MAX */
i__1 = 1, i__2 = *n - *p + 1;
i__3 = max(1,*p);
i__4 = *lwork - *m - np;
dormrq_("Left", "Transpose", p, &c__1, &np, &b_ref(max(i__1,i__2), 1),
ldb, &work[*m + 1], &y[1], &i__3, &work[*m + np + 1], &i__4, info);
/* Computing MAX */
i__1 = lopt, i__2 = (integer) work[*m + np + 1];
work[1] = (doublereal) (*m + np + max(i__1,i__2));
return 0;
/* End of DGGGLM */
} /* dggglm_ */
/* Subroutine */ int dggevx_(char *balanc, char *jobvl, char *jobvr, char *
sense, integer *n, doublereal *a, integer *lda, doublereal *b,
integer *ldb, doublereal *alphar, doublereal *alphai, doublereal *
beta, doublereal *vl, integer *ldvl, doublereal *vr, integer *ldvr,
integer *ilo, integer *ihi, doublereal *lscale, doublereal *rscale,
doublereal *abnrm, doublereal *bbnrm, doublereal *rconde, doublereal *
rcondv, doublereal *work, integer *lwork, integer *iwork, 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
=======
DGGEVX computes for a pair of N-by-N real nonsymmetric matrices (A,B)
the generalized eigenvalues, and optionally, the left and/or right
generalized eigenvectors.
Optionally also, it computes a balancing transformation to improve
the conditioning of the eigenvalues and eigenvectors (ILO, IHI,
LSCALE, RSCALE, ABNRM, and BBNRM), reciprocal condition numbers for
the eigenvalues (RCONDE), and reciprocal condition numbers for the
right eigenvectors (RCONDV).
A generalized eigenvalue for a pair of matrices (A,B) is a scalar
lambda or a ratio alpha/beta = lambda, such that A - lambda*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.
The right eigenvector v(j) corresponding to the eigenvalue lambda(j)
of (A,B) satisfies
A * v(j) = lambda(j) * B * v(j) .
The left eigenvector u(j) corresponding to the eigenvalue lambda(j)
of (A,B) satisfies
u(j)**H * A = lambda(j) * u(j)**H * B.
where u(j)**H is the conjugate-transpose of u(j).
Arguments
=========
BALANC (input) CHARACTER*1
Specifies the balance option to be performed.
= 'N': do not diagonally scale or permute;
= 'P': permute only;
= 'S': scale only;
= 'B': both permute and scale.
Computed reciprocal condition numbers will be for the
matrices after permuting and/or balancing. Permuting does
not change condition numbers (in exact arithmetic), but
balancing does.
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.
SENSE (input) CHARACTER*1
Determines which reciprocal condition numbers are computed.
= 'N': none are computed;
= 'E': computed for eigenvalues only;
= 'V': computed for eigenvectors only;
= 'B': computed for eigenvalues and eigenvectors.
N (input) INTEGER
The order of the matrices A, B, VL, and VR. N >= 0.
A (input/output) DOUBLE PRECISION array, dimension (LDA, N)
On entry, the matrix A in the pair (A,B).
On exit, A has been overwritten. If JOBVL='V' or JOBVR='V'
or both, then A contains the first part of the real Schur
form of the "balanced" versions of the input A and B.
LDA (input) INTEGER
The leading dimension of A. LDA >= max(1,N).
B (input/output) DOUBLE PRECISION array, dimension (LDB, N)
On entry, the matrix B in the pair (A,B).
On exit, B has been overwritten. If JOBVL='V' or JOBVR='V'
or both, then B contains the second part of the real Schur
form of the "balanced" versions of the input A and B.
LDB (input) INTEGER
The leading dimension of B. LDB >= max(1,N).
ALPHAR (output) DOUBLE PRECISION array, dimension (N)
ALPHAI (output) DOUBLE PRECISION array, dimension (N)
BETA (output) DOUBLE PRECISION 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) DOUBLE PRECISION 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 the j-th eigenvalue is real, then
u(j) = VL(:,j), the j-th column of VL. If the j-th and
(j+1)-th 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).
Each eigenvector will be scaled so the largest component have
abs(real part) + abs(imag. part) = 1.
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) DOUBLE PRECISION 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 the j-th eigenvalue is real, then
v(j) = VR(:,j), the j-th column of VR. If the j-th and
(j+1)-th 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).
Each eigenvector will be scaled so the largest component have
abs(real part) + abs(imag. part) = 1.
Not referenced if JOBVR = 'N'.
LDVR (input) INTEGER
The leading dimension of the matrix VR. LDVR >= 1, and
if JOBVR = 'V', LDVR >= N.
ILO,IHI (output) INTEGER
ILO and IHI are integer values such that on exit
A(i,j) = 0 and B(i,j) = 0 if i > j and
j = 1,...,ILO-1 or i = IHI+1,...,N.
If BALANC = 'N' or 'S', ILO = 1 and IHI = N.
LSCALE (output) DOUBLE PRECISION array, dimension (N)
Details of the permutations and scaling factors applied
to the left side of A and B. If PL(j) is the index of the
row interchanged with row j, and DL(j) is the scaling
factor applied to row j, then
LSCALE(j) = PL(j) for j = 1,...,ILO-1
= DL(j) for j = ILO,...,IHI
= PL(j) for j = IHI+1,...,N.
The order in which the interchanges are made is N to IHI+1,
then 1 to ILO-1.
RSCALE (output) DOUBLE PRECISION array, dimension (N)
Details of the permutations and scaling factors applied
to the right side of A and B. If PR(j) is the index of the
column interchanged with column j, and DR(j) is the scaling
factor applied to column j, then
RSCALE(j) = PR(j) for j = 1,...,ILO-1
= DR(j) for j = ILO,...,IHI
= PR(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) DOUBLE PRECISION
The one-norm of the balanced matrix A.
BBNRM (output) DOUBLE PRECISION
The one-norm of the balanced matrix B.
RCONDE (output) DOUBLE PRECISION array, dimension (N)
If SENSE = 'E' or 'B', the reciprocal condition numbers of
the selected eigenvalues, stored in consecutive elements of
the array. For a complex conjugate pair of eigenvalues two
consecutive elements of RCONDE are set to the same value.
Thus RCONDE(j), RCONDV(j), and the j-th columns of VL and VR
all correspond to the same eigenpair (but not in general the
j-th eigenpair, unless all eigenpairs are selected).
If SENSE = 'V', RCONDE is not referenced.
RCONDV (output) DOUBLE PRECISION array, dimension (N)
If SENSE = 'V' or 'B', the estimated reciprocal condition
numbers of the selected eigenvectors, stored in consecutive
elements of the array. For a complex eigenvector two
consecutive elements of RCONDV are set to the same value. If
the eigenvalues cannot be reordered to compute RCONDV(j),
RCONDV(j) is set to 0; this can only occur when the true
value would be very small anyway.
If SENSE = 'E', RCONDV is not referenced.
WORK (workspace/output) DOUBLE PRECISION 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(1,6*N).
If SENSE = 'E', LWORK >= 12*N.
If SENSE = 'V' or 'B', LWORK >= 2*N*N+12*N+16.
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 (N+6)
If SENSE = 'E', IWORK is not referenced.
BWORK (workspace) LOGICAL array, dimension (N)
If SENSE = 'N', BWORK is not referenced.
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: =N+1: other than QZ iteration failed in DHGEQZ.
=N+2: error return from DTGEVC.
Further Details
===============
Balancing a matrix pair (A,B) includes, first, permuting rows and
columns to isolate eigenvalues, second, applying diagonal similarity
transformation to the rows and columns to make the rows and columns
as close in norm as possible. 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 4.11.1.2 of LAPACK Users' Guide.
An approximate error bound on the chordal distance between the i-th
computed generalized eigenvalue w and the corresponding exact
eigenvalue lambda is
chord(w, lambda) <= EPS * norm(ABNRM, BBNRM) / RCONDE(I)
An approximate error bound for the angle between the i-th computed
eigenvector VL(i) or VR(i) is given by
EPS * norm(ABNRM, BBNRM) / DIF(i).
For further explanation of the reciprocal condition numbers RCONDE
and RCONDV, see section 4.11 of LAPACK User's Guide.
=====================================================================
Decode the input arguments
Parameter adjustments */
/* Table of constant values */
static integer c__1 = 1;
static integer c__0 = 0;
static doublereal c_b47 = 0.;
static doublereal c_b48 = 1.;
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, vl_dim1, vl_offset, vr_dim1,
vr_offset, i__1, i__2;
doublereal d__1, d__2, d__3, d__4;
/* Builtin functions */
double sqrt(doublereal);
/* Local variables */
static logical pair;
static doublereal anrm, bnrm;
static integer ierr, itau;
static doublereal temp;
static logical ilvl, ilvr;
static integer iwrk, iwrk1, i__, j, m;
extern logical lsame_(char *, char *);
static integer icols, irows;
extern /* Subroutine */ int dlabad_(doublereal *, doublereal *);
static integer jc;
extern /* Subroutine */ int dggbak_(char *, char *, integer *, integer *,
integer *, doublereal *, doublereal *, integer *, doublereal *,
integer *, integer *), dggbal_(char *, integer *,
doublereal *, integer *, doublereal *, integer *, integer *,
integer *, doublereal *, doublereal *, doublereal *, integer *);
static integer in;
extern doublereal dlamch_(char *);
static integer mm;
extern doublereal dlange_(char *, integer *, integer *, doublereal *,
integer *, doublereal *);
static integer jr;
extern /* Subroutine */ int dgghrd_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, integer *,
doublereal *, integer *, doublereal *, integer *, integer *), dlascl_(char *, integer *, integer *, doublereal
*, doublereal *, integer *, integer *, doublereal *, integer *,
integer *);
static logical ilascl, ilbscl;
extern /* Subroutine */ int dgeqrf_(integer *, integer *, doublereal *,
integer *, doublereal *, doublereal *, integer *, integer *),
dlacpy_(char *, integer *, integer *, doublereal *, integer *,
doublereal *, integer *);
static logical ldumma[1];
static char chtemp[1];
static doublereal bignum;
extern /* Subroutine */ int dhgeqz_(char *, char *, char *, integer *,
integer *, integer *, doublereal *, integer *, doublereal *,
integer *, doublereal *, doublereal *, doublereal *, doublereal *,
integer *, doublereal *, integer *, doublereal *, integer *,
integer *), dlaset_(char *, integer *,
integer *, doublereal *, doublereal *, doublereal *, integer *);
static integer ijobvl;
extern /* Subroutine */ int dtgevc_(char *, char *, logical *, integer *,
doublereal *, integer *, doublereal *, integer *, doublereal *,
integer *, doublereal *, integer *, integer *, integer *,
doublereal *, integer *), dtgsna_(char *, char *,
logical *, integer *, doublereal *, integer *, doublereal *,
integer *, doublereal *, integer *, doublereal *, integer *,
doublereal *, doublereal *, integer *, integer *, doublereal *,
integer *, integer *, integer *), xerbla_(char *,
integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *, ftnlen, ftnlen);
static integer ijobvr;
static logical wantsb;
extern /* Subroutine */ int dorgqr_(integer *, integer *, integer *,
doublereal *, integer *, doublereal *, doublereal *, integer *,
integer *);
static doublereal anrmto;
static logical wantse;
static doublereal bnrmto;
extern /* Subroutine */ int dormqr_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, doublereal *,
integer *, doublereal *, integer *, integer *);
