/* Subroutine */
int slapll_(integer *n, real *x, integer *incx, real *y, integer *incy, real *ssmin)
{
/* System generated locals */
integer i__1;
/* Local variables */
real c__, a11, a12, a22, tau;
extern real sdot_(integer *, real *, integer *, real *, integer *);
extern /* Subroutine */
int slas2_(real *, real *, real *, real *, real *) ;
real ssmax;
extern /* Subroutine */
int saxpy_(integer *, real *, real *, integer *, real *, integer *), slarfg_(integer *, real *, real *, integer *, real *);
/* -- LAPACK auxiliary routine (version 3.4.2) -- */
/* -- LAPACK is a software package provided by Univ. of Tennessee, -- */
/* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- */
/* September 2012 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Executable Statements .. */
/* Quick return if possible */
/* Parameter adjustments */
--y;
--x;
/* Function Body */
if (*n <= 1)
{
*ssmin = 0.f;
return 0;
}
/* Compute the QR factorization of the N-by-2 matrix ( X Y ) */
slarfg_(n, &x[1], &x[*incx + 1], incx, &tau);
a11 = x[1];
x[1] = 1.f;
c__ = -tau * sdot_(n, &x[1], incx, &y[1], incy);
saxpy_(n, &c__, &x[1], incx, &y[1], incy);
i__1 = *n - 1;
slarfg_(&i__1, &y[*incy + 1], &y[(*incy << 1) + 1], incy, &tau);
a12 = y[1];
a22 = y[*incy + 1];
/* Compute the SVD of 2-by-2 Upper triangular matrix. */
slas2_(&a11, &a12, &a22, ssmin, &ssmax);
return 0;
/* End of SLAPLL */
}
/* Subroutine */ int sgerq2_(integer *m, integer *n, real *a, integer *lda,
real *tau, real *work, integer *info)
{
/* -- LAPACK routine (version 2.0) --
Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd.,
Courant Institute, Argonne National Lab, and Rice University
February 29, 1992
Purpose
=======
SGERQ2 computes an RQ factorization of a real m by n matrix A:
A = R * Q.
Arguments
=========
M (input) INTEGER
The number of rows of the matrix A. M >= 0.
N (input) INTEGER
The number of columns of the matrix A. N >= 0.
A (input/output) REAL array, dimension (LDA,N)
On entry, the m by n matrix A.
On exit, if m <= n, the upper triangle of the subarray
A(1:m,n-m+1:n) contains the m by m upper triangular matrix R;
if m >= n, the elements on and above the (m-n)-th subdiagonal
contain the m by n upper trapezoidal matrix R; the remaining
elements, with the array TAU, represent the orthogonal matrix
Q as a product of elementary reflectors (see Further
Details).
LDA (input) INTEGER
The leading dimension of the array A. LDA >= max(1,M).
TAU (output) REAL array, dimension (min(M,N))
The scalar factors of the elementary reflectors (see Further
Details).
WORK (workspace) REAL array, dimension (M)
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(m,n).
Each H(i) has the form
H(i) = I - tau * v * v'
where tau is a real scalar, and v is a real vector with
v(n-k+i+1:n) = 0 and v(n-k+i) = 1; v(1:n-k+i-1) is stored on exit in
A(m-k+i,1:n-k+i-1), and tau in TAU(i).
=====================================================================
Test the input arguments
Parameter adjustments
Function Body */
/* System generated locals */
integer a_dim1, a_offset, i__1, i__2;
/* Local variables */
static integer i, k;
extern /* Subroutine */ int slarf_(char *, integer *, integer *, real *,
integer *, real *, real *, integer *, real *), xerbla_(
char *, integer *), slarfg_(integer *, real *, real *,
integer *, real *);
static real aii;
#define TAU(I) tau[(I)-1]
#define WORK(I) work[(I)-1]
#define A(I,J) a[(I)-1 + ((J)-1)* ( *lda)]
*info = 0;
if (*m < 0) {
*info = -1;
} else if (*n < 0) {
*info = -2;
} else if (*lda < max(1,*m)) {
*info = -4;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SGERQ2", &i__1);
return 0;
}
k = min(*m,*n);
for (i = k; i >= 1; --i) {
/* Generate elementary reflector H(i) to annihilate
A(m-k+i,1:n-k+i-1) */
i__1 = *n - k + i;
slarfg_(&i__1, &A(*m-k+i,*n-k+i), &A(*m-k+i,1), lda, &TAU(i));
/* Apply H(i) to A(1:m-k+i-1,1:n-k+i) from the right */
aii = A(*m-k+i,*n-k+i);
A(*m-k+i,*n-k+i) = 1.f;
i__1 = *m - k + i - 1;
i__2 = *n - k + i;
slarf_("Right", &i__1, &i__2, &A(*m-k+i,1), lda, &TAU(i), &
A(1,1), lda, &WORK(1));
A(*m-k+i,*n-k+i) = aii;
/* L10: */
}
return 0;
/* End of SGERQ2 */
} /* sgerq2_ */
/* Subroutine */ int ssytd2_(char *uplo, integer *n, real *a, integer *lda,
real *d, real *e, real *tau, integer *info)
{
/* -- LAPACK routine (version 2.0) --
Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd.,
Courant Institute, Argonne National Lab, and Rice University
October 31, 1992
Purpose
=======
SSYTD2 reduces a real symmetric matrix A to symmetric tridiagonal
form T by an orthogonal similarity transformation: Q' * A * Q = T.
Arguments
=========
UPLO (input) CHARACTER*1
Specifies whether the upper or lower triangular part of the
symmetric matrix A is stored:
= 'U': Upper triangular
= 'L': Lower triangular
N (input) INTEGER
The order of the matrix A. N >= 0.
A (input/output) REAL array, dimension (LDA,N)
On entry, the symmetric matrix A. If UPLO = 'U', the leading
n-by-n upper triangular part of A contains the upper
triangular part of the matrix A, and the strictly lower
triangular part of A is not referenced. If UPLO = 'L', the
leading n-by-n lower triangular part of A contains the lower
triangular part of the matrix A, and the strictly upper
triangular part of A is not referenced.
On exit, if UPLO = 'U', the diagonal and first superdiagonal
of A are overwritten by the corresponding elements of the
tridiagonal matrix T, and the elements above the first
superdiagonal, with the array TAU, represent the orthogonal
matrix Q as a product of elementary reflectors; if UPLO
= 'L', the diagonal and first subdiagonal of A are over-
written by the corresponding elements of the tridiagonal
matrix T, and the elements below the first subdiagonal, with
the array TAU, represent the orthogonal matrix Q as a product
of elementary reflectors. See Further Details.
LDA (input) INTEGER
The leading dimension of the array A. LDA >= max(1,N).
D (output) REAL array, dimension (N)
The diagonal elements of the tridiagonal matrix T:
D(i) = A(i,i).
E (output) REAL array, dimension (N-1)
The off-diagonal elements of the tridiagonal matrix T:
E(i) = A(i,i+1) if UPLO = 'U', E(i) = A(i+1,i) if UPLO = 'L'.
TAU (output) REAL array, dimension (N-1)
The scalar factors of the elementary reflectors (see Further
Details).
INFO (output) INTEGER
= 0: successful exit
< 0: if INFO = -i, the i-th argument had an illegal value.
Further Details
===============
If UPLO = 'U', the matrix Q is represented as a product of elementary
reflectors
Q = H(n-1) . . . H(2) H(1).
Each H(i) has the form
H(i) = I - tau * v * v'
where tau is a real scalar, and v is a real vector with
v(i+1:n) = 0 and v(i) = 1; v(1:i-1) is stored on exit in
A(1:i-1,i+1), and tau in TAU(i).
If UPLO = 'L', the matrix Q is represented as a product of elementary
reflectors
Q = H(1) H(2) . . . H(n-1).
Each H(i) has the form
H(i) = I - tau * v * v'
where tau is a real scalar, and v is a real vector with
v(1:i) = 0 and v(i+1) = 1; v(i+2:n) is stored on exit in A(i+2:n,i),
and tau in TAU(i).
The contents of A on exit are illustrated by the following examples
with n = 5:
if UPLO = 'U': if UPLO = 'L':
( d e v2 v3 v4 ) ( d )
( d e v3 v4 ) ( e d )
( d e v4 ) ( v1 e d )
( d e ) ( v1 v2 e d )
( d ) ( v1 v2 v3 e d )
where d and e denote diagonal and off-diagonal elements of T, and vi
denotes an element of the vector defining H(i).
=====================================================================
Test the input parameters
Parameter adjustments
Function Body */
/* Table of constant values */
static integer c__1 = 1;
static real c_b8 = 0.f;
static real c_b14 = -1.f;
/* System generated locals */
integer a_dim1, a_offset, i__1, i__2, i__3;
/* Local variables */
static real taui;
extern doublereal sdot_(integer *, real *, integer *, real *, integer *);
static integer i;
extern /* Subroutine */ int ssyr2_(char *, integer *, real *, real *,
integer *, real *, integer *, real *, integer *);
static real alpha;
extern logical lsame_(char *, char *);
static logical upper;
extern /* Subroutine */ int saxpy_(integer *, real *, real *, integer *,
real *, integer *), ssymv_(char *, integer *, real *, real *,
integer *, real *, integer *, real *, real *, integer *),
xerbla_(char *, integer *), slarfg_(integer *, real *,
real *, integer *, real *);
#define D(I) d[(I)-1]
#define E(I) e[(I)-1]
#define TAU(I) tau[(I)-1]
#define A(I,J) a[(I)-1 + ((J)-1)* ( *lda)]
*info = 0;
upper = lsame_(uplo, "U");
if (! upper && ! lsame_(uplo, "L")) {
*info = -1;
} else if (*n < 0) {
*info = -2;
} else if (*lda < max(1,*n)) {
*info = -4;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SSYTD2", &i__1);
return 0;
}
/* Quick return if possible */
if (*n <= 0) {
return 0;
}
if (upper) {
/* Reduce the upper triangle of A */
for (i = *n - 1; i >= 1; --i) {
/* Generate elementary reflector H(i) = I - tau * v * v'
to annihilate A(1:i-1,i+1) */
slarfg_(&i, &A(i,i+1), &A(1,i+1), &
c__1, &taui);
E(i) = A(i,i+1);
if (taui != 0.f) {
/* Apply H(i) from both sides to A(1:i,1:i) */
A(i,i+1) = 1.f;
/* Compute x := tau * A * v storing x in TAU(1:
i) */
ssymv_(uplo, &i, &taui, &A(1,1), lda, &A(1,i+1), &c__1, &c_b8, &TAU(1), &c__1);
/* Compute w := x - 1/2 * tau * (x'*v) * v */
alpha = taui * -.5f * sdot_(&i, &TAU(1), &c__1, &A(1,i+1), &c__1);
saxpy_(&i, &alpha, &A(1,i+1), &c__1, &TAU(1), &
c__1);
/* Apply the transformation as a rank-2 update:
A := A - v * w' - w * v' */
ssyr2_(uplo, &i, &c_b14, &A(1,i+1), &c__1, &
TAU(1), &c__1, &A(1,1), lda);
A(i,i+1) = E(i);
}
D(i + 1) = A(i+1,i+1);
TAU(i) = taui;
/* L10: */
}
D(1) = A(1,1);
} else {
/* Reduce the lower triangle of A */
i__1 = *n - 1;
for (i = 1; i <= *n-1; ++i) {
/* Generate elementary reflector H(i) = I - tau * v * v'
to annihilate A(i+2:n,i) */
i__2 = *n - i;
/* Computing MIN */
i__3 = i + 2;
slarfg_(&i__2, &A(i+1,i), &A(min(i+2,*n),i), &c__1, &taui);
E(i) = A(i+1,i);
if (taui != 0.f) {
/* Apply H(i) from both sides to A(i+1:n,i+1:n)
*/
A(i+1,i) = 1.f;
/* Compute x := tau * A * v storing y in TAU(i:
n-1) */
i__2 = *n - i;
ssymv_(uplo, &i__2, &taui, &A(i+1,i+1), lda,
&A(i+1,i), &c__1, &c_b8, &TAU(i), &c__1);
/* Compute w := x - 1/2 * tau * (x'*v) * v */
i__2 = *n - i;
alpha = taui * -.5f * sdot_(&i__2, &TAU(i), &c__1, &A(i+1,i), &c__1);
i__2 = *n - i;
saxpy_(&i__2, &alpha, &A(i+1,i), &c__1, &TAU(i),
&c__1);
/* Apply the transformation as a rank-2 update:
A := A - v * w' - w * v' */
i__2 = *n - i;
ssyr2_(uplo, &i__2, &c_b14, &A(i+1,i), &c__1, &
TAU(i), &c__1, &A(i+1,i+1), lda);
A(i+1,i) = E(i);
}
D(i) = A(i,i);
TAU(i) = taui;
/* L20: */
}
D(*n) = A(*n,*n);
}
return 0;
/* End of SSYTD2 */
} /* ssytd2_ */
/* Subroutine */ int ssytd2_(char *uplo, integer *n, real *a, integer *lda,
real *d__, real *e, real *tau, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, i__1, i__2, i__3;
/* Local variables */
integer i__;
real taui;
extern doublereal sdot_(integer *, real *, integer *, real *, integer *);
extern /* Subroutine */ int ssyr2_(char *, integer *, real *, real *,
integer *, real *, integer *, real *, integer *);
real alpha;
extern logical lsame_(char *, char *);
logical upper;
extern /* Subroutine */ int saxpy_(integer *, real *, real *, integer *,
real *, integer *), ssymv_(char *, integer *, real *, real *,
integer *, real *, integer *, real *, real *, integer *),
xerbla_(char *, integer *), slarfg_(integer *, real *,
real *, integer *, real *);
/* -- LAPACK routine (version 3.1) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* SSYTD2 reduces a real symmetric matrix A to symmetric tridiagonal */
/* form T by an orthogonal similarity transformation: Q' * A * Q = T. */
/* Arguments */
/* ========= */
/* UPLO (input) CHARACTER*1 */
/* Specifies whether the upper or lower triangular part of the */
/* symmetric matrix A is stored: */
/* = 'U': Upper triangular */
/* = 'L': Lower triangular */
/* N (input) INTEGER */
/* The order of the matrix A. N >= 0. */
/* A (input/output) REAL array, dimension (LDA,N) */
/* On entry, the symmetric matrix A. If UPLO = 'U', the leading */
/* n-by-n upper triangular part of A contains the upper */
/* triangular part of the matrix A, and the strictly lower */
/* triangular part of A is not referenced. If UPLO = 'L', the */
/* leading n-by-n lower triangular part of A contains the lower */
/* triangular part of the matrix A, and the strictly upper */
/* triangular part of A is not referenced. */
/* On exit, if UPLO = 'U', the diagonal and first superdiagonal */
/* of A are overwritten by the corresponding elements of the */
/* tridiagonal matrix T, and the elements above the first */
/* superdiagonal, with the array TAU, represent the orthogonal */
/* matrix Q as a product of elementary reflectors; if UPLO */
/* = 'L', the diagonal and first subdiagonal of A are over- */
/* written by the corresponding elements of the tridiagonal */
/* matrix T, and the elements below the first subdiagonal, with */
/* the array TAU, represent the orthogonal matrix Q as a product */
/* of elementary reflectors. See Further Details. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,N). */
/* D (output) REAL array, dimension (N) */
/* The diagonal elements of the tridiagonal matrix T: */
/* D(i) = A(i,i). */
/* E (output) REAL array, dimension (N-1) */
/* The off-diagonal elements of the tridiagonal matrix T: */
/* E(i) = A(i,i+1) if UPLO = 'U', E(i) = A(i+1,i) if UPLO = 'L'. */
/* TAU (output) REAL array, dimension (N-1) */
/* The scalar factors of the elementary reflectors (see Further */
/* Details). */
/* INFO (output) INTEGER */
/* = 0: successful exit */
/* < 0: if INFO = -i, the i-th argument had an illegal value. */
/* Further Details */
/* =============== */
/* If UPLO = 'U', the matrix Q is represented as a product of elementary */
/* reflectors */
/* Q = H(n-1) . . . H(2) H(1). */
/* Each H(i) has the form */
/* H(i) = I - tau * v * v' */
/* where tau is a real scalar, and v is a real vector with */
/* v(i+1:n) = 0 and v(i) = 1; v(1:i-1) is stored on exit in */
/* A(1:i-1,i+1), and tau in TAU(i). */
/* If UPLO = 'L', the matrix Q is represented as a product of elementary */
/* reflectors */
/* Q = H(1) H(2) . . . H(n-1). */
/* Each H(i) has the form */
/* H(i) = I - tau * v * v' */
/* where tau is a real scalar, and v is a real vector with */
/* v(1:i) = 0 and v(i+1) = 1; v(i+2:n) is stored on exit in A(i+2:n,i), */
/* and tau in TAU(i). */
/* The contents of A on exit are illustrated by the following examples */
/* with n = 5: */
/* if UPLO = 'U': if UPLO = 'L': */
/* ( d e v2 v3 v4 ) ( d ) */
/* ( d e v3 v4 ) ( e d ) */
/* ( d e v4 ) ( v1 e d ) */
/* ( d e ) ( v1 v2 e d ) */
/* ( d ) ( v1 v2 v3 e d ) */
/* where d and e denote diagonal and off-diagonal elements of T, and vi */
/* denotes an element of the vector defining H(i). */
/* ===================================================================== */
/* .. 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;
--d__;
--e;
--tau;
/* Function Body */
*info = 0;
upper = lsame_(uplo, "U");
if (! upper && ! lsame_(uplo, "L")) {
*info = -1;
} else if (*n < 0) {
*info = -2;
} else if (*lda < max(1,*n)) {
*info = -4;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SSYTD2", &i__1);
return 0;
}
/* Quick return if possible */
if (*n <= 0) {
return 0;
}
if (upper) {
/* Reduce the upper triangle of A */
for (i__ = *n - 1; i__ >= 1; --i__) {
/* Generate elementary reflector H(i) = I - tau * v * v' */
/* to annihilate A(1:i-1,i+1) */
slarfg_(&i__, &a[i__ + (i__ + 1) * a_dim1], &a[(i__ + 1) * a_dim1
+ 1], &c__1, &taui);
e[i__] = a[i__ + (i__ + 1) * a_dim1];
if (taui != 0.f) {
/* Apply H(i) from both sides to A(1:i,1:i) */
a[i__ + (i__ + 1) * a_dim1] = 1.f;
/* Compute x := tau * A * v storing x in TAU(1:i) */
ssymv_(uplo, &i__, &taui, &a[a_offset], lda, &a[(i__ + 1) *
a_dim1 + 1], &c__1, &c_b8, &tau[1], &c__1);
/* Compute w := x - 1/2 * tau * (x'*v) * v */
alpha = taui * -.5f * sdot_(&i__, &tau[1], &c__1, &a[(i__ + 1)
* a_dim1 + 1], &c__1);
saxpy_(&i__, &alpha, &a[(i__ + 1) * a_dim1 + 1], &c__1, &tau[
1], &c__1);
/* Apply the transformation as a rank-2 update: */
/* A := A - v * w' - w * v' */
ssyr2_(uplo, &i__, &c_b14, &a[(i__ + 1) * a_dim1 + 1], &c__1,
&tau[1], &c__1, &a[a_offset], lda);
a[i__ + (i__ + 1) * a_dim1] = e[i__];
}
d__[i__ + 1] = a[i__ + 1 + (i__ + 1) * a_dim1];
tau[i__] = taui;
/* L10: */
}
d__[1] = a[a_dim1 + 1];
} else {
/* Reduce the lower triangle of A */
i__1 = *n - 1;
for (i__ = 1; i__ <= i__1; ++i__) {
/* Generate elementary reflector H(i) = I - tau * v * v' */
/* to annihilate A(i+2:n,i) */
i__2 = *n - i__;
/* Computing MIN */
i__3 = i__ + 2;
slarfg_(&i__2, &a[i__ + 1 + i__ * a_dim1], &a[min(i__3, *n)+ i__ *
a_dim1], &c__1, &taui);
e[i__] = a[i__ + 1 + i__ * a_dim1];
if (taui != 0.f) {
/* Apply H(i) from both sides to A(i+1:n,i+1:n) */
a[i__ + 1 + i__ * a_dim1] = 1.f;
/* Compute x := tau * A * v storing y in TAU(i:n-1) */
i__2 = *n - i__;
ssymv_(uplo, &i__2, &taui, &a[i__ + 1 + (i__ + 1) * a_dim1],
lda, &a[i__ + 1 + i__ * a_dim1], &c__1, &c_b8, &tau[
i__], &c__1);
/* Compute w := x - 1/2 * tau * (x'*v) * v */
i__2 = *n - i__;
alpha = taui * -.5f * sdot_(&i__2, &tau[i__], &c__1, &a[i__ +
1 + i__ * a_dim1], &c__1);
i__2 = *n - i__;
saxpy_(&i__2, &alpha, &a[i__ + 1 + i__ * a_dim1], &c__1, &tau[
i__], &c__1);
/* Apply the transformation as a rank-2 update: */
/* A := A - v * w' - w * v' */
i__2 = *n - i__;
ssyr2_(uplo, &i__2, &c_b14, &a[i__ + 1 + i__ * a_dim1], &c__1,
&tau[i__], &c__1, &a[i__ + 1 + (i__ + 1) * a_dim1],
lda);
a[i__ + 1 + i__ * a_dim1] = e[i__];
}
d__[i__] = a[i__ + i__ * a_dim1];
tau[i__] = taui;
/* L20: */
}
d__[*n] = a[*n + *n * a_dim1];
}
return 0;
/* End of SSYTD2 */
} /* ssytd2_ */
/* Subroutine */ int stzrqf_(integer *m, integer *n, real *a, integer *lda,
real *tau, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, i__1, i__2;
real r__1;
/* Local variables */
integer i__, k, m1;
extern /* Subroutine */ int sger_(integer *, integer *, real *, real *,
integer *, real *, integer *, real *, integer *), sgemv_(char *,
integer *, integer *, real *, real *, integer *, real *, integer *
, real *, real *, integer *), scopy_(integer *, real *,
integer *, real *, integer *), saxpy_(integer *, real *, real *,
integer *, real *, integer *), xerbla_(char *, integer *),
slarfg_(integer *, real *, real *, integer *, real *);
/* -- LAPACK routine (version 3.1) -- */
/* 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 STZRZF. */
/* STZRQF reduces the M-by-N ( M<=N ) real upper trapezoidal matrix A */
/* to upper triangular form by means of orthogonal transformations. */
/* The upper trapezoidal matrix A is factored as */
/* A = ( R 0 ) * Z, */
/* where Z is an N-by-N orthogonal matrix and R is an M-by-M upper */
/* triangular matrix. */
/* 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 >= M. */
/* A (input/output) REAL array, dimension (LDA,N) */
/* On entry, the leading M-by-N upper trapezoidal part of the */
/* array A must contain the matrix to be factorized. */
/* On exit, the leading M-by-M upper triangular part of A */
/* contains the upper triangular matrix R, and elements M+1 to */
/* N of the first M rows of A, with the array TAU, represent the */
/* orthogonal matrix Z as a product of M elementary reflectors. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,M). */
/* TAU (output) REAL array, dimension (M) */
/* The scalar factors of the elementary reflectors. */
/* INFO (output) INTEGER */
/* = 0: successful exit */
/* < 0: if INFO = -i, the i-th argument had an illegal value */
/* Further Details */
/* =============== */
/* The factorization is obtained by Householder's method. The kth */
/* transformation matrix, Z( k ), which is used to introduce zeros into */
/* the ( m - k + 1 )th row of A, is given in the form */
/* Z( k ) = ( I 0 ), */
/* ( 0 T( k ) ) */
/* where */
/* T( k ) = I - tau*u( k )*u( k )', u( k ) = ( 1 ), */
/* ( 0 ) */
/* ( z( k ) ) */
/* tau is a scalar and z( k ) is an ( n - m ) element vector. */
/* tau and z( k ) are chosen to annihilate the elements of the kth row */
/* of X. */
/* The scalar tau is returned in the kth element of TAU and the vector */
/* u( k ) in the kth row of A, such that the elements of z( k ) are */
/* in a( k, m + 1 ), ..., a( k, n ). The elements of R are returned in */
/* the upper triangular part of A. */
/* Z is given by */
/* Z = Z( 1 ) * Z( 2 ) * ... * Z( m ). */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Executable Statements .. */
/* Test the input parameters. */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
--tau;
/* Function Body */
*info = 0;
if (*m < 0) {
*info = -1;
} else if (*n < *m) {
*info = -2;
} else if (*lda < max(1,*m)) {
*info = -4;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("STZRQF", &i__1);
return 0;
}
/* Perform the factorization. */
if (*m == 0) {
return 0;
}
if (*m == *n) {
i__1 = *n;
for (i__ = 1; i__ <= i__1; ++i__) {
tau[i__] = 0.f;
/* L10: */
}
} else {
/* Computing MIN */
i__1 = *m + 1;
m1 = min(i__1,*n);
for (k = *m; k >= 1; --k) {
/* Use a Householder reflection to zero the kth row of A. */
/* First set up the reflection. */
i__1 = *n - *m + 1;
slarfg_(&i__1, &a[k + k * a_dim1], &a[k + m1 * a_dim1], lda, &tau[
k]);
if (tau[k] != 0.f && k > 1) {
/* We now perform the operation A := A*P( k ). */
/* Use the first ( k - 1 ) elements of TAU to store a( k ), */
/* where a( k ) consists of the first ( k - 1 ) elements of */
/* the kth column of A. Also let B denote the first */
/* ( k - 1 ) rows of the last ( n - m ) columns of A. */
i__1 = k - 1;
scopy_(&i__1, &a[k * a_dim1 + 1], &c__1, &tau[1], &c__1);
/* Form w = a( k ) + B*z( k ) in TAU. */
i__1 = k - 1;
i__2 = *n - *m;
sgemv_("No transpose", &i__1, &i__2, &c_b8, &a[m1 * a_dim1 +
1], lda, &a[k + m1 * a_dim1], lda, &c_b8, &tau[1], &
c__1);
/* Now form a( k ) := a( k ) - tau*w */
/* and B := B - tau*w*z( k )'. */
i__1 = k - 1;
r__1 = -tau[k];
saxpy_(&i__1, &r__1, &tau[1], &c__1, &a[k * a_dim1 + 1], &
c__1);
i__1 = k - 1;
i__2 = *n - *m;
r__1 = -tau[k];
sger_(&i__1, &i__2, &r__1, &tau[1], &c__1, &a[k + m1 * a_dim1]
, lda, &a[m1 * a_dim1 + 1], lda);
}
/* L20: */
}
}
return 0;
/* End of STZRQF */
} /* stzrqf_ */
/* Subroutine */
int slaqr2_(logical *wantt, logical *wantz, integer *n, integer *ktop, integer *kbot, integer *nw, real *h__, integer *ldh, integer *iloz, integer *ihiz, real *z__, integer *ldz, integer *ns, integer *nd, real *sr, real *si, real *v, integer *ldv, integer *nh, real *t, integer *ldt, integer *nv, real *wv, integer *ldwv, real * work, integer *lwork)
{
/* System generated locals */
integer h_dim1, h_offset, t_dim1, t_offset, v_dim1, v_offset, wv_dim1, wv_offset, z_dim1, z_offset, i__1, i__2, i__3, i__4;
real r__1, r__2, r__3, r__4, r__5, r__6;
/* Builtin functions */
double sqrt(doublereal);
/* Local variables */
integer i__, j, k;
real s, aa, bb, cc, dd, cs, sn;
integer jw;
real evi, evk, foo;
integer kln;
real tau, ulp;
integer lwk1, lwk2;
real beta;
integer kend, kcol, info, ifst, ilst, ltop, krow;
logical bulge;
extern /* Subroutine */
int slarf_(char *, integer *, integer *, real *, integer *, real *, real *, integer *, real *), sgemm_( char *, char *, integer *, integer *, integer *, real *, real *, integer *, real *, integer *, real *, real *, integer *);
integer infqr;
extern /* Subroutine */
int scopy_(integer *, real *, integer *, real *, integer *);
integer kwtop;
extern /* Subroutine */
int slanv2_(real *, real *, real *, real *, real * , real *, real *, real *, real *, real *), slabad_(real *, real *) ;
extern real slamch_(char *);
extern /* Subroutine */
int sgehrd_(integer *, integer *, integer *, real *, integer *, real *, real *, integer *, integer *);
real safmin;
extern /* Subroutine */
int slarfg_(integer *, real *, real *, integer *, real *);
real safmax;
extern /* Subroutine */
int slahqr_(logical *, logical *, integer *, integer *, integer *, real *, integer *, real *, real *, integer * , integer *, real *, integer *, integer *), slacpy_(char *, integer *, integer *, real *, integer *, real *, integer *), slaset_(char *, integer *, integer *, real *, real *, real *, integer *);
logical sorted;
extern /* Subroutine */
int strexc_(char *, integer *, real *, integer *, real *, integer *, integer *, integer *, real *, integer *), sormhr_(char *, char *, integer *, integer *, integer *, integer *, real *, integer *, real *, real *, integer *, real *, integer *, integer *);
real smlnum;
integer lwkopt;
/* -- LAPACK auxiliary routine (version 3.4.2) -- */
/* -- LAPACK is a software package provided by Univ. of Tennessee, -- */
/* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- */
/* September 2012 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* ================================================================ */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* ==== Estimate optimal workspace. ==== */
/* Parameter adjustments */
h_dim1 = *ldh;
h_offset = 1 + h_dim1;
h__ -= h_offset;
z_dim1 = *ldz;
z_offset = 1 + z_dim1;
z__ -= z_offset;
--sr;
--si;
v_dim1 = *ldv;
v_offset = 1 + v_dim1;
v -= v_offset;
t_dim1 = *ldt;
t_offset = 1 + t_dim1;
t -= t_offset;
wv_dim1 = *ldwv;
wv_offset = 1 + wv_dim1;
wv -= wv_offset;
--work;
/* Function Body */
/* Computing MIN */
i__1 = *nw;
i__2 = *kbot - *ktop + 1; // , expr subst
jw = min(i__1,i__2);
if (jw <= 2)
{
lwkopt = 1;
}
else
{
/* ==== Workspace query call to SGEHRD ==== */
i__1 = jw - 1;
sgehrd_(&jw, &c__1, &i__1, &t[t_offset], ldt, &work[1], &work[1], & c_n1, &info);
lwk1 = (integer) work[1];
/* ==== Workspace query call to SORMHR ==== */
i__1 = jw - 1;
sormhr_("R", "N", &jw, &jw, &c__1, &i__1, &t[t_offset], ldt, &work[1], &v[v_offset], ldv, &work[1], &c_n1, &info);
lwk2 = (integer) work[1];
/* ==== Optimal workspace ==== */
lwkopt = jw + max(lwk1,lwk2);
}
/* ==== Quick return in case of workspace query. ==== */
if (*lwork == -1)
{
work[1] = (real) lwkopt;
return 0;
}
/* ==== Nothing to do ... */
/* ... for an empty active block ... ==== */
*ns = 0;
*nd = 0;
work[1] = 1.f;
if (*ktop > *kbot)
{
return 0;
}
/* ... nor for an empty deflation window. ==== */
if (*nw < 1)
{
return 0;
}
/* ==== Machine constants ==== */
safmin = slamch_("SAFE MINIMUM");
safmax = 1.f / safmin;
slabad_(&safmin, &safmax);
ulp = slamch_("PRECISION");
smlnum = safmin * ((real) (*n) / ulp);
/* ==== Setup deflation window ==== */
/* Computing MIN */
i__1 = *nw;
i__2 = *kbot - *ktop + 1; // , expr subst
jw = min(i__1,i__2);
kwtop = *kbot - jw + 1;
if (kwtop == *ktop)
{
s = 0.f;
}
else
{
s = h__[kwtop + (kwtop - 1) * h_dim1];
}
if (*kbot == kwtop)
{
/* ==== 1-by-1 deflation window: not much to do ==== */
sr[kwtop] = h__[kwtop + kwtop * h_dim1];
si[kwtop] = 0.f;
*ns = 1;
*nd = 0;
/* Computing MAX */
r__2 = smlnum;
r__3 = ulp * (r__1 = h__[kwtop + kwtop * h_dim1], abs( r__1)); // , expr subst
if (abs(s) <= max(r__2,r__3))
{
*ns = 0;
*nd = 1;
if (kwtop > *ktop)
{
h__[kwtop + (kwtop - 1) * h_dim1] = 0.f;
}
}
work[1] = 1.f;
return 0;
}
/* ==== Convert to spike-triangular form. (In case of a */
/* . rare QR failure, this routine continues to do */
/* . aggressive early deflation using that part of */
/* . the deflation window that converged using INFQR */
/* . here and there to keep track.) ==== */
slacpy_("U", &jw, &jw, &h__[kwtop + kwtop * h_dim1], ldh, &t[t_offset], ldt);
i__1 = jw - 1;
i__2 = *ldh + 1;
i__3 = *ldt + 1;
scopy_(&i__1, &h__[kwtop + 1 + kwtop * h_dim1], &i__2, &t[t_dim1 + 2], & i__3);
slaset_("A", &jw, &jw, &c_b12, &c_b13, &v[v_offset], ldv);
slahqr_(&c_true, &c_true, &jw, &c__1, &jw, &t[t_offset], ldt, &sr[kwtop], &si[kwtop], &c__1, &jw, &v[v_offset], ldv, &infqr);
/* ==== STREXC needs a clean margin near the diagonal ==== */
i__1 = jw - 3;
for (j = 1;
j <= i__1;
++j)
{
t[j + 2 + j * t_dim1] = 0.f;
t[j + 3 + j * t_dim1] = 0.f;
/* L10: */
}
if (jw > 2)
{
t[jw + (jw - 2) * t_dim1] = 0.f;
}
/* ==== Deflation detection loop ==== */
*ns = jw;
ilst = infqr + 1;
L20:
if (ilst <= *ns)
{
if (*ns == 1)
{
bulge = FALSE_;
}
else
{
bulge = t[*ns + (*ns - 1) * t_dim1] != 0.f;
}
/* ==== Small spike tip test for deflation ==== */
if (! bulge)
{
/* ==== Real eigenvalue ==== */
foo = (r__1 = t[*ns + *ns * t_dim1], abs(r__1));
if (foo == 0.f)
{
foo = abs(s);
}
/* Computing MAX */
r__2 = smlnum;
r__3 = ulp * foo; // , expr subst
if ((r__1 = s * v[*ns * v_dim1 + 1], abs(r__1)) <= max(r__2,r__3))
{
/* ==== Deflatable ==== */
--(*ns);
}
else
{
/* ==== Undeflatable. Move it up out of the way. */
/* . (STREXC can not fail in this case.) ==== */
ifst = *ns;
strexc_("V", &jw, &t[t_offset], ldt, &v[v_offset], ldv, &ifst, &ilst, &work[1], &info);
++ilst;
}
}
else
{
/* ==== Complex conjugate pair ==== */
foo = (r__3 = t[*ns + *ns * t_dim1], abs(r__3)) + sqrt((r__1 = t[* ns + (*ns - 1) * t_dim1], abs(r__1))) * sqrt((r__2 = t[* ns - 1 + *ns * t_dim1], abs(r__2)));
if (foo == 0.f)
{
foo = abs(s);
}
/* Computing MAX */
r__3 = (r__1 = s * v[*ns * v_dim1 + 1], abs(r__1));
r__4 = (r__2 = s * v[(*ns - 1) * v_dim1 + 1], abs(r__2)); // , expr subst
/* Computing MAX */
r__5 = smlnum;
r__6 = ulp * foo; // , expr subst
if (max(r__3,r__4) <= max(r__5,r__6))
{
/* ==== Deflatable ==== */
*ns += -2;
}
else
{
/* ==== Undeflatable. Move them up out of the way. */
/* . Fortunately, STREXC does the right thing with */
/* . ILST in case of a rare exchange failure. ==== */
ifst = *ns;
strexc_("V", &jw, &t[t_offset], ldt, &v[v_offset], ldv, &ifst, &ilst, &work[1], &info);
ilst += 2;
}
}
/* ==== End deflation detection loop ==== */
goto L20;
}
/* ==== Return to Hessenberg form ==== */
if (*ns == 0)
{
s = 0.f;
}
if (*ns < jw)
{
/* ==== sorting diagonal blocks of T improves accuracy for */
/* . graded matrices. Bubble sort deals well with */
/* . exchange failures. ==== */
sorted = FALSE_;
i__ = *ns + 1;
L30:
if (sorted)
{
goto L50;
}
sorted = TRUE_;
kend = i__ - 1;
i__ = infqr + 1;
if (i__ == *ns)
{
k = i__ + 1;
}
else if (t[i__ + 1 + i__ * t_dim1] == 0.f)
{
k = i__ + 1;
}
else
{
k = i__ + 2;
}
L40:
if (k <= kend)
{
if (k == i__ + 1)
{
evi = (r__1 = t[i__ + i__ * t_dim1], abs(r__1));
}
else
{
evi = (r__3 = t[i__ + i__ * t_dim1], abs(r__3)) + sqrt((r__1 = t[i__ + 1 + i__ * t_dim1], abs(r__1))) * sqrt((r__2 = t[i__ + (i__ + 1) * t_dim1], abs(r__2)));
}
if (k == kend)
{
evk = (r__1 = t[k + k * t_dim1], abs(r__1));
}
else if (t[k + 1 + k * t_dim1] == 0.f)
{
evk = (r__1 = t[k + k * t_dim1], abs(r__1));
}
else
{
evk = (r__3 = t[k + k * t_dim1], abs(r__3)) + sqrt((r__1 = t[ k + 1 + k * t_dim1], abs(r__1))) * sqrt((r__2 = t[k + (k + 1) * t_dim1], abs(r__2)));
}
if (evi >= evk)
{
i__ = k;
}
else
{
sorted = FALSE_;
ifst = i__;
ilst = k;
strexc_("V", &jw, &t[t_offset], ldt, &v[v_offset], ldv, &ifst, &ilst, &work[1], &info);
if (info == 0)
{
i__ = ilst;
}
else
{
i__ = k;
}
}
if (i__ == kend)
{
k = i__ + 1;
}
else if (t[i__ + 1 + i__ * t_dim1] == 0.f)
{
k = i__ + 1;
}
else
{
k = i__ + 2;
}
goto L40;
}
goto L30;
L50:
;
}
/* ==== Restore shift/eigenvalue array from T ==== */
i__ = jw;
L60:
if (i__ >= infqr + 1)
{
if (i__ == infqr + 1)
{
sr[kwtop + i__ - 1] = t[i__ + i__ * t_dim1];
si[kwtop + i__ - 1] = 0.f;
--i__;
}
else if (t[i__ + (i__ - 1) * t_dim1] == 0.f)
{
sr[kwtop + i__ - 1] = t[i__ + i__ * t_dim1];
si[kwtop + i__ - 1] = 0.f;
--i__;
}
else
{
aa = t[i__ - 1 + (i__ - 1) * t_dim1];
cc = t[i__ + (i__ - 1) * t_dim1];
bb = t[i__ - 1 + i__ * t_dim1];
dd = t[i__ + i__ * t_dim1];
slanv2_(&aa, &bb, &cc, &dd, &sr[kwtop + i__ - 2], &si[kwtop + i__ - 2], &sr[kwtop + i__ - 1], &si[kwtop + i__ - 1], &cs, & sn);
i__ += -2;
}
goto L60;
}
if (*ns < jw || s == 0.f)
{
if (*ns > 1 && s != 0.f)
{
/* ==== Reflect spike back into lower triangle ==== */
scopy_(ns, &v[v_offset], ldv, &work[1], &c__1);
beta = work[1];
slarfg_(ns, &beta, &work[2], &c__1, &tau);
work[1] = 1.f;
i__1 = jw - 2;
i__2 = jw - 2;
slaset_("L", &i__1, &i__2, &c_b12, &c_b12, &t[t_dim1 + 3], ldt);
slarf_("L", ns, &jw, &work[1], &c__1, &tau, &t[t_offset], ldt, & work[jw + 1]);
slarf_("R", ns, ns, &work[1], &c__1, &tau, &t[t_offset], ldt, & work[jw + 1]);
slarf_("R", &jw, ns, &work[1], &c__1, &tau, &v[v_offset], ldv, & work[jw + 1]);
i__1 = *lwork - jw;
sgehrd_(&jw, &c__1, ns, &t[t_offset], ldt, &work[1], &work[jw + 1] , &i__1, &info);
}
/* ==== Copy updated reduced window into place ==== */
if (kwtop > 1)
{
h__[kwtop + (kwtop - 1) * h_dim1] = s * v[v_dim1 + 1];
}
slacpy_("U", &jw, &jw, &t[t_offset], ldt, &h__[kwtop + kwtop * h_dim1] , ldh);
i__1 = jw - 1;
i__2 = *ldt + 1;
i__3 = *ldh + 1;
scopy_(&i__1, &t[t_dim1 + 2], &i__2, &h__[kwtop + 1 + kwtop * h_dim1], &i__3);
/* ==== Accumulate orthogonal matrix in order update */
/* . H and Z, if requested. ==== */
if (*ns > 1 && s != 0.f)
{
i__1 = *lwork - jw;
sormhr_("R", "N", &jw, ns, &c__1, ns, &t[t_offset], ldt, &work[1], &v[v_offset], ldv, &work[jw + 1], &i__1, &info);
}
/* ==== Update vertical slab in H ==== */
if (*wantt)
{
ltop = 1;
}
else
{
ltop = *ktop;
}
i__1 = kwtop - 1;
i__2 = *nv;
for (krow = ltop;
i__2 < 0 ? krow >= i__1 : krow <= i__1;
krow += i__2)
{
/* Computing MIN */
i__3 = *nv;
i__4 = kwtop - krow; // , expr subst
kln = min(i__3,i__4);
sgemm_("N", "N", &kln, &jw, &jw, &c_b13, &h__[krow + kwtop * h_dim1], ldh, &v[v_offset], ldv, &c_b12, &wv[wv_offset], ldwv);
slacpy_("A", &kln, &jw, &wv[wv_offset], ldwv, &h__[krow + kwtop * h_dim1], ldh);
/* L70: */
}
/* ==== Update horizontal slab in H ==== */
if (*wantt)
{
i__2 = *n;
i__1 = *nh;
for (kcol = *kbot + 1;
i__1 < 0 ? kcol >= i__2 : kcol <= i__2;
kcol += i__1)
{
/* Computing MIN */
i__3 = *nh;
i__4 = *n - kcol + 1; // , expr subst
kln = min(i__3,i__4);
sgemm_("C", "N", &jw, &kln, &jw, &c_b13, &v[v_offset], ldv, & h__[kwtop + kcol * h_dim1], ldh, &c_b12, &t[t_offset], ldt);
slacpy_("A", &jw, &kln, &t[t_offset], ldt, &h__[kwtop + kcol * h_dim1], ldh);
/* L80: */
}
}
/* ==== Update vertical slab in Z ==== */
if (*wantz)
{
i__1 = *ihiz;
i__2 = *nv;
for (krow = *iloz;
i__2 < 0 ? krow >= i__1 : krow <= i__1;
krow += i__2)
{
/* Computing MIN */
i__3 = *nv;
i__4 = *ihiz - krow + 1; // , expr subst
kln = min(i__3,i__4);
sgemm_("N", "N", &kln, &jw, &jw, &c_b13, &z__[krow + kwtop * z_dim1], ldz, &v[v_offset], ldv, &c_b12, &wv[ wv_offset], ldwv);
slacpy_("A", &kln, &jw, &wv[wv_offset], ldwv, &z__[krow + kwtop * z_dim1], ldz);
/* L90: */
}
}
}
/* ==== Return the number of deflations ... ==== */
*nd = jw - *ns;
/* ==== ... and the number of shifts. (Subtracting */
/* . INFQR from the spike length takes care */
/* . of the case of a rare QR failure while */
/* . calculating eigenvalues of the deflation */
/* . window.) ==== */
*ns -= infqr;
/* ==== Return optimal workspace. ==== */
work[1] = (real) lwkopt;
/* ==== End of SLAQR2 ==== */
return 0;
}
/* Subroutine */ int sgehd2_(integer *n, integer *ilo, integer *ihi, real *a,
integer *lda, real *tau, real *work, integer *info)
{
/* -- LAPACK routine (version 3.0) --
Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd.,
Courant Institute, Argonne National Lab, and Rice University
October 31, 1992
Purpose
=======
SGEHD2 reduces a real general matrix A to upper Hessenberg form H by
an orthogonal similarity transformation: Q' * A * Q = H .