static integer minwrk, maxwrk;
static logical wantsn;
static doublereal smlnum;
static logical lquery, wantsv;
static doublereal 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;
--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;
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 (! (lsame_(balanc, "N") || lsame_(balanc, "S") || lsame_(balanc, "P")
|| 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.) */
minwrk = 1;
if (*info == 0 && (*lwork >= 1 || lquery)) {
maxwrk = *n * 5 + *n * ilaenv_(&c__1, "DGEQRF", " ", n, &c__1, n, &
c__0, (ftnlen)6, (ftnlen)1);
/* Computing MAX */
i__1 = 1, i__2 = *n * 6;
minwrk = max(i__1,i__2);
if (wantse) {
/* Computing MAX */
i__1 = 1, i__2 = *n * 12;
minwrk = max(i__1,i__2);
} else if (wantsv || wantsb) {
minwrk = (*n << 1) * *n + *n * 12 + 16;
/* Computing MAX */
i__1 = maxwrk, i__2 = (*n << 1) * *n + *n * 12 + 16;
maxwrk = max(i__1,i__2);
}
work[1] = (doublereal) maxwrk;
}
if (*lwork < minwrk && ! lquery) {
*info = -26;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DGGEVX", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
if (*n == 0) {
return 0;
}
/* Get machine constants */
eps = dlamch_("P");
smlnum = dlamch_("S");
bignum = 1. / smlnum;
dlabad_(&smlnum, &bignum);
smlnum = sqrt(smlnum) / eps;
bignum = 1. / smlnum;
/* Scale A if max element outside range [SMLNUM,BIGNUM] */
anrm = dlange_("M", n, n, &a[a_offset], lda, &work[1]);
ilascl = FALSE_;
if (anrm > 0. && anrm < smlnum) {
anrmto = smlnum;
ilascl = TRUE_;
} else if (anrm > bignum) {
anrmto = bignum;
ilascl = TRUE_;
}
if (ilascl) {
dlascl_("G", &c__0, &c__0, &anrm, &anrmto, n, n, &a[a_offset], lda, &
ierr);
}
/* Scale B if max element outside range [SMLNUM,BIGNUM] */
bnrm = dlange_("M", n, n, &b[b_offset], ldb, &work[1]);
ilbscl = FALSE_;
if (bnrm > 0. && bnrm < smlnum) {
bnrmto = smlnum;
ilbscl = TRUE_;
} else if (bnrm > bignum) {
bnrmto = bignum;
ilbscl = TRUE_;
}
if (ilbscl) {
dlascl_("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) */
dggbal_(balanc, n, &a[a_offset], lda, &b[b_offset], ldb, ilo, ihi, &
lscale[1], &rscale[1], &work[1], &ierr);
/* Compute ABNRM and BBNRM */
*abnrm = dlange_("1", n, n, &a[a_offset], lda, &work[1]);
if (ilascl) {
work[1] = *abnrm;
dlascl_("G", &c__0, &c__0, &anrmto, &anrm, &c__1, &c__1, &work[1], &
c__1, &ierr);
*abnrm = work[1];
}
*bbnrm = dlange_("1", n, n, &b[b_offset], ldb, &work[1]);
if (ilbscl) {
work[1] = *bbnrm;
dlascl_("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;
dgeqrf_(&irows, &icols, &b_ref(*ilo, *ilo), 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;
dormqr_("L", "T", &irows, &icols, &irows, &b_ref(*ilo, *ilo), ldb, &work[
itau], &a_ref(*ilo, *ilo), lda, &work[iwrk], &i__1, &ierr);
/* Initialize VL and/or VR
(Workspace: need N, prefer N*NB) */
if (ilvl) {
dlaset_("Full", n, n, &c_b47, &c_b48, &vl[vl_offset], ldvl)
;
i__1 = irows - 1;
i__2 = irows - 1;
dlacpy_("L", &i__1, &i__2, &b_ref(*ilo + 1, *ilo), ldb, &vl_ref(*ilo
+ 1, *ilo), ldvl);
i__1 = *lwork + 1 - iwrk;
dorgqr_(&irows, &irows, &irows, &vl_ref(*ilo, *ilo), ldvl, &work[itau]
, &work[iwrk], &i__1, &ierr);
}
if (ilvr) {
dlaset_("Full", n, n, &c_b47, &c_b48, &vr[vr_offset], ldvr)
;
}
/* Reduce to generalized Hessenberg form
(Workspace: none needed) */
if (ilv || ! wantsn) {
/* Eigenvectors requested -- work on whole matrix. */
dgghrd_(jobvl, jobvr, n, ilo, ihi, &a[a_offset], lda, &b[b_offset],
ldb, &vl[vl_offset], ldvl, &vr[vr_offset], ldvr, &ierr);
} else {
dgghrd_("N", "N", &irows, &c__1, &irows, &a_ref(*ilo, *ilo), lda, &
b_ref(*ilo, *ilo), 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';
}
dhgeqz_(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: DTGEVC: need 6*N
DTGSNA: 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';
}
dtgevc_(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 (DTGEVC) and estimate condition
numbers (DTGSNA). 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_ref(i__ + 1, i__) != 0.) {
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) {
dtgevc_("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;
dtgsna_(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) {
dggbak_(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.) {
goto L70;
}
temp = 0.;
if (alphai[jc] == 0.) {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
/* Computing MAX */
d__2 = temp, d__3 = (d__1 = vl_ref(jr, jc), abs(d__1));
temp = max(d__2,d__3);
/* L30: */
}
} else {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
/* Computing MAX */
d__3 = temp, d__4 = (d__1 = vl_ref(jr, jc), abs(d__1)) + (
d__2 = vl_ref(jr, jc + 1), abs(d__2));
temp = max(d__3,d__4);
/* L40: */
}
}
if (temp < smlnum) {
goto L70;
}
temp = 1. / temp;
if (alphai[jc] == 0.) {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
vl_ref(jr, jc) = vl_ref(jr, jc) * temp;
/* L50: */
}
} 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;
/* L60: */
}
}
L70:
;
}
}
if (ilvr) {
dggbak_(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.) {
goto L120;
}
temp = 0.;
if (alphai[jc] == 0.) {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
/* Computing MAX */
d__2 = temp, d__3 = (d__1 = vr_ref(jr, jc), abs(d__1));
temp = max(d__2,d__3);
/* L80: */
}
} else {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
/* Computing MAX */
d__3 = temp, d__4 = (d__1 = vr_ref(jr, jc), abs(d__1)) + (
d__2 = vr_ref(jr, jc + 1), abs(d__2));
temp = max(d__3,d__4);
/* L90: */
}
}
if (temp < smlnum) {
goto L120;
}
temp = 1. / temp;
if (alphai[jc] == 0.) {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
vr_ref(jr, jc) = vr_ref(jr, jc) * temp;
/* L100: */
}
} 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;
/* L110: */
}
}
L120:
;
}
}
/* Undo scaling if necessary */
if (ilascl) {
dlascl_("G", &c__0, &c__0, &anrmto, &anrm, n, &c__1, &alphar[1], n, &
ierr);
dlascl_("G", &c__0, &c__0, &anrmto, &anrm, n, &c__1, &alphai[1], n, &
ierr);
}
if (ilbscl) {
dlascl_("G", &c__0, &c__0, &bnrmto, &bnrm, n, &c__1, &beta[1], n, &
ierr);
}
L130:
work[1] = (doublereal) maxwrk;
return 0;
/* End of DGGEVX */
} /* dggevx_ */
/* Subroutine */ int dgeqrs_(integer *m, integer *n, integer *nrhs,
doublereal *a, integer *lda, doublereal *tau, doublereal *b, integer *
ldb, doublereal *work, integer *lwork, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, i__1;
/* Local variables */
extern /* Subroutine */ int dtrsm_(char *, char *, char *, char *,
integer *, integer *, doublereal *, doublereal *, integer *,
doublereal *, integer *), xerbla_(
char *, integer *), dormqr_(char *, char *, integer *,
integer *, integer *, doublereal *, integer *, doublereal *,
doublereal *, integer *, doublereal *, integer *, integer *);
/* -- LAPACK routine (version 3.1) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* Solve the least squares problem */
/* min || A*X - B || */
/* using the QR factorization */
/* A = Q*R */
/* computed by DGEQRF. */
/* 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. M >= N >= 0. */
/* NRHS (input) INTEGER */
/* The number of columns of B. NRHS >= 0. */
/* A (input) DOUBLE PRECISION array, dimension (LDA,N) */
/* Details of the QR factorization of the original matrix A as */
/* returned by DGEQRF. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= M. */
/* TAU (input) DOUBLE PRECISION array, dimension (N) */
/* Details of the orthogonal matrix Q. */
/* B (input/output) DOUBLE PRECISION array, dimension (LDB,NRHS) */
/* On entry, the m-by-nrhs right hand side matrix B. */
/* On exit, the n-by-nrhs solution matrix X. */
/* LDB (input) INTEGER */
/* The leading dimension of the array B. LDB >= M. */
/* WORK (workspace) DOUBLE PRECISION array, dimension (LWORK) */
/* LWORK (input) INTEGER */
/* The length of the array WORK. LWORK must be at least NRHS, */
/* and should be at least NRHS*NB, where NB is the block size */
/* for this environment. */
/* INFO (output) INTEGER */
/* = 0: successful exit */
/* < 0: if INFO = -i, the i-th argument had an illegal value */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Test the input arguments. */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
--tau;
b_dim1 = *ldb;
b_offset = 1 + b_dim1;
b -= b_offset;
--work;
/* Function Body */
*info = 0;
if (*m < 0) {
*info = -1;
} else if (*n < 0 || *n > *m) {
*info = -2;
} else if (*nrhs < 0) {
*info = -3;
} else if (*lda < max(1,*m)) {
*info = -5;
} else if (*ldb < max(1,*m)) {
*info = -8;
} else if (*lwork < 1 || *lwork < *nrhs && *m > 0 && *n > 0) {
*info = -10;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DGEQRS", &i__1);
return 0;
}
/* Quick return if possible */
if (*n == 0 || *nrhs == 0 || *m == 0) {
return 0;
}
/* B := Q' * B */
dormqr_("Left", "Transpose", m, nrhs, n, &a[a_offset], lda, &tau[1], &b[
b_offset], ldb, &work[1], lwork, info);
/* Solve R*X = B(1:n,:) */
dtrsm_("Left", "Upper", "No transpose", "Non-unit", n, nrhs, &c_b9, &a[
a_offset], lda, &b[b_offset], ldb);
return 0;
/* End of DGEQRS */
} /* dgeqrs_ */
/* Subroutine */ int dgels_(char *trans, integer *m, integer *n, integer *
nrhs, doublereal *a, integer *lda, doublereal *b, integer *ldb,
doublereal *work, integer *lwork, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, i__1, i__2;
/* Local variables */
integer i__, j, nb, mn;
doublereal anrm, bnrm;
integer brow;
logical tpsd;
integer iascl, ibscl;
extern logical lsame_(char *, char *);
integer wsize;
doublereal rwork[1];
extern /* Subroutine */ int dlabad_(doublereal *, doublereal *);
extern doublereal dlamch_(char *), dlange_(char *, integer *,
integer *, doublereal *, integer *, doublereal *);
extern /* Subroutine */ int dgelqf_(integer *, integer *, doublereal *,
integer *, doublereal *, doublereal *, integer *, integer *),
dlascl_(char *, integer *, integer *, doublereal *, doublereal *,
integer *, integer *, doublereal *, integer *, integer *),
dgeqrf_(integer *, integer *, doublereal *, integer *,
doublereal *, doublereal *, integer *, integer *), dlaset_(char *,
integer *, integer *, doublereal *, doublereal *, doublereal *,
integer *), xerbla_(char *, integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *);
integer scllen;
doublereal bignum;
extern /* Subroutine */ int dormlq_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, doublereal *,
integer *, doublereal *, integer *, integer *),
dormqr_(char *, char *, integer *, integer *, integer *,
doublereal *, integer *, doublereal *, doublereal *, integer *,
doublereal *, integer *, integer *);
doublereal smlnum;
logical lquery;
extern /* Subroutine */ int dtrtrs_(char *, char *, char *, integer *,
integer *, doublereal *, integer *, doublereal *, 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 */
/* ======= */
/* DGELS 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*1 */
/* = '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) DOUBLE PRECISION 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 DGEQRF; */
/* if M < N, A is overwritten by details of its LQ */
/* factorization as returned by DGELQF. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,M). */
/* B (input/output) DOUBLE PRECISION 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, if INFO = 0, 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) DOUBLE PRECISION 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, 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 */
/* > 0: if INFO = i, the i-th diagonal element of the */
/* triangular factor of A is zero, so that A does not have */
/* full rank; the least squares solution could not be */
/* computed. */
/* ===================================================================== */
/* .. 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") || lsame_(trans, "T"))) {
*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")) {
tpsd = FALSE_;
}
if (*m >= *n) {
nb = ilaenv_(&c__1, "DGEQRF", " ", m, n, &c_n1, &c_n1);
if (tpsd) {
/* Computing MAX */
i__1 = nb, i__2 = ilaenv_(&c__1, "DORMQR", "LN", m, nrhs, n, &
c_n1);
nb = max(i__1,i__2);
} else {
/* Computing MAX */
i__1 = nb, i__2 = ilaenv_(&c__1, "DORMQR", "LT", m, nrhs, n, &
c_n1);
nb = max(i__1,i__2);
}
} else {
nb = ilaenv_(&c__1, "DGELQF", " ", m, n, &c_n1, &c_n1);
if (tpsd) {
/* Computing MAX */
i__1 = nb, i__2 = ilaenv_(&c__1, "DORMLQ", "LT", n, nrhs, m, &
c_n1);
nb = max(i__1,i__2);
} else {
/* Computing MAX */
i__1 = nb, i__2 = ilaenv_(&c__1, "DORMLQ", "LN", n, nrhs, m, &
c_n1);
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] = (doublereal) wsize;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DGELS ", &i__1);
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);
dlaset_("Full", &i__1, nrhs, &c_b33, &c_b33, &b[b_offset], ldb);
return 0;
}
/* Get machine parameters */
smlnum = dlamch_("S") / dlamch_("P");
bignum = 1. / smlnum;
dlabad_(&smlnum, &bignum);
/* Scale A, B if max element outside range [SMLNUM,BIGNUM] */
anrm = dlange_("M", m, n, &a[a_offset], lda, rwork);
iascl = 0;
if (anrm > 0. && anrm < smlnum) {
/* Scale matrix norm up to SMLNUM */
dlascl_("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 */
dlascl_("G", &c__0, &c__0, &anrm, &bignum, m, n, &a[a_offset], lda,
info);
iascl = 2;
} else if (anrm == 0.) {
/* Matrix all zero. Return zero solution. */
i__1 = max(*m,*n);
dlaset_("F", &i__1, nrhs, &c_b33, &c_b33, &b[b_offset], ldb);
goto L50;
}
brow = *m;
if (tpsd) {
brow = *n;
}
bnrm = dlange_("M", &brow, nrhs, &b[b_offset], ldb, rwork);
ibscl = 0;
if (bnrm > 0. && bnrm < smlnum) {
/* Scale matrix norm up to SMLNUM */
dlascl_("G", &c__0, &c__0, &bnrm, &smlnum, &brow, nrhs, &b[b_offset],
ldb, info);
ibscl = 1;
} else if (bnrm > bignum) {
/* Scale matrix norm down to BIGNUM */
dlascl_("G", &c__0, &c__0, &bnrm, &bignum, &brow, nrhs, &b[b_offset],
ldb, info);
ibscl = 2;
}
if (*m >= *n) {
/* compute QR factorization of A */
i__1 = *lwork - mn;
dgeqrf_(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;
dormqr_("Left", "Transpose", m, nrhs, n, &a[a_offset], lda, &work[
1], &b[b_offset], ldb, &work[mn + 1], &i__1, info);
/* workspace at least NRHS, optimally NRHS*NB */
/* B(1:N,1:NRHS) := inv(R) * B(1:N,1:NRHS) */
dtrtrs_("Upper", "No transpose", "Non-unit", n, nrhs, &a[a_offset]
, lda, &b[b_offset], ldb, info);
if (*info > 0) {
return 0;
}
scllen = *n;
} else {
/* Overdetermined system of equations A' * X = B */
/* B(1:N,1:NRHS) := inv(R') * B(1:N,1:NRHS) */
dtrtrs_("Upper", "Transpose", "Non-unit", n, nrhs, &a[a_offset],
lda, &b[b_offset], ldb, info);
if (*info > 0) {
return 0;
}
/* 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.;
/* L10: */
}
/* L20: */
}
/* B(1:M,1:NRHS) := Q(1:N,:) * B(1:N,1:NRHS) */
i__1 = *lwork - mn;
dormqr_("Left", "No transpose", m, nrhs, n, &a[a_offset], lda, &
work[1], &b[b_offset], ldb, &work[mn + 1], &i__1, info);
/* workspace at least NRHS, optimally NRHS*NB */
scllen = *m;
}
} else {
/* Compute LQ factorization of A */
i__1 = *lwork - mn;
dgelqf_(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) */
dtrtrs_("Lower", "No transpose", "Non-unit", m, nrhs, &a[a_offset]
, lda, &b[b_offset], ldb, info);
if (*info > 0) {
return 0;
}
/* 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.;
/* L30: */
}
/* L40: */
}
/* B(1:N,1:NRHS) := Q(1:N,:)' * B(1:M,1:NRHS) */
i__1 = *lwork - mn;
dormlq_("Left", "Transpose", n, nrhs, m, &a[a_offset], lda, &work[
1], &b[b_offset], ldb, &work[mn + 1], &i__1, info);
/* 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;
dormlq_("Left", "No transpose", n, nrhs, m, &a[a_offset], lda, &
work[1], &b[b_offset], ldb, &work[mn + 1], &i__1, info);
/* workspace at least NRHS, optimally NRHS*NB */
/* B(1:M,1:NRHS) := inv(L') * B(1:M,1:NRHS) */
dtrtrs_("Lower", "Transpose", "Non-unit", m, nrhs, &a[a_offset],
lda, &b[b_offset], ldb, info);
if (*info > 0) {
return 0;
}
scllen = *m;
}
}
/* Undo scaling */
if (iascl == 1) {
dlascl_("G", &c__0, &c__0, &anrm, &smlnum, &scllen, nrhs, &b[b_offset]
, ldb, info);
} else if (iascl == 2) {
dlascl_("G", &c__0, &c__0, &anrm, &bignum, &scllen, nrhs, &b[b_offset]
, ldb, info);
}
if (ibscl == 1) {
dlascl_("G", &c__0, &c__0, &smlnum, &bnrm, &scllen, nrhs, &b[b_offset]
, ldb, info);
} else if (ibscl == 2) {
dlascl_("G", &c__0, &c__0, &bignum, &bnrm, &scllen, nrhs, &b[b_offset]
, ldb, info);
}
L50:
work[1] = (doublereal) wsize;
return 0;
/* End of DGELS */
} /* dgels_ */
/* Subroutine */ int dgges_(char *jobvsl, char *jobvsr, char *sort, L_fp
selctg, integer *n, doublereal *a, integer *lda, doublereal *b,
integer *ldb, integer *sdim, doublereal *alphar, doublereal *alphai,
doublereal *beta, doublereal *vsl, integer *ldvsl, doublereal *vsr,
integer *ldvsr, doublereal *work, integer *lwork, logical *bwork,
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;
doublereal d__1;
/* Local variables */
integer i__, ip;
doublereal dif[2];
integer ihi, ilo;
doublereal eps, anrm, bnrm;
integer idum[1], ierr, itau, iwrk;
doublereal pvsl, pvsr;
integer ileft, icols;
logical cursl, ilvsl, ilvsr;
integer irows;
logical lst2sl;
logical ilascl, ilbscl;
doublereal safmin;
doublereal safmax;
doublereal bignum;
integer ijobvl, iright;
integer ijobvr;
doublereal anrmto, bnrmto;
logical lastsl;
integer minwrk, maxwrk;
doublereal smlnum;
logical wantst, lquery;
/* -- LAPACK driver routine (version 3.2) -- */
/* November 2006 */
/* Purpose */
/* ======= */
/* DGGES computes for a pair of N-by-N real nonsymmetric matrices (A,B), */
/* the generalized eigenvalues, the generalized real Schur form (S,T), */
/* 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.The */
/* leading columns of VSL and VSR then form an orthonormal basis for the */
/* corresponding left and right eigenspaces (deflating subspaces). */
/* (If only the generalized eigenvalues are needed, use the driver */
/* DGGEV instead, which is faster.) */
/* 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 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 (external procedure) LOGICAL FUNCTION of three DOUBLE PRECISION 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 in the ill-conditioned case, a selected complex */
/* eigenvalue may no longer satisfy SELCTG(ALPHAR(j),ALPHAI(j), */
/* BETA(j)) = .TRUE. after ordering. INFO is to be set to N+2 */
/* in this case. */
/* N (input) INTEGER */
/* The order of the matrices A, B, VSL, and VSR. N >= 0. */
/* A (input/output) DOUBLE PRECISION 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) DOUBLE PRECISION 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) DOUBLE PRECISION array, dimension (N) */
/* ALPHAI (output) DOUBLE PRECISION array, dimension (N) */
/* BETA (output) DOUBLE PRECISION array, dimension (N) */
/* be the generalized eigenvalues. ALPHAR(j) + ALPHAI(j)*i, */
/* 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) DOUBLE PRECISION 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) DOUBLE PRECISION 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. */
/* WORK (workspace/output) DOUBLE PRECISION 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 N = 0, LWORK >= 1, else LWORK >= 8*N+16. */
/* 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. */
/* 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. */
/* The QZ iteration failed. (A,B) are not in Schur */
/* form, but ALPHAR(j), ALPHAI(j), and BETA(j) should */
/* > N: =N+1: other than QZ iteration failed in DHGEQZ. */
/* =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 DTGSEN. */
/* ===================================================================== */
/* 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;
--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");
/* Test the input arguments */
*info = 0;
lquery = *lwork == -1;
if (ijobvl <= 0) {
*info = -1;
} else if (ijobvr <= 0) {
*info = -2;
} else if (! wantst && ! lsame_(sort, "N")) {
*info = -3;
} else if (*n < 0) {
*info = -5;
} else if (*lda < max(1,*n)) {
*info = -7;
} else if (*ldb < max(1,*n)) {
*info = -9;
} else if (*ldvsl < 1 || ilvsl && *ldvsl < *n) {
*info = -15;
} else if (*ldvsr < 1 || ilvsr && *ldvsr < *n) {
*info = -17;
}
/* 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) {
if (*n > 0) {
/* Computing MAX */
i__1 = *n << 3, i__2 = *n * 6 + 16;
minwrk = max(i__1,i__2);
maxwrk = minwrk - *n + *n * ilaenv_(&c__1, "DGEQRF", " ", n, &
c__1, n, &c__0);
/* Computing MAX */
i__1 = maxwrk, i__2 = minwrk - *n + *n * ilaenv_(&c__1, "DORMQR",
" ", n, &c__1, n, &c_n1);
maxwrk = max(i__1,i__2);
if (ilvsl) {
/* Computing MAX */
i__1 = maxwrk, i__2 = minwrk - *n + *n * ilaenv_(&c__1, "DOR"
"GQR", " ", n, &c__1, n, &c_n1);
maxwrk = max(i__1,i__2);
}
} else {
minwrk = 1;
maxwrk = 1;
}
work[1] = (doublereal) maxwrk;
if (*lwork < minwrk && ! lquery) {
*info = -19;
}
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DGGES ", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
if (*n == 0) {
*sdim = 0;
return 0;
}
/* Get machine constants */
eps = dlamch_("P");
safmin = dlamch_("S");
safmax = 1. / safmin;
dlabad_(&safmin, &safmax);
smlnum = sqrt(safmin) / eps;
bignum = 1. / smlnum;
/* Scale A if max element outside range [SMLNUM,BIGNUM] */
anrm = dlange_("M", n, n, &a[a_offset], lda, &work[1]);
ilascl = FALSE_;
if (anrm > 0. && anrm < smlnum) {
anrmto = smlnum;
ilascl = TRUE_;
} else if (anrm > bignum) {
anrmto = bignum;
ilascl = TRUE_;
}
if (ilascl) {
dlascl_("G", &c__0, &c__0, &anrm, &anrmto, n, n, &a[a_offset], lda, &
ierr);
}
/* Scale B if max element outside range [SMLNUM,BIGNUM] */
bnrm = dlange_("M", n, n, &b[b_offset], ldb, &work[1]);
ilbscl = FALSE_;
if (bnrm > 0. && bnrm < smlnum) {
bnrmto = smlnum;
ilbscl = TRUE_;
} else if (bnrm > bignum) {
bnrmto = bignum;
ilbscl = TRUE_;
}
if (ilbscl) {
dlascl_("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 space for storing balancing factors) */
ileft = 1;
iright = *n + 1;
iwrk = iright + *n;
dggbal_("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;
dgeqrf_(&irows, &icols, &b[ilo + ilo * b_dim1], 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;
dormqr_("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 VSL */
/* (Workspace: need N, prefer N*NB) */
if (ilvsl) {
dlaset_("Full", n, n, &c_b38, &c_b39, &vsl[vsl_offset], ldvsl);
if (irows > 1) {
i__1 = irows - 1;
i__2 = irows - 1;