Arguments
=========
N (input) INTEGER
The order of the matrix A. N >= 0.
ILO (input) INTEGER
IHI (input) INTEGER
It is assumed that A is already upper triangular in rows
and columns 1:ILO-1 and IHI+1:N. ILO and IHI are normally
set by a previous call to SGEBAL; otherwise they should be
set to 1 and N respectively. See Further Details.
1 <= ILO <= IHI <= max(1,N).
A (input/output) REAL array, dimension (LDA,N)
On entry, the n by n general matrix to be reduced.
On exit, the upper triangle and the first subdiagonal of A
are overwritten with the upper Hessenberg matrix H, and the
elements below the first subdiagonal, with the array TAU,
represent the orthogonal matrix Q as a product of elementary
reflectors. See Further Details.
LDA (input) INTEGER
The leading dimension of the array A. LDA >= max(1,N).
TAU (output) REAL array, dimension (N-1)
The scalar factors of the elementary reflectors (see Further
Details).
WORK (workspace) REAL array, dimension (N)
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 (ihi-ilo) elementary
reflectors
Q = H(ilo) H(ilo+1) . . . H(ihi-1).
Each H(i) has the form
H(i) = I - tau * v * v'
where tau is a real scalar, and v is a real vector with
v(1:i) = 0, v(i+1) = 1 and v(ihi+1:n) = 0; v(i+2:ihi) is stored on
exit in A(i+2:ihi,i), and tau in TAU(i).
The contents of A are illustrated by the following example, with
n = 7, ilo = 2 and ihi = 6:
on entry, on exit,
( a a a a a a a ) ( a a h h h h a )
( a a a a a a ) ( a h h h h a )
( a a a a a a ) ( h h h h h h )
( a a a a a a ) ( v2 h h h h h )
( a a a a a a ) ( v2 v3 h h h h )
( a a a a a a ) ( v2 v3 v4 h h h )
( a ) ( a )
where a denotes an element of the original matrix A, h denotes a
modified element of the upper Hessenberg matrix H, and vi denotes an
element of the vector defining H(i).
=====================================================================
Test the input parameters
Parameter adjustments */
/* Table of constant values */
static integer c__1 = 1;
/* System generated locals */
integer a_dim1, a_offset, i__1, i__2, i__3;
/* Local variables */
static integer i__;
extern /* Subroutine */ int slarf_(char *, integer *, integer *, real *,
integer *, real *, real *, integer *, real *), xerbla_(
char *, integer *), slarfg_(integer *, real *, real *,
integer *, real *);
static real aii;
#define a_ref(a_1,a_2) a[(a_2)*a_dim1 + a_1]
a_dim1 = *lda;
a_offset = 1 + a_dim1 * 1;
a -= a_offset;
--tau;
--work;
/* Function Body */
*info = 0;
if (*n < 0) {
*info = -1;
} else if (*ilo < 1 || *ilo > max(1,*n)) {
*info = -2;
} else if (*ihi < min(*ilo,*n) || *ihi > *n) {
*info = -3;
} else if (*lda < max(1,*n)) {
*info = -5;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SGEHD2", &i__1);
return 0;
}
i__1 = *ihi - 1;
for (i__ = *ilo; i__ <= i__1; ++i__) {
/* Compute elementary reflector H(i) to annihilate A(i+2:ihi,i)
Computing MIN */
i__2 = i__ + 2;
i__3 = *ihi - i__;
slarfg_(&i__3, &a_ref(i__ + 1, i__), &a_ref(min(i__2,*n), i__), &c__1,
&tau[i__]);
aii = a_ref(i__ + 1, i__);
a_ref(i__ + 1, i__) = 1.f;
/* Apply H(i) to A(1:ihi,i+1:ihi) from the right */
i__2 = *ihi - i__;
slarf_("Right", ihi, &i__2, &a_ref(i__ + 1, i__), &c__1, &tau[i__], &
a_ref(1, i__ + 1), lda, &work[1]);
/* Apply H(i) to A(i+1:ihi,i+1:n) from the left */
i__2 = *ihi - i__;
i__3 = *n - i__;
slarf_("Left", &i__2, &i__3, &a_ref(i__ + 1, i__), &c__1, &tau[i__], &
a_ref(i__ + 1, i__ + 1), lda, &work[1]);
a_ref(i__ + 1, i__) = aii;
/* L10: */
}
return 0;
/* End of SGEHD2 */
} /* sgehd2_ */
/* Subroutine */ int slaqp2_(integer *m, integer *n, integer *offset, real *a,
integer *lda, integer *jpvt, real *tau, real *vn1, real *vn2, real *
work)
{
/* System generated locals */
integer a_dim1, a_offset, i__1, i__2, i__3;
real r__1, r__2;
/* Builtin functions */
double sqrt(doublereal);
/* Local variables */
static integer i__, j, mn;
static real aii;
static integer pvt;
static real temp, temp2;
extern doublereal snrm2_(integer *, real *, integer *);
static integer offpi;
extern /* Subroutine */ int slarf_(char *, integer *, integer *, real *,
integer *, real *, real *, integer *, real *, ftnlen);
static integer itemp;
extern /* Subroutine */ int sswap_(integer *, real *, integer *, real *,
integer *), slarfg_(integer *, real *, real *, integer *, real *);
extern integer isamax_(integer *, real *, integer *);
/* -- LAPACK auxiliary 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 */
/* ======= */
/* SLAQP2 computes a QR factorization with column pivoting of */
/* the block A(OFFSET+1:M,1:N). */
/* The block A(1:OFFSET,1:N) is accordingly pivoted, but not factorized. */
/* 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. */
/* OFFSET (input) INTEGER */
/* The number of rows of the matrix A that must be pivoted */
/* but no factorized. OFFSET >= 0. */
/* A (input/output) REAL array, dimension (LDA,N) */
/* On entry, the M-by-N matrix A. */
/* On exit, the upper triangle of block A(OFFSET+1:M,1:N) is */
/* the triangular factor obtained; the elements in block */
/* A(OFFSET+1:M,1:N) below the diagonal, together with the */
/* array TAU, represent the orthogonal matrix Q as a product of */
/* elementary reflectors. Block A(1:OFFSET,1:N) has been */
/* accordingly pivoted, but no factorized. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,M). */
/* 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 A*P (a leading column); if JPVT(i) = 0, */
/* the i-th column of A is a free column. */
/* On exit, if JPVT(i) = k, then the i-th column of A*P */
/* was the k-th column of A. */
/* TAU (output) REAL array, dimension (min(M,N)) */
/* The scalar factors of the elementary reflectors. */
/* VN1 (input/output) REAL array, dimension (N) */
/* The vector with the partial column norms. */
/* VN2 (input/output) REAL array, dimension (N) */
/* The vector with the exact column norms. */
/* WORK (workspace) REAL array, dimension (N) */
/* Further Details */
/* =============== */
/* Based on contributions by */
/* G. Quintana-Orti, Depto. de Informatica, Universidad Jaime I, Spain */
/* X. Sun, Computer Science Dept., Duke University, USA */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
--jpvt;
--tau;
--vn1;
--vn2;
--work;
/* Function Body */
/* Computing MIN */
i__1 = *m - *offset;
mn = min(i__1,*n);
/* Compute factorization. */
i__1 = mn;
for (i__ = 1; i__ <= i__1; ++i__) {
offpi = *offset + i__;
/* Determine ith pivot column and swap if necessary. */
i__2 = *n - i__ + 1;
pvt = i__ - 1 + isamax_(&i__2, &vn1[i__], &c__1);
if (pvt != i__) {
sswap_(m, &a[pvt * a_dim1 + 1], &c__1, &a[i__ * a_dim1 + 1], &
c__1);
itemp = jpvt[pvt];
jpvt[pvt] = jpvt[i__];
jpvt[i__] = itemp;
vn1[pvt] = vn1[i__];
vn2[pvt] = vn2[i__];
}
/* Generate elementary reflector H(i). */
if (offpi < *m) {
i__2 = *m - offpi + 1;
slarfg_(&i__2, &a[offpi + i__ * a_dim1], &a[offpi + 1 + i__ *
a_dim1], &c__1, &tau[i__]);
} else {
slarfg_(&c__1, &a[*m + i__ * a_dim1], &a[*m + i__ * a_dim1], &
c__1, &tau[i__]);
}
if (i__ < *n) {
/* Apply H(i)' to A(offset+i:m,i+1:n) from the left. */
aii = a[offpi + i__ * a_dim1];
a[offpi + i__ * a_dim1] = 1.f;
i__2 = *m - offpi + 1;
i__3 = *n - i__;
slarf_("Left", &i__2, &i__3, &a[offpi + i__ * a_dim1], &c__1, &
tau[i__], &a[offpi + (i__ + 1) * a_dim1], lda, &work[1], (
ftnlen)4);
a[offpi + i__ * a_dim1] = aii;
}
/* Update partial column norms. */
i__2 = *n;
for (j = i__ + 1; j <= i__2; ++j) {
if (vn1[j] != 0.f) {
/* Computing 2nd power */
r__2 = (r__1 = a[offpi + j * a_dim1], dabs(r__1)) / vn1[j];
temp = 1.f - r__2 * r__2;
temp = dmax(temp,0.f);
/* Computing 2nd power */
r__1 = vn1[j] / vn2[j];
temp2 = temp * .05f * (r__1 * r__1) + 1.f;
if (temp2 == 1.f) {
if (offpi < *m) {
i__3 = *m - offpi;
vn1[j] = snrm2_(&i__3, &a[offpi + 1 + j * a_dim1], &
c__1);
vn2[j] = vn1[j];
} else {
vn1[j] = 0.f;
vn2[j] = 0.f;
}
} else {
vn1[j] *= sqrt(temp);
}
}
/* L10: */
}
/* L20: */
}
return 0;
/* End of SLAQP2 */
} /* slaqp2_ */
/* Subroutine */ int slabrd_(integer *m, integer *n, integer *nb, real *a,
integer *lda, real *d__, real *e, real *tauq, real *taup, real *x,
integer *ldx, real *y, integer *ldy)
{
/* -- LAPACK auxiliary routine (version 3.0) --
Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd.,
Courant Institute, Argonne National Lab, and Rice University
February 29, 1992
Purpose
=======
SLABRD reduces the first NB rows and columns of a real general
m by n matrix A to upper or lower bidiagonal form by an orthogonal
transformation Q' * A * P, and returns the matrices X and Y which
are needed to apply the transformation to the unreduced part of A.
If m >= n, A is reduced to upper bidiagonal form; if m < n, to lower
bidiagonal form.
This is an auxiliary routine called by SGEBRD
Arguments
=========
M (input) INTEGER
The number of rows in the matrix A.
N (input) INTEGER
The number of columns in the matrix A.
NB (input) INTEGER
The number of leading rows and columns of A to be reduced.
A (input/output) REAL array, dimension (LDA,N)
On entry, the m by n general matrix to be reduced.
On exit, the first NB rows and columns of the matrix are
overwritten; the rest of the array is unchanged.
If m >= n, elements on and below the diagonal in the first NB
columns, with the array TAUQ, represent the orthogonal
matrix Q as a product of elementary reflectors; and
elements above the diagonal in the first NB rows, with the
array TAUP, represent the orthogonal matrix P as a product
of elementary reflectors.
If m < n, elements below the diagonal in the first NB
columns, with the array TAUQ, represent the orthogonal
matrix Q as a product of elementary reflectors, and
elements on and above the diagonal in the first NB rows,
with the array TAUP, represent the orthogonal matrix P as
a product of elementary reflectors.
See Further Details.
LDA (input) INTEGER
The leading dimension of the array A. LDA >= max(1,M).
D (output) REAL array, dimension (NB)
The diagonal elements of the first NB rows and columns of
the reduced matrix. D(i) = A(i,i).
E (output) REAL array, dimension (NB)
The off-diagonal elements of the first NB rows and columns of
the reduced matrix.
TAUQ (output) REAL array dimension (NB)
The scalar factors of the elementary reflectors which
represent the orthogonal matrix Q. See Further Details.
TAUP (output) REAL array, dimension (NB)
The scalar factors of the elementary reflectors which
represent the orthogonal matrix P. See Further Details.
X (output) REAL array, dimension (LDX,NB)
The m-by-nb matrix X required to update the unreduced part
of A.
LDX (input) INTEGER
The leading dimension of the array X. LDX >= M.
Y (output) REAL array, dimension (LDY,NB)
The n-by-nb matrix Y required to update the unreduced part
of A.
LDY (output) INTEGER
The leading dimension of the array Y. LDY >= N.
Further Details
===============
The matrices Q and P are represented as products of elementary
reflectors:
Q = H(1) H(2) . . . H(nb) and P = G(1) G(2) . . . G(nb)
Each H(i) and G(i) has the form:
H(i) = I - tauq * v * v' and G(i) = I - taup * u * u'
where tauq and taup are real scalars, and v and u are real vectors.
If m >= n, v(1:i-1) = 0, v(i) = 1, and v(i:m) is stored on exit in
A(i:m,i); u(1:i) = 0, u(i+1) = 1, and u(i+1:n) is stored on exit in
A(i,i+1:n); tauq is stored in TAUQ(i) and taup in TAUP(i).
If m < n, v(1:i) = 0, v(i+1) = 1, and v(i+1:m) is stored on exit in
A(i+2:m,i); u(1:i-1) = 0, u(i) = 1, and u(i:n) is stored on exit in
A(i,i+1:n); tauq is stored in TAUQ(i) and taup in TAUP(i).
The elements of the vectors v and u together form the m-by-nb matrix
V and the nb-by-n matrix U' which are needed, with X and Y, to apply
the transformation to the unreduced part of the matrix, using a block
update of the form: A := A - V*Y' - X*U'.
The contents of A on exit are illustrated by the following examples
with nb = 2:
m = 6 and n = 5 (m > n): m = 5 and n = 6 (m < n):
( 1 1 u1 u1 u1 ) ( 1 u1 u1 u1 u1 u1 )
( v1 1 1 u2 u2 ) ( 1 1 u2 u2 u2 u2 )
( v1 v2 a a a ) ( v1 1 a a a a )
( v1 v2 a a a ) ( v1 v2 a a a a )
( v1 v2 a a a ) ( v1 v2 a a a a )
( v1 v2 a a a )
where a denotes an element of the original matrix which is unchanged,
vi denotes an element of the vector defining H(i), and ui an element
of the vector defining G(i).
=====================================================================
Quick return if possible
Parameter adjustments */
/* Table of constant values */
static real c_b4 = -1.f;
static real c_b5 = 1.f;
static integer c__1 = 1;
static real c_b16 = 0.f;
/* System generated locals */
integer a_dim1, a_offset, x_dim1, x_offset, y_dim1, y_offset, i__1, i__2,
i__3;
/* Local variables */
static integer i__;
extern /* Subroutine */ int sscal_(integer *, real *, real *, integer *),
sgemv_(char *, integer *, integer *, real *, real *, integer *,
real *, integer *, real *, real *, integer *), slarfg_(
integer *, real *, real *, integer *, real *);
#define a_ref(a_1,a_2) a[(a_2)*a_dim1 + a_1]
#define x_ref(a_1,a_2) x[(a_2)*x_dim1 + a_1]
#define y_ref(a_1,a_2) y[(a_2)*y_dim1 + a_1]
a_dim1 = *lda;
a_offset = 1 + a_dim1 * 1;
a -= a_offset;
--d__;
--e;
--tauq;
--taup;
x_dim1 = *ldx;
x_offset = 1 + x_dim1 * 1;
x -= x_offset;
y_dim1 = *ldy;
y_offset = 1 + y_dim1 * 1;
y -= y_offset;
/* Function Body */
if (*m <= 0 || *n <= 0) {
return 0;
}
if (*m >= *n) {
/* Reduce to upper bidiagonal form */
i__1 = *nb;
for (i__ = 1; i__ <= i__1; ++i__) {
/* Update A(i:m,i) */
i__2 = *m - i__ + 1;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &a_ref(i__, 1), lda, &
y_ref(i__, 1), ldy, &c_b5, &a_ref(i__, i__), &c__1);
i__2 = *m - i__ + 1;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &x_ref(i__, 1), ldx, &
a_ref(1, i__), &c__1, &c_b5, &a_ref(i__, i__), &c__1);
/* Generate reflection Q(i) to annihilate A(i+1:m,i)
Computing MIN */
i__2 = i__ + 1;
i__3 = *m - i__ + 1;
slarfg_(&i__3, &a_ref(i__, i__), &a_ref(min(i__2,*m), i__), &c__1,
&tauq[i__]);
d__[i__] = a_ref(i__, i__);
if (i__ < *n) {
a_ref(i__, i__) = 1.f;
/* Compute Y(i+1:n,i) */
i__2 = *m - i__ + 1;
i__3 = *n - i__;
sgemv_("Transpose", &i__2, &i__3, &c_b5, &a_ref(i__, i__ + 1),
lda, &a_ref(i__, i__), &c__1, &c_b16, &y_ref(i__ + 1,
i__), &c__1);
i__2 = *m - i__ + 1;
i__3 = i__ - 1;
sgemv_("Transpose", &i__2, &i__3, &c_b5, &a_ref(i__, 1), lda,
&a_ref(i__, i__), &c__1, &c_b16, &y_ref(1, i__), &
c__1);
i__2 = *n - i__;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &y_ref(i__ + 1, 1)
, ldy, &y_ref(1, i__), &c__1, &c_b5, &y_ref(i__ + 1,
i__), &c__1);
i__2 = *m - i__ + 1;
i__3 = i__ - 1;
sgemv_("Transpose", &i__2, &i__3, &c_b5, &x_ref(i__, 1), ldx,
&a_ref(i__, i__), &c__1, &c_b16, &y_ref(1, i__), &
c__1);
i__2 = i__ - 1;
i__3 = *n - i__;
sgemv_("Transpose", &i__2, &i__3, &c_b4, &a_ref(1, i__ + 1),
lda, &y_ref(1, i__), &c__1, &c_b5, &y_ref(i__ + 1,
i__), &c__1);
i__2 = *n - i__;
sscal_(&i__2, &tauq[i__], &y_ref(i__ + 1, i__), &c__1);
/* Update A(i,i+1:n) */
i__2 = *n - i__;
sgemv_("No transpose", &i__2, &i__, &c_b4, &y_ref(i__ + 1, 1),
ldy, &a_ref(i__, 1), lda, &c_b5, &a_ref(i__, i__ + 1)
, lda);
i__2 = i__ - 1;
i__3 = *n - i__;
sgemv_("Transpose", &i__2, &i__3, &c_b4, &a_ref(1, i__ + 1),
lda, &x_ref(i__, 1), ldx, &c_b5, &a_ref(i__, i__ + 1),
lda);
/* Generate reflection P(i) to annihilate A(i,i+2:n)
Computing MIN */
i__2 = i__ + 2;
i__3 = *n - i__;
slarfg_(&i__3, &a_ref(i__, i__ + 1), &a_ref(i__, min(i__2,*n))
, lda, &taup[i__]);
e[i__] = a_ref(i__, i__ + 1);
a_ref(i__, i__ + 1) = 1.f;
/* Compute X(i+1:m,i) */
i__2 = *m - i__;
i__3 = *n - i__;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &a_ref(i__ + 1,
i__ + 1), lda, &a_ref(i__, i__ + 1), lda, &c_b16, &
x_ref(i__ + 1, i__), &c__1);
i__2 = *n - i__;
sgemv_("Transpose", &i__2, &i__, &c_b5, &y_ref(i__ + 1, 1),
ldy, &a_ref(i__, i__ + 1), lda, &c_b16, &x_ref(1, i__)
, &c__1);
i__2 = *m - i__;
sgemv_("No transpose", &i__2, &i__, &c_b4, &a_ref(i__ + 1, 1),
lda, &x_ref(1, i__), &c__1, &c_b5, &x_ref(i__ + 1,
i__), &c__1);
i__2 = i__ - 1;
i__3 = *n - i__;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &a_ref(1, i__ + 1)
, lda, &a_ref(i__, i__ + 1), lda, &c_b16, &x_ref(1,
i__), &c__1);
i__2 = *m - i__;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &x_ref(i__ + 1, 1)
, ldx, &x_ref(1, i__), &c__1, &c_b5, &x_ref(i__ + 1,
i__), &c__1);
i__2 = *m - i__;
sscal_(&i__2, &taup[i__], &x_ref(i__ + 1, i__), &c__1);
}
/* L10: */
}
} else {
/* Reduce to lower bidiagonal form */
i__1 = *nb;
for (i__ = 1; i__ <= i__1; ++i__) {
/* Update A(i,i:n) */
i__2 = *n - i__ + 1;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &y_ref(i__, 1), ldy, &
a_ref(i__, 1), lda, &c_b5, &a_ref(i__, i__), lda);
i__2 = i__ - 1;
i__3 = *n - i__ + 1;
sgemv_("Transpose", &i__2, &i__3, &c_b4, &a_ref(1, i__), lda, &
x_ref(i__, 1), ldx, &c_b5, &a_ref(i__, i__), lda);
/* Generate reflection P(i) to annihilate A(i,i+1:n)
Computing MIN */
i__2 = i__ + 1;
i__3 = *n - i__ + 1;
slarfg_(&i__3, &a_ref(i__, i__), &a_ref(i__, min(i__2,*n)), lda, &
taup[i__]);
d__[i__] = a_ref(i__, i__);
if (i__ < *m) {
a_ref(i__, i__) = 1.f;
/* Compute X(i+1:m,i) */
i__2 = *m - i__;
i__3 = *n - i__ + 1;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &a_ref(i__ + 1,
i__), lda, &a_ref(i__, i__), lda, &c_b16, &x_ref(i__
+ 1, i__), &c__1);
i__2 = *n - i__ + 1;
i__3 = i__ - 1;
sgemv_("Transpose", &i__2, &i__3, &c_b5, &y_ref(i__, 1), ldy,
&a_ref(i__, i__), lda, &c_b16, &x_ref(1, i__), &c__1);
i__2 = *m - i__;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &a_ref(i__ + 1, 1)
, lda, &x_ref(1, i__), &c__1, &c_b5, &x_ref(i__ + 1,
i__), &c__1);
i__2 = i__ - 1;
i__3 = *n - i__ + 1;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &a_ref(1, i__),
lda, &a_ref(i__, i__), lda, &c_b16, &x_ref(1, i__), &
c__1);
i__2 = *m - i__;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &x_ref(i__ + 1, 1)
, ldx, &x_ref(1, i__), &c__1, &c_b5, &x_ref(i__ + 1,
i__), &c__1);
i__2 = *m - i__;
sscal_(&i__2, &taup[i__], &x_ref(i__ + 1, i__), &c__1);
/* Update A(i+1:m,i) */
i__2 = *m - i__;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &a_ref(i__ + 1, 1)
, lda, &y_ref(i__, 1), ldy, &c_b5, &a_ref(i__ + 1,
i__), &c__1);
i__2 = *m - i__;
sgemv_("No transpose", &i__2, &i__, &c_b4, &x_ref(i__ + 1, 1),
ldx, &a_ref(1, i__), &c__1, &c_b5, &a_ref(i__ + 1,
i__), &c__1);
/* Generate reflection Q(i) to annihilate A(i+2:m,i)
Computing MIN */
i__2 = i__ + 2;
i__3 = *m - i__;
slarfg_(&i__3, &a_ref(i__ + 1, i__), &a_ref(min(i__2,*m), i__)
, &c__1, &tauq[i__]);
e[i__] = a_ref(i__ + 1, i__);
a_ref(i__ + 1, i__) = 1.f;
/* Compute Y(i+1:n,i) */
i__2 = *m - i__;
i__3 = *n - i__;
sgemv_("Transpose", &i__2, &i__3, &c_b5, &a_ref(i__ + 1, i__
+ 1), lda, &a_ref(i__ + 1, i__), &c__1, &c_b16, &
y_ref(i__ + 1, i__), &c__1);
i__2 = *m - i__;
i__3 = i__ - 1;
sgemv_("Transpose", &i__2, &i__3, &c_b5, &a_ref(i__ + 1, 1),
lda, &a_ref(i__ + 1, i__), &c__1, &c_b16, &y_ref(1,
i__), &c__1);
i__2 = *n - i__;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &y_ref(i__ + 1, 1)
, ldy, &y_ref(1, i__), &c__1, &c_b5, &y_ref(i__ + 1,
i__), &c__1);
i__2 = *m - i__;
sgemv_("Transpose", &i__2, &i__, &c_b5, &x_ref(i__ + 1, 1),
ldx, &a_ref(i__ + 1, i__), &c__1, &c_b16, &y_ref(1,
i__), &c__1);
i__2 = *n - i__;
sgemv_("Transpose", &i__, &i__2, &c_b4, &a_ref(1, i__ + 1),
lda, &y_ref(1, i__), &c__1, &c_b5, &y_ref(i__ + 1,
i__), &c__1);
i__2 = *n - i__;
sscal_(&i__2, &tauq[i__], &y_ref(i__ + 1, i__), &c__1);
}
/* L20: */
}
}
return 0;
/* End of SLABRD */
} /* slabrd_ */
/* Subroutine */ int sgehd2_(integer *n, integer *ilo, integer *ihi, real *a,
integer *lda, real *tau, real *work, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, i__1, i__2, i__3;
/* Local variables */
integer i__;
real aii;
extern /* Subroutine */ int slarf_(char *, integer *, integer *, real *,
integer *, real *, real *, integer *, real *), xerbla_(
char *, integer *), slarfg_(integer *, real *, real *,
integer *, real *);
/* -- LAPACK routine (version 3.2) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* SGEHD2 reduces a real general matrix A to upper Hessenberg form H by */
/* an orthogonal similarity transformation: Q' * A * Q = H . */
/* Arguments */
/* ========= */
/* N (input) INTEGER */
/* The order of the matrix A. N >= 0. */
/* ILO (input) INTEGER */
/* IHI (input) INTEGER */
/* It is assumed that A is already upper triangular in rows */
/* and columns 1:ILO-1 and IHI+1:N. ILO and IHI are normally */
/* set by a previous call to SGEBAL; otherwise they should be */
/* set to 1 and N respectively. See Further Details. */
/* 1 <= ILO <= IHI <= max(1,N). */
/* A (input/output) REAL array, dimension (LDA,N) */
/* On entry, the n by n general matrix to be reduced. */
/* On exit, the upper triangle and the first subdiagonal of A */
/* are overwritten with the upper Hessenberg matrix H, and the */
/* elements below the first subdiagonal, with the array TAU, */
/* represent the orthogonal matrix Q as a product of elementary */
/* reflectors. See Further Details. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,N). */
/* TAU (output) REAL array, dimension (N-1) */
/* The scalar factors of the elementary reflectors (see Further */
/* Details). */
/* WORK (workspace) REAL array, dimension (N) */
/* 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 (ihi-ilo) elementary */
/* reflectors */
/* Q = H(ilo) H(ilo+1) . . . H(ihi-1). */
/* Each H(i) has the form */
/* H(i) = I - tau * v * v' */
/* where tau is a real scalar, and v is a real vector with */
/* v(1:i) = 0, v(i+1) = 1 and v(ihi+1:n) = 0; v(i+2:ihi) is stored on */
/* exit in A(i+2:ihi,i), and tau in TAU(i). */
/* The contents of A are illustrated by the following example, with */
/* n = 7, ilo = 2 and ihi = 6: */
/* on entry, on exit, */
/* ( a a a a a a a ) ( a a h h h h a ) */
/* ( a a a a a a ) ( a h h h h a ) */
/* ( a a a a a a ) ( h h h h h h ) */
/* ( a a a a a a ) ( v2 h h h h h ) */
/* ( a a a a a a ) ( v2 v3 h h h h ) */
/* ( a a a a a a ) ( v2 v3 v4 h h h ) */
/* ( a ) ( a ) */
/* where a denotes an element of the original matrix A, h denotes a */
/* modified element of the upper Hessenberg matrix H, and vi denotes an */
/* element of the vector defining H(i). */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Test the input parameters */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
--tau;
--work;
/* Function Body */
*info = 0;
if (*n < 0) {
*info = -1;
} else if (*ilo < 1 || *ilo > max(1,*n)) {
*info = -2;
} else if (*ihi < min(*ilo,*n) || *ihi > *n) {
*info = -3;
} else if (*lda < max(1,*n)) {
*info = -5;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SGEHD2", &i__1);
return 0;
}
i__1 = *ihi - 1;
for (i__ = *ilo; i__ <= i__1; ++i__) {
/* Compute elementary reflector H(i) to annihilate A(i+2:ihi,i) */
i__2 = *ihi - i__;
/* Computing MIN */
i__3 = i__ + 2;
slarfg_(&i__2, &a[i__ + 1 + i__ * a_dim1], &a[min(i__3, *n)+ i__ *
a_dim1], &c__1, &tau[i__]);
aii = a[i__ + 1 + i__ * a_dim1];
a[i__ + 1 + i__ * a_dim1] = 1.f;
/* Apply H(i) to A(1:ihi,i+1:ihi) from the right */
i__2 = *ihi - i__;
slarf_("Right", ihi, &i__2, &a[i__ + 1 + i__ * a_dim1], &c__1, &tau[
i__], &a[(i__ + 1) * a_dim1 + 1], lda, &work[1]);
/* Apply H(i) to A(i+1:ihi,i+1:n) from the left */
i__2 = *ihi - i__;
i__3 = *n - i__;
slarf_("Left", &i__2, &i__3, &a[i__ + 1 + i__ * a_dim1], &c__1, &tau[
i__], &a[i__ + 1 + (i__ + 1) * a_dim1], lda, &work[1]);
a[i__ + 1 + i__ * a_dim1] = aii;
/* L10: */
}
return 0;
/* End of SGEHD2 */
} /* sgehd2_ */
/* Subroutine */ int slaexc_(logical *wantq, integer *n, real *t, integer *
ldt, real *q, integer *ldq, integer *j1, integer *n1, integer *n2,
real *work, integer *info)
{
/* -- LAPACK auxiliary routine (version 3.0) --
Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd.,
Courant Institute, Argonne National Lab, and Rice University
February 29, 1992
Purpose
=======
SLAEXC swaps adjacent diagonal blocks T11 and T22 of order 1 or 2 in
an upper quasi-triangular matrix T by an orthogonal similarity
transformation.
T must be in Schur canonical form, that is, block upper triangular
with 1-by-1 and 2-by-2 diagonal blocks; each 2-by-2 diagonal block
has its diagonal elemnts equal and its off-diagonal elements of
opposite sign.
Arguments
=========
WANTQ (input) LOGICAL
= .TRUE. : accumulate the transformation in the matrix Q;
= .FALSE.: do not accumulate the transformation.
N (input) INTEGER
The order of the matrix T. N >= 0.
T (input/output) REAL array, dimension (LDT,N)
On entry, the upper quasi-triangular matrix T, in Schur
canonical form.
On exit, the updated matrix T, again in Schur canonical form.
LDT (input) INTEGER
The leading dimension of the array T. LDT >= max(1,N).
Q (input/output) REAL array, dimension (LDQ,N)
On entry, if WANTQ is .TRUE., the orthogonal matrix Q.
On exit, if WANTQ is .TRUE., the updated matrix Q.
If WANTQ is .FALSE., Q is not referenced.
LDQ (input) INTEGER
The leading dimension of the array Q.
LDQ >= 1; and if WANTQ is .TRUE., LDQ >= N.
J1 (input) INTEGER
The index of the first row of the first block T11.
N1 (input) INTEGER
The order of the first block T11. N1 = 0, 1 or 2.
N2 (input) INTEGER
The order of the second block T22. N2 = 0, 1 or 2.
WORK (workspace) REAL array, dimension (N)
INFO (output) INTEGER
= 0: successful exit
= 1: the transformed matrix T would be too far from Schur
form; the blocks are not swapped and T and Q are
unchanged.