dlacpy_("L", &i__1, &i__2, &b[ilo + 1 + ilo * b_dim1], ldb, &vsl[
ilo + 1 + ilo * vsl_dim1], ldvsl);
}
i__1 = *lwork + 1 - iwrk;
dorgqr_(&irows, &irows, &irows, &vsl[ilo + ilo * vsl_dim1], ldvsl, &
work[itau], &work[iwrk], &i__1, &ierr);
}
/* Initialize VSR */
if (ilvsr) {
dlaset_("Full", n, n, &c_b38, &c_b39, &vsr[vsr_offset], ldvsr);
}
/* Reduce to generalized Hessenberg form */
/* (Workspace: none needed) */
dgghrd_(jobvsl, jobvsr, n, &ilo, &ihi, &a[a_offset], lda, &b[b_offset],
ldb, &vsl[vsl_offset], ldvsl, &vsr[vsr_offset], ldvsr, &ierr);
/* Perform QZ algorithm, computing Schur vectors if desired */
/* (Workspace: need N) */
iwrk = itau;
i__1 = *lwork + 1 - iwrk;
dhgeqz_("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 if desired */
/* (Workspace: need 4*N+16 ) */
*sdim = 0;
if (wantst) {
/* Undo scaling on eigenvalues before SELCTGing */
if (ilascl) {
dlascl_("G", &c__0, &c__0, &anrmto, &anrm, n, &c__1, &alphar[1],
n, &ierr);
dlascl_("G", &c__0, &c__0, &anrmto, &anrm, n, &c__1, &alphai[1],
n, &ierr);
}
if (ilbscl) {
dlascl_("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__]);
}
i__1 = *lwork - iwrk + 1;
dtgsen_(&c__0, &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, &pvsl, &
pvsr, dif, &work[iwrk], &i__1, idum, &c__1, &ierr);
if (ierr == 1) {
*info = *n + 3;
}
}
/* Apply back-permutation to VSL and VSR */
/* (Workspace: none needed) */
if (ilvsl) {
dggbak_("P", "L", n, &ilo, &ihi, &work[ileft], &work[iright], n, &vsl[
vsl_offset], ldvsl, &ierr);
}
if (ilvsr) {
dggbak_("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.) {
if (alphar[i__] / safmax > anrmto / anrm || safmin / alphar[
i__] > anrm / anrmto) {
work[1] = (d__1 = a[i__ + i__ * a_dim1] / alphar[i__],
abs(d__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] = (d__1 = a[i__ + (i__ + 1) * a_dim1] / alphai[
i__], abs(d__1));
beta[i__] *= work[1];
alphar[i__] *= work[1];
alphai[i__] *= work[1];
}
}
}
}
if (ilbscl) {
i__1 = *n;
for (i__ = 1; i__ <= i__1; ++i__) {
if (alphai[i__] != 0.) {
if (beta[i__] / safmax > bnrmto / bnrm || safmin / beta[i__]
> bnrm / bnrmto) {
work[1] = (d__1 = b[i__ + i__ * b_dim1] / beta[i__], abs(
d__1));
beta[i__] *= work[1];
alphar[i__] *= work[1];
alphai[i__] *= work[1];
}
}
}
}
/* Undo scaling */
if (ilascl) {
dlascl_("H", &c__0, &c__0, &anrmto, &anrm, n, n, &a[a_offset], lda, &
ierr);
dlascl_("G", &c__0, &c__0, &anrmto, &anrm, n, &c__1, &alphar[1], n, &
ierr);
dlascl_("G", &c__0, &c__0, &anrmto, &anrm, n, &c__1, &alphai[1], n, &
ierr);
}
if (ilbscl) {
dlascl_("U", &c__0, &c__0, &bnrmto, &bnrm, n, n, &b[b_offset], ldb, &
ierr);
dlascl_("G", &c__0, &c__0, &bnrmto, &bnrm, n, &c__1, &beta[1], n, &
ierr);
}
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.) {
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;
}
}
L50:
work[1] = (doublereal) maxwrk;
return 0;
/* End of DGGES */
} /* dgges_ */
/* Subroutine */ int dgegv_(char *jobvl, char *jobvr, integer *n, doublereal *
a, integer *lda, doublereal *b, integer *ldb, doublereal *alphar,
doublereal *alphai, doublereal *beta, doublereal *vl, integer *ldvl,
doublereal *vr, integer *ldvr, doublereal *work, integer *lwork,
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;
doublereal d__1, d__2, d__3, d__4;
/* Local variables */
integer jc, nb, in, jr, nb1, nb2, nb3, ihi, ilo;
doublereal eps;
logical ilv;
doublereal absb, anrm, bnrm;
integer itau;
doublereal temp;
logical ilvl, ilvr;
integer lopt;
doublereal anrm1, anrm2, bnrm1, bnrm2, absai, scale, absar, sbeta;
integer ileft, iinfo, icols, iwork, irows;
doublereal salfai;
doublereal salfar;
doublereal safmin;
doublereal safmax;
char chtemp[1];
logical ldumma[1];
integer ijobvl, iright;
logical ilimit;
integer ijobvr;
doublereal onepls;
integer lwkmin;
integer lwkopt;
logical lquery;
/* -- LAPACK driver routine (version 3.2) -- */
/* November 2006 */
/* Purpose */
/* ======= */
/* This routine is deprecated and has been replaced by routine DGGEV. */
/* DGEGV computes the eigenvalues and, optionally, the left and/or right */
/* eigenvectors of a real matrix pair (A,B). */
/* Given two square matrices A and B, */
/* the generalized nonsymmetric eigenvalue problem (GNEP) is to find the */
/* eigenvalues lambda and corresponding (non-zero) eigenvectors x such */
/* that */
/* A*x = lambda*B*x. */
/* An alternate form is to find the eigenvalues mu and corresponding */
/* eigenvectors y such that */
/* mu*A*y = B*y. */
/* These two forms are equivalent with mu = 1/lambda and x = y if */
/* neither lambda nor mu is zero. In order to deal with the case that */
/* lambda or mu is zero or small, two values alpha and beta are returned */
/* for each eigenvalue, such that lambda = alpha/beta and */
/* mu = beta/alpha. */
/* The vectors x and y in the above equations are right eigenvectors of */
/* the matrix pair (A,B). Vectors u and v satisfying */
/* u**H*A = lambda*u**H*B or mu*v**H*A = v**H*B */
/* are left eigenvectors of (A,B). */
/* 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 (returned */
/* in VL). */
/* JOBVR (input) CHARACTER*1 */
/* = 'N': do not compute the right generalized eigenvectors; */
/* = 'V': compute the right generalized eigenvectors (returned */
/* in VR). */
/* N (input) INTEGER */
/* The order of the matrices A, B, VL, and VR. N >= 0. */
/* A (input/output) DOUBLE PRECISION array, dimension (LDA, N) */
/* On entry, the matrix A. */
/* If JOBVL = 'V' or JOBVR = 'V', then on exit A */
/* contains the real Schur form of A from the generalized Schur */
/* factorization of the pair (A,B) after balancing. */
/* If no eigenvectors were computed, then only the diagonal */
/* blocks from the Schur form will be correct. See DGGHRD and */
/* DHGEQZ for details. */
/* LDA (input) INTEGER */
/* The leading dimension of A. LDA >= max(1,N). */
/* B (input/output) DOUBLE PRECISION array, dimension (LDB, N) */
/* On entry, the matrix B. */
/* If JOBVL = 'V' or JOBVR = 'V', then on exit B contains the */
/* upper triangular matrix obtained from B in the generalized */
/* Schur factorization of the pair (A,B) after balancing. */
/* If no eigenvectors were computed, then only those elements of */
/* B corresponding to the diagonal blocks from the Schur form of */
/* A will be correct. See DGGHRD and DHGEQZ for details. */
/* LDB (input) INTEGER */
/* The leading dimension of B. LDB >= max(1,N). */
/* ALPHAR (output) DOUBLE PRECISION array, dimension (N) */
/* The real parts of each scalar alpha defining an eigenvalue of */
/* GNEP. */
/* ALPHAI (output) DOUBLE PRECISION 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) DOUBLE PRECISION 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. */
/* VL (output) DOUBLE PRECISION array, dimension (LDVL,N) */
/* If JOBVL = 'V', the left eigenvectors u(j) are stored */
/* in the columns of VL, in the same order as their eigenvalues. */
/* If the j-th eigenvalue is real, then u(j) = VL(:,j). */
/* 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). */
/* Each eigenvector is scaled so that its largest component has */
/* abs(real part) + abs(imag. part) = 1, except for eigenvectors */
/* corresponding to an eigenvalue with alpha = beta = 0, which */
/* are set to zero. */
/* 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) DOUBLE PRECISION array, dimension (LDVR,N) */
/* If JOBVR = 'V', the right eigenvectors x(j) are stored */
/* in the columns of VR, in the same order as their eigenvalues. */
/* If the j-th eigenvalue is real, then x(j) = VR(:,j). */
/* If the j-th and (j+1)-st eigenvalues form a complex conjugate */
/* pair, then */
/* x(j) = VR(:,j) + i*VR(:,j+1) */
/* and */
/* x(j+1) = VR(:,j) - i*VR(:,j+1). */
/* Each eigenvector is scaled so that its largest component has */
/* abs(real part) + abs(imag. part) = 1, except for eigenvalues */
/* corresponding to an eigenvalue with alpha = beta = 0, which */
/* are set to zero. */
/* 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) DOUBLE PRECISION 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(1,8*N). */
/* For good performance, LWORK must generally be larger. */
/* To compute the optimal value of LWORK, call ILAENV to get */
/* blocksizes (for DGEQRF, DORMQR, and DORGQR.) Then compute: */
/* NB -- MAX of the blocksizes for DGEQRF, DORMQR, and DORGQR; */
/* 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. */
/* The QZ iteration failed. No eigenvectors have been */
/* calculated, but ALPHAR(j), ALPHAI(j), and BETA(j) */
/* > N: errors that usually indicate LAPACK problems: */
/* =N+1: error return from DGGBAL */
/* =N+2: error return from DGEQRF */
/* =N+3: error return from DORMQR */
/* =N+4: error return from DORGQR */
/* =N+5: error return from DGGHRD */
/* =N+6: error return from DHGEQZ (other than failed */
/* iteration) */
/* =N+7: error return from DTGEVC */
/* =N+8: error return from DGGBAK (computing VL) */
/* =N+9: error return from DGGBAK (computing VR) */
/* =N+10: error return from DLASCL (various calls) */
/* Further Details */
/* =============== */
/* Balancing */
/* --------- */
/* This driver calls DGGBAL 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, DGGBAK 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 DHGEQZ, DGEGS, or read the book "Matrix Computations", */
/* by Golub & van Loan, pub. by Johns Hopkins U. Press. */
/* ===================================================================== */
/* 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;
--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] = (doublereal) 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, "DGEQRF", " ", n, n, &c_n1, &c_n1);
nb2 = ilaenv_(&c__1, "DORMQR", " ", n, n, n, &c_n1);
nb3 = ilaenv_(&c__1, "DORGQR", " ", n, n, n, &c_n1);
/* 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] = (doublereal) lopt;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DGEGV ", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
if (*n == 0) {
return 0;
}
/* Get machine constants */
eps = dlamch_("E") * dlamch_("B");
safmin = dlamch_("S");
safmin += safmin;
safmax = 1. / safmin;
onepls = eps * 4 + 1.;
/* Scale A */
anrm = dlange_("M", n, n, &a[a_offset], lda, &work[1]);
anrm1 = anrm;
anrm2 = 1.;
if (anrm < 1.) {
if (safmax * anrm < 1.) {
anrm1 = safmin;
anrm2 = safmax * anrm;
}
}
if (anrm > 0.) {
dlascl_("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 = dlange_("M", n, n, &b[b_offset], ldb, &work[1]);
bnrm1 = bnrm;
bnrm2 = 1.;
if (bnrm < 1.) {
if (safmax * bnrm < 1.) {
bnrm1 = safmin;
bnrm2 = safmax * bnrm;
}
}
if (bnrm > 0.) {
dlascl_("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) */
ileft = 1;
iright = *n + 1;
iwork = iright + *n;
dggbal_("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 */
irows = ihi + 1 - ilo;
if (ilv) {
icols = *n + 1 - ilo;
} else {
icols = irows;
}
itau = iwork;
iwork = itau + irows;
i__1 = *lwork + 1 - iwork;
dgeqrf_(&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 L120;
}
i__1 = *lwork + 1 - iwork;
dormqr_("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 L120;
}
if (ilvl) {
dlaset_("Full", n, n, &c_b38, &c_b27, &vl[vl_offset], ldvl)
;
i__1 = irows - 1;
i__2 = irows - 1;
dlacpy_("L", &i__1, &i__2, &b[ilo + 1 + ilo * b_dim1], ldb, &vl[ilo +
1 + ilo * vl_dim1], ldvl);
i__1 = *lwork + 1 - iwork;
dorgqr_(&irows, &irows, &irows, &vl[ilo + ilo * vl_dim1], 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) {
dlaset_("Full", n, n, &c_b38, &c_b27, &vr[vr_offset], ldvr)
;
}
/* Reduce to generalized Hessenberg form */
if (ilv) {
/* Eigenvectors requested -- work on whole matrix. */
dgghrd_(jobvl, jobvr, n, &ilo, &ihi, &a[a_offset], lda, &b[b_offset],
ldb, &vl[vl_offset], ldvl, &vr[vr_offset], ldvr, &iinfo);
} else {
dgghrd_("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, &iinfo);
}
if (iinfo != 0) {
*info = *n + 5;
goto L120;
}
/* Perform QZ algorithm */
iwork = itau;
if (ilv) {
*(unsigned char *)chtemp = 'S';
} else {
*(unsigned char *)chtemp = 'E';
}
i__1 = *lwork + 1 - iwork;
dhgeqz_(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 (DTGEVC requires 6*N words of workspace) */
if (ilvl) {
if (ilvr) {
*(unsigned char *)chtemp = 'B';
} else {
*(unsigned char *)chtemp = 'L';
}
} else {
*(unsigned char *)chtemp = 'R';
}
dtgevc_(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) {
dggbak_("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.) {
goto L50;
}
temp = 0.;
if (alphai[jc] == 0.) {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
/* Computing MAX */
d__2 = temp, d__3 = (d__1 = vl[jr + jc * vl_dim1],
abs(d__1));
temp = max(d__2,d__3);
}
} else {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
/* Computing MAX */
d__3 = temp, d__4 = (d__1 = vl[jr + jc * vl_dim1],
abs(d__1)) + (d__2 = vl[jr + (jc + 1) *
vl_dim1], abs(d__2));
temp = max(d__3,d__4);
}
}
if (temp < safmin) {
goto L50;
}
temp = 1. / temp;
if (alphai[jc] == 0.) {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
vl[jr + jc * vl_dim1] *= temp;
}
} else {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
vl[jr + jc * vl_dim1] *= temp;
vl[jr + (jc + 1) * vl_dim1] *= temp;
}
}
L50:
;
}
}
if (ilvr) {
dggbak_("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.) {
goto L100;
}
temp = 0.;
if (alphai[jc] == 0.) {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
/* Computing MAX */
d__2 = temp, d__3 = (d__1 = vr[jr + jc * vr_dim1],
abs(d__1));
temp = max(d__2,d__3);
}
} else {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
/* Computing MAX */
d__3 = temp, d__4 = (d__1 = vr[jr + jc * vr_dim1],
abs(d__1)) + (d__2 = vr[jr + (jc + 1) *
vr_dim1], abs(d__2));
temp = max(d__3,d__4);
}
}
if (temp < safmin) {
goto L100;
}
temp = 1. / temp;
if (alphai[jc] == 0.) {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
vr[jr + jc * vr_dim1] *= temp;
}
} else {
i__2 = *n;
for (jr = 1; jr <= i__2; ++jr) {
vr[jr + jc * vr_dim1] *= temp;
vr[jr + (jc + 1) * vr_dim1] *= temp;
}
}
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 = (d__1 = alphar[jc], abs(d__1));
absai = (d__1 = alphai[jc], abs(d__1));
absb = (d__1 = beta[jc], abs(d__1));
salfar = anrm * alphar[jc];
salfai = anrm * alphai[jc];
sbeta = bnrm * beta[jc];
ilimit = FALSE_;
scale = 1.;
/* Check for significant underflow in ALPHAI */
/* Computing MAX */
d__1 = safmin, d__2 = eps * absar, d__1 = max(d__1,d__2), d__2 = eps *
absb;
if (abs(salfai) < safmin && absai >= max(d__1,d__2)) {
ilimit = TRUE_;
/* Computing MAX */
d__1 = onepls * safmin, d__2 = anrm2 * absai;
scale = onepls * safmin / anrm1 / max(d__1,d__2);
} else if (salfai == 0.) {
/* If insignificant underflow in ALPHAI, then make the */
/* conjugate eigenvalue real. */
if (alphai[jc] < 0. && jc > 1) {
alphai[jc - 1] = 0.;
} else if (alphai[jc] > 0. && jc < *n) {
alphai[jc + 1] = 0.;
}
}
/* Check for significant underflow in ALPHAR */
/* Computing MAX */
d__1 = safmin, d__2 = eps * absai, d__1 = max(d__1,d__2), d__2 = eps *
absb;
if (abs(salfar) < safmin && absar >= max(d__1,d__2)) {
ilimit = TRUE_;
/* Computing MAX */
/* Computing MAX */
d__3 = onepls * safmin, d__4 = anrm2 * absar;
d__1 = scale, d__2 = onepls * safmin / anrm1 / max(d__3,d__4);
scale = max(d__1,d__2);
}
/* Check for significant underflow in BETA */
/* Computing MAX */
d__1 = safmin, d__2 = eps * absar, d__1 = max(d__1,d__2), d__2 = eps *
absai;
if (abs(sbeta) < safmin && absb >= max(d__1,d__2)) {
ilimit = TRUE_;
/* Computing MAX */
/* Computing MAX */
d__3 = onepls * safmin, d__4 = bnrm2 * absb;
d__1 = scale, d__2 = onepls * safmin / bnrm1 / max(d__3,d__4);
scale = max(d__1,d__2);
}
/* Check for possible overflow when limiting scaling */
if (ilimit) {
/* Computing MAX */
d__1 = abs(salfar), d__2 = abs(salfai), d__1 = max(d__1,d__2),
d__2 = abs(sbeta);
temp = scale * safmin * max(d__1,d__2);
if (temp > 1.) {
scale /= temp;
}
if (scale < 1.) {
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;
}
L120:
work[1] = (doublereal) lwkopt;
return 0;
/* End of DGEGV */
} /* dgegv_ */
/* Subroutine */ int dormhr_(char *side, char *trans, integer *m, integer *n,
integer *ilo, integer *ihi, doublereal *a, integer *lda, doublereal *
tau, doublereal *c__, integer *ldc, doublereal *work, integer *lwork,
integer *info)
{
/* -- LAPACK 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
=======
DORMHR overwrites the general real M-by-N matrix C with
SIDE = 'L' SIDE = 'R'
TRANS = 'N': Q * C C * Q
TRANS = 'T': Q**T * C C * Q**T
where Q is a real orthogonal matrix of order nq, with nq = m if
SIDE = 'L' and nq = n if SIDE = 'R'. Q is defined as the product of
IHI-ILO elementary reflectors, as returned by DGEHRD:
Q = H(ilo) H(ilo+1) . . . H(ihi-1).
Arguments
=========
SIDE (input) CHARACTER*1
= 'L': apply Q or Q**T from the Left;
= 'R': apply Q or Q**T from the Right.
TRANS (input) CHARACTER*1
= 'N': No transpose, apply Q;
= 'T': Transpose, apply Q**T.
M (input) INTEGER
The number of rows of the matrix C. M >= 0.
N (input) INTEGER
The number of columns of the matrix C. N >= 0.
ILO (input) INTEGER
IHI (input) INTEGER
ILO and IHI must have the same values as in the previous call
of DGEHRD. Q is equal to the unit matrix except in the
submatrix Q(ilo+1:ihi,ilo+1:ihi).
If SIDE = 'L', then 1 <= ILO <= IHI <= M, if M > 0, and
ILO = 1 and IHI = 0, if M = 0;
if SIDE = 'R', then 1 <= ILO <= IHI <= N, if N > 0, and
ILO = 1 and IHI = 0, if N = 0.
A (input) DOUBLE PRECISION array, dimension
(LDA,M) if SIDE = 'L'
(LDA,N) if SIDE = 'R'
The vectors which define the elementary reflectors, as
returned by DGEHRD.
LDA (input) INTEGER
The leading dimension of the array A.
LDA >= max(1,M) if SIDE = 'L'; LDA >= max(1,N) if SIDE = 'R'.
TAU (input) DOUBLE PRECISION array, dimension
(M-1) if SIDE = 'L'
(N-1) if SIDE = 'R'
TAU(i) must contain the scalar factor of the elementary
reflector H(i), as returned by DGEHRD.
C (input/output) DOUBLE PRECISION array, dimension (LDC,N)
On entry, the M-by-N matrix C.
On exit, C is overwritten by Q*C or Q**T*C or C*Q**T or C*Q.
LDC (input) INTEGER
The leading dimension of the array C. LDC >= max(1,M).
WORK (workspace/output) DOUBLE PRECISION array, dimension (LWORK)
On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
LWORK (input) INTEGER
The dimension of the array WORK.
If SIDE = 'L', LWORK >= max(1,N);
if SIDE = 'R', LWORK >= max(1,M).
For optimum performance LWORK >= N*NB if SIDE = 'L', and
LWORK >= M*NB if SIDE = 'R', where NB is the optimal
blocksize.
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
=====================================================================
Test the input arguments
Parameter adjustments */
/* Table of constant values */
static integer c__1 = 1;
static integer c_n1 = -1;
static integer c__2 = 2;
/* System generated locals */
address a__1[2];
integer a_dim1, a_offset, c_dim1, c_offset, i__1[2], i__2;
char ch__1[2];
/* Builtin functions
Subroutine */ int s_cat(char *, char **, integer *, integer *, ftnlen);
/* Local variables */
static logical left;
extern logical lsame_(char *, char *);
static integer iinfo, i1, i2, nb, mi, nh, ni, nq, nw;
extern /* Subroutine */ int xerbla_(char *, integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *, ftnlen, ftnlen);
extern /* Subroutine */ int dormqr_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, doublereal *,
integer *, doublereal *, integer *, integer *);
static integer lwkopt;
static logical lquery;
#define a_ref(a_1,a_2) a[(a_2)*a_dim1 + a_1]
#define c___ref(a_1,a_2) c__[(a_2)*c_dim1 + a_1]
a_dim1 = *lda;
a_offset = 1 + a_dim1 * 1;
a -= a_offset;
--tau;
c_dim1 = *ldc;
c_offset = 1 + c_dim1 * 1;
c__ -= c_offset;
--work;
/* Function Body */
*info = 0;
nh = *ihi - *ilo;
left = lsame_(side, "L");
lquery = *lwork == -1;
/* NQ is the order of Q and NW is the minimum dimension of WORK */
if (left) {
nq = *m;
nw = *n;
} else {
nq = *n;
nw = *m;
}
if (! left && ! lsame_(side, "R")) {
*info = -1;
} else if (! lsame_(trans, "N") && ! lsame_(trans,
"T")) {
*info = -2;
} else if (*m < 0) {
*info = -3;
} else if (*n < 0) {
*info = -4;
} else if (*ilo < 1 || *ilo > max(1,nq)) {
*info = -5;
} else if (*ihi < min(*ilo,nq) || *ihi > nq) {
*info = -6;
} else if (*lda < max(1,nq)) {
*info = -8;
} else if (*ldc < max(1,*m)) {
*info = -11;
} else if (*lwork < max(1,nw) && ! lquery) {
*info = -13;
}
if (*info == 0) {
if (left) {
/* Writing concatenation */
i__1[0] = 1, a__1[0] = side;
i__1[1] = 1, a__1[1] = trans;
s_cat(ch__1, a__1, i__1, &c__2, (ftnlen)2);
nb = ilaenv_(&c__1, "DORMQR", ch__1, &nh, n, &nh, &c_n1, (ftnlen)
6, (ftnlen)2);
} else {
/* Writing concatenation */
i__1[0] = 1, a__1[0] = side;
i__1[1] = 1, a__1[1] = trans;
s_cat(ch__1, a__1, i__1, &c__2, (ftnlen)2);
nb = ilaenv_(&c__1, "DORMQR", ch__1, m, &nh, &nh, &c_n1, (ftnlen)
6, (ftnlen)2);
}
lwkopt = max(1,nw) * nb;
work[1] = (doublereal) lwkopt;
}
if (*info != 0) {
i__2 = -(*info);
xerbla_("DORMHR", &i__2);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
if (*m == 0 || *n == 0 || nh == 0) {
work[1] = 1.;
return 0;
}
if (left) {
mi = nh;
ni = *n;
i1 = *ilo + 1;
i2 = 1;
} else {
mi = *m;
ni = nh;
i1 = 1;
i2 = *ilo + 1;
}
dormqr_(side, trans, &mi, &ni, &nh, &a_ref(*ilo + 1, *ilo), lda, &tau[*
ilo], &c___ref(i1, i2), ldc, &work[1], lwork, &iinfo);
work[1] = (doublereal) lwkopt;
return 0;
/* End of DORMHR */
} /* dormhr_ */
/* Subroutine */ int dormbr_(char *vect, char *side, char *trans, integer *m,
integer *n, integer *k, doublereal *a, integer *lda, doublereal *tau,
doublereal *c, integer *ldc, doublereal *work, integer *lwork,
integer *info)
{
/* -- LAPACK routine (version 2.0) --
Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd.,
Courant Institute, Argonne National Lab, and Rice University
September 30, 1994
Purpose
=======
If VECT = 'Q', DORMBR overwrites the general real M-by-N matrix C
with
SIDE = 'L' SIDE = 'R'
TRANS = 'N': Q * C C * Q
TRANS = 'T': Q**T * C C * Q**T
If VECT = 'P', DORMBR overwrites the general real M-by-N matrix C
with
SIDE = 'L' SIDE = 'R'
TRANS = 'N': P * C C * P
TRANS = 'T': P**T * C C * P**T
Here Q and P**T are the orthogonal matrices determined by DGEBRD when
reducing a real matrix A to bidiagonal form: A = Q * B * P**T. Q and
P**T are defined as products of elementary reflectors H(i) and G(i)
respectively.