=====================================================================
Parameter adjustments */
/* Table of constant values */
static integer c__1 = 1;
static integer c__4 = 4;
static logical c_false = FALSE_;
static integer c_n1 = -1;
static integer c__2 = 2;
static integer c__3 = 3;
/* System generated locals */
integer q_dim1, q_offset, t_dim1, t_offset, i__1;
real r__1, r__2, r__3, r__4, r__5, r__6;
/* Local variables */
static integer ierr;
static real temp;
extern /* Subroutine */ int srot_(integer *, real *, integer *, real *,
integer *, real *, real *);
static real d__[16] /* was [4][4] */;
static integer k;
static real u[3], scale, x[4] /* was [2][2] */, dnorm;
static integer j2, j3, j4;
static real xnorm, u1[3], u2[3];
extern /* Subroutine */ int slanv2_(real *, real *, real *, real *, real *
, real *, real *, real *, real *, real *), slasy2_(logical *,
logical *, integer *, integer *, integer *, real *, integer *,
real *, integer *, real *, integer *, real *, real *, integer *,
real *, integer *);
static integer nd;
static real cs, t11, t22, t33, sn;
extern doublereal slamch_(char *), slange_(char *, integer *,
integer *, real *, integer *, real *);
extern /* Subroutine */ int slarfg_(integer *, real *, real *, integer *,
real *), slacpy_(char *, integer *, integer *, real *, integer *,
real *, integer *), slartg_(real *, real *, real *, real *
, real *);
static real thresh;
extern /* Subroutine */ int slarfx_(char *, integer *, integer *, real *,
real *, real *, integer *, real *);
static real smlnum, wi1, wi2, wr1, wr2, eps, tau, tau1, tau2;
#define d___ref(a_1,a_2) d__[(a_2)*4 + a_1 - 5]
#define q_ref(a_1,a_2) q[(a_2)*q_dim1 + a_1]
#define t_ref(a_1,a_2) t[(a_2)*t_dim1 + a_1]
#define x_ref(a_1,a_2) x[(a_2)*2 + a_1 - 3]
t_dim1 = *ldt;
t_offset = 1 + t_dim1 * 1;
t -= t_offset;
q_dim1 = *ldq;
q_offset = 1 + q_dim1 * 1;
q -= q_offset;
--work;
/* Function Body */
*info = 0;
/* Quick return if possible */
if (*n == 0 || *n1 == 0 || *n2 == 0) {
return 0;
}
if (*j1 + *n1 > *n) {
return 0;
}
j2 = *j1 + 1;
j3 = *j1 + 2;
j4 = *j1 + 3;
if (*n1 == 1 && *n2 == 1) {
/* Swap two 1-by-1 blocks. */
t11 = t_ref(*j1, *j1);
t22 = t_ref(j2, j2);
/* Determine the transformation to perform the interchange. */
r__1 = t22 - t11;
slartg_(&t_ref(*j1, j2), &r__1, &cs, &sn, &temp);
/* Apply transformation to the matrix T. */
if (j3 <= *n) {
i__1 = *n - *j1 - 1;
srot_(&i__1, &t_ref(*j1, j3), ldt, &t_ref(j2, j3), ldt, &cs, &sn);
}
i__1 = *j1 - 1;
srot_(&i__1, &t_ref(1, *j1), &c__1, &t_ref(1, j2), &c__1, &cs, &sn);
t_ref(*j1, *j1) = t22;
t_ref(j2, j2) = t11;
if (*wantq) {
/* Accumulate transformation in the matrix Q. */
srot_(n, &q_ref(1, *j1), &c__1, &q_ref(1, j2), &c__1, &cs, &sn);
}
} else {
/* Swapping involves at least one 2-by-2 block.
Copy the diagonal block of order N1+N2 to the local array D
and compute its norm. */
nd = *n1 + *n2;
slacpy_("Full", &nd, &nd, &t_ref(*j1, *j1), ldt, d__, &c__4);
dnorm = slange_("Max", &nd, &nd, d__, &c__4, &work[1]);
/* Compute machine-dependent threshold for test for accepting
swap. */
eps = slamch_("P");
smlnum = slamch_("S") / eps;
/* Computing MAX */
r__1 = eps * 10.f * dnorm;
thresh = dmax(r__1,smlnum);
/* Solve T11*X - X*T22 = scale*T12 for X. */
slasy2_(&c_false, &c_false, &c_n1, n1, n2, d__, &c__4, &d___ref(*n1 +
1, *n1 + 1), &c__4, &d___ref(1, *n1 + 1), &c__4, &scale, x, &
c__2, &xnorm, &ierr);
/* Swap the adjacent diagonal blocks. */
k = *n1 + *n1 + *n2 - 3;
switch (k) {
case 1: goto L10;
case 2: goto L20;
case 3: goto L30;
}
L10:
/* N1 = 1, N2 = 2: generate elementary reflector H so that:
( scale, X11, X12 ) H = ( 0, 0, * ) */
u[0] = scale;
u[1] = x_ref(1, 1);
u[2] = x_ref(1, 2);
slarfg_(&c__3, &u[2], u, &c__1, &tau);
u[2] = 1.f;
t11 = t_ref(*j1, *j1);
/* Perform swap provisionally on diagonal block in D. */
slarfx_("L", &c__3, &c__3, u, &tau, d__, &c__4, &work[1]);
slarfx_("R", &c__3, &c__3, u, &tau, d__, &c__4, &work[1]);
/* Test whether to reject swap.
Computing MAX */
r__4 = (r__1 = d___ref(3, 1), dabs(r__1)), r__5 = (r__2 = d___ref(3,
2), dabs(r__2)), r__4 = max(r__4,r__5), r__5 = (r__3 =
d___ref(3, 3) - t11, dabs(r__3));
if (dmax(r__4,r__5) > thresh) {
goto L50;
}
/* Accept swap: apply transformation to the entire matrix T. */
i__1 = *n - *j1 + 1;
slarfx_("L", &c__3, &i__1, u, &tau, &t_ref(*j1, *j1), ldt, &work[1]);
slarfx_("R", &j2, &c__3, u, &tau, &t_ref(1, *j1), ldt, &work[1]);
t_ref(j3, *j1) = 0.f;
t_ref(j3, j2) = 0.f;
t_ref(j3, j3) = t11;
if (*wantq) {
/* Accumulate transformation in the matrix Q. */
slarfx_("R", n, &c__3, u, &tau, &q_ref(1, *j1), ldq, &work[1]);
}
goto L40;
L20:
/* N1 = 2, N2 = 1: generate elementary reflector H so that:
H ( -X11 ) = ( * )
( -X21 ) = ( 0 )
( scale ) = ( 0 ) */
u[0] = -x_ref(1, 1);
u[1] = -x_ref(2, 1);
u[2] = scale;
slarfg_(&c__3, u, &u[1], &c__1, &tau);
u[0] = 1.f;
t33 = t_ref(j3, j3);
/* Perform swap provisionally on diagonal block in D. */
slarfx_("L", &c__3, &c__3, u, &tau, d__, &c__4, &work[1]);
slarfx_("R", &c__3, &c__3, u, &tau, d__, &c__4, &work[1]);
/* Test whether to reject swap.
Computing MAX */
r__4 = (r__1 = d___ref(2, 1), dabs(r__1)), r__5 = (r__2 = d___ref(3,
1), dabs(r__2)), r__4 = max(r__4,r__5), r__5 = (r__3 =
d___ref(1, 1) - t33, dabs(r__3));
if (dmax(r__4,r__5) > thresh) {
goto L50;
}
/* Accept swap: apply transformation to the entire matrix T. */
slarfx_("R", &j3, &c__3, u, &tau, &t_ref(1, *j1), ldt, &work[1]);
i__1 = *n - *j1;
slarfx_("L", &c__3, &i__1, u, &tau, &t_ref(*j1, j2), ldt, &work[1]);
t_ref(*j1, *j1) = t33;
t_ref(j2, *j1) = 0.f;
t_ref(j3, *j1) = 0.f;
if (*wantq) {
/* Accumulate transformation in the matrix Q. */
slarfx_("R", n, &c__3, u, &tau, &q_ref(1, *j1), ldq, &work[1]);
}
goto L40;
L30:
/* N1 = 2, N2 = 2: generate elementary reflectors H(1) and H(2) so
that:
H(2) H(1) ( -X11 -X12 ) = ( * * )
( -X21 -X22 ) ( 0 * )
( scale 0 ) ( 0 0 )
( 0 scale ) ( 0 0 ) */
u1[0] = -x_ref(1, 1);
u1[1] = -x_ref(2, 1);
u1[2] = scale;
slarfg_(&c__3, u1, &u1[1], &c__1, &tau1);
u1[0] = 1.f;
temp = -tau1 * (x_ref(1, 2) + u1[1] * x_ref(2, 2));
u2[0] = -temp * u1[1] - x_ref(2, 2);
u2[1] = -temp * u1[2];
u2[2] = scale;
slarfg_(&c__3, u2, &u2[1], &c__1, &tau2);
u2[0] = 1.f;
/* Perform swap provisionally on diagonal block in D. */
slarfx_("L", &c__3, &c__4, u1, &tau1, d__, &c__4, &work[1])
;
slarfx_("R", &c__4, &c__3, u1, &tau1, d__, &c__4, &work[1])
;
slarfx_("L", &c__3, &c__4, u2, &tau2, &d___ref(2, 1), &c__4, &work[1]);
slarfx_("R", &c__4, &c__3, u2, &tau2, &d___ref(1, 2), &c__4, &work[1]);
/* Test whether to reject swap.
Computing MAX */
r__5 = (r__1 = d___ref(3, 1), dabs(r__1)), r__6 = (r__2 = d___ref(3,
2), dabs(r__2)), r__5 = max(r__5,r__6), r__6 = (r__3 =
d___ref(4, 1), dabs(r__3)), r__5 = max(r__5,r__6), r__6 = (
r__4 = d___ref(4, 2), dabs(r__4));
if (dmax(r__5,r__6) > thresh) {
goto L50;
}
/* Accept swap: apply transformation to the entire matrix T. */
i__1 = *n - *j1 + 1;
slarfx_("L", &c__3, &i__1, u1, &tau1, &t_ref(*j1, *j1), ldt, &work[1]);
slarfx_("R", &j4, &c__3, u1, &tau1, &t_ref(1, *j1), ldt, &work[1]);
i__1 = *n - *j1 + 1;
slarfx_("L", &c__3, &i__1, u2, &tau2, &t_ref(j2, *j1), ldt, &work[1]);
slarfx_("R", &j4, &c__3, u2, &tau2, &t_ref(1, j2), ldt, &work[1]);
t_ref(j3, *j1) = 0.f;
t_ref(j3, j2) = 0.f;
t_ref(j4, *j1) = 0.f;
t_ref(j4, j2) = 0.f;
if (*wantq) {
/* Accumulate transformation in the matrix Q. */
slarfx_("R", n, &c__3, u1, &tau1, &q_ref(1, *j1), ldq, &work[1]);
slarfx_("R", n, &c__3, u2, &tau2, &q_ref(1, j2), ldq, &work[1]);
}
L40:
if (*n2 == 2) {
/* Standardize new 2-by-2 block T11 */
slanv2_(&t_ref(*j1, *j1), &t_ref(*j1, j2), &t_ref(j2, *j1), &
t_ref(j2, j2), &wr1, &wi1, &wr2, &wi2, &cs, &sn);
i__1 = *n - *j1 - 1;
srot_(&i__1, &t_ref(*j1, *j1 + 2), ldt, &t_ref(j2, *j1 + 2), ldt,
&cs, &sn);
i__1 = *j1 - 1;
srot_(&i__1, &t_ref(1, *j1), &c__1, &t_ref(1, j2), &c__1, &cs, &
sn);
if (*wantq) {
srot_(n, &q_ref(1, *j1), &c__1, &q_ref(1, j2), &c__1, &cs, &
sn);
}
}
if (*n1 == 2) {
/* Standardize new 2-by-2 block T22 */
j3 = *j1 + *n2;
j4 = j3 + 1;
slanv2_(&t_ref(j3, j3), &t_ref(j3, j4), &t_ref(j4, j3), &t_ref(j4,
j4), &wr1, &wi1, &wr2, &wi2, &cs, &sn);
if (j3 + 2 <= *n) {
i__1 = *n - j3 - 1;
srot_(&i__1, &t_ref(j3, j3 + 2), ldt, &t_ref(j4, j3 + 2), ldt,
&cs, &sn);
}
i__1 = j3 - 1;
srot_(&i__1, &t_ref(1, j3), &c__1, &t_ref(1, j4), &c__1, &cs, &sn)
;
if (*wantq) {
srot_(n, &q_ref(1, j3), &c__1, &q_ref(1, j4), &c__1, &cs, &sn)
;
}
}
}
return 0;
/* Exit with INFO = 1 if swap was rejected. */
L50:
*info = 1;
return 0;
/* End of SLAEXC */
} /* slaexc_ */
/* Subroutine */ int slapll_(integer *n, real *x, integer *incx, real *y,
integer *incy, real *ssmin)
{
/* -- LAPACK auxiliary routine (version 3.0) --
Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd.,
Courant Institute, Argonne National Lab, and Rice University
March 31, 1993
Purpose
=======
Given two column vectors X and Y, let
A = ( X Y ).
The subroutine first computes the QR factorization of A = Q*R,
and then computes the SVD of the 2-by-2 upper triangular matrix R.
The smaller singular value of R is returned in SSMIN, which is used
as the measurement of the linear dependency of the vectors X and Y.
Arguments
=========
N (input) INTEGER
The length of the vectors X and Y.
X (input/output) REAL array,
dimension (1+(N-1)*INCX)
On entry, X contains the N-vector X.
On exit, X is overwritten.
INCX (input) INTEGER
The increment between successive elements of X. INCX > 0.
Y (input/output) REAL array,
dimension (1+(N-1)*INCY)
On entry, Y contains the N-vector Y.
On exit, Y is overwritten.
INCY (input) INTEGER
The increment between successive elements of Y. INCY > 0.
SSMIN (output) REAL
The smallest singular value of the N-by-2 matrix A = ( X Y ).
=====================================================================
Quick return if possible
Parameter adjustments */
/* System generated locals */
integer i__1;
/* Local variables */
extern doublereal sdot_(integer *, real *, integer *, real *, integer *);
extern /* Subroutine */ int slas2_(real *, real *, real *, real *, real *)
;
static real c__, ssmax;
extern /* Subroutine */ int saxpy_(integer *, real *, real *, integer *,
real *, integer *);
static real a11, a12, a22;
extern /* Subroutine */ int slarfg_(integer *, real *, real *, integer *,
real *);
static real tau;
--y;
--x;
/* Function Body */
if (*n <= 1) {
*ssmin = 0.f;
return 0;
}
/* Compute the QR factorization of the N-by-2 matrix ( X Y ) */
slarfg_(n, &x[1], &x[*incx + 1], incx, &tau);
a11 = x[1];
x[1] = 1.f;
c__ = -tau * sdot_(n, &x[1], incx, &y[1], incy);
saxpy_(n, &c__, &x[1], incx, &y[1], incy);
i__1 = *n - 1;
slarfg_(&i__1, &y[*incy + 1], &y[(*incy << 1) + 1], incy, &tau);
a12 = y[1];
a22 = y[*incy + 1];
/* Compute the SVD of 2-by-2 Upper triangular matrix. */
slas2_(&a11, &a12, &a22, ssmin, &ssmax);
return 0;
/* End of SLAPLL */
} /* slapll_ */
/* Subroutine */ int ssptrd_(char *uplo, integer *n, real *ap, real *d, real *
e, real *tau, integer *info)
{
/* -- LAPACK routine (version 2.0) --
Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd.,
Courant Institute, Argonne National Lab, and Rice University
March 31, 1993
Purpose
=======
SSPTRD reduces a real symmetric matrix A stored in packed form to
symmetric tridiagonal form T by an orthogonal similarity
transformation: Q**T * A * Q = T.
Arguments
=========
UPLO (input) CHARACTER*1
= 'U': Upper triangle of A is stored;
= 'L': Lower triangle of A is stored.
N (input) INTEGER
The order of the matrix A. N >= 0.
AP (input/output) REAL array, dimension (N*(N+1)/2)
On entry, the upper or lower triangle of the symmetric matrix
A, packed columnwise in a linear array. The j-th column of A
is stored in the array AP as follows:
if UPLO = 'U', AP(i + (j-1)*j/2) = A(i,j) for 1<=i<=j;
if UPLO = 'L', AP(i + (j-1)*(2*n-j)/2) = A(i,j) for j<=i<=n.
On exit, if UPLO = 'U', the diagonal and first superdiagonal
of A are overwritten by the corresponding elements of the
tridiagonal matrix T, and the elements above the first
superdiagonal, with the array TAU, represent the orthogonal
matrix Q as a product of elementary reflectors; if UPLO
= 'L', the diagonal and first subdiagonal of A are over-
written by the corresponding elements of the tridiagonal
matrix T, and the elements below the first subdiagonal, with
the array TAU, represent the orthogonal matrix Q as a product
of elementary reflectors. See Further Details.
D (output) REAL array, dimension (N)
The diagonal elements of the tridiagonal matrix T:
D(i) = A(i,i).
E (output) REAL array, dimension (N-1)
The off-diagonal elements of the tridiagonal matrix T:
E(i) = A(i,i+1) if UPLO = 'U', E(i) = A(i+1,i) if UPLO = 'L'.
TAU (output) REAL array, dimension (N-1)
The scalar factors of the elementary reflectors (see Further
Details).
INFO (output) INTEGER
= 0: successful exit
< 0: if INFO = -i, the i-th argument had an illegal value
Further Details
===============
If UPLO = 'U', the matrix Q is represented as a product of elementary
reflectors
Q = H(n-1) . . . H(2) H(1).
Each H(i) has the form
H(i) = I - tau * v * v'
where tau is a real scalar, and v is a real vector with
v(i+1:n) = 0 and v(i) = 1; v(1:i-1) is stored on exit in AP,
overwriting A(1:i-1,i+1), and tau is stored in TAU(i).
If UPLO = 'L', the matrix Q is represented as a product of elementary
reflectors
Q = H(1) H(2) . . . H(n-1).
Each H(i) has the form
H(i) = I - tau * v * v'
where tau is a real scalar, and v is a real vector with
v(1:i) = 0 and v(i+1) = 1; v(i+2:n) is stored on exit in AP,
overwriting A(i+2:n,i), and tau is stored in TAU(i).
=====================================================================
Test the input parameters
Parameter adjustments
Function Body */
/* Table of constant values */
static integer c__1 = 1;
static real c_b8 = 0.f;
static real c_b14 = -1.f;
/* System generated locals */
integer i__1, i__2;
/* Local variables */
static real taui;
extern doublereal sdot_(integer *, real *, integer *, real *, integer *);
static integer i;
extern /* Subroutine */ int sspr2_(char *, integer *, real *, real *,
integer *, real *, integer *, real *);
static real alpha;
extern logical lsame_(char *, char *);
static integer i1;
static logical upper;
extern /* Subroutine */ int saxpy_(integer *, real *, real *, integer *,
real *, integer *), sspmv_(char *, integer *, real *, real *,
real *, integer *, real *, real *, integer *);
static integer ii;
extern /* Subroutine */ int xerbla_(char *, integer *), slarfg_(
integer *, real *, real *, integer *, real *);
static integer i1i1;
#define TAU(I) tau[(I)-1]
#define E(I) e[(I)-1]
#define D(I) d[(I)-1]
#define AP(I) ap[(I)-1]
*info = 0;
upper = lsame_(uplo, "U");
if (! upper && ! lsame_(uplo, "L")) {
*info = -1;
} else if (*n < 0) {
*info = -2;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SSPTRD", &i__1);
return 0;
}
/* Quick return if possible */
if (*n <= 0) {
return 0;
}
if (upper) {
/* Reduce the upper triangle of A.
I1 is the index in AP of A(1,I+1). */
i1 = *n * (*n - 1) / 2 + 1;
for (i = *n - 1; i >= 1; --i) {
/* Generate elementary reflector H(i) = I - tau * v * v'
to annihilate A(1:i-1,i+1) */
slarfg_(&i, &AP(i1 + i - 1), &AP(i1), &c__1, &taui);
E(i) = AP(i1 + i - 1);
if (taui != 0.f) {
/* Apply H(i) from both sides to A(1:i,1:i) */
AP(i1 + i - 1) = 1.f;
/* Compute y := tau * A * v storing y in TAU(1:
i) */
sspmv_(uplo, &i, &taui, &AP(1), &AP(i1), &c__1, &c_b8, &TAU(1)
, &c__1);
/* Compute w := y - 1/2 * tau * (y'*v) * v */
alpha = taui * -.5f * sdot_(&i, &TAU(1), &c__1, &AP(i1), &
c__1);
saxpy_(&i, &alpha, &AP(i1), &c__1, &TAU(1), &c__1);
/* Apply the transformation as a rank-2 update:
A := A - v * w' - w * v' */
sspr2_(uplo, &i, &c_b14, &AP(i1), &c__1, &TAU(1), &c__1, &AP(
1));
AP(i1 + i - 1) = E(i);
}
D(i + 1) = AP(i1 + i);
TAU(i) = taui;
i1 -= i;
/* L10: */
}
D(1) = AP(1);
} else {
/* Reduce the lower triangle of A. II is the index in AP of
A(i,i) and I1I1 is the index of A(i+1,i+1). */
ii = 1;
i__1 = *n - 1;
for (i = 1; i <= *n-1; ++i) {
i1i1 = ii + *n - i + 1;
/* Generate elementary reflector H(i) = I - tau * v * v'
to annihilate A(i+2:n,i) */
i__2 = *n - i;
slarfg_(&i__2, &AP(ii + 1), &AP(ii + 2), &c__1, &taui);
E(i) = AP(ii + 1);
if (taui != 0.f) {
/* Apply H(i) from both sides to A(i+1:n,i+1:n)
*/
AP(ii + 1) = 1.f;
/* Compute y := tau * A * v storing y in TAU(i:
n-1) */
i__2 = *n - i;
sspmv_(uplo, &i__2, &taui, &AP(i1i1), &AP(ii + 1), &c__1, &
c_b8, &TAU(i), &c__1);
/* Compute w := y - 1/2 * tau * (y'*v) * v */
i__2 = *n - i;
alpha = taui * -.5f * sdot_(&i__2, &TAU(i), &c__1, &AP(ii + 1)
, &c__1);
i__2 = *n - i;
saxpy_(&i__2, &alpha, &AP(ii + 1), &c__1, &TAU(i), &c__1);
/* Apply the transformation as a rank-2 update:
A := A - v * w' - w * v' */
i__2 = *n - i;
sspr2_(uplo, &i__2, &c_b14, &AP(ii + 1), &c__1, &TAU(i), &
c__1, &AP(i1i1));
AP(ii + 1) = E(i);
}
D(i) = AP(ii);
TAU(i) = taui;
ii = i1i1;
/* L20: */
}
D(*n) = AP(ii);
}
return 0;
/* End of SSPTRD */
} /* ssptrd_ */
/* Subroutine */ int snapps_(integer *n, integer *kev, integer *np, real *
shiftr, real *shifti, real *v, integer *ldv, real *h__, integer *ldh,
real *resid, real *q, integer *ldq, real *workl, real *workd)
{
/* Initialized data */
static logical first = TRUE_;
/* System generated locals */
integer h_dim1, h_offset, v_dim1, v_offset, q_dim1, q_offset, i__1, i__2,
i__3, i__4;
real r__1, r__2;
/* Local variables */
static real c__, f, g;
static integer i__, j;
static real r__, s, t, u[3], t0, t1, h11, h12, h21, h22, h32;
static integer jj, ir, nr;
static real tau, ulp, tst1;
static integer iend;
static real unfl, ovfl;
static logical cconj;
extern /* Subroutine */ int sscal_(integer *, real *, real *, integer *),
slarf_(char *, integer *, integer *, real *, integer *, real *,
real *, integer *, real *, ftnlen), sgemv_(char *, integer *,
integer *, real *, real *, integer *, real *, integer *, real *,
real *, integer *, ftnlen), scopy_(integer *, real *, integer *,
real *, integer *), saxpy_(integer *, real *, real *, integer *,
real *, integer *), ivout_(integer *, integer *, integer *,
integer *, char *, ftnlen), smout_(integer *, integer *, integer *
, real *, integer *, integer *, char *, ftnlen), svout_(integer *,
integer *, real *, integer *, char *, ftnlen);
extern doublereal slapy2_(real *, real *);
extern /* Subroutine */ int slabad_(real *, real *);
extern doublereal slamch_(char *, ftnlen);
static real sigmai;
extern /* Subroutine */ int second_(real *);
static real sigmar;
static integer istart, kplusp, msglvl;
static real smlnum;
extern /* Subroutine */ int slacpy_(char *, integer *, integer *, real *,
integer *, real *, integer *, ftnlen), slarfg_(integer *, real *,
real *, integer *, real *), slaset_(char *, integer *, integer *,
real *, real *, real *, integer *, ftnlen), slartg_(real *, real *
, real *, real *, real *);
extern doublereal slanhs_(char *, integer *, real *, integer *, real *,
ftnlen);
/* %----------------------------------------------------% */
/* | Include files for debugging and timing information | */
/* %----------------------------------------------------% */
/* \SCCS Information: @(#) */
/* FILE: debug.h SID: 2.3 DATE OF SID: 11/16/95 RELEASE: 2 */
/* %---------------------------------% */
/* | See debug.doc for documentation | */
/* %---------------------------------% */
/* %------------------% */
/* | Scalar Arguments | */
/* %------------------% */
/* %--------------------------------% */
/* | See stat.doc for documentation | */
/* %--------------------------------% */
/* \SCCS Information: @(#) */
/* FILE: stat.h SID: 2.2 DATE OF SID: 11/16/95 RELEASE: 2 */
/* %-----------------% */
/* | Array Arguments | */
/* %-----------------% */
/* %------------% */
/* | Parameters | */
/* %------------% */
/* %------------------------% */
/* | Local Scalars & Arrays | */
/* %------------------------% */
/* %----------------------% */
/* | External Subroutines | */
/* %----------------------% */
/* %--------------------% */
/* | External Functions | */
/* %--------------------% */
/* %----------------------% */
/* | Intrinsics Functions | */
/* %----------------------% */
/* %----------------% */
/* | Data statments | */
/* %----------------% */
/* Parameter adjustments */
--workd;
--resid;
--workl;
--shifti;
--shiftr;
v_dim1 = *ldv;
v_offset = 1 + v_dim1;
v -= v_offset;
h_dim1 = *ldh;
h_offset = 1 + h_dim1;
h__ -= h_offset;
q_dim1 = *ldq;
q_offset = 1 + q_dim1;
q -= q_offset;
/* Function Body */
/* %-----------------------% */
/* | Executable Statements | */
/* %-----------------------% */
if (first) {
/* %-----------------------------------------------% */
/* | Set machine-dependent constants for the | */
/* | stopping criterion. If norm(H) <= sqrt(OVFL), | */
/* | overflow should not occur. | */
/* | REFERENCE: LAPACK subroutine slahqr | */
/* %-----------------------------------------------% */
unfl = slamch_("safe minimum", (ftnlen)12);
ovfl = 1.f / unfl;
slabad_(&unfl, &ovfl);
ulp = slamch_("precision", (ftnlen)9);
smlnum = unfl * (*n / ulp);
first = FALSE_;
}
/* %-------------------------------% */
/* | Initialize timing statistics | */
/* | & message level for debugging | */
/* %-------------------------------% */
second_(&t0);
msglvl = debug_1.mnapps;
kplusp = *kev + *np;
/* %--------------------------------------------% */
/* | Initialize Q to the identity to accumulate | */
/* | the rotations and reflections | */
/* %--------------------------------------------% */
slaset_("All", &kplusp, &kplusp, &c_b5, &c_b6, &q[q_offset], ldq, (ftnlen)
3);
/* %----------------------------------------------% */
/* | Quick return if there are no shifts to apply | */
/* %----------------------------------------------% */
if (*np == 0) {
goto L9000;
}
/* %----------------------------------------------% */
/* | Chase the bulge with the application of each | */
/* | implicit shift. Each shift is applied to the | */
/* | whole matrix including each block. | */
/* %----------------------------------------------% */
cconj = FALSE_;
i__1 = *np;
for (jj = 1; jj <= i__1; ++jj) {
sigmar = shiftr[jj];
sigmai = shifti[jj];
if (msglvl > 2) {
ivout_(&debug_1.logfil, &c__1, &jj, &debug_1.ndigit, "_napps: sh"
"ift number.", (ftnlen)21);
svout_(&debug_1.logfil, &c__1, &sigmar, &debug_1.ndigit, "_napps"
": The real part of the shift ", (ftnlen)35);
svout_(&debug_1.logfil, &c__1, &sigmai, &debug_1.ndigit, "_napps"
": The imaginary part of the shift ", (ftnlen)40);
}
/* %-------------------------------------------------% */
/* | The following set of conditionals is necessary | */
/* | in order that complex conjugate pairs of shifts | */
/* | are applied together or not at all. | */
/* %-------------------------------------------------% */
if (cconj) {
/* %-----------------------------------------% */
/* | cconj = .true. means the previous shift | */
/* | had non-zero imaginary part. | */
/* %-----------------------------------------% */
cconj = FALSE_;
goto L110;
} else if (jj < *np && dabs(sigmai) > 0.f) {
/* %------------------------------------% */
/* | Start of a complex conjugate pair. | */
/* %------------------------------------% */
cconj = TRUE_;
} else if (jj == *np && dabs(sigmai) > 0.f) {
/* %----------------------------------------------% */
/* | The last shift has a nonzero imaginary part. | */
/* | Don't apply it; thus the order of the | */
/* | compressed H is order KEV+1 since only np-1 | */
/* | were applied. | */
/* %----------------------------------------------% */
++(*kev);
goto L110;
}
istart = 1;
L20:
/* %--------------------------------------------------% */
/* | if sigmai = 0 then | */
/* | Apply the jj-th shift ... | */
/* | else | */
/* | Apply the jj-th and (jj+1)-th together ... | */
/* | (Note that jj < np at this point in the code) | */
/* | end | */
/* | to the current block of H. The next do loop | */
/* | determines the current block ; | */
/* %--------------------------------------------------% */
i__2 = kplusp - 1;
for (i__ = istart; i__ <= i__2; ++i__) {
/* %----------------------------------------% */
/* | Check for splitting and deflation. Use | */
/* | a standard test as in the QR algorithm | */
/* | REFERENCE: LAPACK subroutine slahqr | */
/* %----------------------------------------% */
tst1 = (r__1 = h__[i__ + i__ * h_dim1], dabs(r__1)) + (r__2 = h__[
i__ + 1 + (i__ + 1) * h_dim1], dabs(r__2));
if (tst1 == 0.f) {
i__3 = kplusp - jj + 1;
tst1 = slanhs_("1", &i__3, &h__[h_offset], ldh, &workl[1], (
ftnlen)1);
}
/* Computing MAX */
r__2 = ulp * tst1;
if ((r__1 = h__[i__ + 1 + i__ * h_dim1], dabs(r__1)) <= dmax(r__2,
smlnum)) {
if (msglvl > 0) {
ivout_(&debug_1.logfil, &c__1, &i__, &debug_1.ndigit,
"_napps: matrix splitting at row/column no.", (
ftnlen)42);
ivout_(&debug_1.logfil, &c__1, &jj, &debug_1.ndigit,
"_napps: matrix splitting with shift number.", (
ftnlen)43);
svout_(&debug_1.logfil, &c__1, &h__[i__ + 1 + i__ *
h_dim1], &debug_1.ndigit, "_napps: off diagonal "
"element.", (ftnlen)29);
}
iend = i__;
h__[i__ + 1 + i__ * h_dim1] = 0.f;
goto L40;
}
/* L30: */
}
iend = kplusp;
L40:
if (msglvl > 2) {
ivout_(&debug_1.logfil, &c__1, &istart, &debug_1.ndigit, "_napps"
": Start of current block ", (ftnlen)31);
ivout_(&debug_1.logfil, &c__1, &iend, &debug_1.ndigit, "_napps: "
"End of current block ", (ftnlen)29);
}
/* %------------------------------------------------% */
/* | No reason to apply a shift to block of order 1 | */
/* %------------------------------------------------% */
if (istart == iend) {
goto L100;
}
/* %------------------------------------------------------% */
/* | If istart + 1 = iend then no reason to apply a | */
/* | complex conjugate pair of shifts on a 2 by 2 matrix. | */
/* %------------------------------------------------------% */
if (istart + 1 == iend && dabs(sigmai) > 0.f) {
goto L100;
}
h11 = h__[istart + istart * h_dim1];
h21 = h__[istart + 1 + istart * h_dim1];
if (dabs(sigmai) <= 0.f) {
/* %---------------------------------------------% */
/* | Real-valued shift ==> apply single shift QR | */
/* %---------------------------------------------% */
f = h11 - sigmar;
g = h21;
i__2 = iend - 1;
for (i__ = istart; i__ <= i__2; ++i__) {
/* %-----------------------------------------------------% */
/* | Contruct the plane rotation G to zero out the bulge | */
/* %-----------------------------------------------------% */
slartg_(&f, &g, &c__, &s, &r__);
if (i__ > istart) {
/* %-------------------------------------------% */
/* | The following ensures that h(1:iend-1,1), | */
/* | the first iend-2 off diagonal of elements | */
/* | H, remain non negative. | */
/* %-------------------------------------------% */
if (r__ < 0.f) {
r__ = -r__;
c__ = -c__;
s = -s;
}
h__[i__ + (i__ - 1) * h_dim1] = r__;
h__[i__ + 1 + (i__ - 1) * h_dim1] = 0.f;
}
/* %---------------------------------------------% */
/* | Apply rotation to the left of H; H <- G'*H | */
/* %---------------------------------------------% */
i__3 = kplusp;
for (j = i__; j <= i__3; ++j) {
t = c__ * h__[i__ + j * h_dim1] + s * h__[i__ + 1 + j *
h_dim1];
h__[i__ + 1 + j * h_dim1] = -s * h__[i__ + j * h_dim1] +
c__ * h__[i__ + 1 + j * h_dim1];
h__[i__ + j * h_dim1] = t;
/* L50: */
}
/* %---------------------------------------------% */
/* | Apply rotation to the right of H; H <- H*G | */
/* %---------------------------------------------% */
/* Computing MIN */
i__4 = i__ + 2;
i__3 = min(i__4,iend);
for (j = 1; j <= i__3; ++j) {
t = c__ * h__[j + i__ * h_dim1] + s * h__[j + (i__ + 1) *
h_dim1];
h__[j + (i__ + 1) * h_dim1] = -s * h__[j + i__ * h_dim1]
+ c__ * h__[j + (i__ + 1) * h_dim1];
h__[j + i__ * h_dim1] = t;
/* L60: */
}
/* %----------------------------------------------------% */
/* | Accumulate the rotation in the matrix Q; Q <- Q*G | */
/* %----------------------------------------------------% */
/* Computing MIN */
i__4 = i__ + jj;
i__3 = min(i__4,kplusp);
for (j = 1; j <= i__3; ++j) {
t = c__ * q[j + i__ * q_dim1] + s * q[j + (i__ + 1) *
q_dim1];
q[j + (i__ + 1) * q_dim1] = -s * q[j + i__ * q_dim1] +
c__ * q[j + (i__ + 1) * q_dim1];
q[j + i__ * q_dim1] = t;
/* L70: */
}
/* %---------------------------% */
/* | Prepare for next rotation | */
/* %---------------------------% */
if (i__ < iend - 1) {
f = h__[i__ + 1 + i__ * h_dim1];
g = h__[i__ + 2 + i__ * h_dim1];
}
/* L80: */
}
/* %-----------------------------------% */
/* | Finished applying the real shift. | */
/* %-----------------------------------% */
} else {
/* %----------------------------------------------------% */
/* | Complex conjugate shifts ==> apply double shift QR | */
/* %----------------------------------------------------% */
h12 = h__[istart + (istart + 1) * h_dim1];
h22 = h__[istart + 1 + (istart + 1) * h_dim1];
h32 = h__[istart + 2 + (istart + 1) * h_dim1];
/* %---------------------------------------------------------% */
/* | Compute 1st column of (H - shift*I)*(H - conj(shift)*I) | */
/* %---------------------------------------------------------% */
s = sigmar * 2.f;
t = slapy2_(&sigmar, &sigmai);
u[0] = (h11 * (h11 - s) + t * t) / h21 + h12;
u[1] = h11 + h22 - s;
u[2] = h32;
i__2 = iend - 1;
for (i__ = istart; i__ <= i__2; ++i__) {
/* Computing MIN */
i__3 = 3, i__4 = iend - i__ + 1;
nr = min(i__3,i__4);
/* %-----------------------------------------------------% */
/* | Construct Householder reflector G to zero out u(1). | */
/* | G is of the form I - tau*( 1 u )' * ( 1 u' ). | */
/* %-----------------------------------------------------% */
slarfg_(&nr, u, &u[1], &c__1, &tau);
if (i__ > istart) {
h__[i__ + (i__ - 1) * h_dim1] = u[0];
h__[i__ + 1 + (i__ - 1) * h_dim1] = 0.f;
if (i__ < iend - 1) {
h__[i__ + 2 + (i__ - 1) * h_dim1] = 0.f;
}
}
u[0] = 1.f;
/* %--------------------------------------% */
/* | Apply the reflector to the left of H | */
/* %--------------------------------------% */
i__3 = kplusp - i__ + 1;
slarf_("Left", &nr, &i__3, u, &c__1, &tau, &h__[i__ + i__ *
h_dim1], ldh, &workl[1], (ftnlen)4);
/* %---------------------------------------% */
/* | Apply the reflector to the right of H | */
/* %---------------------------------------% */
/* Computing MIN */
i__3 = i__ + 3;
ir = min(i__3,iend);
slarf_("Right", &ir, &nr, u, &c__1, &tau, &h__[i__ * h_dim1 +
1], ldh, &workl[1], (ftnlen)5);
/* %-----------------------------------------------------% */
/* | Accumulate the reflector in the matrix Q; Q <- Q*G | */
/* %-----------------------------------------------------% */
slarf_("Right", &kplusp, &nr, u, &c__1, &tau, &q[i__ * q_dim1
+ 1], ldq, &workl[1], (ftnlen)5);
/* %----------------------------% */
/* | Prepare for next reflector | */
/* %----------------------------% */
if (i__ < iend - 1) {
u[0] = h__[i__ + 1 + i__ * h_dim1];
u[1] = h__[i__ + 2 + i__ * h_dim1];
if (i__ < iend - 2) {
u[2] = h__[i__ + 3 + i__ * h_dim1];
}
}
/* L90: */
}
/* %--------------------------------------------% */
/* | Finished applying a complex pair of shifts | */
/* | to the current block | */
/* %--------------------------------------------% */
}
L100:
/* %---------------------------------------------------------% */
/* | Apply the same shift to the next block if there is any. | */
/* %---------------------------------------------------------% */
istart = iend + 1;
if (iend < kplusp) {
goto L20;
}
/* %---------------------------------------------% */
/* | Loop back to the top to get the next shift. | */
/* %---------------------------------------------% */
L110:
;
}
/* %--------------------------------------------------% */
/* | Perform a similarity transformation that makes | */
/* | sure that H will have non negative sub diagonals | */
/* %--------------------------------------------------% */
i__1 = *kev;
for (j = 1; j <= i__1; ++j) {
if (h__[j + 1 + j * h_dim1] < 0.f) {
i__2 = kplusp - j + 1;
sscal_(&i__2, &c_b43, &h__[j + 1 + j * h_dim1], ldh);
/* Computing MIN */
i__3 = j + 2;
i__2 = min(i__3,kplusp);
sscal_(&i__2, &c_b43, &h__[(j + 1) * h_dim1 + 1], &c__1);
/* Computing MIN */
i__3 = j + *np + 1;
i__2 = min(i__3,kplusp);
sscal_(&i__2, &c_b43, &q[(j + 1) * q_dim1 + 1], &c__1);
}
/* L120: */
}
i__1 = *kev;
for (i__ = 1; i__ <= i__1; ++i__) {
/* %--------------------------------------------% */
/* | Final check for splitting and deflation. | */
/* | Use a standard test as in the QR algorithm | */
/* | REFERENCE: LAPACK subroutine slahqr | */
/* %--------------------------------------------% */
tst1 = (r__1 = h__[i__ + i__ * h_dim1], dabs(r__1)) + (r__2 = h__[i__
+ 1 + (i__ + 1) * h_dim1], dabs(r__2));
if (tst1 == 0.f) {
tst1 = slanhs_("1", kev, &h__[h_offset], ldh, &workl[1], (ftnlen)
1);
}
/* Computing MAX */
r__1 = ulp * tst1;
if (h__[i__ + 1 + i__ * h_dim1] <= dmax(r__1,smlnum)) {
h__[i__ + 1 + i__ * h_dim1] = 0.f;
}
/* L130: */
}
/* %-------------------------------------------------% */
/* | Compute the (kev+1)-st column of (V*Q) and | */
/* | temporarily store the result in WORKD(N+1:2*N). | */
/* | This is needed in the residual update since we | */
/* | cannot GUARANTEE that the corresponding entry | */
/* | of H would be zero as in exact arithmetic. | */
/* %-------------------------------------------------% */
if (h__[*kev + 1 + *kev * h_dim1] > 0.f) {
sgemv_("N", n, &kplusp, &c_b6, &v[v_offset], ldv, &q[(*kev + 1) *
q_dim1 + 1], &c__1, &c_b5, &workd[*n + 1], &c__1, (ftnlen)1);
}
/* %----------------------------------------------------------% */
/* | Compute column 1 to kev of (V*Q) in backward order | */
/* | taking advantage of the upper Hessenberg structure of Q. | */
/* %----------------------------------------------------------% */
i__1 = *kev;
for (i__ = 1; i__ <= i__1; ++i__) {
i__2 = kplusp - i__ + 1;
sgemv_("N", n, &i__2, &c_b6, &v[v_offset], ldv, &q[(*kev - i__ + 1) *
q_dim1 + 1], &c__1, &c_b5, &workd[1], &c__1, (ftnlen)1);
scopy_(n, &workd[1], &c__1, &v[(kplusp - i__ + 1) * v_dim1 + 1], &
c__1);
/* L140: */
}
/* %-------------------------------------------------% */
/* | Move v(:,kplusp-kev+1:kplusp) into v(:,1:kev). | */
/* %-------------------------------------------------% */
slacpy_("A", n, kev, &v[(kplusp - *kev + 1) * v_dim1 + 1], ldv, &v[
v_offset], ldv, (ftnlen)1);
/* %--------------------------------------------------------------% */
/* | Copy the (kev+1)-st column of (V*Q) in the appropriate place | */
/* %--------------------------------------------------------------% */
if (h__[*kev + 1 + *kev * h_dim1] > 0.f) {
scopy_(n, &workd[*n + 1], &c__1, &v[(*kev + 1) * v_dim1 + 1], &c__1);
}
/* %-------------------------------------% */
/* | Update the residual vector: | */
/* | r <- sigmak*r + betak*v(:,kev+1) | */
/* | where | */
/* | sigmak = (e_{kplusp}'*Q)*e_{kev} | */
/* | betak = e_{kev+1}'*H*e_{kev} | */
/* %-------------------------------------% */
sscal_(n, &q[kplusp + *kev * q_dim1], &resid[1], &c__1);
if (h__[*kev + 1 + *kev * h_dim1] > 0.f) {
saxpy_(n, &h__[*kev + 1 + *kev * h_dim1], &v[(*kev + 1) * v_dim1 + 1],
&c__1, &resid[1], &c__1);
}
if (msglvl > 1) {
svout_(&debug_1.logfil, &c__1, &q[kplusp + *kev * q_dim1], &
debug_1.ndigit, "_napps: sigmak = (e_{kev+p}^T*Q)*e_{kev}", (
ftnlen)40);
svout_(&debug_1.logfil, &c__1, &h__[*kev + 1 + *kev * h_dim1], &
debug_1.ndigit, "_napps: betak = e_{kev+1}^T*H*e_{kev}", (
ftnlen)37);
ivout_(&debug_1.logfil, &c__1, kev, &debug_1.ndigit, "_napps: Order "
"of the final Hessenberg matrix ", (ftnlen)45);
if (msglvl > 2) {
smout_(&debug_1.logfil, kev, kev, &h__[h_offset], ldh, &
debug_1.ndigit, "_napps: updated Hessenberg matrix H for"
" next iteration", (ftnlen)54);
}
}
L9000:
second_(&t1);
timing_1.tnapps += t1 - t0;
return 0;
/* %---------------% */
/* | End of snapps | */
/* %---------------% */
} /* snapps_ */
/* Subroutine */ int sgeqpf_(integer *m, integer *n, real *a, integer *lda,
integer *jpvt, real *tau, real *work, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, i__1, i__2, i__3;
real r__1, r__2;
/* Builtin functions */
double sqrt(doublereal);
/* Local variables */
integer i__, j, ma, mn;
real aii;
integer pvt;
real temp, temp2;
extern doublereal snrm2_(integer *, real *, integer *);
real tol3z;
extern /* Subroutine */ int slarf_(char *, integer *, integer *, real *,
integer *, real *, real *, integer *, real *);
integer itemp;
extern /* Subroutine */ int sswap_(integer *, real *, integer *, real *,
integer *), sgeqr2_(integer *, integer *, real *, integer *, real
*, real *, integer *), sorm2r_(char *, char *, integer *, integer
*, integer *, real *, integer *, real *, real *, integer *, real *
, integer *);
extern doublereal slamch_(char *);
extern /* Subroutine */ int xerbla_(char *, integer *), slarfg_(
integer *, real *, real *, integer *, real *);
extern integer isamax_(integer *, real *, integer *);
/* -- LAPACK deprecated driver routine (version 3.1) -- */
/* 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 SGEQP3. */
/* SGEQPF computes a QR factorization with column pivoting of a */
/* real M-by-N matrix A: A*P = Q*R. */
/* 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 */
/* A (input/output) REAL array, dimension (LDA,N) */
/* On entry, the M-by-N matrix A. */
/* On exit, the upper triangle of the array contains the */
/* min(M,N)-by-N upper triangular matrix R; the elements */
/* below the diagonal, together with the array TAU, */
/* represent the orthogonal matrix Q as a product of */
/* min(m,n) elementary reflectors. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,M). */
/* 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 A*P (a leading column); if JPVT(i) = 0, */
/* the i-th column of A is a free column. */
/* On exit, if JPVT(i) = k, then the i-th column of A*P */
/* was the k-th column of A. */
/* TAU (output) REAL array, dimension (min(M,N)) */
/* The scalar factors of the elementary reflectors. */
/* WORK (workspace) REAL array, dimension (3*N) */
/* 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(n) */
/* Each H(i) has the form */
/* H = I - tau * v * v' */
/* where tau is a real scalar, and v is a real vector with */
/* v(1:i-1) = 0 and v(i) = 1; v(i+1:m) is stored on exit in A(i+1:m,i). */
/* The matrix P is represented in jpvt as follows: If */
/* jpvt(j) = i */
/* then the jth column of P is the ith canonical unit vector. */
/* Partial column norm updating strategy modified by */
/* Z. Drmac and Z. Bujanovic, Dept. of Mathematics, */
/* University of Zagreb, Croatia. */
/* June 2006. */
/* For more details see LAPACK Working Note 176. */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Test the input arguments */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
--jpvt;
--tau;
--work;
/* Function Body */
*info = 0;
if (*m < 0) {
*info = -1;
} else if (*n < 0) {
*info = -2;
} else if (*lda < max(1,*m)) {
*info = -4;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SGEQPF", &i__1);
return 0;
}
mn = min(*m,*n);
tol3z = sqrt(slamch_("Epsilon"));
/* Move initial columns up front */
itemp = 1;
i__1 = *n;
for (i__ = 1; i__ <= i__1; ++i__) {
if (jpvt[i__] != 0) {
if (i__ != itemp) {
sswap_(m, &a[i__ * a_dim1 + 1], &c__1, &a[itemp * a_dim1 + 1],
&c__1);
jpvt[i__] = jpvt[itemp];
jpvt[itemp] = i__;
} else {
jpvt[i__] = i__;
}
++itemp;
} else {
jpvt[i__] = i__;
}
/* L10: */
}
--itemp;
/* Compute the QR factorization and update remaining columns */
if (itemp > 0) {
ma = min(itemp,*m);
sgeqr2_(m, &ma, &a[a_offset], lda, &tau[1], &work[1], info);
if (ma < *n) {
i__1 = *n - ma;
sorm2r_("Left", "Transpose", m, &i__1, &ma, &a[a_offset], lda, &
tau[1], &a[(ma + 1) * a_dim1 + 1], lda, &work[1], info);
}
}
if (itemp < mn) {
/* Initialize partial column norms. The first n elements of */
/* work store the exact column norms. */
i__1 = *n;
for (i__ = itemp + 1; i__ <= i__1; ++i__) {
i__2 = *m - itemp;
work[i__] = snrm2_(&i__2, &a[itemp + 1 + i__ * a_dim1], &c__1);
work[*n + i__] = work[i__];
/* L20: */
}
/* Compute factorization */
i__1 = mn;
for (i__ = itemp + 1; i__ <= i__1; ++i__) {
/* Determine ith pivot column and swap if necessary */
i__2 = *n - i__ + 1;
pvt = i__ - 1 + isamax_(&i__2, &work[i__], &c__1);
if (pvt != i__) {
sswap_(m, &a[pvt * a_dim1 + 1], &c__1, &a[i__ * a_dim1 + 1], &
c__1);
itemp = jpvt[pvt];
jpvt[pvt] = jpvt[i__];
jpvt[i__] = itemp;
work[pvt] = work[i__];
work[*n + pvt] = work[*n + i__];
}
/* Generate elementary reflector H(i) */
if (i__ < *m) {
i__2 = *m - i__ + 1;
slarfg_(&i__2, &a[i__ + i__ * a_dim1], &a[i__ + 1 + i__ *
a_dim1], &c__1, &tau[i__]);
} else {
slarfg_(&c__1, &a[*m + *m * a_dim1], &a[*m + *m * a_dim1], &
c__1, &tau[*m]);
}
if (i__ < *n) {
/* Apply H(i) to A(i:m,i+1:n) from the left */
aii = a[i__ + i__ * a_dim1];
a[i__ + i__ * a_dim1] = 1.f;
i__2 = *m - i__ + 1;
i__3 = *n - i__;
slarf_("LEFT", &i__2, &i__3, &a[i__ + i__ * a_dim1], &c__1, &
tau[i__], &a[i__ + (i__ + 1) * a_dim1], lda, &work[(*
n << 1) + 1]);
a[i__ + i__ * a_dim1] = aii;
}
/* Update partial column norms */
i__2 = *n;
for (j = i__ + 1; j <= i__2; ++j) {
if (work[j] != 0.f) {
/* NOTE: The following 4 lines follow from the analysis in */
/* Lapack Working Note 176. */
temp = (r__1 = a[i__ + j * a_dim1], dabs(r__1)) / work[j];
/* Computing MAX */
r__1 = 0.f, r__2 = (temp + 1.f) * (1.f - temp);
temp = dmax(r__1,r__2);
/* Computing 2nd power */
r__1 = work[j] / work[*n + j];
temp2 = temp * (r__1 * r__1);
if (temp2 <= tol3z) {
if (*m - i__ > 0) {
i__3 = *m - i__;
work[j] = snrm2_(&i__3, &a[i__ + 1 + j * a_dim1],
&c__1);
work[*n + j] = work[j];
} else {
work[j] = 0.f;
work[*n + j] = 0.f;
}
} else {
work[j] *= sqrt(temp);
}
}
/* L30: */
}
/* L40: */
}
}
return 0;
/* End of SGEQPF */
} /* sgeqpf_ */
/* Subroutine */ int slatme_(integer *n, char *dist, integer *iseed, real *
d__, integer *mode, real *cond, real *dmax__, char *ei, char *rsign,
char *upper, char *sim, real *ds, integer *modes, real *conds,
integer *kl, integer *ku, real *anorm, real *a, integer *lda, real *
work, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, i__1, i__2;
real r__1, r__2, r__3;
/* Local variables */
static logical bads;
extern /* Subroutine */ int sger_(integer *, integer *, real *, real *,
integer *, real *, integer *, real *, integer *);
static integer isim;
static real temp;
static logical badei;
static integer i__, j;
static real alpha;
extern logical lsame_(char *, char *);
static integer iinfo;
extern /* Subroutine */ int sscal_(integer *, real *, real *, integer *);
static real tempa[1];
static integer icols;
static logical useei;
static integer idist;
extern /* Subroutine */ int sgemv_(char *, integer *, integer *, real *,
real *, integer *, real *, integer *, real *, real *, integer *), scopy_(integer *, real *, integer *, real *, integer *);
static integer irows;
extern /* Subroutine */ int slatm1_(integer *, real *, integer *, integer
*, integer *, real *, integer *, integer *);
static integer ic, jc, ir, jr;
extern doublereal slange_(char *, integer *, integer *, real *, integer *,
real *);
extern /* Subroutine */ int slarge_(integer *, real *, integer *, integer
*, real *, integer *), slarfg_(integer *, real *, real *, integer
*, real *), xerbla_(char *, integer *);
extern doublereal slaran_(integer *);
static integer irsign;
extern /* Subroutine */ int slaset_(char *, integer *, integer *, real *,
real *, real *, integer *);
static integer iupper;
extern /* Subroutine */ int slarnv_(integer *, integer *, integer *, real
*);
static real xnorms;
static integer jcr;
static real tau;
#define a_ref(a_1,a_2) a[(a_2)*a_dim1 + a_1]
/* -- LAPACK test routine (version 3.0) --
Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd.,
Courant Institute, Argonne National Lab, and Rice University
September 30, 1994
Purpose
=======
SLATME generates random non-symmetric square matrices with
specified eigenvalues for testing LAPACK programs.
SLATME operates by applying the following sequence of
operations:
1. Set the diagonal to D, where D may be input or
computed according to MODE, COND, DMAX, and RSIGN
as described below.
2. If complex conjugate pairs are desired (MODE=0 and EI(1)='R',
or MODE=5), certain pairs of adjacent elements of D are
interpreted as the real and complex parts of a complex
conjugate pair; A thus becomes block diagonal, with 1x1
and 2x2 blocks.
3. If UPPER='T', the upper triangle of A is set to random values
out of distribution DIST.
4. If SIM='T', A is multiplied on the left by a random matrix
X, whose singular values are specified by DS, MODES, and
CONDS, and on the right by X inverse.
5. If KL < N-1, the lower bandwidth is reduced to KL using
Householder transformations. If KU < N-1, the upper
bandwidth is reduced to KU.
6. If ANORM is not negative, the matrix is scaled to have
maximum-element-norm ANORM.
(Note: since the matrix cannot be reduced beyond Hessenberg form,
no packing options are available.)
Arguments
=========
N - INTEGER
The number of columns (or rows) of A. Not modified.
DIST - CHARACTER*1
On entry, DIST specifies the type of distribution to be used
to generate the random eigen-/singular values, and for the
upper triangle (see UPPER).
'U' => UNIFORM( 0, 1 ) ( 'U' for uniform )
'S' => UNIFORM( -1, 1 ) ( 'S' for symmetric )
'N' => NORMAL( 0, 1 ) ( 'N' for normal )
Not modified.
ISEED - INTEGER array, dimension ( 4 )
On entry ISEED specifies the seed of the random number
generator. They should lie between 0 and 4095 inclusive,
and ISEED(4) should be odd. The random number generator
uses a linear congruential sequence limited to small
integers, and so should produce machine independent
random numbers. The values of ISEED are changed on
exit, and can be used in the next call to SLATME
to continue the same random number sequence.
Changed on exit.
D - REAL array, dimension ( N )
This array is used to specify the eigenvalues of A. If
MODE=0, then D is assumed to contain the eigenvalues (but
see the description of EI), otherwise they will be
computed according to MODE, COND, DMAX, and RSIGN and
placed in D.
Modified if MODE is nonzero.
MODE - INTEGER
On entry this describes how the eigenvalues are to
be specified:
MODE = 0 means use D (with EI) as input
MODE = 1 sets D(1)=1 and D(2:N)=1.0/COND
MODE = 2 sets D(1:N-1)=1 and D(N)=1.0/COND
MODE = 3 sets D(I)=COND**(-(I-1)/(N-1))
MODE = 4 sets D(i)=1 - (i-1)/(N-1)*(1 - 1/COND)
MODE = 5 sets D to random numbers in the range
( 1/COND , 1 ) such that their logarithms
are uniformly distributed. Each odd-even pair
of elements will be either used as two real
eigenvalues or as the real and imaginary part
of a complex conjugate pair of eigenvalues;
the choice of which is done is random, with
50-50 probability, for each pair.
MODE = 6 set D to random numbers from same distribution
as the rest of the matrix.
MODE < 0 has the same meaning as ABS(MODE), except that
the order of the elements of D is reversed.
Thus if MODE is between 1 and 4, D has entries ranging
from 1 to 1/COND, if between -1 and -4, D has entries
ranging from 1/COND to 1,
Not modified.
COND - REAL
On entry, this is used as described under MODE above.
If used, it must be >= 1. Not modified.
DMAX - REAL
If MODE is neither -6, 0 nor 6, the contents of D, as
computed according to MODE and COND, will be scaled by
DMAX / max(abs(D(i))). Note that DMAX need not be
positive: if DMAX is negative (or zero), D will be
scaled by a negative number (or zero).
Not modified.
EI - CHARACTER*1 array, dimension ( N )
If MODE is 0, and EI(1) is not ' ' (space character),
this array specifies which elements of D (on input) are
real eigenvalues and which are the real and imaginary parts
of a complex conjugate pair of eigenvalues. The elements
of EI may then only have the values 'R' and 'I'. If
EI(j)='R' and EI(j+1)='I', then the j-th eigenvalue is
CMPLX( D(j) , D(j+1) ), and the (j+1)-th is the complex
conjugate thereof. If EI(j)=EI(j+1)='R', then the j-th
eigenvalue is D(j) (i.e., real). EI(1) may not be 'I',
nor may two adjacent elements of EI both have the value 'I'.
If MODE is not 0, then EI is ignored. If MODE is 0 and
EI(1)=' ', then the eigenvalues will all be real.
Not modified.
RSIGN - CHARACTER*1
If MODE is not 0, 6, or -6, and RSIGN='T', then the
elements of D, as computed according to MODE and COND, will
be multiplied by a random sign (+1 or -1). If RSIGN='F',
they will not be. RSIGN may only have the values 'T' or
'F'.
Not modified.
UPPER - CHARACTER*1
If UPPER='T', then the elements of A above the diagonal
(and above the 2x2 diagonal blocks, if A has complex
eigenvalues) will be set to random numbers out of DIST.
If UPPER='F', they will not. UPPER may only have the
values 'T' or 'F'.
Not modified.
SIM - CHARACTER*1
If SIM='T', then A will be operated on by a "similarity
transform", i.e., multiplied on the left by a matrix X and
on the right by X inverse. X = U S V, where U and V are
random unitary matrices and S is a (diagonal) matrix of
singular values specified by DS, MODES, and CONDS. If
SIM='F', then A will not be transformed.
Not modified.
DS - REAL array, dimension ( N )
This array is used to specify the singular values of X,
in the same way that D specifies the eigenvalues of A.
If MODE=0, the DS contains the singular values, which
may not be zero.
Modified if MODE is nonzero.
MODES - INTEGER
CONDS - REAL
Same as MODE and COND, but for specifying the diagonal
of S. MODES=-6 and +6 are not allowed (since they would
result in randomly ill-conditioned eigenvalues.)
KL - INTEGER
This specifies the lower bandwidth of the matrix. KL=1
specifies upper Hessenberg form. If KL is at least N-1,
then A will have full lower bandwidth. KL must be at
least 1.
Not modified.
KU - INTEGER
This specifies the upper bandwidth of the matrix. KU=1
specifies lower Hessenberg form. If KU is at least N-1,
then A will have full upper bandwidth; if KU and KL
are both at least N-1, then A will be dense. Only one of
KU and KL may be less than N-1. KU must be at least 1.
Not modified.
ANORM - REAL
If ANORM is not negative, then A will be scaled by a non-
negative real number to make the maximum-element-norm of A
to be ANORM.
Not modified.
A - REAL array, dimension ( LDA, N )
On exit A is the desired test matrix.
Modified.
LDA - INTEGER
LDA specifies the first dimension of A as declared in the
calling program. LDA must be at least N.
Not modified.
WORK - REAL array, dimension ( 3*N )
Workspace.
Modified.
INFO - INTEGER
Error code. On exit, INFO will be set to one of the
following values:
0 => normal return
-1 => N negative
-2 => DIST illegal string
-5 => MODE not in range -6 to 6
-6 => COND less than 1.0, and MODE neither -6, 0 nor 6
-8 => EI(1) is not ' ' or 'R', EI(j) is not 'R' or 'I', or
two adjacent elements of EI are 'I'.
-9 => RSIGN is not 'T' or 'F'
-10 => UPPER is not 'T' or 'F'
-11 => SIM is not 'T' or 'F'
-12 => MODES=0 and DS has a zero singular value.
-13 => MODES is not in the range -5 to 5.
-14 => MODES is nonzero and CONDS is less than 1.
-15 => KL is less than 1.
-16 => KU is less than 1, or KL and KU are both less than
N-1.
-19 => LDA is less than N.
1 => Error return from SLATM1 (computing D)
2 => Cannot scale to DMAX (max. eigenvalue is 0)
3 => Error return from SLATM1 (computing DS)
4 => Error return from SLARGE
5 => Zero singular value from SLATM1.
=====================================================================
1) Decode and Test the input parameters.
Initialize flags & seed.
Parameter adjustments */
--iseed;
--d__;
--ei;
--ds;
a_dim1 = *lda;
a_offset = 1 + a_dim1 * 1;
a -= a_offset;
--work;
/* Function Body */
*info = 0;
/* Quick return if possible */
if (*n == 0) {
return 0;
}
/* Decode DIST */
if (lsame_(dist, "U")) {
idist = 1;
} else if (lsame_(dist, "S")) {
idist = 2;
} else if (lsame_(dist, "N")) {
idist = 3;
} else {
idist = -1;
}
/* Check EI */
useei = TRUE_;
badei = FALSE_;
if (lsame_(ei + 1, " ") || *mode != 0) {
useei = FALSE_;
} else {
if (lsame_(ei + 1, "R")) {
i__1 = *n;
for (j = 2; j <= i__1; ++j) {
if (lsame_(ei + j, "I")) {
if (lsame_(ei + (j - 1), "I")) {
badei = TRUE_;
}
} else {
if (! lsame_(ei + j, "R")) {
badei = TRUE_;
}
}
/* L10: */
}
} else {
badei = TRUE_;
}
}
/* Decode RSIGN */
if (lsame_(rsign, "T")) {
irsign = 1;
} else if (lsame_(rsign, "F")) {
irsign = 0;
} else {
irsign = -1;
}
/* Decode UPPER */
if (lsame_(upper, "T")) {
iupper = 1;
} else if (lsame_(upper, "F")) {
iupper = 0;
} else {
iupper = -1;
}
/* Decode SIM */
if (lsame_(sim, "T")) {
isim = 1;
} else if (lsame_(sim, "F")) {
isim = 0;
} else {
isim = -1;
}
/* Check DS, if MODES=0 and ISIM=1 */
bads = FALSE_;
if (*modes == 0 && isim == 1) {
i__1 = *n;
for (j = 1; j <= i__1; ++j) {
if (ds[j] == 0.f) {
bads = TRUE_;
}
/* L20: */
}
}
/* Set INFO if an error */
if (*n < 0) {
*info = -1;
} else if (idist == -1) {
*info = -2;
} else if (abs(*mode) > 6) {
*info = -5;
} else if (*mode != 0 && abs(*mode) != 6 && *cond < 1.f) {
*info = -6;
} else if (badei) {
*info = -8;
} else if (irsign == -1) {
*info = -9;
} else if (iupper == -1) {
*info = -10;
} else if (isim == -1) {
*info = -11;
} else if (bads) {
*info = -12;
} else if (isim == 1 && abs(*modes) > 5) {
*info = -13;
} else if (isim == 1 && *modes != 0 && *conds < 1.f) {
*info = -14;
} else if (*kl < 1) {
*info = -15;
} else if (*ku < 1 || *ku < *n - 1 && *kl < *n - 1) {
*info = -16;
} else if (*lda < max(1,*n)) {
*info = -19;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SLATME", &i__1);
return 0;
}
/* Initialize random number generator */
for (i__ = 1; i__ <= 4; ++i__) {
iseed[i__] = (i__1 = iseed[i__], abs(i__1)) % 4096;
/* L30: */
}
if (iseed[4] % 2 != 1) {
++iseed[4];
}
/* 2) Set up diagonal of A
Compute D according to COND and MODE */
slatm1_(mode, cond, &irsign, &idist, &iseed[1], &d__[1], n, &iinfo);
if (iinfo != 0) {
*info = 1;
return 0;
}
if (*mode != 0 && abs(*mode) != 6) {
/* Scale by DMAX */
temp = dabs(d__[1]);
i__1 = *n;
for (i__ = 2; i__ <= i__1; ++i__) {
/* Computing MAX */
r__2 = temp, r__3 = (r__1 = d__[i__], dabs(r__1));
temp = dmax(r__2,r__3);
/* L40: */
}
if (temp > 0.f) {
alpha = *dmax__ / temp;
} else if (*dmax__ != 0.f) {
*info = 2;
return 0;
} else {
alpha = 0.f;
}
sscal_(n, &alpha, &d__[1], &c__1);
}
slaset_("Full", n, n, &c_b23, &c_b23, &a[a_offset], lda);
i__1 = *lda + 1;
scopy_(n, &d__[1], &c__1, &a[a_offset], &i__1);
/* Set up complex conjugate pairs */
if (*mode == 0) {
if (useei) {
i__1 = *n;
for (j = 2; j <= i__1; ++j) {
if (lsame_(ei + j, "I")) {
a_ref(j - 1, j) = a_ref(j, j);
a_ref(j, j - 1) = -a_ref(j, j);
a_ref(j, j) = a_ref(j - 1, j - 1);
}
/* L50: */
}
}
} else if (abs(*mode) == 5) {
i__1 = *n;
for (j = 2; j <= i__1; j += 2) {
if (slaran_(&iseed[1]) > .5f) {
a_ref(j - 1, j) = a_ref(j, j);
a_ref(j, j - 1) = -a_ref(j, j);
a_ref(j, j) = a_ref(j - 1, j - 1);
}
/* L60: */
}
}
/* 3) If UPPER='T', set upper triangle of A to random numbers.
(but don't modify the corners of 2x2 blocks.) */
if (iupper != 0) {
i__1 = *n;
for (jc = 2; jc <= i__1; ++jc) {
if (a_ref(jc - 1, jc) != 0.f) {
jr = jc - 2;
} else {
jr = jc - 1;
}
slarnv_(&idist, &iseed[1], &jr, &a_ref(1, jc));
/* L70: */
}
}
/* 4) If SIM='T', apply similarity transformation.