Let nq = m if SIDE = 'L' and nq = n if SIDE = 'R'. Thus nq is the
order of the orthogonal matrix Q or P**T that is applied.
If VECT = 'Q', A is assumed to have been an NQ-by-K matrix:
if nq >= k, Q = H(1) H(2) . . . H(k);
if nq < k, Q = H(1) H(2) . . . H(nq-1).
If VECT = 'P', A is assumed to have been a K-by-NQ matrix:
if k < nq, P = G(1) G(2) . . . G(k);
if k >= nq, P = G(1) G(2) . . . G(nq-1).
Arguments
=========
VECT (input) CHARACTER*1
= 'Q': apply Q or Q**T;
= 'P': apply P or P**T.
SIDE (input) CHARACTER*1
= 'L': apply Q, Q**T, P or P**T from the Left;
= 'R': apply Q, Q**T, P or P**T from the Right.
TRANS (input) CHARACTER*1
= 'N': No transpose, apply Q or P;
= 'T': Transpose, apply Q**T or P**T.
M (input) INTEGER
The number of rows of the matrix C. M >= 0.
N (input) INTEGER
The number of columns of the matrix C. N >= 0.
K (input) INTEGER
If VECT = 'Q', the number of columns in the original
matrix reduced by DGEBRD.
If VECT = 'P', the number of rows in the original
matrix reduced by DGEBRD.
K >= 0.
A (input) DOUBLE PRECISION array, dimension
(LDA,min(nq,K)) if VECT = 'Q'
(LDA,nq) if VECT = 'P'
The vectors which define the elementary reflectors H(i) and
G(i), whose products determine the matrices Q and P, as
returned by DGEBRD.
LDA (input) INTEGER
The leading dimension of the array A.
If VECT = 'Q', LDA >= max(1,nq);
if VECT = 'P', LDA >= max(1,min(nq,K)).
TAU (input) DOUBLE PRECISION array, dimension (min(nq,K))
TAU(i) must contain the scalar factor of the elementary
reflector H(i) or G(i) which determines Q or P, as returned
by DGEBRD in the array argument TAUQ or TAUP.
C (input/output) DOUBLE PRECISION array, dimension (LDC,N)
On entry, the M-by-N matrix C.
On exit, C is overwritten by Q*C or Q**T*C or C*Q**T or C*Q
or P*C or P**T*C or C*P or C*P**T.
LDC (input) INTEGER
The leading dimension of the array C. LDC >= max(1,M).
WORK (workspace/output) DOUBLE PRECISION array, dimension (LWORK)
On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
LWORK (input) INTEGER
The dimension of the array WORK.
If SIDE = 'L', LWORK >= max(1,N);
if SIDE = 'R', LWORK >= max(1,M).
For optimum performance LWORK >= N*NB if SIDE = 'L', and
LWORK >= M*NB if SIDE = 'R', where NB is the optimal
blocksize.
INFO (output) INTEGER
= 0: successful exit
< 0: if INFO = -i, the i-th argument had an illegal value
=====================================================================
Test the input arguments
Parameter adjustments
Function Body */
/* System generated locals */
integer a_dim1, a_offset, c_dim1, c_offset, i__1, i__2;
/* Local variables */
static logical left;
extern logical lsame_(char *, char *);
static integer iinfo, i1, i2, mi, ni, nq, nw;
extern /* Subroutine */ int xerbla_(char *, integer *), dormlq_(
char *, char *, integer *, integer *, integer *, doublereal *,
integer *, doublereal *, doublereal *, integer *, doublereal *,
integer *, integer *);
static logical notran;
extern /* Subroutine */ int dormqr_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, doublereal *,
integer *, doublereal *, integer *, integer *);
static logical applyq;
static char transt[1];
#define TAU(I) tau[(I)-1]
#define WORK(I) work[(I)-1]
#define A(I,J) a[(I)-1 + ((J)-1)* ( *lda)]
#define C(I,J) c[(I)-1 + ((J)-1)* ( *ldc)]
*info = 0;
applyq = lsame_(vect, "Q");
left = lsame_(side, "L");
notran = lsame_(trans, "N");
/* NQ is the order of Q or P and NW is the minimum dimension of WORK
*/
if (left) {
nq = *m;
nw = *n;
} else {
nq = *n;
nw = *m;
}
if (! applyq && ! lsame_(vect, "P")) {
*info = -1;
} else if (! left && ! lsame_(side, "R")) {
*info = -2;
} else if (! notran && ! lsame_(trans, "T")) {
*info = -3;
} else if (*m < 0) {
*info = -4;
} else if (*n < 0) {
*info = -5;
} else if (*k < 0) {
*info = -6;
} else /* if(complicated condition) */ {
/* Computing MAX */
i__1 = 1, i__2 = min(nq,*k);
if (applyq && *lda < max(1,nq) || ! applyq && *lda < max(i__1,i__2)) {
*info = -8;
} else if (*ldc < max(1,*m)) {
*info = -11;
} else if (*lwork < max(1,nw)) {
*info = -13;
}
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DORMBR", &i__1);
return 0;
}
/* Quick return if possible */
WORK(1) = 1.;
if (*m == 0 || *n == 0) {
return 0;
}
if (applyq) {
/* Apply Q */
if (nq >= *k) {
/* Q was determined by a call to DGEBRD with nq >= k */
dormqr_(side, trans, m, n, k, &A(1,1), lda, &TAU(1), &C(1,1), ldc, &WORK(1), lwork, &iinfo);
} else if (nq > 1) {
/* Q was determined by a call to DGEBRD with nq < k */
if (left) {
mi = *m - 1;
ni = *n;
i1 = 2;
i2 = 1;
} else {
mi = *m;
ni = *n - 1;
i1 = 1;
i2 = 2;
}
i__1 = nq - 1;
dormqr_(side, trans, &mi, &ni, &i__1, &A(2,1), lda, &TAU(1)
, &C(i1,i2), ldc, &WORK(1), lwork, &iinfo);
}
} else {
/* Apply P */
if (notran) {
*(unsigned char *)transt = 'T';
} else {
*(unsigned char *)transt = 'N';
}
if (nq > *k) {
/* P was determined by a call to DGEBRD with nq > k */
dormlq_(side, transt, m, n, k, &A(1,1), lda, &TAU(1), &C(1,1), ldc, &WORK(1), lwork, &iinfo);
} else if (nq > 1) {
/* P was determined by a call to DGEBRD with nq <= k */
if (left) {
mi = *m - 1;
ni = *n;
i1 = 2;
i2 = 1;
} else {
mi = *m;
ni = *n - 1;
i1 = 1;
i2 = 2;
}
i__1 = nq - 1;
dormlq_(side, transt, &mi, &ni, &i__1, &A(1,2), lda,
&TAU(1), &C(i1,i2), ldc, &WORK(1), lwork, &
iinfo);
}
}
return 0;
/* End of DORMBR */
} /* dormbr_ */
/* Subroutine */ int dgelsy_(integer *m, integer *n, integer *nrhs,
doublereal *a, integer *lda, doublereal *b, integer *ldb, integer *
jpvt, doublereal *rcond, integer *rank, doublereal *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
=======
DGELSY 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.
This routine is basically identical to the original xGELSX except
three differences:
o The call to the subroutine xGEQPF has been substituted by the
the call to the subroutine xGEQP3. This subroutine is a Blas-3
version of the QR factorization with column pivoting.
o Matrix B (the right hand side) is updated with Blas-3.
o The permutation of matrix B (the right hand side) is faster and
more simple.
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) DOUBLE PRECISION 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) DOUBLE PRECISION array, dimension (LDB,NRHS)
On entry, the M-by-NRHS right hand side matrix B.
On exit, the N-by-NRHS solution matrix X.
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 permuted
to the front of AP, otherwise column i is a free column.
On exit, if JPVT(i) = k, then the i-th column of AP
was the k-th column of A.
RCOND (input) DOUBLE PRECISION
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/output) DOUBLE PRECISION array, dimension (LWORK)
On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
LWORK (input) INTEGER
The dimension of the array WORK.
The unblocked strategy requires that:
LWORK >= MAX( MN+3*N+1, 2*MN+NRHS ),
where MN = min( M, N ).
The block algorithm requires that:
LWORK >= MAX( MN+2*N+NB*(N+1), 2*MN+NB*NRHS ),
where NB is an upper bound on the blocksize returned
by ILAENV for the routines DGEQP3, DTZRZF, STZRQF, DORMQR,
and DORMRZ.
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.
Further Details
===============
Based on contributions by
A. Petitet, Computer Science Dept., Univ. of Tenn., Knoxville, USA
E. Quintana-Orti, Depto. de Informatica, Universidad Jaime I, Spain
G. Quintana-Orti, Depto. de Informatica, Universidad Jaime I, Spain
=====================================================================
Parameter adjustments */
/* Table of constant values */
static integer c__1 = 1;
static integer c_n1 = -1;
static integer c__0 = 0;
static doublereal c_b31 = 0.;
static integer c__2 = 2;
static doublereal c_b54 = 1.;
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, i__1, i__2;
doublereal d__1, d__2;
/* Local variables */
static doublereal anrm, bnrm, smin, smax;
static integer i__, j, iascl, ibscl;
extern /* Subroutine */ int dcopy_(integer *, doublereal *, integer *,
doublereal *, integer *);
static integer ismin, ismax;
static doublereal c1, c2;
extern /* Subroutine */ int dtrsm_(char *, char *, char *, char *,
integer *, integer *, doublereal *, doublereal *, integer *,
doublereal *, integer *), dlaic1_(
integer *, integer *, doublereal *, doublereal *, doublereal *,
doublereal *, doublereal *, doublereal *, doublereal *);
static doublereal wsize, s1, s2;
extern /* Subroutine */ int dgeqp3_(integer *, integer *, doublereal *,
integer *, integer *, doublereal *, doublereal *, integer *,
integer *), dlabad_(doublereal *, doublereal *);
static integer nb;
extern doublereal dlamch_(char *), dlange_(char *, integer *,
integer *, doublereal *, integer *, doublereal *);
static integer mn;
extern /* Subroutine */ int dlascl_(char *, integer *, integer *,
doublereal *, doublereal *, integer *, integer *, doublereal *,
integer *, integer *), dlaset_(char *, integer *, integer
*, doublereal *, doublereal *, doublereal *, integer *),
xerbla_(char *, integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *, ftnlen, ftnlen);
static doublereal bignum;
static integer nb1, nb2, nb3, nb4;
extern /* Subroutine */ int dormqr_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, doublereal *,
integer *, doublereal *, integer *, integer *);
static doublereal sminpr, smaxpr, smlnum;
extern /* Subroutine */ int dormrz_(char *, char *, integer *, integer *,
integer *, integer *, doublereal *, integer *, doublereal *,
doublereal *, integer *, doublereal *, integer *, integer *);
static integer lwkopt;
static logical lquery;
extern /* Subroutine */ int dtzrzf_(integer *, integer *, doublereal *,
integer *, doublereal *, doublereal *, integer *, integer *);
#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]
a_dim1 = *lda;
a_offset = 1 + a_dim1 * 1;
a -= a_offset;
b_dim1 = *ldb;
b_offset = 1 + b_dim1 * 1;
b -= b_offset;
--jpvt;
--work;
/* Function Body */
mn = min(*m,*n);
ismin = mn + 1;
ismax = (mn << 1) + 1;
/* Test the input arguments. */
*info = 0;
nb1 = ilaenv_(&c__1, "DGEQRF", " ", m, n, &c_n1, &c_n1, (ftnlen)6, (
ftnlen)1);
nb2 = ilaenv_(&c__1, "DGERQF", " ", m, n, &c_n1, &c_n1, (ftnlen)6, (
ftnlen)1);
nb3 = ilaenv_(&c__1, "DORMQR", " ", m, n, nrhs, &c_n1, (ftnlen)6, (ftnlen)
1);
nb4 = ilaenv_(&c__1, "DORMRQ", " ", m, n, nrhs, &c_n1, (ftnlen)6, (ftnlen)
1);
/* Computing MAX */
i__1 = max(nb1,nb2), i__1 = max(i__1,nb3);
nb = max(i__1,nb4);
/* Computing MAX */
i__1 = 1, i__2 = mn + (*n << 1) + nb * (*n + 1), i__1 = max(i__1,i__2),
i__2 = (mn << 1) + nb * *nrhs;
lwkopt = max(i__1,i__2);
work[1] = (doublereal) lwkopt;
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(complicated condition) */ {
/* Computing MAX */
i__1 = max(1,*m);
if (*ldb < max(i__1,*n)) {
*info = -7;
} else /* if(complicated condition) */ {
/* Computing MAX */
i__1 = 1, i__2 = mn + *n * 3 + 1, i__1 = max(i__1,i__2), i__2 = (
mn << 1) + *nrhs;
if (*lwork < max(i__1,i__2) && ! lquery) {
*info = -12;
}
}
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DGELSY", &i__1);
return 0;
} else if (lquery) {
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 = dlamch_("S") / dlamch_("P");
bignum = 1. / smlnum;
dlabad_(&smlnum, &bignum);
/* Scale A, B if max entries outside range [SMLNUM,BIGNUM] */
anrm = dlange_("M", m, n, &a[a_offset], lda, &work[1]);
iascl = 0;
if (anrm > 0. && anrm < smlnum) {
/* Scale matrix norm up to SMLNUM */
dlascl_("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 */
dlascl_("G", &c__0, &c__0, &anrm, &bignum, m, n, &a[a_offset], lda,
info);
iascl = 2;
} else if (anrm == 0.) {
/* Matrix all zero. Return zero solution. */
i__1 = max(*m,*n);
dlaset_("F", &i__1, nrhs, &c_b31, &c_b31, &b[b_offset], ldb);
*rank = 0;
goto L70;
}
bnrm = dlange_("M", m, nrhs, &b[b_offset], ldb, &work[1]);
ibscl = 0;
if (bnrm > 0. && bnrm < smlnum) {
/* Scale matrix norm up to SMLNUM */
dlascl_("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 */
dlascl_("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 */
i__1 = *lwork - mn;
dgeqp3_(m, n, &a[a_offset], lda, &jpvt[1], &work[1], &work[mn + 1], &i__1,
info);
wsize = mn + work[mn + 1];
/* workspace: MN+2*N+NB*(N+1).