-1
Transform is X A X , where X = U S V, thus
it is U S V A V' (1/S) U' */
if (isim != 0) {
/* Compute S (singular values of the eigenvector matrix)
according to CONDS and MODES */
slatm1_(modes, conds, &c__0, &c__0, &iseed[1], &ds[1], n, &iinfo);
if (iinfo != 0) {
*info = 3;
return 0;
}
/* Multiply by V and V' */
slarge_(n, &a[a_offset], lda, &iseed[1], &work[1], &iinfo);
if (iinfo != 0) {
*info = 4;
return 0;
}
/* Multiply by S and (1/S) */
i__1 = *n;
for (j = 1; j <= i__1; ++j) {
sscal_(n, &ds[j], &a_ref(j, 1), lda);
if (ds[j] != 0.f) {
r__1 = 1.f / ds[j];
sscal_(n, &r__1, &a_ref(1, j), &c__1);
} else {
*info = 5;
return 0;
}
/* L80: */
}
/* Multiply by U and U' */
slarge_(n, &a[a_offset], lda, &iseed[1], &work[1], &iinfo);
if (iinfo != 0) {
*info = 4;
return 0;
}
}
/* 5) Reduce the bandwidth. */
if (*kl < *n - 1) {
/* Reduce bandwidth -- kill column */
i__1 = *n - 1;
for (jcr = *kl + 1; jcr <= i__1; ++jcr) {
ic = jcr - *kl;
irows = *n + 1 - jcr;
icols = *n + *kl - jcr;
scopy_(&irows, &a_ref(jcr, ic), &c__1, &work[1], &c__1);
xnorms = work[1];
slarfg_(&irows, &xnorms, &work[2], &c__1, &tau);
work[1] = 1.f;
sgemv_("T", &irows, &icols, &c_b39, &a_ref(jcr, ic + 1), lda, &
work[1], &c__1, &c_b23, &work[irows + 1], &c__1);
r__1 = -tau;
sger_(&irows, &icols, &r__1, &work[1], &c__1, &work[irows + 1], &
c__1, &a_ref(jcr, ic + 1), lda);
sgemv_("N", n, &irows, &c_b39, &a_ref(1, jcr), lda, &work[1], &
c__1, &c_b23, &work[irows + 1], &c__1);
r__1 = -tau;
sger_(n, &irows, &r__1, &work[irows + 1], &c__1, &work[1], &c__1,
&a_ref(1, jcr), lda);
a_ref(jcr, ic) = xnorms;
i__2 = irows - 1;
slaset_("Full", &i__2, &c__1, &c_b23, &c_b23, &a_ref(jcr + 1, ic),
lda);
/* L90: */
}
} else if (*ku < *n - 1) {
/* Reduce upper bandwidth -- kill a row at a time. */
i__1 = *n - 1;
for (jcr = *ku + 1; jcr <= i__1; ++jcr) {
ir = jcr - *ku;
irows = *n + *ku - jcr;
icols = *n + 1 - jcr;
scopy_(&icols, &a_ref(ir, jcr), lda, &work[1], &c__1);
xnorms = work[1];
slarfg_(&icols, &xnorms, &work[2], &c__1, &tau);
work[1] = 1.f;
sgemv_("N", &irows, &icols, &c_b39, &a_ref(ir + 1, jcr), lda, &
work[1], &c__1, &c_b23, &work[icols + 1], &c__1);
r__1 = -tau;
sger_(&irows, &icols, &r__1, &work[icols + 1], &c__1, &work[1], &
c__1, &a_ref(ir + 1, jcr), lda);
sgemv_("C", &icols, n, &c_b39, &a_ref(jcr, 1), lda, &work[1], &
c__1, &c_b23, &work[icols + 1], &c__1);
r__1 = -tau;
sger_(&icols, n, &r__1, &work[1], &c__1, &work[icols + 1], &c__1,
&a_ref(jcr, 1), lda);
a_ref(ir, jcr) = xnorms;
i__2 = icols - 1;
slaset_("Full", &c__1, &i__2, &c_b23, &c_b23, &a_ref(ir, jcr + 1),
lda);
/* L100: */
}
}
/* Scale the matrix to have norm ANORM */
if (*anorm >= 0.f) {
temp = slange_("M", n, n, &a[a_offset], lda, tempa);
if (temp > 0.f) {
alpha = *anorm / temp;
i__1 = *n;
for (j = 1; j <= i__1; ++j) {
sscal_(n, &alpha, &a_ref(1, j), &c__1);
/* L110: */
}
}
}
return 0;
/* End of SLATME */
} /* slatme_ */
/* Subroutine */ int slaqr2_(logical *wantt, logical *wantz, integer *n,
integer *ktop, integer *kbot, integer *nw, real *h__, integer *ldh,
integer *iloz, integer *ihiz, real *z__, integer *ldz, integer *ns,
integer *nd, real *sr, real *si, real *v, integer *ldv, integer *nh,
real *t, integer *ldt, integer *nv, real *wv, integer *ldwv, real *
work, integer *lwork)
{
/* System generated locals */
integer h_dim1, h_offset, t_dim1, t_offset, v_dim1, v_offset, wv_dim1,
wv_offset, z_dim1, z_offset, i__1, i__2, i__3, i__4;
real r__1, r__2, r__3, r__4, r__5, r__6;
/* Local variables */
integer i__, j, k;
real s, aa, bb, cc, dd, cs, sn;
integer jw;
real evi, evk, foo;
integer kln;
real tau, ulp;
integer lwk1, lwk2;
real beta;
integer kend, kcol, info, ifst, ilst, ltop, krow;
logical bulge;
integer infqr;
integer kwtop;
real safmin;
real safmax;
logical sorted;
real smlnum;
integer lwkopt;
/* -- LAPACK auxiliary routine (version 3.2.1) -- */
/* -- April 2009 -- */
/* This subroutine is identical to SLAQR3 except that it avoids */
/* recursion by calling SLAHQR instead of SLAQR4. */
/* ****************************************************************** */
/* Aggressive early deflation: */
/* This subroutine accepts as input an upper Hessenberg matrix */
/* H and performs an orthogonal similarity transformation */
/* designed to detect and deflate fully converged eigenvalues from */
/* a trailing principal submatrix. On output H has been over- */
/* written by a new Hessenberg matrix that is a perturbation of */
/* an orthogonal similarity transformation of H. It is to be */
/* hoped that the final version of H has many zero subdiagonal */
/* entries. */
/* ****************************************************************** */
/* WANTT (input) LOGICAL */
/* If .TRUE., then the Hessenberg matrix H is fully updated */
/* so that the quasi-triangular Schur factor may be */
/* computed (in cooperation with the calling subroutine). */
/* If .FALSE., then only enough of H is updated to preserve */
/* the eigenvalues. */
/* WANTZ (input) LOGICAL */
/* If .TRUE., then the orthogonal matrix Z is updated so */
/* so that the orthogonal Schur factor may be computed */
/* (in cooperation with the calling subroutine). */
/* If .FALSE., then Z is not referenced. */
/* N (input) INTEGER */
/* The order of the matrix H and (if WANTZ is .TRUE.) the */
/* order of the orthogonal matrix Z. */
/* KTOP (input) INTEGER */
/* It is assumed that either KTOP = 1 or H(KTOP,KTOP-1)=0. */
/* KBOT and KTOP together determine an isolated block */
/* along the diagonal of the Hessenberg matrix. */
/* KBOT (input) INTEGER */
/* It is assumed without a check that either */
/* KBOT = N or H(KBOT+1,KBOT)=0. KBOT and KTOP together */
/* determine an isolated block along the diagonal of the */
/* Hessenberg matrix. */
/* NW (input) INTEGER */
/* Deflation window size. 1 .LE. NW .LE. (KBOT-KTOP+1). */
/* H (input/output) REAL array, dimension (LDH,N) */
/* On input the initial N-by-N section of H stores the */
/* Hessenberg matrix undergoing aggressive early deflation. */
/* On output H has been transformed by an orthogonal */
/* similarity transformation, perturbed, and the returned */
/* to Hessenberg form that (it is to be hoped) has some */
/* zero subdiagonal entries. */
/* LDH (input) integer */
/* Leading dimension of H just as declared in the calling */
/* subroutine. N .LE. LDH */
/* ILOZ (input) INTEGER */
/* IHIZ (input) INTEGER */
/* Specify the rows of Z to which transformations must be */
/* Z (input/output) REAL array, dimension (LDZ,N) */
/* IF WANTZ is .TRUE., then on output, the orthogonal */
/* similarity transformation mentioned above has been */
/* accumulated into Z(ILOZ:IHIZ,ILO:IHI) from the right. */
/* If WANTZ is .FALSE., then Z is unreferenced. */
/* LDZ (input) integer */
/* The leading dimension of Z just as declared in the */
/* calling subroutine. 1 .LE. LDZ. */
/* NS (output) integer */
/* The number of unconverged (ie approximate) eigenvalues */
/* returned in SR and SI that may be used as shifts by the */
/* calling subroutine. */
/* ND (output) integer */
/* The number of converged eigenvalues uncovered by this */
/* subroutine. */
/* SR (output) REAL array, dimension KBOT */
/* SI (output) REAL array, dimension KBOT */
/* On output, the real and imaginary parts of approximate */
/* eigenvalues that may be used for shifts are stored in */
/* SR(KBOT-ND-NS+1) through SR(KBOT-ND) and */
/* SI(KBOT-ND-NS+1) through SI(KBOT-ND), respectively. */
/* The real and imaginary parts of converged eigenvalues */
/* are stored in SR(KBOT-ND+1) through SR(KBOT) and */
/* SI(KBOT-ND+1) through SI(KBOT), respectively. */
/* V (workspace) REAL array, dimension (LDV,NW) */
/* An NW-by-NW work array. */
/* LDV (input) integer scalar */
/* The leading dimension of V just as declared in the */
/* calling subroutine. NW .LE. LDV */
/* NH (input) integer scalar */
/* The number of columns of T. NH.GE.NW. */
/* T (workspace) REAL array, dimension (LDT,NW) */
/* LDT (input) integer */
/* The leading dimension of T just as declared in the */
/* calling subroutine. NW .LE. LDT */
/* NV (input) integer */
/* The number of rows of work array WV available for */
/* workspace. NV.GE.NW. */
/* WV (workspace) REAL array, dimension (LDWV,NW) */
/* LDWV (input) integer */
/* The leading dimension of W just as declared in the */
/* calling subroutine. NW .LE. LDV */
/* WORK (workspace) REAL array, dimension LWORK. */
/* On exit, WORK(1) is set to an estimate of the optimal value */
/* of LWORK for the given values of N, NW, KTOP and KBOT. */
/* LWORK (input) integer */
/* The dimension of the work array WORK. LWORK = 2*NW */
/* suffices, but greater efficiency may result from larger */
/* values of LWORK. */
/* If LWORK = -1, then a workspace query is assumed; SLAQR2 */
/* only estimates the optimal workspace size for the given */
/* values of N, NW, KTOP and KBOT. The estimate is returned */
/* in WORK(1). No error message related to LWORK is issued */
/* by XERBLA. Neither H nor Z are accessed. */
/* ================================================================ */
/* Based on contributions by */
/* Karen Braman and Ralph Byers, Department of Mathematics, */
/* University of Kansas, USA */
/* ================================================================ */
/* ==== Estimate optimal workspace. ==== */
/* Parameter adjustments */
h_dim1 = *ldh;
h_offset = 1 + h_dim1;
h__ -= h_offset;
z_dim1 = *ldz;
z_offset = 1 + z_dim1;
z__ -= z_offset;
--sr;
--si;
v_dim1 = *ldv;
v_offset = 1 + v_dim1;
v -= v_offset;
t_dim1 = *ldt;
t_offset = 1 + t_dim1;
t -= t_offset;
wv_dim1 = *ldwv;
wv_offset = 1 + wv_dim1;
wv -= wv_offset;
--work;
/* Function Body */
/* Computing MIN */
i__1 = *nw, i__2 = *kbot - *ktop + 1;
jw = min(i__1,i__2);
if (jw <= 2) {
lwkopt = 1;
} else {
/* ==== Workspace query call to SGEHRD ==== */
i__1 = jw - 1;
sgehrd_(&jw, &c__1, &i__1, &t[t_offset], ldt, &work[1], &work[1], &
c_n1, &info);
lwk1 = (integer) work[1];
/* ==== Workspace query call to SORMHR ==== */
i__1 = jw - 1;
sormhr_("R", "N", &jw, &jw, &c__1, &i__1, &t[t_offset], ldt, &work[1],
&v[v_offset], ldv, &work[1], &c_n1, &info);
lwk2 = (integer) work[1];
/* ==== Optimal workspace ==== */
lwkopt = jw + max(lwk1,lwk2);
}
/* ==== Quick return in case of workspace query. ==== */
if (*lwork == -1) {
work[1] = (real) lwkopt;
return 0;
}
*ns = 0;
*nd = 0;
work[1] = 1.f;
if (*ktop > *kbot) {
return 0;
}
if (*nw < 1) {
return 0;
}
/* ==== Machine constants ==== */
safmin = slamch_("SAFE MINIMUM");
safmax = 1.f / safmin;
slabad_(&safmin, &safmax);
ulp = slamch_("PRECISION");
smlnum = safmin * ((real) (*n) / ulp);
/* ==== Setup deflation window ==== */
/* Computing MIN */
i__1 = *nw, i__2 = *kbot - *ktop + 1;
jw = min(i__1,i__2);
kwtop = *kbot - jw + 1;
if (kwtop == *ktop) {
s = 0.f;
} else {
s = h__[kwtop + (kwtop - 1) * h_dim1];
}
if (*kbot == kwtop) {
/* ==== 1-by-1 deflation window: not much to do ==== */
sr[kwtop] = h__[kwtop + kwtop * h_dim1];
si[kwtop] = 0.f;
*ns = 1;
*nd = 0;
/* Computing MAX */
r__2 = smlnum, r__3 = ulp * (r__1 = h__[kwtop + kwtop * h_dim1], dabs(
r__1));
if (dabs(s) <= dmax(r__2,r__3)) {
*ns = 0;
*nd = 1;
if (kwtop > *ktop) {
h__[kwtop + (kwtop - 1) * h_dim1] = 0.f;
}
}
work[1] = 1.f;
return 0;
}
/* ==== Convert to spike-triangular form. (In case of a */
/* . rare QR failure, this routine continues to do */
/* . aggressive early deflation using that part of */
/* . the deflation window that converged using INFQR */
/* . here and there to keep track.) ==== */
slacpy_("U", &jw, &jw, &h__[kwtop + kwtop * h_dim1], ldh, &t[t_offset],
ldt);
i__1 = jw - 1;
i__2 = *ldh + 1;
i__3 = *ldt + 1;
scopy_(&i__1, &h__[kwtop + 1 + kwtop * h_dim1], &i__2, &t[t_dim1 + 2], &
i__3);
slaset_("A", &jw, &jw, &c_b12, &c_b13, &v[v_offset], ldv);
slahqr_(&c_true, &c_true, &jw, &c__1, &jw, &t[t_offset], ldt, &sr[kwtop],
&si[kwtop], &c__1, &jw, &v[v_offset], ldv, &infqr);
/* ==== STREXC needs a clean margin near the diagonal ==== */
i__1 = jw - 3;
for (j = 1; j <= i__1; ++j) {
t[j + 2 + j * t_dim1] = 0.f;
t[j + 3 + j * t_dim1] = 0.f;
}
if (jw > 2) {
t[jw + (jw - 2) * t_dim1] = 0.f;
}
/* ==== Deflation detection loop ==== */
*ns = jw;
ilst = infqr + 1;
L20:
if (ilst <= *ns) {
if (*ns == 1) {
bulge = FALSE_;
} else {
bulge = t[*ns + (*ns - 1) * t_dim1] != 0.f;
}
/* ==== Small spike tip test for deflation ==== */
if (! bulge) {
/* ==== Real eigenvalue ==== */
foo = (r__1 = t[*ns + *ns * t_dim1], dabs(r__1));
if (foo == 0.f) {
foo = dabs(s);
}
/* Computing MAX */
r__2 = smlnum, r__3 = ulp * foo;
if ((r__1 = s * v[*ns * v_dim1 + 1], dabs(r__1)) <= dmax(r__2,
r__3)) {
/* ==== Deflatable ==== */
--(*ns);
} else {
/* ==== Undeflatable. Move it up out of the way. */
/* . (STREXC can not fail in this case.) ==== */
ifst = *ns;
strexc_("V", &jw, &t[t_offset], ldt, &v[v_offset], ldv, &ifst,
&ilst, &work[1], &info);
++ilst;
}
} else {
/* ==== Complex conjugate pair ==== */
foo = (r__3 = t[*ns + *ns * t_dim1], dabs(r__3)) + sqrt((r__1 = t[
*ns + (*ns - 1) * t_dim1], dabs(r__1))) * sqrt((r__2 = t[*
ns - 1 + *ns * t_dim1], dabs(r__2)));
if (foo == 0.f) {
foo = dabs(s);
}
/* Computing MAX */
r__3 = (r__1 = s * v[*ns * v_dim1 + 1], dabs(r__1)), r__4 = (r__2
= s * v[(*ns - 1) * v_dim1 + 1], dabs(r__2));
/* Computing MAX */
r__5 = smlnum, r__6 = ulp * foo;
if (dmax(r__3,r__4) <= dmax(r__5,r__6)) {
/* ==== Deflatable ==== */
*ns += -2;
} else {
/* ==== Undeflatable. Move them up out of the way. */
/* . Fortunately, STREXC does the right thing with */
/* . ILST in case of a rare exchange failure. ==== */
ifst = *ns;
strexc_("V", &jw, &t[t_offset], ldt, &v[v_offset], ldv, &ifst,
&ilst, &work[1], &info);
ilst += 2;
}
}
/* ==== End deflation detection loop ==== */
goto L20;
}
/* ==== Return to Hessenberg form ==== */
if (*ns == 0) {
s = 0.f;
}
if (*ns < jw) {
/* ==== sorting diagonal blocks of T improves accuracy for */
/* . graded matrices. Bubble sort deals well with */
/* . exchange failures. ==== */
sorted = FALSE_;
i__ = *ns + 1;
L30:
if (sorted) {
goto L50;
}
sorted = TRUE_;
kend = i__ - 1;
i__ = infqr + 1;
if (i__ == *ns) {
k = i__ + 1;
} else if (t[i__ + 1 + i__ * t_dim1] == 0.f) {
k = i__ + 1;
} else {
k = i__ + 2;
}
L40:
if (k <= kend) {
if (k == i__ + 1) {
evi = (r__1 = t[i__ + i__ * t_dim1], dabs(r__1));
} else {
evi = (r__3 = t[i__ + i__ * t_dim1], dabs(r__3)) + sqrt((r__1
= t[i__ + 1 + i__ * t_dim1], dabs(r__1))) * sqrt((
r__2 = t[i__ + (i__ + 1) * t_dim1], dabs(r__2)));
}
if (k == kend) {
evk = (r__1 = t[k + k * t_dim1], dabs(r__1));
} else if (t[k + 1 + k * t_dim1] == 0.f) {
evk = (r__1 = t[k + k * t_dim1], dabs(r__1));
} else {
evk = (r__3 = t[k + k * t_dim1], dabs(r__3)) + sqrt((r__1 = t[
k + 1 + k * t_dim1], dabs(r__1))) * sqrt((r__2 = t[k
+ (k + 1) * t_dim1], dabs(r__2)));
}
if (evi >= evk) {
i__ = k;
} else {
sorted = FALSE_;
ifst = i__;
ilst = k;
strexc_("V", &jw, &t[t_offset], ldt, &v[v_offset], ldv, &ifst,
&ilst, &work[1], &info);
if (info == 0) {
i__ = ilst;
} else {
i__ = k;
}
}
if (i__ == kend) {
k = i__ + 1;
} else if (t[i__ + 1 + i__ * t_dim1] == 0.f) {
k = i__ + 1;
} else {
k = i__ + 2;
}
goto L40;
}
goto L30;
L50:
;
}
/* ==== Restore shift/eigenvalue array from T ==== */
i__ = jw;
L60:
if (i__ >= infqr + 1) {
if (i__ == infqr + 1) {
sr[kwtop + i__ - 1] = t[i__ + i__ * t_dim1];
si[kwtop + i__ - 1] = 0.f;
--i__;
} else if (t[i__ + (i__ - 1) * t_dim1] == 0.f) {
sr[kwtop + i__ - 1] = t[i__ + i__ * t_dim1];
si[kwtop + i__ - 1] = 0.f;
--i__;
} else {
aa = t[i__ - 1 + (i__ - 1) * t_dim1];
cc = t[i__ + (i__ - 1) * t_dim1];
bb = t[i__ - 1 + i__ * t_dim1];
dd = t[i__ + i__ * t_dim1];
slanv2_(&aa, &bb, &cc, &dd, &sr[kwtop + i__ - 2], &si[kwtop + i__
- 2], &sr[kwtop + i__ - 1], &si[kwtop + i__ - 1], &cs, &
sn);
i__ += -2;
}
goto L60;
}
if (*ns < jw || s == 0.f) {
if (*ns > 1 && s != 0.f) {
/* ==== Reflect spike back into lower triangle ==== */
scopy_(ns, &v[v_offset], ldv, &work[1], &c__1);
beta = work[1];
slarfg_(ns, &beta, &work[2], &c__1, &tau);
work[1] = 1.f;
i__1 = jw - 2;
i__2 = jw - 2;
slaset_("L", &i__1, &i__2, &c_b12, &c_b12, &t[t_dim1 + 3], ldt);
slarf_("L", ns, &jw, &work[1], &c__1, &tau, &t[t_offset], ldt, &
work[jw + 1]);
slarf_("R", ns, ns, &work[1], &c__1, &tau, &t[t_offset], ldt, &
work[jw + 1]);
slarf_("R", &jw, ns, &work[1], &c__1, &tau, &v[v_offset], ldv, &
work[jw + 1]);
i__1 = *lwork - jw;
sgehrd_(&jw, &c__1, ns, &t[t_offset], ldt, &work[1], &work[jw + 1]
, &i__1, &info);
}
/* ==== Copy updated reduced window into place ==== */
if (kwtop > 1) {
h__[kwtop + (kwtop - 1) * h_dim1] = s * v[v_dim1 + 1];
}
slacpy_("U", &jw, &jw, &t[t_offset], ldt, &h__[kwtop + kwtop * h_dim1]
, ldh);
i__1 = jw - 1;
i__2 = *ldt + 1;
i__3 = *ldh + 1;
scopy_(&i__1, &t[t_dim1 + 2], &i__2, &h__[kwtop + 1 + kwtop * h_dim1],
&i__3);
/* ==== Accumulate orthogonal matrix in order update */
/* . H and Z, if requested. ==== */
if (*ns > 1 && s != 0.f) {
i__1 = *lwork - jw;
sormhr_("R", "N", &jw, ns, &c__1, ns, &t[t_offset], ldt, &work[1],
&v[v_offset], ldv, &work[jw + 1], &i__1, &info);
}
/* ==== Update vertical slab in H ==== */
if (*wantt) {
ltop = 1;
} else {
ltop = *ktop;
}
i__1 = kwtop - 1;
i__2 = *nv;
for (krow = ltop; i__2 < 0 ? krow >= i__1 : krow <= i__1; krow +=
i__2) {
/* Computing MIN */
i__3 = *nv, i__4 = kwtop - krow;
kln = min(i__3,i__4);
sgemm_("N", "N", &kln, &jw, &jw, &c_b13, &h__[krow + kwtop *
h_dim1], ldh, &v[v_offset], ldv, &c_b12, &wv[wv_offset],
ldwv);
slacpy_("A", &kln, &jw, &wv[wv_offset], ldwv, &h__[krow + kwtop *
h_dim1], ldh);
}
/* ==== Update horizontal slab in H ==== */
if (*wantt) {
i__2 = *n;
i__1 = *nh;
for (kcol = *kbot + 1; i__1 < 0 ? kcol >= i__2 : kcol <= i__2;
kcol += i__1) {
/* Computing MIN */
i__3 = *nh, i__4 = *n - kcol + 1;
kln = min(i__3,i__4);
sgemm_("C", "N", &jw, &kln, &jw, &c_b13, &v[v_offset], ldv, &
h__[kwtop + kcol * h_dim1], ldh, &c_b12, &t[t_offset],
ldt);
slacpy_("A", &jw, &kln, &t[t_offset], ldt, &h__[kwtop + kcol *
h_dim1], ldh);
}
}
/* ==== Update vertical slab in Z ==== */
if (*wantz) {
i__1 = *ihiz;
i__2 = *nv;
for (krow = *iloz; i__2 < 0 ? krow >= i__1 : krow <= i__1; krow +=
i__2) {
/* Computing MIN */
i__3 = *nv, i__4 = *ihiz - krow + 1;
kln = min(i__3,i__4);
sgemm_("N", "N", &kln, &jw, &jw, &c_b13, &z__[krow + kwtop *
z_dim1], ldz, &v[v_offset], ldv, &c_b12, &wv[
wv_offset], ldwv);
slacpy_("A", &kln, &jw, &wv[wv_offset], ldwv, &z__[krow +
kwtop * z_dim1], ldz);
}
}
}
*nd = jw - *ns;
/* . INFQR from the spike length takes care */
/* . of the case of a rare QR failure while */
/* . calculating eigenvalues of the deflation */
/* . window.) ==== */
*ns -= infqr;
/* ==== Return optimal workspace. ==== */
work[1] = (real) lwkopt;
/* ==== End of SLAQR2 ==== */
return 0;
} /* slaqr2_ */
/* Subroutine */ int slaqrb_(logical *wantt, integer *n, integer *ilo,
integer *ihi, real *h__, integer *ldh, real *wr, real *wi, real *z__,
integer *info)
{
/* System generated locals */
integer h_dim1, h_offset, i__1, i__2, i__3, i__4;
real r__1, r__2;
/* Local variables */
static integer i__, j, k, l, m;
static real s, v[3];
static integer i1, i2;
static real t1, t2, t3, v1, v2, v3, h00, h10, h11, h12, h21, h22, h33,
h44;
static integer nh;
static real cs;
static integer nr;
static real sn, h33s, h44s;
static integer itn, its;
static real ulp, sum, tst1, h43h34, unfl, ovfl, work[1];
extern /* Subroutine */ int srot_(integer *, real *, integer *, real *,
integer *, real *, real *), scopy_(integer *, real *, integer *,
real *, integer *), slanv2_(real *, real *, real *, real *, real *
, real *, real *, real *, real *, real *), slabad_(real *, real *)
;
extern doublereal slamch_(char *, ftnlen);
extern /* Subroutine */ int slarfg_(integer *, real *, real *, integer *,
real *);
extern doublereal slanhs_(char *, integer *, real *, integer *, real *,
ftnlen);
static real smlnum;
/* %------------------% */
/* | Scalar Arguments | */
/* %------------------% */
/* %-----------------% */
/* | Array Arguments | */
/* %-----------------% */
/* %------------% */
/* | Parameters | */
/* %------------% */
/* %------------------------% */
/* | Local Scalars & Arrays | */
/* %------------------------% */
/* %--------------------% */
/* | External Functions | */
/* %--------------------% */
/* %----------------------% */
/* | External Subroutines | */
/* %----------------------% */
/* %-----------------------% */
/* | Executable Statements | */
/* %-----------------------% */
/* Parameter adjustments */
h_dim1 = *ldh;
h_offset = 1 + h_dim1;
h__ -= h_offset;
--wr;
--wi;
--z__;
/* Function Body */
*info = 0;
/* %--------------------------% */
/* | Quick return if possible | */
/* %--------------------------% */
if (*n == 0) {
return 0;
}
if (*ilo == *ihi) {
wr[*ilo] = h__[*ilo + *ilo * h_dim1];
wi[*ilo] = 0.f;
return 0;
}
/* %---------------------------------------------% */
/* | Initialize the vector of last components of | */
/* | the Schur vectors for accumulation. | */
/* %---------------------------------------------% */
i__1 = *n - 1;
for (j = 1; j <= i__1; ++j) {
z__[j] = 0.f;
/* L5: */
}
z__[*n] = 1.f;
nh = *ihi - *ilo + 1;
/* %-------------------------------------------------------------% */
/* | Set machine-dependent constants for the stopping criterion. | */
/* | If norm(H) <= sqrt(OVFL), overflow should not occur. | */
/* %-------------------------------------------------------------% */
unfl = slamch_("safe minimum", (ftnlen)12);
ovfl = 1.f / unfl;
slabad_(&unfl, &ovfl);
ulp = slamch_("precision", (ftnlen)9);
smlnum = unfl * (nh / ulp);
/* %---------------------------------------------------------------% */
/* | I1 and I2 are the indices of the first row and last column | */
/* | of H to which transformations must be applied. If eigenvalues | */
/* | only are computed, I1 and I2 are set inside the main loop. | */
/* | Zero out H(J+2,J) = ZERO for J=1:N if WANTT = .TRUE. | */
/* | else H(J+2,J) for J=ILO:IHI-ILO-1 if WANTT = .FALSE. | */
/* %---------------------------------------------------------------% */
if (*wantt) {
i1 = 1;
i2 = *n;
i__1 = i2 - 2;
for (i__ = 1; i__ <= i__1; ++i__) {
h__[i1 + i__ + 1 + i__ * h_dim1] = 0.f;
/* L8: */
}
} else {
i__1 = *ihi - *ilo - 1;
for (i__ = 1; i__ <= i__1; ++i__) {
h__[*ilo + i__ + 1 + (*ilo + i__ - 1) * h_dim1] = 0.f;
/* L9: */
}
}
/* %---------------------------------------------------% */
/* | ITN is the total number of QR iterations allowed. | */
/* %---------------------------------------------------% */
itn = nh * 30;
/* ------------------------------------------------------------------ */
/* The main loop begins here. I is the loop index and decreases from */
/* IHI to ILO in steps of 1 or 2. Each iteration of the loop works */
/* with the active submatrix in rows and columns L to I. */
/* Eigenvalues I+1 to IHI have already converged. Either L = ILO or */
/* H(L,L-1) is negligible so that the matrix splits. */
/* ------------------------------------------------------------------ */
i__ = *ihi;
L10:
l = *ilo;
if (i__ < *ilo) {
goto L150;
}
/* %--------------------------------------------------------------% */
/* | Perform QR iterations on rows and columns ILO to I until a | */
/* | submatrix of order 1 or 2 splits off at the bottom because a | */
/* | subdiagonal element has become negligible. | */
/* %--------------------------------------------------------------% */
i__1 = itn;
for (its = 0; its <= i__1; ++its) {
/* %----------------------------------------------% */
/* | Look for a single small subdiagonal element. | */
/* %----------------------------------------------% */
i__2 = l + 1;
for (k = i__; k >= i__2; --k) {
tst1 = (r__1 = h__[k - 1 + (k - 1) * h_dim1], dabs(r__1)) + (r__2
= h__[k + k * h_dim1], dabs(r__2));
if (tst1 == 0.f) {
i__3 = i__ - l + 1;
tst1 = slanhs_("1", &i__3, &h__[l + l * h_dim1], ldh, work, (
ftnlen)1);
}
/* Computing MAX */
r__2 = ulp * tst1;
if ((r__1 = h__[k + (k - 1) * h_dim1], dabs(r__1)) <= dmax(r__2,
smlnum)) {
goto L30;
}
/* L20: */
}
L30:
l = k;
if (l > *ilo) {
/* %------------------------% */
/* | H(L,L-1) is negligible | */
/* %------------------------% */
h__[l + (l - 1) * h_dim1] = 0.f;
}
/* %-------------------------------------------------------------% */
/* | Exit from loop if a submatrix of order 1 or 2 has split off | */
/* %-------------------------------------------------------------% */
if (l >= i__ - 1) {
goto L140;
}
/* %---------------------------------------------------------% */
/* | Now the active submatrix is in rows and columns L to I. | */
/* | If eigenvalues only are being computed, only the active | */
/* | submatrix need be transformed. | */
/* %---------------------------------------------------------% */
if (! (*wantt)) {
i1 = l;
i2 = i__;
}
if (its == 10 || its == 20) {
/* %-------------------% */
/* | Exceptional shift | */
/* %-------------------% */
s = (r__1 = h__[i__ + (i__ - 1) * h_dim1], dabs(r__1)) + (r__2 =
h__[i__ - 1 + (i__ - 2) * h_dim1], dabs(r__2));
h44 = s * .75f;
h33 = h44;
h43h34 = s * -.4375f * s;
} else {
/* %-----------------------------------------% */
/* | Prepare to use Wilkinson's double shift | */
/* %-----------------------------------------% */
h44 = h__[i__ + i__ * h_dim1];
h33 = h__[i__ - 1 + (i__ - 1) * h_dim1];
h43h34 = h__[i__ + (i__ - 1) * h_dim1] * h__[i__ - 1 + i__ *
h_dim1];
}
/* %-----------------------------------------------------% */
/* | Look for two consecutive small subdiagonal elements | */
/* %-----------------------------------------------------% */
i__2 = l;
for (m = i__ - 2; m >= i__2; --m) {
/* %---------------------------------------------------------% */
/* | Determine the effect of starting the double-shift QR | */
/* | iteration at row M, and see if this would make H(M,M-1) | */
/* | negligible. | */
/* %---------------------------------------------------------% */
h11 = h__[m + m * h_dim1];
h22 = h__[m + 1 + (m + 1) * h_dim1];
h21 = h__[m + 1 + m * h_dim1];
h12 = h__[m + (m + 1) * h_dim1];
h44s = h44 - h11;
h33s = h33 - h11;
v1 = (h33s * h44s - h43h34) / h21 + h12;
v2 = h22 - h11 - h33s - h44s;
v3 = h__[m + 2 + (m + 1) * h_dim1];
s = dabs(v1) + dabs(v2) + dabs(v3);
v1 /= s;
v2 /= s;
v3 /= s;
v[0] = v1;
v[1] = v2;
v[2] = v3;
if (m == l) {
goto L50;
}
h00 = h__[m - 1 + (m - 1) * h_dim1];
h10 = h__[m + (m - 1) * h_dim1];
tst1 = dabs(v1) * (dabs(h00) + dabs(h11) + dabs(h22));
if (dabs(h10) * (dabs(v2) + dabs(v3)) <= ulp * tst1) {
goto L50;
}
/* L40: */
}
L50:
/* %----------------------% */
/* | Double-shift QR step | */
/* %----------------------% */
i__2 = i__ - 1;
for (k = m; k <= i__2; ++k) {
/* ------------------------------------------------------------ */
/* The first iteration of this loop determines a reflection G */
/* from the vector V and applies it from left and right to H, */
/* thus creating a nonzero bulge below the subdiagonal. */
/* Each subsequent iteration determines a reflection G to */
/* restore the Hessenberg form in the (K-1)th column, and thus */
/* chases the bulge one step toward the bottom of the active */
/* submatrix. NR is the order of G. */
/* ------------------------------------------------------------ */
/* Computing MIN */
i__3 = 3, i__4 = i__ - k + 1;
nr = min(i__3,i__4);
if (k > m) {
scopy_(&nr, &h__[k + (k - 1) * h_dim1], &c__1, v, &c__1);
}
slarfg_(&nr, v, &v[1], &c__1, &t1);
if (k > m) {
h__[k + (k - 1) * h_dim1] = v[0];
h__[k + 1 + (k - 1) * h_dim1] = 0.f;
if (k < i__ - 1) {
h__[k + 2 + (k - 1) * h_dim1] = 0.f;
}
} else if (m > l) {
h__[k + (k - 1) * h_dim1] = -h__[k + (k - 1) * h_dim1];
}
v2 = v[1];
t2 = t1 * v2;
if (nr == 3) {
v3 = v[2];
t3 = t1 * v3;
/* %------------------------------------------------% */
/* | Apply G from the left to transform the rows of | */
/* | the matrix in columns K to I2. | */
/* %------------------------------------------------% */
i__3 = i2;
for (j = k; j <= i__3; ++j) {
sum = h__[k + j * h_dim1] + v2 * h__[k + 1 + j * h_dim1]
+ v3 * h__[k + 2 + j * h_dim1];
h__[k + j * h_dim1] -= sum * t1;
h__[k + 1 + j * h_dim1] -= sum * t2;
h__[k + 2 + j * h_dim1] -= sum * t3;
/* L60: */
}
/* %----------------------------------------------------% */
/* | Apply G from the right to transform the columns of | */
/* | the matrix in rows I1 to min(K+3,I). | */
/* %----------------------------------------------------% */
/* Computing MIN */
i__4 = k + 3;
i__3 = min(i__4,i__);
for (j = i1; j <= i__3; ++j) {
sum = h__[j + k * h_dim1] + v2 * h__[j + (k + 1) * h_dim1]
+ v3 * h__[j + (k + 2) * h_dim1];
h__[j + k * h_dim1] -= sum * t1;
h__[j + (k + 1) * h_dim1] -= sum * t2;
h__[j + (k + 2) * h_dim1] -= sum * t3;
/* L70: */
}
/* %----------------------------------% */
/* | Accumulate transformations for Z | */
/* %----------------------------------% */
sum = z__[k] + v2 * z__[k + 1] + v3 * z__[k + 2];
z__[k] -= sum * t1;
z__[k + 1] -= sum * t2;
z__[k + 2] -= sum * t3;
} else if (nr == 2) {
/* %------------------------------------------------% */
/* | Apply G from the left to transform the rows of | */
/* | the matrix in columns K to I2. | */
/* %------------------------------------------------% */
i__3 = i2;
for (j = k; j <= i__3; ++j) {
sum = h__[k + j * h_dim1] + v2 * h__[k + 1 + j * h_dim1];
h__[k + j * h_dim1] -= sum * t1;
h__[k + 1 + j * h_dim1] -= sum * t2;
/* L90: */
}
/* %----------------------------------------------------% */
/* | Apply G from the right to transform the columns of | */
/* | the matrix in rows I1 to min(K+3,I). | */
/* %----------------------------------------------------% */
i__3 = i__;
for (j = i1; j <= i__3; ++j) {
sum = h__[j + k * h_dim1] + v2 * h__[j + (k + 1) * h_dim1]
;
h__[j + k * h_dim1] -= sum * t1;
h__[j + (k + 1) * h_dim1] -= sum * t2;
/* L100: */
}
/* %----------------------------------% */
/* | Accumulate transformations for Z | */
/* %----------------------------------% */
sum = z__[k] + v2 * z__[k + 1];
z__[k] -= sum * t1;
z__[k + 1] -= sum * t2;
}
/* L120: */
}
/* L130: */
}
/* %-------------------------------------------------------% */
/* | Failure to converge in remaining number of iterations | */
/* %-------------------------------------------------------% */
*info = i__;
return 0;
L140:
if (l == i__) {
/* %------------------------------------------------------% */
/* | H(I,I-1) is negligible: one eigenvalue has converged | */
/* %------------------------------------------------------% */
wr[i__] = h__[i__ + i__ * h_dim1];
wi[i__] = 0.f;
} else if (l == i__ - 1) {
/* %--------------------------------------------------------% */
/* | H(I-1,I-2) is negligible; | */
/* | a pair of eigenvalues have converged. | */
/* | | */
/* | Transform the 2-by-2 submatrix to standard Schur form, | */
/* | and compute and store the eigenvalues. | */
/* %--------------------------------------------------------% */
slanv2_(&h__[i__ - 1 + (i__ - 1) * h_dim1], &h__[i__ - 1 + i__ *
h_dim1], &h__[i__ + (i__ - 1) * h_dim1], &h__[i__ + i__ *
h_dim1], &wr[i__ - 1], &wi[i__ - 1], &wr[i__], &wi[i__], &cs,
&sn);
if (*wantt) {
/* %-----------------------------------------------------% */
/* | Apply the transformation to the rest of H and to Z, | */
/* | as required. | */
/* %-----------------------------------------------------% */
if (i2 > i__) {
i__1 = i2 - i__;
srot_(&i__1, &h__[i__ - 1 + (i__ + 1) * h_dim1], ldh, &h__[
i__ + (i__ + 1) * h_dim1], ldh, &cs, &sn);
}
i__1 = i__ - i1 - 1;
srot_(&i__1, &h__[i1 + (i__ - 1) * h_dim1], &c__1, &h__[i1 + i__ *
h_dim1], &c__1, &cs, &sn);
sum = cs * z__[i__ - 1] + sn * z__[i__];
z__[i__] = cs * z__[i__] - sn * z__[i__ - 1];
z__[i__ - 1] = sum;
}
}
/* %---------------------------------------------------------% */
/* | Decrement number of remaining iterations, and return to | */
/* | start of the main loop with new value of I. | */
/* %---------------------------------------------------------% */
itn -= its;
i__ = l - 1;
goto L10;
L150:
return 0;
/* %---------------% */
/* | End of slaqrb | */
/* %---------------% */
} /* slaqrb_ */
/* Subroutine */
int sgehd2_(integer *n, integer *ilo, integer *ihi, real *a, integer *lda, real *tau, real *work, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, i__1, i__2, i__3;
/* Local variables */
integer i__;
real aii;
extern /* Subroutine */
int slarf_(char *, integer *, integer *, real *, integer *, real *, real *, integer *, real *), xerbla_( char *, integer *), slarfg_(integer *, real *, real *, integer *, real *);
/* -- LAPACK computational routine (version 3.4.2) -- */
/* -- LAPACK is a software package provided by Univ. of Tennessee, -- */
/* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- */
/* September 2012 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Test the input parameters */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
--tau;
--work;
/* Function Body */
*info = 0;
if (*n < 0)
{
*info = -1;
}
else if (*ilo < 1 || *ilo > max(1,*n))
{
*info = -2;
}
else if (*ihi < min(*ilo,*n) || *ihi > *n)
{
*info = -3;
}
else if (*lda < max(1,*n))
{
*info = -5;
}
if (*info != 0)
{
i__1 = -(*info);
xerbla_("SGEHD2", &i__1);
return 0;
}
i__1 = *ihi - 1;
for (i__ = *ilo;
i__ <= i__1;
++i__)
{
/* Compute elementary reflector H(i) to annihilate A(i+2:ihi,i) */
i__2 = *ihi - i__;
/* Computing MIN */
i__3 = i__ + 2;
slarfg_(&i__2, &a[i__ + 1 + i__ * a_dim1], &a[min(i__3,*n) + i__ * a_dim1], &c__1, &tau[i__]);
aii = a[i__ + 1 + i__ * a_dim1];
a[i__ + 1 + i__ * a_dim1] = 1.f;
/* Apply H(i) to A(1:ihi,i+1:ihi) from the right */
i__2 = *ihi - i__;
slarf_("Right", ihi, &i__2, &a[i__ + 1 + i__ * a_dim1], &c__1, &tau[ i__], &a[(i__ + 1) * a_dim1 + 1], lda, &work[1]);
/* Apply H(i) to A(i+1:ihi,i+1:n) from the left */
i__2 = *ihi - i__;
i__3 = *n - i__;
slarf_("Left", &i__2, &i__3, &a[i__ + 1 + i__ * a_dim1], &c__1, &tau[ i__], &a[i__ + 1 + (i__ + 1) * a_dim1], lda, &work[1]);
a[i__ + 1 + i__ * a_dim1] = aii;
/* L10: */
}
return 0;
/* End of SGEHD2 */
}
/* Subroutine */ int shseqr_(char *job, char *compz, integer *n, integer *ilo,
integer *ihi, real *h__, integer *ldh, real *wr, real *wi, real *z__,
integer *ldz, real *work, integer *lwork, integer *info, ftnlen
job_len, ftnlen compz_len)
{
/* System generated locals */
address a__1[2];
integer h_dim1, h_offset, z_dim1, z_offset, i__1, i__2, i__3[2], i__4,
i__5;
real r__1, r__2;
char ch__1[2];
/* Builtin functions */
/* Subroutine */ int s_cat(char *, char **, integer *, integer *, ftnlen);
/* Local variables */
static integer i__, j, k, l;
static real s[225] /* was [15][15] */, v[16];
static integer i1, i2, ii, nh, nr, ns, nv;
static real vv[16];
static integer itn;
static real tau;
static integer its;
static real ulp, tst1;
static integer maxb;
static real absw;
static integer ierr;
static real unfl, temp, ovfl;
extern logical lsame_(char *, char *, ftnlen, ftnlen);
extern /* Subroutine */ int sscal_(integer *, real *, real *, integer *);
static integer itemp;
extern /* Subroutine */ int sgemv_(char *, integer *, integer *, real *,
real *, integer *, real *, integer *, real *, real *, integer *,
ftnlen);
static logical initz, wantt;
extern /* Subroutine */ int scopy_(integer *, real *, integer *, real *,
integer *);
static logical wantz;
extern doublereal slapy2_(real *, real *);
extern /* Subroutine */ int slabad_(real *, real *);
extern doublereal slamch_(char *, ftnlen);
extern /* Subroutine */ int xerbla_(char *, integer *, ftnlen);
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *, ftnlen, ftnlen);
extern /* Subroutine */ int slarfg_(integer *, real *, real *, integer *,
real *);
extern integer isamax_(integer *, real *, integer *);
extern doublereal slanhs_(char *, integer *, real *, integer *, real *,
ftnlen);
extern /* Subroutine */ int slahqr_(logical *, logical *, integer *,
integer *, integer *, real *, integer *, real *, real *, integer *
, integer *, real *, integer *, integer *), slacpy_(char *,
integer *, integer *, real *, integer *, real *, integer *,
ftnlen), slaset_(char *, integer *, integer *, real *, real *,
real *, integer *, ftnlen), slarfx_(char *, integer *, integer *,
real *, real *, real *, integer *, real *, ftnlen);
static real smlnum;
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 */
/* ======= */
/* SHSEQR computes the eigenvalues of a real upper Hessenberg matrix H */
/* and, optionally, the matrices T and Z from the Schur decomposition */
/* H = Z T Z**T, where T is an upper quasi-triangular matrix (the Schur */
/* form), and Z is the orthogonal matrix of Schur vectors. */
/* Optionally Z may be postmultiplied into an input orthogonal matrix Q, */
/* so that this routine can give the Schur factorization of a matrix A */
/* which has been reduced to the Hessenberg form H by the orthogonal */
/* matrix Q: A = Q*H*Q**T = (QZ)*T*(QZ)**T. */
/* Arguments */
/* ========= */
/* JOB (input) CHARACTER*1 */
/* = 'E': compute eigenvalues only; */
/* = 'S': compute eigenvalues and the Schur form T. */
/* COMPZ (input) CHARACTER*1 */
/* = 'N': no Schur vectors are computed; */
/* = 'I': Z is initialized to the unit matrix and the matrix Z */
/* of Schur vectors of H is returned; */
/* = 'V': Z must contain an orthogonal matrix Q on entry, and */
/* the product Q*Z is returned. */
/* N (input) INTEGER */
/* The order of the matrix H. N >= 0. */
/* ILO (input) INTEGER */
/* IHI (input) INTEGER */
/* It is assumed that H is already upper triangular in rows */
/* and columns 1:ILO-1 and IHI+1:N. ILO and IHI are normally */
/* set by a previous call to SGEBAL, and then passed to SGEHRD */
/* when the matrix output by SGEBAL is reduced to Hessenberg */
/* form. Otherwise ILO and IHI should be set to 1 and N */
/* respectively. */
/* 1 <= ILO <= IHI <= N, if N > 0; ILO=1 and IHI=0, if N=0. */
/* H (input/output) REAL array, dimension (LDH,N) */
/* On entry, the upper Hessenberg matrix H. */
/* On exit, if JOB = 'S', H contains the upper quasi-triangular */
/* matrix T from the Schur decomposition (the Schur form); */
/* 2-by-2 diagonal blocks (corresponding to complex conjugate */
/* pairs of eigenvalues) are returned in standard form, with */
/* H(i,i) = H(i+1,i+1) and H(i+1,i)*H(i,i+1) < 0. If JOB = 'E', */
/* the contents of H are unspecified on exit. */
/* LDH (input) INTEGER */
/* The leading dimension of the array H. LDH >= max(1,N). */
/* WR (output) REAL array, dimension (N) */
/* WI (output) REAL array, dimension (N) */
/* The real and imaginary parts, respectively, of the computed */
/* eigenvalues. If two eigenvalues are computed as a complex */
/* conjugate pair, they are stored in consecutive elements of */
/* WR and WI, say the i-th and (i+1)th, with WI(i) > 0 and */
/* WI(i+1) < 0. If JOB = 'S', the eigenvalues are stored in the */
/* same order as on the diagonal of the Schur form returned in */
/* H, with WR(i) = H(i,i) and, if H(i:i+1,i:i+1) is a 2-by-2 */
/* diagonal block, WI(i) = sqrt(H(i+1,i)*H(i,i+1)) and */
/* WI(i+1) = -WI(i). */
/* Z (input/output) REAL array, dimension (LDZ,N) */
/* If COMPZ = 'N': Z is not referenced. */
/* If COMPZ = 'I': on entry, Z need not be set, and on exit, Z */
/* contains the orthogonal matrix Z of the Schur vectors of H. */
/* If COMPZ = 'V': on entry Z must contain an N-by-N matrix Q, */
/* which is assumed to be equal to the unit matrix except for */
/* the submatrix Z(ILO:IHI,ILO:IHI); on exit Z contains Q*Z. */
/* Normally Q is the orthogonal matrix generated by SORGHR after */
/* the call to SGEHRD which formed the Hessenberg matrix H. */
/* LDZ (input) INTEGER */
/* The leading dimension of the array Z. */
/* LDZ >= max(1,N) if COMPZ = 'I' or 'V'; LDZ >= 1 otherwise. */
/* WORK (workspace/output) REAL array, dimension (LWORK) */
/* On exit, if INFO = 0, WORK(1) returns the optimal LWORK. */
/* LWORK (input) INTEGER */
/* The dimension of the array WORK. LWORK >= max(1,N). */
/* 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, SHSEQR failed to compute all of the */
/* eigenvalues in a total of 30*(IHI-ILO+1) iterations; */
/* elements 1:ilo-1 and i+1:n of WR and WI contain those */
/* eigenvalues which have been successfully computed. */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. Local Arrays .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Decode and test the input parameters */
/* Parameter adjustments */
h_dim1 = *ldh;
h_offset = 1 + h_dim1;
h__ -= h_offset;
--wr;
--wi;
z_dim1 = *ldz;
z_offset = 1 + z_dim1;
z__ -= z_offset;
--work;
/* Function Body */
wantt = lsame_(job, "S", (ftnlen)1, (ftnlen)1);
initz = lsame_(compz, "I", (ftnlen)1, (ftnlen)1);
wantz = initz || lsame_(compz, "V", (ftnlen)1, (ftnlen)1);
*info = 0;
work[1] = (real) max(1,*n);
lquery = *lwork == -1;
if (! lsame_(job, "E", (ftnlen)1, (ftnlen)1) && ! wantt) {
*info = -1;
} else if (! lsame_(compz, "N", (ftnlen)1, (ftnlen)1) && ! wantz) {
*info = -2;
} else if (*n < 0) {
*info = -3;
} else if (*ilo < 1 || *ilo > max(1,*n)) {
*info = -4;
} else if (*ihi < min(*ilo,*n) || *ihi > *n) {
*info = -5;
} else if (*ldh < max(1,*n)) {
*info = -7;
} else if (*ldz < 1 || wantz && *ldz < max(1,*n)) {
*info = -11;
} else if (*lwork < max(1,*n) && ! lquery) {
*info = -13;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SHSEQR", &i__1, (ftnlen)6);
return 0;
} else if (lquery) {
return 0;
}
/* Initialize Z, if necessary */
if (initz) {
slaset_("Full", n, n, &c_b9, &c_b10, &z__[z_offset], ldz, (ftnlen)4);
}
/* Store the eigenvalues isolated by SGEBAL. */
i__1 = *ilo - 1;
for (i__ = 1; i__ <= i__1; ++i__) {
wr[i__] = h__[i__ + i__ * h_dim1];
wi[i__] = 0.f;
/* L10: */
}
i__1 = *n;
for (i__ = *ihi + 1; i__ <= i__1; ++i__) {
wr[i__] = h__[i__ + i__ * h_dim1];
wi[i__] = 0.f;
/* L20: */
}
/* Quick return if possible. */
if (*n == 0) {
return 0;
}
if (*ilo == *ihi) {
wr[*ilo] = h__[*ilo + *ilo * h_dim1];
wi[*ilo] = 0.f;
return 0;
}
/* Set rows and columns ILO to IHI to zero below the first */
/* subdiagonal. */
i__1 = *ihi - 2;
for (j = *ilo; j <= i__1; ++j) {
i__2 = *n;
for (i__ = j + 2; i__ <= i__2; ++i__) {
h__[i__ + j * h_dim1] = 0.f;
/* L30: */
}
/* L40: */
}
nh = *ihi - *ilo + 1;
/* Determine the order of the multi-shift QR algorithm to be used. */
/* Writing concatenation */
i__3[0] = 1, a__1[0] = job;
i__3[1] = 1, a__1[1] = compz;
s_cat(ch__1, a__1, i__3, &c__2, (ftnlen)2);
ns = ilaenv_(&c__4, "SHSEQR", ch__1, n, ilo, ihi, &c_n1, (ftnlen)6, (
ftnlen)2);
/* Writing concatenation */
i__3[0] = 1, a__1[0] = job;
i__3[1] = 1, a__1[1] = compz;
s_cat(ch__1, a__1, i__3, &c__2, (ftnlen)2);
maxb = ilaenv_(&c__8, "SHSEQR", ch__1, n, ilo, ihi, &c_n1, (ftnlen)6, (
ftnlen)2);
if (ns <= 2 || ns > nh || maxb >= nh) {
/* Use the standard double-shift algorithm */
slahqr_(&wantt, &wantz, n, ilo, ihi, &h__[h_offset], ldh, &wr[1], &wi[
1], ilo, ihi, &z__[z_offset], ldz, info);
return 0;
}
maxb = max(3,maxb);
/* Computing MIN */
i__1 = min(ns,maxb);
ns = min(i__1,15);
/* Now 2 < NS <= MAXB < NH. */
/* Set machine-dependent constants for the stopping criterion. */
/* If norm(H) <= sqrt(OVFL), overflow should not occur. */
unfl = slamch_("Safe minimum", (ftnlen)12);
ovfl = 1.f / unfl;
slabad_(&unfl, &ovfl);
ulp = slamch_("Precision", (ftnlen)9);
smlnum = unfl * (nh / ulp);
/* I1 and I2 are the indices of the first row and last column of H */
/* to which transformations must be applied. If eigenvalues only are */
/* being computed, I1 and I2 are set inside the main loop. */
if (wantt) {
i1 = 1;
i2 = *n;
}
/* ITN is the total number of multiple-shift QR iterations allowed. */
itn = nh * 30;
/* The main loop begins here. I is the loop index and decreases from */
/* IHI to ILO in steps of at most MAXB. Each iteration of the loop */
/* works with the active submatrix in rows and columns L to I. */
/* Eigenvalues I+1 to IHI have already converged. Either L = ILO or */
/* H(L,L-1) is negligible so that the matrix splits. */
i__ = *ihi;
L50:
l = *ilo;
if (i__ < *ilo) {
goto L170;
}
/* Perform multiple-shift QR iterations on rows and columns ILO to I */
/* until a submatrix of order at most MAXB splits off at the bottom */
/* because a subdiagonal element has become negligible. */
i__1 = itn;
for (its = 0; its <= i__1; ++its) {
/* Look for a single small subdiagonal element. */
i__2 = l + 1;
for (k = i__; k >= i__2; --k) {
tst1 = (r__1 = h__[k - 1 + (k - 1) * h_dim1], dabs(r__1)) + (r__2
= h__[k + k * h_dim1], dabs(r__2));
if (tst1 == 0.f) {
i__4 = i__ - l + 1;
tst1 = slanhs_("1", &i__4, &h__[l + l * h_dim1], ldh, &work[1]
, (ftnlen)1);
}
/* Computing MAX */
r__2 = ulp * tst1;
if ((r__1 = h__[k + (k - 1) * h_dim1], dabs(r__1)) <= dmax(r__2,
smlnum)) {
goto L70;
}
/* L60: */
}
L70:
l = k;
if (l > *ilo) {
/* H(L,L-1) is negligible. */
h__[l + (l - 1) * h_dim1] = 0.f;
}
/* Exit from loop if a submatrix of order <= MAXB has split off. */
if (l >= i__ - maxb + 1) {
goto L160;
}
/* Now the active submatrix is in rows and columns L to I. If */
/* eigenvalues only are being computed, only the active submatrix */
/* need be transformed. */
if (! wantt) {
i1 = l;
i2 = i__;
}
if (its == 20 || its == 30) {
/* Exceptional shifts. */
i__2 = i__;
for (ii = i__ - ns + 1; ii <= i__2; ++ii) {
wr[ii] = ((r__1 = h__[ii + (ii - 1) * h_dim1], dabs(r__1)) + (
r__2 = h__[ii + ii * h_dim1], dabs(r__2))) * 1.5f;
wi[ii] = 0.f;
/* L80: */
}
} else {
/* Use eigenvalues of trailing submatrix of order NS as shifts. */
slacpy_("Full", &ns, &ns, &h__[i__ - ns + 1 + (i__ - ns + 1) *
h_dim1], ldh, s, &c__15, (ftnlen)4);
slahqr_(&c_false, &c_false, &ns, &c__1, &ns, s, &c__15, &wr[i__ -
ns + 1], &wi[i__ - ns + 1], &c__1, &ns, &z__[z_offset],
ldz, &ierr);
if (ierr > 0) {
/* If SLAHQR failed to compute all NS eigenvalues, use the */
/* unconverged diagonal elements as the remaining shifts. */
i__2 = ierr;
for (ii = 1; ii <= i__2; ++ii) {
wr[i__ - ns + ii] = s[ii + ii * 15 - 16];
wi[i__ - ns + ii] = 0.f;
/* L90: */
}
}
}
/* Form the first column of (G-w(1)) (G-w(2)) . . . (G-w(ns)) */
/* where G is the Hessenberg submatrix H(L:I,L:I) and w is */
/* the vector of shifts (stored in WR and WI). The result is */
/* stored in the local array V. */
v[0] = 1.f;
i__2 = ns + 1;
for (ii = 2; ii <= i__2; ++ii) {
v[ii - 1] = 0.f;
/* L100: */
}
nv = 1;
i__2 = i__;
for (j = i__ - ns + 1; j <= i__2; ++j) {
if (wi[j] >= 0.f) {
if (wi[j] == 0.f) {
/* real shift */
i__4 = nv + 1;
scopy_(&i__4, v, &c__1, vv, &c__1);
i__4 = nv + 1;
r__1 = -wr[j];
sgemv_("No transpose", &i__4, &nv, &c_b10, &h__[l + l *
h_dim1], ldh, vv, &c__1, &r__1, v, &c__1, (ftnlen)
12);
++nv;
} else if (wi[j] > 0.f) {
/* complex conjugate pair of shifts */
i__4 = nv + 1;
scopy_(&i__4, v, &c__1, vv, &c__1);
i__4 = nv + 1;
r__1 = wr[j] * -2.f;
sgemv_("No transpose", &i__4, &nv, &c_b10, &h__[l + l *
h_dim1], ldh, v, &c__1, &r__1, vv, &c__1, (ftnlen)
12);
i__4 = nv + 1;
itemp = isamax_(&i__4, vv, &c__1);
/* Computing MAX */
r__2 = (r__1 = vv[itemp - 1], dabs(r__1));
temp = 1.f / dmax(r__2,smlnum);
i__4 = nv + 1;
sscal_(&i__4, &temp, vv, &c__1);
absw = slapy2_(&wr[j], &wi[j]);
temp = temp * absw * absw;
i__4 = nv + 2;
i__5 = nv + 1;
sgemv_("No transpose", &i__4, &i__5, &c_b10, &h__[l + l *
h_dim1], ldh, vv, &c__1, &temp, v, &c__1, (ftnlen)
12);
nv += 2;
}
/* Scale V(1:NV) so that max(abs(V(i))) = 1. If V is zero, */
/* reset it to the unit vector. */
itemp = isamax_(&nv, v, &c__1);
temp = (r__1 = v[itemp - 1], dabs(r__1));
if (temp == 0.f) {
v[0] = 1.f;
i__4 = nv;
for (ii = 2; ii <= i__4; ++ii) {
v[ii - 1] = 0.f;
/* L110: */
}
} else {
temp = dmax(temp,smlnum);
r__1 = 1.f / temp;
sscal_(&nv, &r__1, v, &c__1);
}
}
/* L120: */
}
/* Multiple-shift QR step */
i__2 = i__ - 1;
for (k = l; k <= i__2; ++k) {
/* The first iteration of this loop determines a reflection G */
/* from the vector V and applies it from left and right to H, */
/* thus creating a nonzero bulge below the subdiagonal. */
/* Each subsequent iteration determines a reflection G to */
/* restore the Hessenberg form in the (K-1)th column, and thus */
/* chases the bulge one step toward the bottom of the active */
/* submatrix. NR is the order of G. */
/* Computing MIN */
i__4 = ns + 1, i__5 = i__ - k + 1;
nr = min(i__4,i__5);
if (k > l) {
scopy_(&nr, &h__[k + (k - 1) * h_dim1], &c__1, v, &c__1);
}
slarfg_(&nr, v, &v[1], &c__1, &tau);
if (k > l) {
h__[k + (k - 1) * h_dim1] = v[0];
i__4 = i__;
for (ii = k + 1; ii <= i__4; ++ii) {
h__[ii + (k - 1) * h_dim1] = 0.f;
/* L130: */
}
}
v[0] = 1.f;
/* Apply G from the left to transform the rows of the matrix in */
/* columns K to I2. */
i__4 = i2 - k + 1;
slarfx_("Left", &nr, &i__4, v, &tau, &h__[k + k * h_dim1], ldh, &
work[1], (ftnlen)4);
/* Apply G from the right to transform the columns of the */
/* matrix in rows I1 to min(K+NR,I). */
/* Computing MIN */
i__5 = k + nr;
i__4 = min(i__5,i__) - i1 + 1;
slarfx_("Right", &i__4, &nr, v, &tau, &h__[i1 + k * h_dim1], ldh,
&work[1], (ftnlen)5);
if (wantz) {
/* Accumulate transformations in the matrix Z */
slarfx_("Right", &nh, &nr, v, &tau, &z__[*ilo + k * z_dim1],
ldz, &work[1], (ftnlen)5);
}
/* L140: */
}
/* L150: */
}
/* Failure to converge in remaining number of iterations */
*info = i__;
return 0;
L160:
/* A submatrix of order <= MAXB in rows and columns L to I has split */
/* off. Use the double-shift QR algorithm to handle it. */
slahqr_(&wantt, &wantz, n, &l, &i__, &h__[h_offset], ldh, &wr[1], &wi[1],
ilo, ihi, &z__[z_offset], ldz, info);
if (*info > 0) {
return 0;
}
/* Decrement number of remaining iterations, and return to start of */
/* the main loop with a new value of I. */
itn -= its;
i__ = l - 1;
goto L50;
L170:
work[1] = (real) max(1,*n);
return 0;
/* End of SHSEQR */
} /* shseqr_ */
/* Subroutine */ int slatrd_(char *uplo, int *n, int *nb, real *a,
int *lda, real *e, real *tau, real *w, int *ldw)
{
/* -- LAPACK auxiliary routine (version 2.0) --
Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd.,
Courant Institute, Argonne National Lab, and Rice University
October 31, 1992
Purpose
=======
SLATRD reduces NB rows and columns of a real symmetric matrix A to
symmetric tridiagonal form by an orthogonal similarity
transformation Q' * A * Q, and returns the matrices V and W which are
needed to apply the transformation to the unreduced part of A.
If UPLO = 'U', SLATRD reduces the last NB rows and columns of a
matrix, of which the upper triangle is supplied;
if UPLO = 'L', SLATRD reduces the first NB rows and columns of a
matrix, of which the lower triangle is supplied.
This is an auxiliary routine called by SSYTRD.
Arguments
=========
UPLO (input) CHARACTER
Specifies whether the upper or lower triangular part of the
symmetric matrix A is stored:
= 'U': Upper triangular
= 'L': Lower triangular
N (input) INTEGER
The order of the matrix A.
NB (input) INTEGER
The number of rows and columns to be reduced.
A (input/output) REAL array, dimension (LDA,N)
On entry, the symmetric matrix A. If UPLO = 'U', the leading
n-by-n upper triangular part of A contains the upper
triangular part of the matrix A, and the strictly lower
triangular part of A is not referenced. If UPLO = 'L', the
leading n-by-n lower triangular part of A contains the lower
triangular part of the matrix A, and the strictly upper
triangular part of A is not referenced.
On exit:
if UPLO = 'U', the last NB columns have been reduced to
tridiagonal form, with the diagonal elements overwriting
the diagonal elements of A; the elements above the diagonal
with the array TAU, represent the orthogonal matrix Q as a
product of elementary reflectors;
if UPLO = 'L', the first NB columns have been reduced to
tridiagonal form, with the diagonal elements overwriting
the diagonal elements of A; the elements below the diagonal
with the array TAU, represent the orthogonal matrix Q as a
product of elementary reflectors.
See Further Details.
LDA (input) INTEGER
The leading dimension of the array A. LDA >= (1,N).
E (output) REAL array, dimension (N-1)
If UPLO = 'U', E(n-nb:n-1) contains the superdiagonal
elements of the last NB columns of the reduced matrix;
if UPLO = 'L', E(1:nb) contains the subdiagonal elements of
the first NB columns of the reduced matrix.
TAU (output) REAL array, dimension (N-1)
The scalar factors of the elementary reflectors, stored in
TAU(n-nb:n-1) if UPLO = 'U', and in TAU(1:nb) if UPLO = 'L'.
See Further Details.
W (output) REAL array, dimension (LDW,NB)
The n-by-nb matrix W required to update the unreduced part
of A.
LDW (input) INTEGER
The leading dimension of the array W. LDW >= max(1,N).
Further Details
===============
If UPLO = 'U', the matrix Q is represented as a product of elementary
reflectors
Q = H(n) H(n-1) . . . H(n-nb+1).
Each H(i) has the form
H(i) = I - tau * v * v'
where tau is a real scalar, and v is a real vector with
v(i:n) = 0 and v(i-1) = 1; v(1:i-1) is stored on exit in A(1:i-1,i),
and tau in TAU(i-1).
If UPLO = 'L', the matrix Q is represented as a product of elementary
reflectors
Q = H(1) H(2) . . . H(nb).
Each H(i) has the form
H(i) = I - tau * v * v'
where tau is a real scalar, and v is a real vector with
v(1:i) = 0 and v(i+1) = 1; v(i+1:n) is stored on exit in A(i+1:n,i),
and tau in TAU(i).
The elements of the vectors v together form the n-by-nb matrix V
which is needed, with W, to apply the transformation to the unreduced
part of the matrix, using a symmetric rank-2k update of the form:
A := A - V*W' - W*V'.
The contents of A on exit are illustrated by the following examples
with n = 5 and nb = 2:
if UPLO = 'U': if UPLO = 'L':
( a a a v4 v5 ) ( d )
( a a v4 v5 ) ( 1 d )
( a 1 v5 ) ( v1 1 a )
( d 1 ) ( v1 v2 a a )
( d ) ( v1 v2 a a a )
where d denotes a diagonal element of the reduced matrix, a denotes
an element of the original matrix that is unchanged, and vi denotes
an element of the vector defining H(i).
=====================================================================
Quick return if possible
Parameter adjustments
Function Body */
/* Table of constant values */
static real c_b5 = -1.f;
static real c_b6 = 1.f;
static int c__1 = 1;
static real c_b16 = 0.f;
/* System generated locals */
/* Unused variables commented out by MDG on 03-09-05
int a_dim1, a_offset, w_dim1, w_offset;
*/
int i__1, i__2, i__3;
/* Local variables */
extern doublereal sdot_(int *, real *, int *, real *, int *);
static int i;
static real alpha;
extern logical lsame_(char *, char *);
extern /* Subroutine */ int sscal_(int *, real *, real *, int *),
sgemv_(char *, int *, int *, real *, real *, int *,
real *, int *, real *, real *, int *), saxpy_(
int *, real *, real *, int *, real *, int *), ssymv_(
char *, int *, real *, real *, int *, real *, int *,
real *, real *, int *);
static int iw;
extern /* Subroutine */ int slarfg_(int *, real *, real *, int *,
real *);
#define E(I) e[(I)-1]
#define TAU(I) tau[(I)-1]
#define A(I,J) a[(I)-1 + ((J)-1)* ( *lda)]
#define W(I,J) w[(I)-1 + ((J)-1)* ( *ldw)]
if (*n <= 0) {
return 0;
}
if (lsame_(uplo, "U")) {
/* Reduce last NB columns of upper triangle */
i__1 = *n - *nb + 1;
for (i = *n; i >= *n-*nb+1; --i) {
iw = i - *n + *nb;
if (i < *n) {
/* Update A(1:i,i) */
i__2 = *n - i;
sgemv_("No transpose", &i, &i__2, &c_b5, &A(1,i+1), lda, &W(i,iw+1), ldw, &c_b6, &A(1,i), &c__1);
i__2 = *n - i;
sgemv_("No transpose", &i, &i__2, &c_b5, &W(1,iw+1), ldw, &A(i,i+1), lda, &c_b6, &A(1,i), &c__1);
}
if (i > 1) {
/* Generate elementary reflector H(i) to annihila
te
A(1:i-2,i) */
i__2 = i - 1;
slarfg_(&i__2, &A(i-1,i), &A(1,i), &
c__1, &TAU(i - 1));
E(i - 1) = A(i-1,i);
A(i-1,i) = 1.f;
/* Compute W(1:i-1,i) */
i__2 = i - 1;
ssymv_("Upper", &i__2, &c_b6, &A(1,1), lda, &A(1,i), &c__1, &c_b16, &W(1,iw), &
c__1);
if (i < *n) {
i__2 = i - 1;
i__3 = *n - i;
sgemv_("Transpose", &i__2, &i__3, &c_b6, &W(1,iw+1), ldw, &A(1,i), &c__1, &
c_b16, &W(i+1,iw), &c__1);
i__2 = i - 1;
i__3 = *n - i;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &A(1,i+1), lda, &W(i+1,iw), &c__1,
&c_b6, &W(1,iw), &c__1);
i__2 = i - 1;
i__3 = *n - i;
sgemv_("Transpose", &i__2, &i__3, &c_b6, &A(1,i+1), lda, &A(1,i), &c__1, &
c_b16, &W(i+1,iw), &c__1);
i__2 = i - 1;
i__3 = *n - i;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &W(1,iw+1), ldw, &W(i+1,iw), &c__1,
&c_b6, &W(1,iw), &c__1);
}
i__2 = i - 1;
sscal_(&i__2, &TAU(i - 1), &W(1,iw), &c__1);
i__2 = i - 1;
alpha = TAU(i - 1) * -.5f * sdot_(&i__2, &W(1,iw),
&c__1, &A(1,i), &c__1);
i__2 = i - 1;
saxpy_(&i__2, &alpha, &A(1,i), &c__1, &W(1,iw), &c__1);
}
/* L10: */
}
} else {
/* Reduce first NB columns of lower triangle */
i__1 = *nb;
for (i = 1; i <= *nb; ++i) {
/* Update A(i:n,i) */
i__2 = *n - i + 1;
i__3 = i - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &A(i,1), lda, &
W(i,1), ldw, &c_b6, &A(i,i), &c__1)
;
i__2 = *n - i + 1;
i__3 = i - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &W(i,1), ldw, &
A(i,1), lda, &c_b6, &A(i,i), &c__1)
;
if (i < *n) {
/* Generate elementary reflector H(i) to annihila
te
A(i+2:n,i) */
i__2 = *n - i;
/* Computing MIN */
i__3 = i + 2;
slarfg_(&i__2, &A(i+1,i), &A(min(i+2,*n),i), &c__1, &TAU(i));
E(i) = A(i+1,i);
A(i+1,i) = 1.f;
/* Compute W(i+1:n,i) */
i__2 = *n - i;
ssymv_("Lower", &i__2, &c_b6, &A(i+1,i+1),
lda, &A(i+1,i), &c__1, &c_b16, &W(i+1,i), &c__1);
i__2 = *n - i;
i__3 = i - 1;
sgemv_("Transpose", &i__2, &i__3, &c_b6, &W(i+1,1),
ldw, &A(i+1,i), &c__1, &c_b16, &W(1,i), &c__1);
i__2 = *n - i;
i__3 = i - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &A(i+1,1)
, lda, &W(1,i), &c__1, &c_b6, &W(i+1,i), &c__1);
i__2 = *n - i;
i__3 = i - 1;
sgemv_("Transpose", &i__2, &i__3, &c_b6, &A(i+1,1),
lda, &A(i+1,i), &c__1, &c_b16, &W(1,i), &c__1);
i__2 = *n - i;
i__3 = i - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &W(i+1,1)
, ldw, &W(1,i), &c__1, &c_b6, &W(i+1,i), &c__1);
i__2 = *n - i;
sscal_(&i__2, &TAU(i), &W(i+1,i), &c__1);
i__2 = *n - i;
alpha = TAU(i) * -.5f * sdot_(&i__2, &W(i+1,i), &
c__1, &A(i+1,i), &c__1);
i__2 = *n - i;
saxpy_(&i__2, &alpha, &A(i+1,i), &c__1, &W(i+1,i), &c__1);
}
/* L20: */
}
}
return 0;
/* End of SLATRD */
} /* slatrd_ */
/* Subroutine */ int sgerq2_(integer *m, integer *n, real *a, integer *lda,
real *tau, real *work, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, i__1, i__2;
/* Local variables */
integer i__, k;
real aii;
extern /* Subroutine */ int slarf_(char *, integer *, integer *, real *,
integer *, real *, real *, integer *, real *), xerbla_(
char *, integer *), slarfg_(integer *, real *, real *,
integer *, real *);
/* -- LAPACK routine (version 3.1) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* SGERQ2 computes an RQ factorization of a real m by n matrix A: */
/* A = R * Q. */
/* Arguments */
/* ========= */
/* M (input) INTEGER */
/* The number of rows of the matrix A. M >= 0. */
/* N (input) INTEGER */
/* The number of columns of the matrix A. N >= 0. */
/* A (input/output) REAL array, dimension (LDA,N) */
/* On entry, the m by n matrix A. */
/* On exit, if m <= n, the upper triangle of the subarray */
/* A(1:m,n-m+1:n) contains the m by m upper triangular matrix R; */
/* if m >= n, the elements on and above the (m-n)-th subdiagonal */
/* contain the m by n upper trapezoidal matrix R; the remaining */
/* elements, with the array TAU, represent the orthogonal matrix */
/* Q as a product of elementary reflectors (see Further */
/* Details). */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,M). */
/* TAU (output) REAL array, dimension (min(M,N)) */
/* The scalar factors of the elementary reflectors (see Further */
/* Details). */
/* WORK (workspace) REAL array, dimension (M) */
/* 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(m,n). */
/* Each H(i) has the form */
/* H(i) = I - tau * v * v' */
/* where tau is a real scalar, and v is a real vector with */
/* v(n-k+i+1:n) = 0 and v(n-k+i) = 1; v(1:n-k+i-1) is stored on exit in */
/* A(m-k+i,1:n-k+i-1), and tau in TAU(i). */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Test the input arguments */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
--tau;
--work;
/* Function Body */
*info = 0;
if (*m < 0) {
*info = -1;
} else if (*n < 0) {
*info = -2;
} else if (*lda < max(1,*m)) {
*info = -4;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SGERQ2", &i__1);
return 0;
}
k = min(*m,*n);
for (i__ = k; i__ >= 1; --i__) {
/* Generate elementary reflector H(i) to annihilate */
/* A(m-k+i,1:n-k+i-1) */
i__1 = *n - k + i__;
slarfg_(&i__1, &a[*m - k + i__ + (*n - k + i__) * a_dim1], &a[*m - k
+ i__ + a_dim1], lda, &tau[i__]);
/* Apply H(i) to A(1:m-k+i-1,1:n-k+i) from the right */
aii = a[*m - k + i__ + (*n - k + i__) * a_dim1];
a[*m - k + i__ + (*n - k + i__) * a_dim1] = 1.f;
i__1 = *m - k + i__ - 1;
i__2 = *n - k + i__;
slarf_("Right", &i__1, &i__2, &a[*m - k + i__ + a_dim1], lda, &tau[
i__], &a[a_offset], lda, &work[1]);
a[*m - k + i__ + (*n - k + i__) * a_dim1] = aii;
/* L10: */
}
return 0;
/* End of SGERQ2 */
} /* sgerq2_ */
/* Subroutine */ int slabrd_(integer *m, integer *n, integer *nb, real *a,
integer *lda, real *d__, real *e, real *tauq, real *taup, real *x,
integer *ldx, real *y, integer *ldy)
{
/* System generated locals */
integer a_dim1, a_offset, x_dim1, x_offset, y_dim1, y_offset, i__1, i__2,
i__3;
/* Local variables */
static integer i__;
extern /* Subroutine */ int sscal_(integer *, real *, real *, integer *),
sgemv_(char *, integer *, integer *, real *, real *, integer *,
real *, integer *, real *, real *, integer *, ftnlen), slarfg_(
integer *, real *, real *, integer *, real *);
/* -- LAPACK auxiliary routine (version 3.0) -- */
/* Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd., */
/* Courant Institute, Argonne National Lab, and Rice University */
/* February 29, 1992 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* SLABRD reduces the first NB rows and columns of a real general */
/* m by n matrix A to upper or lower bidiagonal form by an orthogonal */
/* transformation Q' * A * P, and returns the matrices X and Y which */
/* are needed to apply the transformation to the unreduced part of A. */
/* If m >= n, A is reduced to upper bidiagonal form; if m < n, to lower */
/* bidiagonal form. */
/* This is an auxiliary routine called by SGEBRD */
/* Arguments */
/* ========= */
/* M (input) INTEGER */
/* The number of rows in the matrix A. */
/* N (input) INTEGER */
/* The number of columns in the matrix A. */
/* NB (input) INTEGER */
/* The number of leading rows and columns of A to be reduced. */
/* A (input/output) REAL array, dimension (LDA,N) */
/* On entry, the m by n general matrix to be reduced. */
/* On exit, the first NB rows and columns of the matrix are */
/* overwritten; the rest of the array is unchanged. */
/* If m >= n, elements on and below the diagonal in the first NB */
/* columns, with the array TAUQ, represent the orthogonal */
/* matrix Q as a product of elementary reflectors; and */
/* elements above the diagonal in the first NB rows, with the */
/* array TAUP, represent the orthogonal matrix P as a product */
/* of elementary reflectors. */