Details of Householder rotations stored in WORK(1:MN).
Determine RANK using incremental condition estimation */
work[ismin] = 1.;
work[ismax] = 1.;
smax = (d__1 = a_ref(1, 1), abs(d__1));
smin = smax;
if ((d__1 = a_ref(1, 1), abs(d__1)) == 0.) {
*rank = 0;
i__1 = max(*m,*n);
dlaset_("F", &i__1, nrhs, &c_b31, &c_b31, &b[b_offset], ldb);
goto L70;
} else {
*rank = 1;
}
L10:
if (*rank < mn) {
i__ = *rank + 1;
dlaic1_(&c__2, rank, &work[ismin], &smin, &a_ref(1, i__), &a_ref(i__,
i__), &sminpr, &s1, &c1);
dlaic1_(&c__1, rank, &work[ismax], &smax, &a_ref(1, i__), &a_ref(i__,
i__), &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;
}
}
/* workspace: 3*MN.
Logically partition R = [ R11 R12 ]
[ 0 R22 ]
where R11 = R(1:RANK,1:RANK)
[R11,R12] = [ T11, 0 ] * Y */
if (*rank < *n) {
i__1 = *lwork - (mn << 1);
dtzrzf_(rank, n, &a[a_offset], lda, &work[mn + 1], &work[(mn << 1) +
1], &i__1, info);
}
/* workspace: 2*MN.
Details of Householder rotations stored in WORK(MN+1:2*MN)
B(1:M,1:NRHS) := Q' * B(1:M,1:NRHS) */
i__1 = *lwork - (mn << 1);
dormqr_("Left", "Transpose", m, nrhs, &mn, &a[a_offset], lda, &work[1], &
b[b_offset], ldb, &work[(mn << 1) + 1], &i__1, info);
/* Computing MAX */
d__1 = wsize, d__2 = (mn << 1) + work[(mn << 1) + 1];
wsize = max(d__1,d__2);
/* workspace: 2*MN+NB*NRHS.
B(1:RANK,1:NRHS) := inv(T11) * B(1:RANK,1:NRHS) */
dtrsm_("Left", "Upper", "No transpose", "Non-unit", rank, nrhs, &c_b54, &
a[a_offset], lda, &b[b_offset], ldb);
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
i__2 = *n;
for (i__ = *rank + 1; i__ <= i__2; ++i__) {
b_ref(i__, j) = 0.;
/* L30: */
}
/* L40: */
}
/* B(1:N,1:NRHS) := Y' * B(1:N,1:NRHS) */
if (*rank < *n) {
i__1 = *n - *rank;
i__2 = *lwork - (mn << 1);
dormrz_("Left", "Transpose", n, nrhs, rank, &i__1, &a[a_offset], lda,
&work[mn + 1], &b[b_offset], ldb, &work[(mn << 1) + 1], &i__2,
info);
}
/* workspace: 2*MN+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[jpvt[i__]] = b_ref(i__, j);
/* L50: */
}
dcopy_(n, &work[1], &c__1, &b_ref(1, j), &c__1);
/* L60: */
}
/* workspace: N.
Undo scaling */
if (iascl == 1) {
dlascl_("G", &c__0, &c__0, &anrm, &smlnum, n, nrhs, &b[b_offset], ldb,
info);
dlascl_("U", &c__0, &c__0, &smlnum, &anrm, rank, rank, &a[a_offset],
lda, info);
} else if (iascl == 2) {
dlascl_("G", &c__0, &c__0, &anrm, &bignum, n, nrhs, &b[b_offset], ldb,
info);
dlascl_("U", &c__0, &c__0, &bignum, &anrm, rank, rank, &a[a_offset],
lda, info);
}
if (ibscl == 1) {
dlascl_("G", &c__0, &c__0, &smlnum, &bnrm, n, nrhs, &b[b_offset], ldb,
info);
} else if (ibscl == 2) {
dlascl_("G", &c__0, &c__0, &bignum, &bnrm, n, nrhs, &b[b_offset], ldb,
info);
}
L70:
work[1] = (doublereal) lwkopt;
return 0;
/* End of DGELSY */
} /* dgelsy_ */
/* Subroutine */ int dormtr_(char *side, char *uplo, char *trans, integer *m,
integer *n, doublereal *a, integer *lda, doublereal *tau, doublereal *
c__, integer *ldc, doublereal *work, integer *lwork, integer *info,
ftnlen side_len, ftnlen uplo_len, ftnlen trans_len)
{
/* System generated locals */
address a__1[2];
integer a_dim1, a_offset, c_dim1, c_offset, i__1[2], i__2, i__3;
char ch__1[2];
/* Builtin functions */
/* Subroutine */ int s_cat(char *, char **, integer *, integer *, ftnlen);
/* Local variables */
static integer i1, i2, nb, mi, ni, nq, nw;
static logical left;
extern logical lsame_(char *, char *, ftnlen, ftnlen);
static integer iinfo;
static logical upper;
extern /* Subroutine */ int xerbla_(char *, integer *, ftnlen);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *, ftnlen, ftnlen);
extern /* Subroutine */ int dormql_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, doublereal *,
integer *, doublereal *, integer *, integer *, ftnlen, ftnlen),
dormqr_(char *, char *, integer *, integer *, integer *,
doublereal *, integer *, doublereal *, doublereal *, integer *,
doublereal *, integer *, integer *, ftnlen, ftnlen);
static integer lwkopt;
static logical lquery;
/* -- LAPACK 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 */
/* ======= */
/* DORMTR overwrites the general real M-by-N matrix C with */
/* SIDE = 'L' SIDE = 'R' */
/* TRANS = 'N': Q * C C * Q */
/* TRANS = 'T': Q**T * C C * Q**T */
/* where Q is a real orthogonal matrix of order nq, with nq = m if */
/* SIDE = 'L' and nq = n if SIDE = 'R'. Q is defined as the product of */
/* nq-1 elementary reflectors, as returned by DSYTRD: */
/* if UPLO = 'U', Q = H(nq-1) . . . H(2) H(1); */
/* if UPLO = 'L', Q = H(1) H(2) . . . H(nq-1). */
/* Arguments */
/* ========= */
/* SIDE (input) CHARACTER*1 */
/* = 'L': apply Q or Q**T from the Left; */
/* = 'R': apply Q or Q**T from the Right. */
/* UPLO (input) CHARACTER*1 */
/* = 'U': Upper triangle of A contains elementary reflectors */
/* from DSYTRD; */
/* = 'L': Lower triangle of A contains elementary reflectors */
/* from DSYTRD. */
/* TRANS (input) CHARACTER*1 */
/* = 'N': No transpose, apply Q; */
/* = 'T': Transpose, apply Q**T. */
/* M (input) INTEGER */
/* The number of rows of the matrix C. M >= 0. */
/* N (input) INTEGER */
/* The number of columns of the matrix C. N >= 0. */
/* A (input) DOUBLE PRECISION array, dimension */
/* (LDA,M) if SIDE = 'L' */
/* (LDA,N) if SIDE = 'R' */
/* The vectors which define the elementary reflectors, as */
/* returned by DSYTRD. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. */
/* LDA >= max(1,M) if SIDE = 'L'; LDA >= max(1,N) if SIDE = 'R'. */
/* TAU (input) DOUBLE PRECISION array, dimension */
/* (M-1) if SIDE = 'L' */
/* (N-1) if SIDE = 'R' */
/* TAU(i) must contain the scalar factor of the elementary */
/* reflector H(i), as returned by DSYTRD. */
/* C (input/output) DOUBLE PRECISION array, dimension (LDC,N) */
/* On entry, the M-by-N matrix C. */
/* On exit, C is overwritten by Q*C or Q**T*C or C*Q**T or C*Q. */
/* LDC (input) INTEGER */
/* The leading dimension of the array C. LDC >= max(1,M). */
/* WORK (workspace/output) DOUBLE PRECISION array, dimension (LWORK) */
/* On exit, if INFO = 0, WORK(1) returns the optimal LWORK. */
/* LWORK (input) INTEGER */
/* The dimension of the array WORK. */
/* If SIDE = 'L', LWORK >= max(1,N); */
/* if SIDE = 'R', LWORK >= max(1,M). */
/* For optimum performance LWORK >= N*NB if SIDE = 'L', and */
/* LWORK >= M*NB if SIDE = 'R', where NB is the optimal */
/* blocksize. */
/* 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 */
/* ===================================================================== */
/* .. Local Scalars .. */
/* .. */
/* .. 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;
--tau;
c_dim1 = *ldc;
c_offset = 1 + c_dim1;
c__ -= c_offset;
--work;
/* Function Body */
*info = 0;
left = lsame_(side, "L", (ftnlen)1, (ftnlen)1);
upper = lsame_(uplo, "U", (ftnlen)1, (ftnlen)1);
lquery = *lwork == -1;
/* NQ is the order of Q and NW is the minimum dimension of WORK */
if (left) {
nq = *m;
nw = *n;
} else {
nq = *n;
nw = *m;
}
if (! left && ! lsame_(side, "R", (ftnlen)1, (ftnlen)1)) {
*info = -1;
} else if (! upper && ! lsame_(uplo, "L", (ftnlen)1, (ftnlen)1)) {
*info = -2;
} else if (! lsame_(trans, "N", (ftnlen)1, (ftnlen)1) && ! lsame_(trans,
"T", (ftnlen)1, (ftnlen)1)) {
*info = -3;
} else if (*m < 0) {
*info = -4;
} else if (*n < 0) {
*info = -5;
} else if (*lda < max(1,nq)) {
*info = -7;
} else if (*ldc < max(1,*m)) {
*info = -10;
} else if (*lwork < max(1,nw) && ! lquery) {
*info = -12;
}
if (*info == 0) {
if (upper) {
if (left) {
/* Writing concatenation */
i__1[0] = 1, a__1[0] = side;
i__1[1] = 1, a__1[1] = trans;
s_cat(ch__1, a__1, i__1, &c__2, (ftnlen)2);
i__2 = *m - 1;
i__3 = *m - 1;
nb = ilaenv_(&c__1, "DORMQL", ch__1, &i__2, n, &i__3, &c_n1, (
ftnlen)6, (ftnlen)2);
} else {
/* Writing concatenation */
i__1[0] = 1, a__1[0] = side;
i__1[1] = 1, a__1[1] = trans;
s_cat(ch__1, a__1, i__1, &c__2, (ftnlen)2);
i__2 = *n - 1;
i__3 = *n - 1;
nb = ilaenv_(&c__1, "DORMQL", ch__1, m, &i__2, &i__3, &c_n1, (
ftnlen)6, (ftnlen)2);
}
} else {
if (left) {
/* Writing concatenation */
i__1[0] = 1, a__1[0] = side;
i__1[1] = 1, a__1[1] = trans;
s_cat(ch__1, a__1, i__1, &c__2, (ftnlen)2);
i__2 = *m - 1;
i__3 = *m - 1;
nb = ilaenv_(&c__1, "DORMQR", ch__1, &i__2, n, &i__3, &c_n1, (
ftnlen)6, (ftnlen)2);
} else {
/* Writing concatenation */
i__1[0] = 1, a__1[0] = side;
i__1[1] = 1, a__1[1] = trans;
s_cat(ch__1, a__1, i__1, &c__2, (ftnlen)2);
i__2 = *n - 1;
i__3 = *n - 1;
nb = ilaenv_(&c__1, "DORMQR", ch__1, m, &i__2, &i__3, &c_n1, (
ftnlen)6, (ftnlen)2);
}
}
lwkopt = max(1,nw) * nb;
work[1] = (doublereal) lwkopt;
}
if (*info != 0) {
i__2 = -(*info);
xerbla_("DORMTR", &i__2, (ftnlen)6);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
if (*m == 0 || *n == 0 || nq == 1) {
work[1] = 1.;
return 0;
}
if (left) {
mi = *m - 1;
ni = *n;
} else {
mi = *m;
ni = *n - 1;
}
if (upper) {
/* Q was determined by a call to DSYTRD with UPLO = 'U' */
i__2 = nq - 1;
dormql_(side, trans, &mi, &ni, &i__2, &a[(a_dim1 << 1) + 1], lda, &
tau[1], &c__[c_offset], ldc, &work[1], lwork, &iinfo, (ftnlen)
1, (ftnlen)1);
} else {
/* Q was determined by a call to DSYTRD with UPLO = 'L' */
if (left) {
i1 = 2;
i2 = 1;
} else {
i1 = 1;
i2 = 2;
}
i__2 = nq - 1;
dormqr_(side, trans, &mi, &ni, &i__2, &a[a_dim1 + 2], lda, &tau[1], &
c__[i1 + i2 * c_dim1], ldc, &work[1], lwork, &iinfo, (ftnlen)
1, (ftnlen)1);
}
work[1] = (doublereal) lwkopt;
return 0;
/* End of DORMTR */
} /* dormtr_ */
/* Subroutine */ int dgegs_(char *jobvsl, char *jobvsr, integer *n,
doublereal *a, integer *lda, doublereal *b, integer *ldb, doublereal *
alphar, doublereal *alphai, doublereal *beta, doublereal *vsl,
integer *ldvsl, doublereal *vsr, integer *ldvsr, doublereal *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 DGGES.