/* If m < n, elements below the diagonal in the first NB */
/* columns, with the array TAUQ, represent the orthogonal */
/* matrix Q as a product of elementary reflectors, and */
/* elements on and above the diagonal in the first NB rows, */
/* with the array TAUP, represent the orthogonal matrix P as */
/* a product of elementary reflectors. */
/* See Further Details. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,M). */
/* D (output) REAL array, dimension (NB) */
/* The diagonal elements of the first NB rows and columns of */
/* the reduced matrix. D(i) = A(i,i). */
/* E (output) REAL array, dimension (NB) */
/* The off-diagonal elements of the first NB rows and columns of */
/* the reduced matrix. */
/* TAUQ (output) REAL array dimension (NB) */
/* The scalar factors of the elementary reflectors which */
/* represent the orthogonal matrix Q. See Further Details. */
/* TAUP (output) REAL array, dimension (NB) */
/* The scalar factors of the elementary reflectors which */
/* represent the orthogonal matrix P. See Further Details. */
/* X (output) REAL array, dimension (LDX,NB) */
/* The m-by-nb matrix X required to update the unreduced part */
/* of A. */
/* LDX (input) INTEGER */
/* The leading dimension of the array X. LDX >= M. */
/* Y (output) REAL array, dimension (LDY,NB) */
/* The n-by-nb matrix Y required to update the unreduced part */
/* of A. */
/* LDY (output) INTEGER */
/* The leading dimension of the array Y. LDY >= N. */
/* Further Details */
/* =============== */
/* The matrices Q and P are represented as products of elementary */
/* reflectors: */
/* Q = H(1) H(2) . . . H(nb) and P = G(1) G(2) . . . G(nb) */
/* Each H(i) and G(i) has the form: */
/* H(i) = I - tauq * v * v' and G(i) = I - taup * u * u' */
/* where tauq and taup are real scalars, and v and u are real vectors. */
/* If m >= n, v(1:i-1) = 0, v(i) = 1, and v(i:m) is stored on exit in */
/* A(i:m,i); u(1:i) = 0, u(i+1) = 1, and u(i+1:n) is stored on exit in */
/* A(i,i+1:n); tauq is stored in TAUQ(i) and taup in TAUP(i). */
/* If m < n, v(1:i) = 0, v(i+1) = 1, and v(i+1:m) is stored on exit in */
/* A(i+2:m,i); u(1:i-1) = 0, u(i) = 1, and u(i:n) is stored on exit in */
/* A(i,i+1:n); tauq is stored in TAUQ(i) and taup in TAUP(i). */
/* The elements of the vectors v and u together form the m-by-nb matrix */
/* V and the nb-by-n matrix U' which are needed, with X and Y, to apply */
/* the transformation to the unreduced part of the matrix, using a block */
/* update of the form: A := A - V*Y' - X*U'. */
/* The contents of A on exit are illustrated by the following examples */
/* with nb = 2: */
/* m = 6 and n = 5 (m > n): m = 5 and n = 6 (m < n): */
/* ( 1 1 u1 u1 u1 ) ( 1 u1 u1 u1 u1 u1 ) */
/* ( v1 1 1 u2 u2 ) ( 1 1 u2 u2 u2 u2 ) */
/* ( v1 v2 a a a ) ( v1 1 a a a a ) */
/* ( v1 v2 a a a ) ( v1 v2 a a a a ) */
/* ( v1 v2 a a a ) ( v1 v2 a a a a ) */
/* ( v1 v2 a a a ) */
/* where a denotes an element of the original matrix which is unchanged, */
/* vi denotes an element of the vector defining H(i), and ui an element */
/* of the vector defining G(i). */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Quick return if possible */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
--d__;
--e;
--tauq;
--taup;
x_dim1 = *ldx;
x_offset = 1 + x_dim1;
x -= x_offset;
y_dim1 = *ldy;
y_offset = 1 + y_dim1;
y -= y_offset;
/* Function Body */
if (*m <= 0 || *n <= 0) {
return 0;
}
if (*m >= *n) {
/* Reduce to upper bidiagonal form */
i__1 = *nb;
for (i__ = 1; i__ <= i__1; ++i__) {
/* Update A(i:m,i) */
i__2 = *m - i__ + 1;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &a[i__ + a_dim1], lda,
&y[i__ + y_dim1], ldy, &c_b5, &a[i__ + i__ * a_dim1], &
c__1, (ftnlen)12);
i__2 = *m - i__ + 1;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &x[i__ + x_dim1], ldx,
&a[i__ * a_dim1 + 1], &c__1, &c_b5, &a[i__ + i__ *
a_dim1], &c__1, (ftnlen)12);
/* Generate reflection Q(i) to annihilate A(i+1:m,i) */
i__2 = *m - i__ + 1;
/* Computing MIN */
i__3 = i__ + 1;
slarfg_(&i__2, &a[i__ + i__ * a_dim1], &a[min(i__3,*m) + i__ *
a_dim1], &c__1, &tauq[i__]);
d__[i__] = a[i__ + i__ * a_dim1];
if (i__ < *n) {
a[i__ + i__ * a_dim1] = 1.f;
/* Compute Y(i+1:n,i) */
i__2 = *m - i__ + 1;
i__3 = *n - i__;
sgemv_("Transpose", &i__2, &i__3, &c_b5, &a[i__ + (i__ + 1) *
a_dim1], lda, &a[i__ + i__ * a_dim1], &c__1, &c_b16, &
y[i__ + 1 + i__ * y_dim1], &c__1, (ftnlen)9);
i__2 = *m - i__ + 1;
i__3 = i__ - 1;
sgemv_("Transpose", &i__2, &i__3, &c_b5, &a[i__ + a_dim1],
lda, &a[i__ + i__ * a_dim1], &c__1, &c_b16, &y[i__ *
y_dim1 + 1], &c__1, (ftnlen)9);
i__2 = *n - i__;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &y[i__ + 1 +
y_dim1], ldy, &y[i__ * y_dim1 + 1], &c__1, &c_b5, &y[
i__ + 1 + i__ * y_dim1], &c__1, (ftnlen)12);
i__2 = *m - i__ + 1;
i__3 = i__ - 1;
sgemv_("Transpose", &i__2, &i__3, &c_b5, &x[i__ + x_dim1],
ldx, &a[i__ + i__ * a_dim1], &c__1, &c_b16, &y[i__ *
y_dim1 + 1], &c__1, (ftnlen)9);
i__2 = i__ - 1;
i__3 = *n - i__;
sgemv_("Transpose", &i__2, &i__3, &c_b4, &a[(i__ + 1) *
a_dim1 + 1], lda, &y[i__ * y_dim1 + 1], &c__1, &c_b5,
&y[i__ + 1 + i__ * y_dim1], &c__1, (ftnlen)9);
i__2 = *n - i__;
sscal_(&i__2, &tauq[i__], &y[i__ + 1 + i__ * y_dim1], &c__1);
/* Update A(i,i+1:n) */
i__2 = *n - i__;
sgemv_("No transpose", &i__2, &i__, &c_b4, &y[i__ + 1 +
y_dim1], ldy, &a[i__ + a_dim1], lda, &c_b5, &a[i__ + (
i__ + 1) * a_dim1], lda, (ftnlen)12);
i__2 = i__ - 1;
i__3 = *n - i__;
sgemv_("Transpose", &i__2, &i__3, &c_b4, &a[(i__ + 1) *
a_dim1 + 1], lda, &x[i__ + x_dim1], ldx, &c_b5, &a[
i__ + (i__ + 1) * a_dim1], lda, (ftnlen)9);
/* Generate reflection P(i) to annihilate A(i,i+2:n) */
i__2 = *n - i__;
/* Computing MIN */
i__3 = i__ + 2;
slarfg_(&i__2, &a[i__ + (i__ + 1) * a_dim1], &a[i__ + min(
i__3,*n) * a_dim1], lda, &taup[i__]);
e[i__] = a[i__ + (i__ + 1) * a_dim1];
a[i__ + (i__ + 1) * a_dim1] = 1.f;
/* Compute X(i+1:m,i) */
i__2 = *m - i__;
i__3 = *n - i__;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &a[i__ + 1 + (i__
+ 1) * a_dim1], lda, &a[i__ + (i__ + 1) * a_dim1],
lda, &c_b16, &x[i__ + 1 + i__ * x_dim1], &c__1, (
ftnlen)12);
i__2 = *n - i__;
sgemv_("Transpose", &i__2, &i__, &c_b5, &y[i__ + 1 + y_dim1],
ldy, &a[i__ + (i__ + 1) * a_dim1], lda, &c_b16, &x[
i__ * x_dim1 + 1], &c__1, (ftnlen)9);
i__2 = *m - i__;
sgemv_("No transpose", &i__2, &i__, &c_b4, &a[i__ + 1 +
a_dim1], lda, &x[i__ * x_dim1 + 1], &c__1, &c_b5, &x[
i__ + 1 + i__ * x_dim1], &c__1, (ftnlen)12);
i__2 = i__ - 1;
i__3 = *n - i__;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &a[(i__ + 1) *
a_dim1 + 1], lda, &a[i__ + (i__ + 1) * a_dim1], lda, &
c_b16, &x[i__ * x_dim1 + 1], &c__1, (ftnlen)12);
i__2 = *m - i__;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &x[i__ + 1 +
x_dim1], ldx, &x[i__ * x_dim1 + 1], &c__1, &c_b5, &x[
i__ + 1 + i__ * x_dim1], &c__1, (ftnlen)12);
i__2 = *m - i__;
sscal_(&i__2, &taup[i__], &x[i__ + 1 + i__ * x_dim1], &c__1);
}
/* L10: */
}
} else {
/* Reduce to lower bidiagonal form */
i__1 = *nb;
for (i__ = 1; i__ <= i__1; ++i__) {
/* Update A(i,i:n) */
i__2 = *n - i__ + 1;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &y[i__ + y_dim1], ldy,
&a[i__ + a_dim1], lda, &c_b5, &a[i__ + i__ * a_dim1],
lda, (ftnlen)12);
i__2 = i__ - 1;
i__3 = *n - i__ + 1;
sgemv_("Transpose", &i__2, &i__3, &c_b4, &a[i__ * a_dim1 + 1],
lda, &x[i__ + x_dim1], ldx, &c_b5, &a[i__ + i__ * a_dim1],
lda, (ftnlen)9);
/* Generate reflection P(i) to annihilate A(i,i+1:n) */
i__2 = *n - i__ + 1;
/* Computing MIN */
i__3 = i__ + 1;
slarfg_(&i__2, &a[i__ + i__ * a_dim1], &a[i__ + min(i__3,*n) *
a_dim1], lda, &taup[i__]);
d__[i__] = a[i__ + i__ * a_dim1];
if (i__ < *m) {
a[i__ + i__ * a_dim1] = 1.f;
/* Compute X(i+1:m,i) */
i__2 = *m - i__;
i__3 = *n - i__ + 1;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &a[i__ + 1 + i__ *
a_dim1], lda, &a[i__ + i__ * a_dim1], lda, &c_b16, &
x[i__ + 1 + i__ * x_dim1], &c__1, (ftnlen)12);
i__2 = *n - i__ + 1;
i__3 = i__ - 1;
sgemv_("Transpose", &i__2, &i__3, &c_b5, &y[i__ + y_dim1],
ldy, &a[i__ + i__ * a_dim1], lda, &c_b16, &x[i__ *
x_dim1 + 1], &c__1, (ftnlen)9);
i__2 = *m - i__;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &a[i__ + 1 +
a_dim1], lda, &x[i__ * x_dim1 + 1], &c__1, &c_b5, &x[
i__ + 1 + i__ * x_dim1], &c__1, (ftnlen)12);
i__2 = i__ - 1;
i__3 = *n - i__ + 1;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &a[i__ * a_dim1 +
1], lda, &a[i__ + i__ * a_dim1], lda, &c_b16, &x[i__ *
x_dim1 + 1], &c__1, (ftnlen)12);
i__2 = *m - i__;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &x[i__ + 1 +
x_dim1], ldx, &x[i__ * x_dim1 + 1], &c__1, &c_b5, &x[
i__ + 1 + i__ * x_dim1], &c__1, (ftnlen)12);
i__2 = *m - i__;
sscal_(&i__2, &taup[i__], &x[i__ + 1 + i__ * x_dim1], &c__1);
/* Update A(i+1:m,i) */
i__2 = *m - i__;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &a[i__ + 1 +
a_dim1], lda, &y[i__ + y_dim1], ldy, &c_b5, &a[i__ +
1 + i__ * a_dim1], &c__1, (ftnlen)12);
i__2 = *m - i__;
sgemv_("No transpose", &i__2, &i__, &c_b4, &x[i__ + 1 +
x_dim1], ldx, &a[i__ * a_dim1 + 1], &c__1, &c_b5, &a[
i__ + 1 + i__ * a_dim1], &c__1, (ftnlen)12);
/* Generate reflection Q(i) to annihilate A(i+2:m,i) */
i__2 = *m - i__;
/* Computing MIN */
i__3 = i__ + 2;
slarfg_(&i__2, &a[i__ + 1 + i__ * a_dim1], &a[min(i__3,*m) +
i__ * a_dim1], &c__1, &tauq[i__]);
e[i__] = a[i__ + 1 + i__ * a_dim1];
a[i__ + 1 + i__ * a_dim1] = 1.f;
/* Compute Y(i+1:n,i) */
i__2 = *m - i__;
i__3 = *n - i__;
sgemv_("Transpose", &i__2, &i__3, &c_b5, &a[i__ + 1 + (i__ +
1) * a_dim1], lda, &a[i__ + 1 + i__ * a_dim1], &c__1,
&c_b16, &y[i__ + 1 + i__ * y_dim1], &c__1, (ftnlen)9);
i__2 = *m - i__;
i__3 = i__ - 1;
sgemv_("Transpose", &i__2, &i__3, &c_b5, &a[i__ + 1 + a_dim1],
lda, &a[i__ + 1 + i__ * a_dim1], &c__1, &c_b16, &y[
i__ * y_dim1 + 1], &c__1, (ftnlen)9);
i__2 = *n - i__;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b4, &y[i__ + 1 +
y_dim1], ldy, &y[i__ * y_dim1 + 1], &c__1, &c_b5, &y[
i__ + 1 + i__ * y_dim1], &c__1, (ftnlen)12);
i__2 = *m - i__;
sgemv_("Transpose", &i__2, &i__, &c_b5, &x[i__ + 1 + x_dim1],
ldx, &a[i__ + 1 + i__ * a_dim1], &c__1, &c_b16, &y[
i__ * y_dim1 + 1], &c__1, (ftnlen)9);
i__2 = *n - i__;
sgemv_("Transpose", &i__, &i__2, &c_b4, &a[(i__ + 1) * a_dim1
+ 1], lda, &y[i__ * y_dim1 + 1], &c__1, &c_b5, &y[i__
+ 1 + i__ * y_dim1], &c__1, (ftnlen)9);
i__2 = *n - i__;
sscal_(&i__2, &tauq[i__], &y[i__ + 1 + i__ * y_dim1], &c__1);
}
/* L20: */
}
}
return 0;
/* End of SLABRD */
} /* slabrd_ */
/* Subroutine */ int sgebd2_(integer *m, integer *n, real *a, integer *lda,
real *d__, real *e, real *tauq, real *taup, real *work, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, i__1, i__2, i__3;
/* Local variables */
integer i__;
extern /* Subroutine */ int slarf_(char *, integer *, integer *, real *,
integer *, real *, real *, integer *, real *), xerbla_(
char *, integer *), slarfg_(integer *, real *, real *,
integer *, real *);
/* -- LAPACK routine (version 3.2) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* SGEBD2 reduces a real general m by n matrix A to upper or lower */
/* bidiagonal form B by an orthogonal transformation: Q' * A * P = B. */
/* If m >= n, B is upper bidiagonal; if m < n, B is lower bidiagonal. */
/* Arguments */
/* ========= */
/* M (input) INTEGER */
/* The number of rows in the matrix A. M >= 0. */
/* N (input) INTEGER */
/* The number of columns in the matrix A. N >= 0. */
/* A (input/output) REAL array, dimension (LDA,N) */
/* On entry, the m by n general matrix to be reduced. */
/* On exit, */
/* if m >= n, the diagonal and the first superdiagonal are */
/* overwritten with the upper bidiagonal matrix B; the */
/* elements below the diagonal, with the array TAUQ, represent */
/* the orthogonal matrix Q as a product of elementary */
/* reflectors, and the elements above the first superdiagonal, */
/* with the array TAUP, represent the orthogonal matrix P as */
/* a product of elementary reflectors; */
/* if m < n, the diagonal and the first subdiagonal are */
/* overwritten with the lower bidiagonal matrix B; the */
/* elements below the first subdiagonal, with the array TAUQ, */
/* represent the orthogonal matrix Q as a product of */
/* elementary reflectors, and the elements above the diagonal, */
/* with the array TAUP, represent the orthogonal matrix P as */
/* a product of elementary reflectors. */
/* See Further Details. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,M). */
/* D (output) REAL array, dimension (min(M,N)) */
/* The diagonal elements of the bidiagonal matrix B: */
/* D(i) = A(i,i). */
/* E (output) REAL array, dimension (min(M,N)-1) */
/* The off-diagonal elements of the bidiagonal matrix B: */
/* if m >= n, E(i) = A(i,i+1) for i = 1,2,...,n-1; */
/* if m < n, E(i) = A(i+1,i) for i = 1,2,...,m-1. */
/* TAUQ (output) REAL array dimension (min(M,N)) */
/* The scalar factors of the elementary reflectors which */
/* represent the orthogonal matrix Q. See Further Details. */
/* TAUP (output) REAL array, dimension (min(M,N)) */
/* The scalar factors of the elementary reflectors which */
/* represent the orthogonal matrix P. See Further Details. */
/* WORK (workspace) REAL array, dimension (max(M,N)) */
/* INFO (output) INTEGER */
/* = 0: successful exit. */
/* < 0: if INFO = -i, the i-th argument had an illegal value. */
/* Further Details */
/* =============== */
/* The matrices Q and P are represented as products of elementary */
/* reflectors: */
/* If m >= n, */
/* Q = H(1) H(2) . . . H(n) and P = G(1) G(2) . . . G(n-1) */
/* Each H(i) and G(i) has the form: */
/* H(i) = I - tauq * v * v' and G(i) = I - taup * u * u' */
/* where tauq and taup are real scalars, and v and u are real vectors; */
/* v(1:i-1) = 0, v(i) = 1, and v(i+1:m) is stored on exit in A(i+1:m,i); */
/* u(1:i) = 0, u(i+1) = 1, and u(i+2:n) is stored on exit in A(i,i+2:n); */
/* tauq is stored in TAUQ(i) and taup in TAUP(i). */
/* If m < n, */
/* Q = H(1) H(2) . . . H(m-1) and P = G(1) G(2) . . . G(m) */
/* Each H(i) and G(i) has the form: */
/* H(i) = I - tauq * v * v' and G(i) = I - taup * u * u' */
/* where tauq and taup are real scalars, and v and u are real vectors; */
/* v(1:i) = 0, v(i+1) = 1, and v(i+2:m) is stored on exit in A(i+2:m,i); */
/* u(1:i-1) = 0, u(i) = 1, and u(i+1:n) is stored on exit in A(i,i+1:n); */
/* tauq is stored in TAUQ(i) and taup in TAUP(i). */
/* The contents of A on exit are illustrated by the following examples: */
/* m = 6 and n = 5 (m > n): m = 5 and n = 6 (m < n): */
/* ( d e u1 u1 u1 ) ( d u1 u1 u1 u1 u1 ) */
/* ( v1 d e u2 u2 ) ( e d u2 u2 u2 u2 ) */
/* ( v1 v2 d e u3 ) ( v1 e d u3 u3 u3 ) */
/* ( v1 v2 v3 d e ) ( v1 v2 e d u4 u4 ) */
/* ( v1 v2 v3 v4 d ) ( v1 v2 v3 e d u5 ) */
/* ( v1 v2 v3 v4 v5 ) */
/* where d and e denote diagonal and off-diagonal elements of B, vi */
/* denotes an element of the vector defining H(i), and ui an element of */
/* the vector defining G(i). */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Test the input parameters */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
--d__;
--e;
--tauq;
--taup;
--work;
/* Function Body */
*info = 0;
if (*m < 0) {
*info = -1;
} else if (*n < 0) {
*info = -2;
} else if (*lda < max(1,*m)) {
*info = -4;
}
if (*info < 0) {
i__1 = -(*info);
xerbla_("SGEBD2", &i__1);
return 0;
}
if (*m >= *n) {
/* Reduce to upper bidiagonal form */
i__1 = *n;
for (i__ = 1; i__ <= i__1; ++i__) {
/* Generate elementary reflector H(i) to annihilate A(i+1:m,i) */
i__2 = *m - i__ + 1;
/* Computing MIN */
i__3 = i__ + 1;
slarfg_(&i__2, &a[i__ + i__ * a_dim1], &a[min(i__3, *m)+ i__ *
a_dim1], &c__1, &tauq[i__]);
d__[i__] = a[i__ + i__ * a_dim1];
a[i__ + i__ * a_dim1] = 1.f;
/* Apply H(i) to A(i:m,i+1:n) from the left */
if (i__ < *n) {
i__2 = *m - i__ + 1;
i__3 = *n - i__;
slarf_("Left", &i__2, &i__3, &a[i__ + i__ * a_dim1], &c__1, &
tauq[i__], &a[i__ + (i__ + 1) * a_dim1], lda, &work[1]
);
}
a[i__ + i__ * a_dim1] = d__[i__];
if (i__ < *n) {
/* Generate elementary reflector G(i) to annihilate */
/* A(i,i+2:n) */
i__2 = *n - i__;
/* Computing MIN */
i__3 = i__ + 2;
slarfg_(&i__2, &a[i__ + (i__ + 1) * a_dim1], &a[i__ + min(
i__3, *n)* a_dim1], lda, &taup[i__]);
e[i__] = a[i__ + (i__ + 1) * a_dim1];
a[i__ + (i__ + 1) * a_dim1] = 1.f;
/* Apply G(i) to A(i+1:m,i+1:n) from the right */
i__2 = *m - i__;
i__3 = *n - i__;
slarf_("Right", &i__2, &i__3, &a[i__ + (i__ + 1) * a_dim1],
lda, &taup[i__], &a[i__ + 1 + (i__ + 1) * a_dim1],
lda, &work[1]);
a[i__ + (i__ + 1) * a_dim1] = e[i__];
} else {
taup[i__] = 0.f;
}
/* L10: */
}
} else {
/* Reduce to lower bidiagonal form */
i__1 = *m;
for (i__ = 1; i__ <= i__1; ++i__) {
/* Generate elementary reflector G(i) to annihilate A(i,i+1:n) */
i__2 = *n - i__ + 1;
/* Computing MIN */
i__3 = i__ + 1;
slarfg_(&i__2, &a[i__ + i__ * a_dim1], &a[i__ + min(i__3, *n)*
a_dim1], lda, &taup[i__]);
d__[i__] = a[i__ + i__ * a_dim1];
a[i__ + i__ * a_dim1] = 1.f;
/* Apply G(i) to A(i+1:m,i:n) from the right */
if (i__ < *m) {
i__2 = *m - i__;
i__3 = *n - i__ + 1;
slarf_("Right", &i__2, &i__3, &a[i__ + i__ * a_dim1], lda, &
taup[i__], &a[i__ + 1 + i__ * a_dim1], lda, &work[1]);
}
a[i__ + i__ * a_dim1] = d__[i__];
if (i__ < *m) {
/* Generate elementary reflector H(i) to annihilate */
/* A(i+2:m,i) */
i__2 = *m - i__;
/* Computing MIN */
i__3 = i__ + 2;
slarfg_(&i__2, &a[i__ + 1 + i__ * a_dim1], &a[min(i__3, *m)+
i__ * a_dim1], &c__1, &tauq[i__]);
e[i__] = a[i__ + 1 + i__ * a_dim1];
a[i__ + 1 + i__ * a_dim1] = 1.f;
/* Apply H(i) to A(i+1:m,i+1:n) from the left */
i__2 = *m - i__;
i__3 = *n - i__;
slarf_("Left", &i__2, &i__3, &a[i__ + 1 + i__ * a_dim1], &
c__1, &tauq[i__], &a[i__ + 1 + (i__ + 1) * a_dim1],
lda, &work[1]);
a[i__ + 1 + i__ * a_dim1] = e[i__];
} else {
tauq[i__] = 0.f;
}
/* L20: */
}
}
return 0;
/* End of SGEBD2 */
} /* sgebd2_ */
/* Subroutine */ int slahr2_(integer *n, integer *k, integer *nb, real *a,
integer *lda, real *tau, real *t, integer *ldt, real *y, integer *ldy)
{
/* System generated locals */
integer a_dim1, a_offset, t_dim1, t_offset, y_dim1, y_offset, i__1, i__2,
i__3;
real r__1;
/* Local variables */
integer i__;
real ei;
extern /* Subroutine */ int sscal_(integer *, real *, real *, integer *),
sgemm_(char *, char *, integer *, integer *, integer *, real *,
real *, integer *, real *, integer *, real *, real *, integer *), sgemv_(char *, integer *, integer *, real *,
real *, integer *, real *, integer *, real *, real *, integer *), scopy_(integer *, real *, integer *, real *, integer *),
strmm_(char *, char *, char *, char *, integer *, integer *, real
*, real *, integer *, real *, integer *), saxpy_(integer *, real *, real *, integer *, real *,
integer *), strmv_(char *, char *, char *, integer *, real *,
integer *, real *, integer *), slarfg_(
integer *, real *, real *, integer *, real *), slacpy_(char *,
integer *, integer *, real *, integer *, real *, integer *);
/* -- LAPACK auxiliary routine (version 3.2) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* SLAHR2 reduces the first NB columns of A real general n-BY-(n-k+1) */
/* matrix A so that elements below the k-th subdiagonal are zero. The */
/* reduction is performed by an orthogonal similarity transformation */
/* Q' * A * Q. The routine returns the matrices V and T which determine */
/* Q as a block reflector I - V*T*V', and also the matrix Y = A * V * T. */
/* This is an auxiliary routine called by SGEHRD. */
/* Arguments */
/* ========= */
/* N (input) INTEGER */
/* The order of the matrix A. */
/* K (input) INTEGER */
/* The offset for the reduction. Elements below the k-th */
/* subdiagonal in the first NB columns are reduced to zero. */
/* K < N. */
/* NB (input) INTEGER */
/* The number of columns to be reduced. */
/* A (input/output) REAL array, dimension (LDA,N-K+1) */
/* On entry, the n-by-(n-k+1) general matrix A. */
/* On exit, the elements on and above the k-th subdiagonal in */
/* the first NB columns are overwritten with the corresponding */
/* elements of the reduced matrix; the elements below the k-th */
/* subdiagonal, with the array TAU, represent the matrix Q as a */
/* product of elementary reflectors. The other columns of A are */
/* unchanged. See Further Details. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,N). */
/* TAU (output) REAL array, dimension (NB) */
/* The scalar factors of the elementary reflectors. See Further */
/* Details. */
/* T (output) REAL array, dimension (LDT,NB) */
/* The upper triangular matrix T. */
/* LDT (input) INTEGER */
/* The leading dimension of the array T. LDT >= NB. */
/* Y (output) REAL array, dimension (LDY,NB) */
/* The n-by-nb matrix Y. */
/* LDY (input) INTEGER */
/* The leading dimension of the array Y. LDY >= N. */
/* Further Details */
/* =============== */
/* The matrix Q is represented as a product of nb elementary reflectors */
/* Q = H(1) H(2) . . . H(nb). */
/* Each H(i) has the form */
/* H(i) = I - tau * v * v' */
/* where tau is a real scalar, and v is a real vector with */
/* v(1:i+k-1) = 0, v(i+k) = 1; v(i+k+1:n) is stored on exit in */
/* A(i+k+1:n,i), and tau in TAU(i). */
/* The elements of the vectors v together form the (n-k+1)-by-nb matrix */
/* V which is needed, with T and Y, to apply the transformation to the */
/* unreduced part of the matrix, using an update of the form: */
/* A := (I - V*T*V') * (A - Y*V'). */
/* The contents of A on exit are illustrated by the following example */
/* with n = 7, k = 3 and nb = 2: */
/* ( a a a a a ) */
/* ( a a a a a ) */
/* ( a a a a a ) */
/* ( h h a a a ) */
/* ( v1 h a a a ) */
/* ( v1 v2 a a a ) */
/* ( v1 v2 a a a ) */
/* where a denotes an element of the original matrix A, h denotes a */
/* modified element of the upper Hessenberg matrix H, and vi denotes an */
/* element of the vector defining H(i). */
/* This file is a slight modification of LAPACK-3.0's SLAHRD */
/* incorporating improvements proposed by Quintana-Orti and Van de */
/* Gejin. Note that the entries of A(1:K,2:NB) differ from those */
/* returned by the original LAPACK routine. This function is */
/* not backward compatible with LAPACK3.0. */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Quick return if possible */
/* Parameter adjustments */
--tau;
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
t_dim1 = *ldt;
t_offset = 1 + t_dim1;
t -= t_offset;
y_dim1 = *ldy;
y_offset = 1 + y_dim1;
y -= y_offset;
/* Function Body */
if (*n <= 1) {
return 0;
}
i__1 = *nb;
for (i__ = 1; i__ <= i__1; ++i__) {
if (i__ > 1) {
/* Update A(K+1:N,I) */
/* Update I-th column of A - Y * V' */
i__2 = *n - *k;
i__3 = i__ - 1;
sgemv_("NO TRANSPOSE", &i__2, &i__3, &c_b4, &y[*k + 1 + y_dim1],
ldy, &a[*k + i__ - 1 + a_dim1], lda, &c_b5, &a[*k + 1 +
i__ * a_dim1], &c__1);
/* Apply I - V * T' * V' to this column (call it b) from the */
/* left, using the last column of T as workspace */
/* Let V = ( V1 ) and b = ( b1 ) (first I-1 rows) */
/* ( V2 ) ( b2 ) */
/* where V1 is unit lower triangular */
/* w := V1' * b1 */
i__2 = i__ - 1;
scopy_(&i__2, &a[*k + 1 + i__ * a_dim1], &c__1, &t[*nb * t_dim1 +
1], &c__1);
i__2 = i__ - 1;
strmv_("Lower", "Transpose", "UNIT", &i__2, &a[*k + 1 + a_dim1],
lda, &t[*nb * t_dim1 + 1], &c__1);
/* w := w + V2'*b2 */
i__2 = *n - *k - i__ + 1;
i__3 = i__ - 1;
sgemv_("Transpose", &i__2, &i__3, &c_b5, &a[*k + i__ + a_dim1],
lda, &a[*k + i__ + i__ * a_dim1], &c__1, &c_b5, &t[*nb *
t_dim1 + 1], &c__1);
/* w := T'*w */
i__2 = i__ - 1;
strmv_("Upper", "Transpose", "NON-UNIT", &i__2, &t[t_offset], ldt,
&t[*nb * t_dim1 + 1], &c__1);
/* b2 := b2 - V2*w */
i__2 = *n - *k - i__ + 1;
i__3 = i__ - 1;
sgemv_("NO TRANSPOSE", &i__2, &i__3, &c_b4, &a[*k + i__ + a_dim1],
lda, &t[*nb * t_dim1 + 1], &c__1, &c_b5, &a[*k + i__ +
i__ * a_dim1], &c__1);
/* b1 := b1 - V1*w */
i__2 = i__ - 1;
strmv_("Lower", "NO TRANSPOSE", "UNIT", &i__2, &a[*k + 1 + a_dim1]
, lda, &t[*nb * t_dim1 + 1], &c__1);
i__2 = i__ - 1;
saxpy_(&i__2, &c_b4, &t[*nb * t_dim1 + 1], &c__1, &a[*k + 1 + i__
* a_dim1], &c__1);
a[*k + i__ - 1 + (i__ - 1) * a_dim1] = ei;
}
/* Generate the elementary reflector H(I) to annihilate */
/* A(K+I+1:N,I) */
i__2 = *n - *k - i__ + 1;
/* Computing MIN */
i__3 = *k + i__ + 1;
slarfg_(&i__2, &a[*k + i__ + i__ * a_dim1], &a[min(i__3, *n)+ i__ *
a_dim1], &c__1, &tau[i__]);
ei = a[*k + i__ + i__ * a_dim1];
a[*k + i__ + i__ * a_dim1] = 1.f;
/* Compute Y(K+1:N,I) */
i__2 = *n - *k;
i__3 = *n - *k - i__ + 1;
sgemv_("NO TRANSPOSE", &i__2, &i__3, &c_b5, &a[*k + 1 + (i__ + 1) *
a_dim1], lda, &a[*k + i__ + i__ * a_dim1], &c__1, &c_b38, &y[*
k + 1 + i__ * y_dim1], &c__1);
i__2 = *n - *k - i__ + 1;
i__3 = i__ - 1;
sgemv_("Transpose", &i__2, &i__3, &c_b5, &a[*k + i__ + a_dim1], lda, &
a[*k + i__ + i__ * a_dim1], &c__1, &c_b38, &t[i__ * t_dim1 +
1], &c__1);
i__2 = *n - *k;
i__3 = i__ - 1;
sgemv_("NO TRANSPOSE", &i__2, &i__3, &c_b4, &y[*k + 1 + y_dim1], ldy,
&t[i__ * t_dim1 + 1], &c__1, &c_b5, &y[*k + 1 + i__ * y_dim1],
&c__1);
i__2 = *n - *k;
sscal_(&i__2, &tau[i__], &y[*k + 1 + i__ * y_dim1], &c__1);
/* Compute T(1:I,I) */
i__2 = i__ - 1;
r__1 = -tau[i__];
sscal_(&i__2, &r__1, &t[i__ * t_dim1 + 1], &c__1);
i__2 = i__ - 1;
strmv_("Upper", "No Transpose", "NON-UNIT", &i__2, &t[t_offset], ldt,
&t[i__ * t_dim1 + 1], &c__1)
;
t[i__ + i__ * t_dim1] = tau[i__];
/* L10: */
}
a[*k + *nb + *nb * a_dim1] = ei;
/* Compute Y(1:K,1:NB) */
slacpy_("ALL", k, nb, &a[(a_dim1 << 1) + 1], lda, &y[y_offset], ldy);
strmm_("RIGHT", "Lower", "NO TRANSPOSE", "UNIT", k, nb, &c_b5, &a[*k + 1
+ a_dim1], lda, &y[y_offset], ldy);
if (*n > *k + *nb) {
i__1 = *n - *k - *nb;
sgemm_("NO TRANSPOSE", "NO TRANSPOSE", k, nb, &i__1, &c_b5, &a[(*nb +
2) * a_dim1 + 1], lda, &a[*k + 1 + *nb + a_dim1], lda, &c_b5,
&y[y_offset], ldy);
}
strmm_("RIGHT", "Upper", "NO TRANSPOSE", "NON-UNIT", k, nb, &c_b5, &t[
t_offset], ldt, &y[y_offset], ldy);
return 0;
/* End of SLAHR2 */
} /* slahr2_ */
/* Subroutine */ int slatrd_(char *uplo, integer *n, integer *nb, real *a,
integer *lda, real *e, real *tau, real *w, integer *ldw)
{
/* System generated locals */
integer a_dim1, a_offset, w_dim1, w_offset, i__1, i__2, i__3;
/* Local variables */
integer i__, iw;
real alpha;
/* -- LAPACK auxiliary routine (version 3.2) -- */
/* November 2006 */
/* Purpose */
/* ======= */
/* SLATRD reduces NB rows and columns of a real symmetric matrix A to */
/* symmetric tridiagonal form by an orthogonal similarity */
/* transformation Q' * A * Q, and returns the matrices V and W which are */
/* needed to apply the transformation to the unreduced part of A. */
/* If UPLO = 'U', SLATRD reduces the last NB rows and columns of a */
/* matrix, of which the upper triangle is supplied; */
/* if UPLO = 'L', SLATRD reduces the first NB rows and columns of a */
/* matrix, of which the lower triangle is supplied. */
/* This is an auxiliary routine called by SSYTRD. */
/* Arguments */
/* ========= */
/* UPLO (input) CHARACTER*1 */
/* Specifies whether the upper or lower triangular part of the */
/* symmetric matrix A is stored: */
/* = 'U': Upper triangular */
/* = 'L': Lower triangular */
/* N (input) INTEGER */
/* The order of the matrix A. */
/* NB (input) INTEGER */
/* The number of rows and columns to be reduced. */
/* A (input/output) REAL array, dimension (LDA,N) */
/* On entry, the symmetric matrix A. If UPLO = 'U', the leading */
/* n-by-n upper triangular part of A contains the upper */
/* triangular part of the matrix A, and the strictly lower */
/* triangular part of A is not referenced. If UPLO = 'L', the */
/* leading n-by-n lower triangular part of A contains the lower */
/* triangular part of the matrix A, and the strictly upper */
/* triangular part of A is not referenced. */
/* On exit: */
/* if UPLO = 'U', the last NB columns have been reduced to */
/* tridiagonal form, with the diagonal elements overwriting */
/* the diagonal elements of A; the elements above the diagonal */
/* with the array TAU, represent the orthogonal matrix Q as a */
/* product of elementary reflectors; */
/* if UPLO = 'L', the first NB columns have been reduced to */
/* tridiagonal form, with the diagonal elements overwriting */
/* the diagonal elements of A; the elements below the diagonal */
/* with the array TAU, represent the orthogonal matrix Q as a */
/* product of elementary reflectors. */
/* See Further Details. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= (1,N). */
/* E (output) REAL array, dimension (N-1) */
/* If UPLO = 'U', E(n-nb:n-1) contains the superdiagonal */
/* elements of the last NB columns of the reduced matrix; */
/* if UPLO = 'L', E(1:nb) contains the subdiagonal elements of */
/* the first NB columns of the reduced matrix. */
/* TAU (output) REAL array, dimension (N-1) */
/* The scalar factors of the elementary reflectors, stored in */
/* TAU(n-nb:n-1) if UPLO = 'U', and in TAU(1:nb) if UPLO = 'L'. */
/* See Further Details. */
/* W (output) REAL array, dimension (LDW,NB) */
/* The n-by-nb matrix W required to update the unreduced part */
/* of A. */
/* LDW (input) INTEGER */
/* The leading dimension of the array W. LDW >= max(1,N). */
/* Further Details */
/* =============== */
/* If UPLO = 'U', the matrix Q is represented as a product of elementary */
/* reflectors */
/* Q = H(n) H(n-1) . . . H(n-nb+1). */
/* Each H(i) has the form */
/* H(i) = I - tau * v * v' */
/* where tau is a real scalar, and v is a real vector with */
/* v(i:n) = 0 and v(i-1) = 1; v(1:i-1) is stored on exit in A(1:i-1,i), */
/* and tau in TAU(i-1). */
/* If UPLO = 'L', the matrix Q is represented as a product of elementary */
/* reflectors */
/* Q = H(1) H(2) . . . H(nb). */
/* Each H(i) has the form */
/* H(i) = I - tau * v * v' */
/* where tau is a real scalar, and v is a real vector with */
/* v(1:i) = 0 and v(i+1) = 1; v(i+1:n) is stored on exit in A(i+1:n,i), */
/* and tau in TAU(i). */
/* The elements of the vectors v together form the n-by-nb matrix V */
/* which is needed, with W, to apply the transformation to the unreduced */
/* part of the matrix, using a symmetric rank-2k update of the form: */
/* A := A - V*W' - W*V'. */
/* The contents of A on exit are illustrated by the following examples */
/* with n = 5 and nb = 2: */
/* if UPLO = 'U': if UPLO = 'L': */
/* ( a a a v4 v5 ) ( d ) */
/* ( a a v4 v5 ) ( 1 d ) */
/* ( a 1 v5 ) ( v1 1 a ) */
/* ( d 1 ) ( v1 v2 a a ) */
/* ( d ) ( v1 v2 a a a ) */
/* where d denotes a diagonal element of the reduced matrix, a denotes */
/* an element of the original matrix that is unchanged, and vi denotes */
/* an element of the vector defining H(i). */
/* ===================================================================== */
/* Quick return if possible */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
--e;
--tau;
w_dim1 = *ldw;
w_offset = 1 + w_dim1;
w -= w_offset;
/* Function Body */
if (*n <= 0) {
return 0;
}
if (lsame_(uplo, "U")) {
/* Reduce last NB columns of upper triangle */
i__1 = *n - *nb + 1;
for (i__ = *n; i__ >= i__1; --i__) {
iw = i__ - *n + *nb;
if (i__ < *n) {
/* Update A(1:i,i) */
i__2 = *n - i__;
sgemv_("No transpose", &i__, &i__2, &c_b5, &a[(i__ + 1) *
a_dim1 + 1], lda, &w[i__ + (iw + 1) * w_dim1], ldw, &
c_b6, &a[i__ * a_dim1 + 1], &c__1);
i__2 = *n - i__;
sgemv_("No transpose", &i__, &i__2, &c_b5, &w[(iw + 1) *
w_dim1 + 1], ldw, &a[i__ + (i__ + 1) * a_dim1], lda, &
c_b6, &a[i__ * a_dim1 + 1], &c__1);
}