DGEGS computes for a pair of N-by-N real nonsymmetric matrices A, B:
the generalized eigenvalues (alphar +/- alphai*i, beta), the real
Schur form (A, B), and optionally left and/or right Schur vectors
(VSL and VSR).
(If only the generalized eigenvalues are needed, use the driver DGEGV
instead.)
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)
The (generalized) Schur form of a pair of matrices is the result of
multiplying both matrices on the left by one orthogonal matrix and
both on the right by another orthogonal matrix, these two orthogonal
matrices being chosen so as to bring the pair of matrices into
(real) Schur form.
A pair of matrices A, B is in generalized real Schur form if B is
upper triangular with non-negative diagonal and A 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 A will be
"standardized" by making the corresponding elements of B have the
form:
[ a 0 ]
[ 0 b ]
and the pair of corresponding 2-by-2 blocks in A and B will
have a complex conjugate pair of generalized eigenvalues.
The left and right Schur vectors are the columns of VSL and VSR,
respectively, where VSL and VSR are the orthogonal matrices
which reduce A and B to Schur form:
Schur form of (A,B) = ( (VSL)**T A (VSR), (VSL)**T B (VSR) )
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.
N (input) INTEGER
The order of the matrices A, B, VSL, and VSR. N >= 0.
A (input/output) DOUBLE PRECISION array, dimension (LDA, N)
On entry, the first of the pair of matrices whose generalized
eigenvalues and (optionally) Schur vectors are to be
computed.
On exit, the generalized Schur form of A.
Note: to avoid overflow, the Frobenius norm of the matrix
A should be less than the overflow threshold.
LDA (input) INTEGER
The leading dimension of A. LDA >= max(1,N).
B (input/output) DOUBLE PRECISION array, dimension (LDB, N)
On entry, the second of the pair of matrices whose
generalized eigenvalues and (optionally) Schur vectors are
to be computed.
On exit, the generalized Schur form of B.
Note: to avoid overflow, the Frobenius norm of the matrix
B should be less than the overflow threshold.
LDB (input) INTEGER
The leading dimension of B. LDB >= max(1,N).
ALPHAR (output) DOUBLE PRECISION array, dimension (N)
ALPHAI (output) DOUBLE PRECISION array, dimension (N)
BETA (output) DOUBLE PRECISION 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,
j=1,...,N and BETA(j),j=1,...,N are the diagonals of the
complex Schur form (A,B) 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
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).
VSL (output) DOUBLE PRECISION array, dimension (LDVSL,N)
If JOBVSL = 'V', VSL will contain the left Schur vectors.
(See "Purpose", above.)
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) DOUBLE PRECISION array, dimension (LDVSR,N)
If JOBVSR = 'V', VSR will contain the right Schur vectors.
(See "Purpose", above.)
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) DOUBLE PRECISION 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(1,4*N).
For good performance, LWORK must generally be larger.
To compute the optimal value of LWORK, call ILAENV to get
blocksizes (for DGEQRF, DORMQR, and DORGQR.) Then compute:
NB -- MAX of the blocksizes for DGEQRF, DORMQR, and DORGQR
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 DGGBAL
=N+2: error return from DGEQRF
=N+3: error return from DORMQR
=N+4: error return from DORGQR
=N+5: error return from DGGHRD
=N+6: error return from DHGEQZ (other than failed
iteration)
=N+7: error return from DGGBAK (computing VSL)
=N+8: error return from DGGBAK (computing VSR)
=N+9: error return from DLASCL (various places)
=====================================================================
Decode the input arguments
Parameter adjustments */
/* Table of constant values */
static integer c__1 = 1;
static integer c_n1 = -1;
static doublereal c_b36 = 0.;
static doublereal c_b37 = 1.;
/* 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 */
static doublereal anrm, bnrm;
static integer itau, lopt;
extern logical lsame_(char *, char *);
static integer ileft, iinfo, icols;
static logical ilvsl;
static integer iwork;
static logical ilvsr;
static integer irows;
extern /* Subroutine */ int dggbak_(char *, char *, integer *, integer *,
integer *, doublereal *, doublereal *, integer *, doublereal *,
integer *, integer *);
static integer nb;
extern /* Subroutine */ int dggbal_(char *, integer *, doublereal *,
integer *, doublereal *, integer *, integer *, integer *,
doublereal *, doublereal *, doublereal *, integer *);
extern doublereal dlamch_(char *), dlange_(char *, integer *,
integer *, doublereal *, integer *, doublereal *);
extern /* Subroutine */ int dgghrd_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, integer *,
doublereal *, integer *, doublereal *, integer *, integer *), dlascl_(char *, integer *, integer *, doublereal
*, doublereal *, integer *, integer *, doublereal *, integer *,
integer *);
static logical ilascl, ilbscl;
extern /* Subroutine */ int dgeqrf_(integer *, integer *, doublereal *,
integer *, doublereal *, doublereal *, integer *, integer *),
dlacpy_(char *, integer *, integer *, doublereal *, integer *,
doublereal *, integer *);
static doublereal safmin;
extern /* Subroutine */ int dlaset_(char *, integer *, integer *,
doublereal *, doublereal *, doublereal *, integer *),
xerbla_(char *, integer *);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *, ftnlen, ftnlen);
static doublereal bignum;
extern /* Subroutine */ int dhgeqz_(char *, char *, char *, integer *,
integer *, integer *, doublereal *, integer *, doublereal *,
integer *, doublereal *, doublereal *, doublereal *, doublereal *,
integer *, doublereal *, integer *, doublereal *, integer *,
integer *);
static integer ijobvl, iright, ijobvr;
extern /* Subroutine */ int dorgqr_(integer *, integer *, integer *,
doublereal *, integer *, doublereal *, doublereal *, integer *,
integer *);
static doublereal anrmto;
static integer lwkmin, nb1, nb2, nb3;
static doublereal bnrmto;
extern /* Subroutine */ int dormqr_(char *, char *, integer *, integer *,
integer *, doublereal *, integer *, doublereal *, doublereal *,
integer *, doublereal *, integer *, integer *);
static doublereal smlnum;
static integer lwkopt;
static logical lquery;
static integer ihi, ilo;
static doublereal 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;
--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] = (doublereal) 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, "DGEQRF", " ", n, n, &c_n1, &c_n1, (ftnlen)6, (
ftnlen)1);
nb2 = ilaenv_(&c__1, "DORMQR", " ", n, n, n, &c_n1, (ftnlen)6, (
ftnlen)1);
nb3 = ilaenv_(&c__1, "DORGQR", " ", n, n, n, &c_n1, (ftnlen)6, (
ftnlen)1);
/* Computing MAX */
i__1 = max(nb1,nb2);
nb = max(i__1,nb3);
lopt = (*n << 1) + *n * (nb + 1);
work[1] = (doublereal) lopt;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DGEGS ", &i__1);
return 0;
} else if (lquery) {
return 0;
}
/* Quick return if possible */
if (*n == 0) {
return 0;
}
/* Get machine constants */
eps = dlamch_("E") * dlamch_("B");
safmin = dlamch_("S");
smlnum = *n * safmin / eps;
bignum = 1. / smlnum;
/* Scale A if max element outside range [SMLNUM,BIGNUM] */
anrm = dlange_("M", n, n, &a[a_offset], lda, &work[1]);
ilascl = FALSE_;
if (anrm > 0. && anrm < smlnum) {
anrmto = smlnum;
ilascl = TRUE_;
} else if (anrm > bignum) {
anrmto = bignum;
ilascl = TRUE_;
}
if (ilascl) {
dlascl_("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 = dlange_("M", n, n, &b[b_offset], ldb, &work[1]);
ilbscl = FALSE_;
if (bnrm > 0. && bnrm < smlnum) {
bnrmto = smlnum;
ilbscl = TRUE_;
} else if (bnrm > bignum) {
bnrmto = bignum;
ilbscl = TRUE_;
}
if (ilbscl) {
dlascl_("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;
dggbal_("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;
dgeqrf_(&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 L10;
}
i__1 = *lwork + 1 - iwork;
dormqr_("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 L10;
}
if (ilvsl) {
dlaset_("Full", n, n, &c_b36, &c_b37, &vsl[vsl_offset], ldvsl);
i__1 = irows - 1;
i__2 = irows - 1;
dlacpy_("L", &i__1, &i__2, &b_ref(ilo + 1, ilo), ldb, &vsl_ref(ilo +
1, ilo), ldvsl);
i__1 = *lwork + 1 - iwork;
dorgqr_(&irows, &irows, &irows, &vsl_ref(ilo, ilo), 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) {
dlaset_("Full", n, n, &c_b36, &c_b37, &vsr[vsr_offset], ldvsr);
}
/* Reduce to generalized Hessenberg form */
dgghrd_(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;
dhgeqz_("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) {
dggbak_("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) {
dggbak_("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) {
dlascl_("H", &c_n1, &c_n1, &anrmto, &anrm, n, n, &a[a_offset], lda, &
iinfo);
if (iinfo != 0) {
*info = *n + 9;
return 0;
}
dlascl_("G", &c_n1, &c_n1, &anrmto, &anrm, n, &c__1, &alphar[1], n, &
iinfo);
if (iinfo != 0) {
*info = *n + 9;
return 0;
}
dlascl_("G", &c_n1, &c_n1, &anrmto, &anrm, n, &c__1, &alphai[1], n, &
iinfo);
if (iinfo != 0) {
*info = *n + 9;
return 0;
}
}
if (ilbscl) {
dlascl_("U", &c_n1, &c_n1, &bnrmto, &bnrm, n, n, &b[b_offset], ldb, &
iinfo);
if (iinfo != 0) {
*info = *n + 9;
return 0;
}
dlascl_("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] = (doublereal) lwkopt;
return 0;
/* End of DGEGS */
} /* dgegs_ */