if (i__ > 1) {
/* Generate elementary reflector H(i) to annihilate */
/* A(1:i-2,i) */
i__2 = i__ - 1;
slarfg_(&i__2, &a[i__ - 1 + i__ * a_dim1], &a[i__ * a_dim1 +
1], &c__1, &tau[i__ - 1]);
e[i__ - 1] = a[i__ - 1 + i__ * a_dim1];
a[i__ - 1 + i__ * a_dim1] = 1.f;
/* Compute W(1:i-1,i) */
i__2 = i__ - 1;
ssymv_("Upper", &i__2, &c_b6, &a[a_offset], lda, &a[i__ *
a_dim1 + 1], &c__1, &c_b16, &w[iw * w_dim1 + 1], &
c__1);
if (i__ < *n) {
i__2 = i__ - 1;
i__3 = *n - i__;
sgemv_("Transpose", &i__2, &i__3, &c_b6, &w[(iw + 1) *
w_dim1 + 1], ldw, &a[i__ * a_dim1 + 1], &c__1, &
c_b16, &w[i__ + 1 + iw * w_dim1], &c__1);
i__2 = i__ - 1;
i__3 = *n - i__;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &a[(i__ + 1) *
a_dim1 + 1], lda, &w[i__ + 1 + iw * w_dim1], &
c__1, &c_b6, &w[iw * w_dim1 + 1], &c__1);
i__2 = i__ - 1;
i__3 = *n - i__;
sgemv_("Transpose", &i__2, &i__3, &c_b6, &a[(i__ + 1) *
a_dim1 + 1], lda, &a[i__ * a_dim1 + 1], &c__1, &
c_b16, &w[i__ + 1 + iw * w_dim1], &c__1);
i__2 = i__ - 1;
i__3 = *n - i__;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &w[(iw + 1) *
w_dim1 + 1], ldw, &w[i__ + 1 + iw * w_dim1], &
c__1, &c_b6, &w[iw * w_dim1 + 1], &c__1);
}
i__2 = i__ - 1;
sscal_(&i__2, &tau[i__ - 1], &w[iw * w_dim1 + 1], &c__1);
i__2 = i__ - 1;
alpha = tau[i__ - 1] * -.5f * sdot_(&i__2, &w[iw * w_dim1 + 1]
, &c__1, &a[i__ * a_dim1 + 1], &c__1);
i__2 = i__ - 1;
saxpy_(&i__2, &alpha, &a[i__ * a_dim1 + 1], &c__1, &w[iw *
w_dim1 + 1], &c__1);
}
}
} else {
/* Reduce first NB columns of lower triangle */
i__1 = *nb;
for (i__ = 1; i__ <= i__1; ++i__) {
/* Update A(i:n,i) */
i__2 = *n - i__ + 1;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &a[i__ + a_dim1], lda,
&w[i__ + w_dim1], ldw, &c_b6, &a[i__ + i__ * a_dim1], &
c__1);
i__2 = *n - i__ + 1;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &w[i__ + w_dim1], ldw,
&a[i__ + a_dim1], lda, &c_b6, &a[i__ + i__ * a_dim1], &
c__1);
if (i__ < *n) {
/* Generate elementary reflector H(i) to annihilate */
/* A(i+2:n,i) */
i__2 = *n - i__;
/* Computing MIN */
i__3 = i__ + 2;
slarfg_(&i__2, &a[i__ + 1 + i__ * a_dim1], &a[min(i__3, *n)+
i__ * a_dim1], &c__1, &tau[i__]);
e[i__] = a[i__ + 1 + i__ * a_dim1];
a[i__ + 1 + i__ * a_dim1] = 1.f;
/* Compute W(i+1:n,i) */
i__2 = *n - i__;
ssymv_("Lower", &i__2, &c_b6, &a[i__ + 1 + (i__ + 1) * a_dim1]
, lda, &a[i__ + 1 + i__ * a_dim1], &c__1, &c_b16, &w[
i__ + 1 + i__ * w_dim1], &c__1);
i__2 = *n - i__;
i__3 = i__ - 1;
sgemv_("Transpose", &i__2, &i__3, &c_b6, &w[i__ + 1 + w_dim1],
ldw, &a[i__ + 1 + i__ * a_dim1], &c__1, &c_b16, &w[
i__ * w_dim1 + 1], &c__1);
i__2 = *n - i__;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &a[i__ + 1 +
a_dim1], lda, &w[i__ * w_dim1 + 1], &c__1, &c_b6, &w[
i__ + 1 + i__ * w_dim1], &c__1);
i__2 = *n - i__;
i__3 = i__ - 1;
sgemv_("Transpose", &i__2, &i__3, &c_b6, &a[i__ + 1 + a_dim1],
lda, &a[i__ + 1 + i__ * a_dim1], &c__1, &c_b16, &w[
i__ * w_dim1 + 1], &c__1);
i__2 = *n - i__;
i__3 = i__ - 1;
sgemv_("No transpose", &i__2, &i__3, &c_b5, &w[i__ + 1 +
w_dim1], ldw, &w[i__ * w_dim1 + 1], &c__1, &c_b6, &w[
i__ + 1 + i__ * w_dim1], &c__1);
i__2 = *n - i__;
sscal_(&i__2, &tau[i__], &w[i__ + 1 + i__ * w_dim1], &c__1);
i__2 = *n - i__;
alpha = tau[i__] * -.5f * sdot_(&i__2, &w[i__ + 1 + i__ *
w_dim1], &c__1, &a[i__ + 1 + i__ * a_dim1], &c__1);
i__2 = *n - i__;
saxpy_(&i__2, &alpha, &a[i__ + 1 + i__ * a_dim1], &c__1, &w[
i__ + 1 + i__ * w_dim1], &c__1);
}
}
}
return 0;
/* End of SLATRD */
} /* slatrd_ */
/* Subroutine */ int slahqr_(logical *wantt, logical *wantz, integer *n,
integer *ilo, integer *ihi, real *h__, integer *ldh, real *wr, real *
wi, integer *iloz, integer *ihiz, real *z__, integer *ldz, integer *
info)
{
/* System generated locals */
integer h_dim1, h_offset, z_dim1, z_offset, i__1, i__2, i__3;
real r__1, r__2, r__3, r__4;
/* Local variables */
integer i__, j, k, l, m;
real s, v[3];
integer i1, i2;
real t1, t2, t3, v2, v3, aa, ab, ba, bb, h11, h12, h21, h22, cs;
integer nh;
real sn;
integer nr;
real tr;
integer nz;
real det, h21s;
integer its;
real ulp, sum, tst, rt1i, rt2i, rt1r, rt2r;
real safmin;
real safmax, rtdisc, smlnum;
/* -- LAPACK auxiliary routine (version 3.2) -- */
/* November 2006 */
/* Purpose */
/* ======= */
/* SLAHQR is an auxiliary routine called by SHSEQR to update the */
/* eigenvalues and Schur decomposition already computed by SHSEQR, by */
/* dealing with the Hessenberg submatrix in rows and columns ILO to */
/* IHI. */
/* Arguments */
/* ========= */
/* WANTT (input) LOGICAL */
/* = .TRUE. : the full Schur form T is required; */
/* = .FALSE.: only eigenvalues are required. */
/* WANTZ (input) LOGICAL */
/* = .TRUE. : the matrix of Schur vectors Z is required; */
/* = .FALSE.: Schur vectors are not required. */
/* N (input) INTEGER */
/* The order of the matrix H. N >= 0. */
/* ILO (input) INTEGER */
/* IHI (input) INTEGER */
/* It is assumed that H is already upper quasi-triangular in */
/* rows and columns IHI+1:N, and that H(ILO,ILO-1) = 0 (unless */
/* ILO = 1). SLAHQR works primarily with the Hessenberg */
/* submatrix in rows and columns ILO to IHI, but applies */
/* 1 <= ILO <= max(1,IHI); IHI <= N. */
/* H (input/output) REAL array, dimension (LDH,N) */
/* On entry, the upper Hessenberg matrix H. */
/* On exit, if INFO is zero and if WANTT is .TRUE., H is upper */
/* quasi-triangular in rows and columns ILO:IHI, with any */
/* 2-by-2 diagonal blocks in standard form. If INFO is zero */
/* and WANTT is .FALSE., the contents of H are unspecified on */
/* exit. The output state of H if INFO is nonzero is given */
/* below under the description of INFO. */
/* LDH (input) INTEGER */
/* The leading dimension of the array H. LDH >= max(1,N). */
/* WR (output) REAL array, dimension (N) */
/* WI (output) REAL array, dimension (N) */
/* The real and imaginary parts, respectively, of the computed */
/* eigenvalues ILO to IHI are stored in the corresponding */
/* elements of WR and WI. If two eigenvalues are computed as a */
/* complex conjugate pair, they are stored in consecutive */
/* elements of WR and WI, say the i-th and (i+1)th, with */
/* WI(i) > 0 and WI(i+1) < 0. If WANTT is .TRUE., the */
/* eigenvalues are stored in the same order as on the diagonal */
/* of the Schur form returned in H, with WR(i) = H(i,i), and, if */
/* H(i:i+1,i:i+1) is a 2-by-2 diagonal block, */
/* WI(i) = sqrt(H(i+1,i)*H(i,i+1)) and WI(i+1) = -WI(i). */
/* ILOZ (input) INTEGER */
/* IHIZ (input) INTEGER */
/* Specify the rows of Z to which transformations must be */
/* 1 <= ILOZ <= ILO; IHI <= IHIZ <= N. */
/* Z (input/output) REAL array, dimension (LDZ,N) */
/* If WANTZ is .TRUE., on entry Z must contain the current */
/* matrix Z of transformations accumulated by SHSEQR, and on */
/* exit Z has been updated; transformations are applied only to */
/* the submatrix Z(ILOZ:IHIZ,ILO:IHI). */
/* If WANTZ is .FALSE., Z is not referenced. */
/* LDZ (input) INTEGER */
/* The leading dimension of the array Z. LDZ >= max(1,N). */
/* INFO (output) INTEGER */
/* = 0: successful exit */
/* .GT. 0: If INFO = i, SLAHQR failed to compute all the */
/* eigenvalues ILO to IHI in a total of 30 iterations */
/* per eigenvalue; elements i+1:ihi of WR and WI */
/* contain those eigenvalues which have been */
/* successfully computed. */
/* If INFO .GT. 0 and WANTT is .FALSE., then on exit, */
/* the remaining unconverged eigenvalues are the */
/* eigenvalues of the upper Hessenberg matrix rows */
/* and columns ILO thorugh INFO of the final, output */
/* value of H. */
/* If INFO .GT. 0 and WANTT is .TRUE., then on exit */
/* (*) (initial value of H)*U = U*(final value of H) */
/* where U is an orthognal matrix. The final */
/* value of H is upper Hessenberg and triangular in */
/* rows and columns INFO+1 through IHI. */
/* If INFO .GT. 0 and WANTZ is .TRUE., then on exit */
/* (final value of Z) = (initial value of Z)*U */
/* where U is the orthogonal matrix in (*) */
/* (regardless of the value of WANTT.) */
/* Further Details */
/* =============== */
/* 02-96 Based on modifications by */
/* David Day, Sandia National Laboratory, USA */
/* 12-04 Further modifications by */
/* Ralph Byers, University of Kansas, USA */
/* This is a modified version of SLAHQR from LAPACK version 3.0. */
/* It is (1) more robust against overflow and underflow and */
/* (2) adopts the more conservative Ahues & Tisseur stopping */
/* criterion (LAWN 122, 1997). */
/* ========================================================= */
/* Parameter adjustments */
h_dim1 = *ldh;
h_offset = 1 + h_dim1;
h__ -= h_offset;
--wr;
--wi;
z_dim1 = *ldz;
z_offset = 1 + z_dim1;
z__ -= z_offset;
/* Function Body */
*info = 0;
/* Quick return if possible */
if (*n == 0) {
return 0;
}
if (*ilo == *ihi) {
wr[*ilo] = h__[*ilo + *ilo * h_dim1];
wi[*ilo] = 0.f;
return 0;
}
/* ==== clear out the trash ==== */
i__1 = *ihi - 3;
for (j = *ilo; j <= i__1; ++j) {
h__[j + 2 + j * h_dim1] = 0.f;
h__[j + 3 + j * h_dim1] = 0.f;
}
if (*ilo <= *ihi - 2) {
h__[*ihi + (*ihi - 2) * h_dim1] = 0.f;
}
nh = *ihi - *ilo + 1;
nz = *ihiz - *iloz + 1;
/* Set machine-dependent constants for the stopping criterion. */
safmin = slamch_("SAFE MINIMUM");
safmax = 1.f / safmin;
slabad_(&safmin, &safmax);
ulp = slamch_("PRECISION");
smlnum = safmin * ((real) nh / ulp);
/* I1 and I2 are the indices of the first row and last column of H */
/* to which transformations must be applied. If eigenvalues only are */
/* being computed, I1 and I2 are set inside the main loop. */
if (*wantt) {
i1 = 1;
i2 = *n;
}
/* The main loop begins here. I is the loop index and decreases from */
/* IHI to ILO in steps of 1 or 2. Each iteration of the loop works */
/* with the active submatrix in rows and columns L to I. */
/* Eigenvalues I+1 to IHI have already converged. Either L = ILO or */
/* H(L,L-1) is negligible so that the matrix splits. */
i__ = *ihi;
L20:
l = *ilo;
if (i__ < *ilo) {
goto L160;
}
/* Perform QR iterations on rows and columns ILO to I until a */
/* submatrix of order 1 or 2 splits off at the bottom because a */
/* subdiagonal element has become negligible. */
for (its = 0; its <= 30; ++its) {
/* Look for a single small subdiagonal element. */
i__1 = l + 1;
for (k = i__; k >= i__1; --k) {
if ((r__1 = h__[k + (k - 1) * h_dim1], dabs(r__1)) <= smlnum) {
goto L40;
}
tst = (r__1 = h__[k - 1 + (k - 1) * h_dim1], dabs(r__1)) + (r__2 =
h__[k + k * h_dim1], dabs(r__2));
if (tst == 0.f) {
if (k - 2 >= *ilo) {
tst += (r__1 = h__[k - 1 + (k - 2) * h_dim1], dabs(r__1));
}
if (k + 1 <= *ihi) {
tst += (r__1 = h__[k + 1 + k * h_dim1], dabs(r__1));
}
}
/* ==== The following is a conservative small subdiagonal */
/* . deflation criterion due to Ahues & Tisseur (LAWN 122, */
/* . 1997). It has better mathematical foundation and */
/* . improves accuracy in some cases. ==== */
if ((r__1 = h__[k + (k - 1) * h_dim1], dabs(r__1)) <= ulp * tst) {
/* Computing MAX */
r__3 = (r__1 = h__[k + (k - 1) * h_dim1], dabs(r__1)), r__4 =
(r__2 = h__[k - 1 + k * h_dim1], dabs(r__2));
ab = dmax(r__3,r__4);
/* Computing MIN */
r__3 = (r__1 = h__[k + (k - 1) * h_dim1], dabs(r__1)), r__4 =
(r__2 = h__[k - 1 + k * h_dim1], dabs(r__2));
ba = dmin(r__3,r__4);
/* Computing MAX */
r__3 = (r__1 = h__[k + k * h_dim1], dabs(r__1)), r__4 = (r__2
= h__[k - 1 + (k - 1) * h_dim1] - h__[k + k * h_dim1],
dabs(r__2));
aa = dmax(r__3,r__4);
/* Computing MIN */
r__3 = (r__1 = h__[k + k * h_dim1], dabs(r__1)), r__4 = (r__2
= h__[k - 1 + (k - 1) * h_dim1] - h__[k + k * h_dim1],
dabs(r__2));
bb = dmin(r__3,r__4);
s = aa + ab;
/* Computing MAX */
r__1 = smlnum, r__2 = ulp * (bb * (aa / s));
if (ba * (ab / s) <= dmax(r__1,r__2)) {
goto L40;
}
}
}
L40:
l = k;
if (l > *ilo) {
/* H(L,L-1) is negligible */
h__[l + (l - 1) * h_dim1] = 0.f;
}
/* Exit from loop if a submatrix of order 1 or 2 has split off. */
if (l >= i__ - 1) {
goto L150;
}
/* Now the active submatrix is in rows and columns L to I. If */
/* eigenvalues only are being computed, only the active submatrix */
/* need be transformed. */
if (! (*wantt)) {
i1 = l;
i2 = i__;
}
if (its == 10) {
/* Exceptional shift. */
s = (r__1 = h__[l + 1 + l * h_dim1], dabs(r__1)) + (r__2 = h__[l
+ 2 + (l + 1) * h_dim1], dabs(r__2));
h11 = s * .75f + h__[l + l * h_dim1];
h12 = s * -.4375f;
h21 = s;
h22 = h11;
} else if (its == 20) {
/* Exceptional shift. */
s = (r__1 = h__[i__ + (i__ - 1) * h_dim1], dabs(r__1)) + (r__2 =
h__[i__ - 1 + (i__ - 2) * h_dim1], dabs(r__2));
h11 = s * .75f + h__[i__ + i__ * h_dim1];
h12 = s * -.4375f;
h21 = s;
h22 = h11;
} else {
/* Prepare to use Francis' double shift */
/* (i.e. 2nd degree generalized Rayleigh quotient) */
h11 = h__[i__ - 1 + (i__ - 1) * h_dim1];
h21 = h__[i__ + (i__ - 1) * h_dim1];
h12 = h__[i__ - 1 + i__ * h_dim1];
h22 = h__[i__ + i__ * h_dim1];
}
s = dabs(h11) + dabs(h12) + dabs(h21) + dabs(h22);
if (s == 0.f) {
rt1r = 0.f;
rt1i = 0.f;
rt2r = 0.f;
rt2i = 0.f;
} else {
h11 /= s;
h21 /= s;
h12 /= s;
h22 /= s;
tr = (h11 + h22) / 2.f;
det = (h11 - tr) * (h22 - tr) - h12 * h21;
rtdisc = sqrt((dabs(det)));
if (det >= 0.f) {
/* ==== complex conjugate shifts ==== */
rt1r = tr * s;
rt2r = rt1r;
rt1i = rtdisc * s;
rt2i = -rt1i;
} else {
/* ==== real shifts (use only one of them) ==== */
rt1r = tr + rtdisc;
rt2r = tr - rtdisc;
if ((r__1 = rt1r - h22, dabs(r__1)) <= (r__2 = rt2r - h22,
dabs(r__2))) {
rt1r *= s;
rt2r = rt1r;
} else {
rt2r *= s;
rt1r = rt2r;
}
rt1i = 0.f;
rt2i = 0.f;
}
}
/* Look for two consecutive small subdiagonal elements. */
i__1 = l;
for (m = i__ - 2; m >= i__1; --m) {
/* Determine the effect of starting the double-shift QR */
/* iteration at row M, and see if this would make H(M,M-1) */
/* negligible. (The following uses scaling to avoid */
/* overflows and most underflows.) */
h21s = h__[m + 1 + m * h_dim1];
s = (r__1 = h__[m + m * h_dim1] - rt2r, dabs(r__1)) + dabs(rt2i)
+ dabs(h21s);
h21s = h__[m + 1 + m * h_dim1] / s;
v[0] = h21s * h__[m + (m + 1) * h_dim1] + (h__[m + m * h_dim1] -
rt1r) * ((h__[m + m * h_dim1] - rt2r) / s) - rt1i * (rt2i
/ s);
v[1] = h21s * (h__[m + m * h_dim1] + h__[m + 1 + (m + 1) * h_dim1]
- rt1r - rt2r);
v[2] = h21s * h__[m + 2 + (m + 1) * h_dim1];
s = dabs(v[0]) + dabs(v[1]) + dabs(v[2]);
v[0] /= s;
v[1] /= s;
v[2] /= s;
if (m == l) {
goto L60;
}
if ((r__1 = h__[m + (m - 1) * h_dim1], dabs(r__1)) * (dabs(v[1])
+ dabs(v[2])) <= ulp * dabs(v[0]) * ((r__2 = h__[m - 1 + (
m - 1) * h_dim1], dabs(r__2)) + (r__3 = h__[m + m *
h_dim1], dabs(r__3)) + (r__4 = h__[m + 1 + (m + 1) *
h_dim1], dabs(r__4)))) {
goto L60;
}
}
L60:
/* Double-shift QR step */
i__1 = i__ - 1;
for (k = m; k <= i__1; ++k) {
/* The first iteration of this loop determines a reflection G */
/* from the vector V and applies it from left and right to H, */
/* thus creating a nonzero bulge below the subdiagonal. */
/* Each subsequent iteration determines a reflection G to */
/* restore the Hessenberg form in the (K-1)th column, and thus */
/* chases the bulge one step toward the bottom of the active */
/* submatrix. NR is the order of G. */
/* Computing MIN */
i__2 = 3, i__3 = i__ - k + 1;
nr = min(i__2,i__3);
if (k > m) {
scopy_(&nr, &h__[k + (k - 1) * h_dim1], &c__1, v, &c__1);
}
slarfg_(&nr, v, &v[1], &c__1, &t1);
if (k > m) {
h__[k + (k - 1) * h_dim1] = v[0];
h__[k + 1 + (k - 1) * h_dim1] = 0.f;
if (k < i__ - 1) {
h__[k + 2 + (k - 1) * h_dim1] = 0.f;
}
} else if (m > l) {
/* ==== Use the following instead of */
/* . H( K, K-1 ) = -H( K, K-1 ) to */
/* . avoid a bug when v(2) and v(3) */
/* . underflow. ==== */
h__[k + (k - 1) * h_dim1] *= 1.f - t1;
}
v2 = v[1];
t2 = t1 * v2;
if (nr == 3) {
v3 = v[2];
t3 = t1 * v3;
/* Apply G from the left to transform the rows of the matrix */
/* in columns K to I2. */
i__2 = i2;
for (j = k; j <= i__2; ++j) {
sum = h__[k + j * h_dim1] + v2 * h__[k + 1 + j * h_dim1]
+ v3 * h__[k + 2 + j * h_dim1];
h__[k + j * h_dim1] -= sum * t1;
h__[k + 1 + j * h_dim1] -= sum * t2;
h__[k + 2 + j * h_dim1] -= sum * t3;
}
/* Apply G from the right to transform the columns of the */
/* matrix in rows I1 to min(K+3,I). */
/* Computing MIN */
i__3 = k + 3;
i__2 = min(i__3,i__);
for (j = i1; j <= i__2; ++j) {
sum = h__[j + k * h_dim1] + v2 * h__[j + (k + 1) * h_dim1]
+ v3 * h__[j + (k + 2) * h_dim1];
h__[j + k * h_dim1] -= sum * t1;
h__[j + (k + 1) * h_dim1] -= sum * t2;
h__[j + (k + 2) * h_dim1] -= sum * t3;
}
if (*wantz) {
/* Accumulate transformations in the matrix Z */
i__2 = *ihiz;
for (j = *iloz; j <= i__2; ++j) {
sum = z__[j + k * z_dim1] + v2 * z__[j + (k + 1) *
z_dim1] + v3 * z__[j + (k + 2) * z_dim1];
z__[j + k * z_dim1] -= sum * t1;
z__[j + (k + 1) * z_dim1] -= sum * t2;
z__[j + (k + 2) * z_dim1] -= sum * t3;
}
}
} else if (nr == 2) {
/* Apply G from the left to transform the rows of the matrix */
/* in columns K to I2. */
i__2 = i2;
for (j = k; j <= i__2; ++j) {
sum = h__[k + j * h_dim1] + v2 * h__[k + 1 + j * h_dim1];
h__[k + j * h_dim1] -= sum * t1;
h__[k + 1 + j * h_dim1] -= sum * t2;
}
/* Apply G from the right to transform the columns of the */
/* matrix in rows I1 to min(K+3,I). */
i__2 = i__;
for (j = i1; j <= i__2; ++j) {
sum = h__[j + k * h_dim1] + v2 * h__[j + (k + 1) * h_dim1]
;
h__[j + k * h_dim1] -= sum * t1;
h__[j + (k + 1) * h_dim1] -= sum * t2;
}
if (*wantz) {
/* Accumulate transformations in the matrix Z */
i__2 = *ihiz;
for (j = *iloz; j <= i__2; ++j) {
sum = z__[j + k * z_dim1] + v2 * z__[j + (k + 1) *
z_dim1];
z__[j + k * z_dim1] -= sum * t1;
z__[j + (k + 1) * z_dim1] -= sum * t2;
}
}
}
}
}
/* Failure to converge in remaining number of iterations */
*info = i__;
return 0;
L150:
if (l == i__) {
/* H(I,I-1) is negligible: one eigenvalue has converged. */
wr[i__] = h__[i__ + i__ * h_dim1];
wi[i__] = 0.f;
} else if (l == i__ - 1) {
/* H(I-1,I-2) is negligible: a pair of eigenvalues have converged. */
/* Transform the 2-by-2 submatrix to standard Schur form, */
/* and compute and store the eigenvalues. */
slanv2_(&h__[i__ - 1 + (i__ - 1) * h_dim1], &h__[i__ - 1 + i__ *
h_dim1], &h__[i__ + (i__ - 1) * h_dim1], &h__[i__ + i__ *
h_dim1], &wr[i__ - 1], &wi[i__ - 1], &wr[i__], &wi[i__], &cs,
&sn);
if (*wantt) {
/* Apply the transformation to the rest of H. */
if (i2 > i__) {
i__1 = i2 - i__;
srot_(&i__1, &h__[i__ - 1 + (i__ + 1) * h_dim1], ldh, &h__[
i__ + (i__ + 1) * h_dim1], ldh, &cs, &sn);
}
i__1 = i__ - i1 - 1;
srot_(&i__1, &h__[i1 + (i__ - 1) * h_dim1], &c__1, &h__[i1 + i__ *
h_dim1], &c__1, &cs, &sn);
}
if (*wantz) {
/* Apply the transformation to Z. */
srot_(&nz, &z__[*iloz + (i__ - 1) * z_dim1], &c__1, &z__[*iloz +
i__ * z_dim1], &c__1, &cs, &sn);
}
}
/* return to start of the main loop with new value of I. */
i__ = l - 1;
goto L20;
L160:
return 0;
/* End of SLAHQR */
} /* slahqr_ */
/* Subroutine */ int slaqr5_(logical *wantt, logical *wantz, integer *kacc22,
integer *n, integer *ktop, integer *kbot, integer *nshfts, real *sr,
real *si, real *h__, integer *ldh, integer *iloz, integer *ihiz, real
*z__, integer *ldz, real *v, integer *ldv, real *u, integer *ldu,
integer *nv, real *wv, integer *ldwv, integer *nh, real *wh, integer *
ldwh)
{
/* System generated locals */
integer h_dim1, h_offset, u_dim1, u_offset, v_dim1, v_offset, wh_dim1,
wh_offset, wv_dim1, wv_offset, z_dim1, z_offset, i__1, i__2, i__3,
i__4, i__5, i__6, i__7;
real r__1, r__2, r__3, r__4, r__5;
/* Local variables */
integer i__, j, k, m, i2, j2, i4, j4, k1;
real h11, h12, h21, h22;
integer m22, ns, nu;
real vt[3], scl;
integer kdu, kms;
real ulp;
integer knz, kzs;
real tst1, tst2, beta;
logical blk22, bmp22;
integer mend, jcol, jlen, jbot, mbot;
real swap;
integer jtop, jrow, mtop;
real alpha;
logical accum;
integer ndcol, incol;
extern /* Subroutine */ int sgemm_(char *, char *, integer *, integer *,
integer *, real *, real *, integer *, real *, integer *, real *,
real *, integer *);
integer krcol, nbmps;
extern /* Subroutine */ int strmm_(char *, char *, char *, char *,
integer *, integer *, real *, real *, integer *, real *, integer *
), slaqr1_(integer *, real *,
integer *, real *, real *, real *, real *, real *), slabad_(real *
, real *);
extern doublereal slamch_(char *);
real safmin;
extern /* Subroutine */ int slarfg_(integer *, real *, real *, integer *,
real *);
real safmax;
extern /* Subroutine */ int slacpy_(char *, integer *, integer *, real *,
integer *, real *, integer *), slaset_(char *, integer *,
integer *, real *, real *, real *, integer *);
real refsum;
integer mstart;
real smlnum;
/* -- LAPACK auxiliary routine (version 3.2) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* This auxiliary subroutine called by SLAQR0 performs a */
/* single small-bulge multi-shift QR sweep. */
/* WANTT (input) logical scalar */
/* WANTT = .true. if the quasi-triangular Schur factor */
/* is being computed. WANTT is set to .false. otherwise. */
/* WANTZ (input) logical scalar */
/* WANTZ = .true. if the orthogonal Schur factor is being */
/* computed. WANTZ is set to .false. otherwise. */
/* KACC22 (input) integer with value 0, 1, or 2. */
/* Specifies the computation mode of far-from-diagonal */
/* orthogonal updates. */
/* = 0: SLAQR5 does not accumulate reflections and does not */
/* use matrix-matrix multiply to update far-from-diagonal */
/* matrix entries. */
/* = 1: SLAQR5 accumulates reflections and uses matrix-matrix */
/* multiply to update the far-from-diagonal matrix entries. */
/* = 2: SLAQR5 accumulates reflections, uses matrix-matrix */
/* multiply to update the far-from-diagonal matrix entries, */
/* and takes advantage of 2-by-2 block structure during */
/* matrix multiplies. */
/* N (input) integer scalar */
/* N is the order of the Hessenberg matrix H upon which this */
/* subroutine operates. */
/* KTOP (input) integer scalar */
/* KBOT (input) integer scalar */
/* These are the first and last rows and columns of an */
/* isolated diagonal block upon which the QR sweep is to be */
/* applied. It is assumed without a check that */
/* either KTOP = 1 or H(KTOP,KTOP-1) = 0 */
/* and */
/* either KBOT = N or H(KBOT+1,KBOT) = 0. */
/* NSHFTS (input) integer scalar */
/* NSHFTS gives the number of simultaneous shifts. NSHFTS */
/* must be positive and even. */
/* SR (input/output) REAL array of size (NSHFTS) */
/* SI (input/output) REAL array of size (NSHFTS) */
/* SR contains the real parts and SI contains the imaginary */
/* parts of the NSHFTS shifts of origin that define the */
/* multi-shift QR sweep. On output SR and SI may be */
/* reordered. */
/* H (input/output) REAL array of size (LDH,N) */
/* On input H contains a Hessenberg matrix. On output a */
/* multi-shift QR sweep with shifts SR(J)+i*SI(J) is applied */
/* to the isolated diagonal block in rows and columns KTOP */
/* through KBOT. */
/* LDH (input) integer scalar */
/* LDH is the leading dimension of H just as declared in the */
/* calling procedure. LDH.GE.MAX(1,N). */
/* ILOZ (input) INTEGER */
/* IHIZ (input) INTEGER */
/* Specify the rows of Z to which transformations must be */
/* applied if WANTZ is .TRUE.. 1 .LE. ILOZ .LE. IHIZ .LE. N */
/* Z (input/output) REAL array of size (LDZ,IHI) */
/* If WANTZ = .TRUE., then the QR Sweep orthogonal */
/* similarity transformation is accumulated into */
/* Z(ILOZ:IHIZ,ILO:IHI) from the right. */
/* If WANTZ = .FALSE., then Z is unreferenced. */
/* LDZ (input) integer scalar */
/* LDA is the leading dimension of Z just as declared in */
/* the calling procedure. LDZ.GE.N. */
/* V (workspace) REAL array of size (LDV,NSHFTS/2) */
/* LDV (input) integer scalar */
/* LDV is the leading dimension of V as declared in the */
/* calling procedure. LDV.GE.3. */
/* U (workspace) REAL array of size */
/* (LDU,3*NSHFTS-3) */
/* LDU (input) integer scalar */
/* LDU is the leading dimension of U just as declared in the */
/* in the calling subroutine. LDU.GE.3*NSHFTS-3. */
/* NH (input) integer scalar */
/* NH is the number of columns in array WH available for */
/* workspace. NH.GE.1. */
/* WH (workspace) REAL array of size (LDWH,NH) */
/* LDWH (input) integer scalar */
/* Leading dimension of WH just as declared in the */
/* calling procedure. LDWH.GE.3*NSHFTS-3. */
/* NV (input) integer scalar */
/* NV is the number of rows in WV agailable for workspace. */
/* NV.GE.1. */
/* WV (workspace) REAL array of size */
/* (LDWV,3*NSHFTS-3) */
/* LDWV (input) integer scalar */
/* LDWV is the leading dimension of WV as declared in the */
/* in the calling subroutine. LDWV.GE.NV. */
/* ================================================================ */
/* Based on contributions by */
/* Karen Braman and Ralph Byers, Department of Mathematics, */
/* University of Kansas, USA */
/* ================================================================ */
/* Reference: */
/* K. Braman, R. Byers and R. Mathias, The Multi-Shift QR */
/* Algorithm Part I: Maintaining Well Focused Shifts, and */
/* Level 3 Performance, SIAM Journal of Matrix Analysis, */
/* volume 23, pages 929--947, 2002. */
/* ================================================================ */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Local Arrays .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Executable Statements .. */
/* ==== If there are no shifts, then there is nothing to do. ==== */
/* Parameter adjustments */
--sr;
--si;
h_dim1 = *ldh;
h_offset = 1 + h_dim1;
h__ -= h_offset;
z_dim1 = *ldz;
z_offset = 1 + z_dim1;
z__ -= z_offset;
v_dim1 = *ldv;
v_offset = 1 + v_dim1;
v -= v_offset;
u_dim1 = *ldu;
u_offset = 1 + u_dim1;
u -= u_offset;
wv_dim1 = *ldwv;
wv_offset = 1 + wv_dim1;
wv -= wv_offset;
wh_dim1 = *ldwh;
wh_offset = 1 + wh_dim1;
wh -= wh_offset;
/* Function Body */
if (*nshfts < 2) {
return 0;
}
/* ==== If the active block is empty or 1-by-1, then there */
/* . is nothing to do. ==== */
if (*ktop >= *kbot) {
return 0;
}
/* ==== Shuffle shifts into pairs of real shifts and pairs */
/* . of complex conjugate shifts assuming complex */
/* . conjugate shifts are already adjacent to one */
/* . another. ==== */
i__1 = *nshfts - 2;
for (i__ = 1; i__ <= i__1; i__ += 2) {
if (si[i__] != -si[i__ + 1]) {
swap = sr[i__];
sr[i__] = sr[i__ + 1];
sr[i__ + 1] = sr[i__ + 2];
sr[i__ + 2] = swap;
swap = si[i__];
si[i__] = si[i__ + 1];
si[i__ + 1] = si[i__ + 2];
si[i__ + 2] = swap;
}
/* L10: */
}
/* ==== NSHFTS is supposed to be even, but if it is odd, */
/* . then simply reduce it by one. The shuffle above */
/* . ensures that the dropped shift is real and that */
/* . the remaining shifts are paired. ==== */
ns = *nshfts - *nshfts % 2;
/* ==== Machine constants for deflation ==== */
safmin = slamch_("SAFE MINIMUM");
safmax = 1.f / safmin;
slabad_(&safmin, &safmax);
ulp = slamch_("PRECISION");
smlnum = safmin * ((real) (*n) / ulp);
/* ==== Use accumulated reflections to update far-from-diagonal */
/* . entries ? ==== */
accum = *kacc22 == 1 || *kacc22 == 2;
/* ==== If so, exploit the 2-by-2 block structure? ==== */
blk22 = ns > 2 && *kacc22 == 2;
/* ==== clear trash ==== */
if (*ktop + 2 <= *kbot) {
h__[*ktop + 2 + *ktop * h_dim1] = 0.f;
}
/* ==== NBMPS = number of 2-shift bulges in the chain ==== */
nbmps = ns / 2;
/* ==== KDU = width of slab ==== */
kdu = nbmps * 6 - 3;
/* ==== Create and chase chains of NBMPS bulges ==== */
i__1 = *kbot - 2;
i__2 = nbmps * 3 - 2;
for (incol = (1 - nbmps) * 3 + *ktop - 1; i__2 < 0 ? incol >= i__1 :
incol <= i__1; incol += i__2) {
ndcol = incol + kdu;
if (accum) {
slaset_("ALL", &kdu, &kdu, &c_b7, &c_b8, &u[u_offset], ldu);
}
/* ==== Near-the-diagonal bulge chase. The following loop */
/* . performs the near-the-diagonal part of a small bulge */
/* . multi-shift QR sweep. Each 6*NBMPS-2 column diagonal */
/* . chunk extends from column INCOL to column NDCOL */
/* . (including both column INCOL and column NDCOL). The */
/* . following loop chases a 3*NBMPS column long chain of */
/* . NBMPS bulges 3*NBMPS-2 columns to the right. (INCOL */
/* . may be less than KTOP and and NDCOL may be greater than */
/* . KBOT indicating phantom columns from which to chase */
/* . bulges before they are actually introduced or to which */
/* . to chase bulges beyond column KBOT.) ==== */
/* Computing MIN */
i__4 = incol + nbmps * 3 - 3, i__5 = *kbot - 2;
i__3 = min(i__4,i__5);
for (krcol = incol; krcol <= i__3; ++krcol) {
/* ==== Bulges number MTOP to MBOT are active double implicit */
/* . shift bulges. There may or may not also be small */
/* . 2-by-2 bulge, if there is room. The inactive bulges */
/* . (if any) must wait until the active bulges have moved */
/* . down the diagonal to make room. The phantom matrix */
/* . paradigm described above helps keep track. ==== */
/* Computing MAX */
i__4 = 1, i__5 = (*ktop - 1 - krcol + 2) / 3 + 1;
mtop = max(i__4,i__5);
/* Computing MIN */
i__4 = nbmps, i__5 = (*kbot - krcol) / 3;
mbot = min(i__4,i__5);
m22 = mbot + 1;
bmp22 = mbot < nbmps && krcol + (m22 - 1) * 3 == *kbot - 2;
/* ==== Generate reflections to chase the chain right */
/* . one column. (The minimum value of K is KTOP-1.) ==== */
i__4 = mbot;
for (m = mtop; m <= i__4; ++m) {
k = krcol + (m - 1) * 3;
if (k == *ktop - 1) {
slaqr1_(&c__3, &h__[*ktop + *ktop * h_dim1], ldh, &sr[(m
<< 1) - 1], &si[(m << 1) - 1], &sr[m * 2], &si[m *
2], &v[m * v_dim1 + 1]);
alpha = v[m * v_dim1 + 1];
slarfg_(&c__3, &alpha, &v[m * v_dim1 + 2], &c__1, &v[m *
v_dim1 + 1]);
} else {
beta = h__[k + 1 + k * h_dim1];
v[m * v_dim1 + 2] = h__[k + 2 + k * h_dim1];
v[m * v_dim1 + 3] = h__[k + 3 + k * h_dim1];
slarfg_(&c__3, &beta, &v[m * v_dim1 + 2], &c__1, &v[m *
v_dim1 + 1]);
/* ==== A Bulge may collapse because of vigilant */
/* . deflation or destructive underflow. In the */
/* . underflow case, try the two-small-subdiagonals */
/* . trick to try to reinflate the bulge. ==== */
if (h__[k + 3 + k * h_dim1] != 0.f || h__[k + 3 + (k + 1)
* h_dim1] != 0.f || h__[k + 3 + (k + 2) * h_dim1]
== 0.f) {
/* ==== Typical case: not collapsed (yet). ==== */
h__[k + 1 + k * h_dim1] = beta;
h__[k + 2 + k * h_dim1] = 0.f;
h__[k + 3 + k * h_dim1] = 0.f;
} else {
/* ==== Atypical case: collapsed. Attempt to */
/* . reintroduce ignoring H(K+1,K) and H(K+2,K). */
/* . If the fill resulting from the new */
/* . reflector is too large, then abandon it. */
/* . Otherwise, use the new one. ==== */
slaqr1_(&c__3, &h__[k + 1 + (k + 1) * h_dim1], ldh, &
sr[(m << 1) - 1], &si[(m << 1) - 1], &sr[m *
2], &si[m * 2], vt);
alpha = vt[0];
slarfg_(&c__3, &alpha, &vt[1], &c__1, vt);
refsum = vt[0] * (h__[k + 1 + k * h_dim1] + vt[1] *
h__[k + 2 + k * h_dim1]);
if ((r__1 = h__[k + 2 + k * h_dim1] - refsum * vt[1],
dabs(r__1)) + (r__2 = refsum * vt[2], dabs(
r__2)) > ulp * ((r__3 = h__[k + k * h_dim1],
dabs(r__3)) + (r__4 = h__[k + 1 + (k + 1) *
h_dim1], dabs(r__4)) + (r__5 = h__[k + 2 + (k
+ 2) * h_dim1], dabs(r__5)))) {
/* ==== Starting a new bulge here would */
/* . create non-negligible fill. Use */
/* . the old one with trepidation. ==== */
h__[k + 1 + k * h_dim1] = beta;
h__[k + 2 + k * h_dim1] = 0.f;
h__[k + 3 + k * h_dim1] = 0.f;
} else {
/* ==== Stating a new bulge here would */
/* . create only negligible fill. */
/* . Replace the old reflector with */
/* . the new one. ==== */
h__[k + 1 + k * h_dim1] -= refsum;
h__[k + 2 + k * h_dim1] = 0.f;
h__[k + 3 + k * h_dim1] = 0.f;
v[m * v_dim1 + 1] = vt[0];
v[m * v_dim1 + 2] = vt[1];
v[m * v_dim1 + 3] = vt[2];
}
}
}
/* L20: */
}
/* ==== Generate a 2-by-2 reflection, if needed. ==== */
k = krcol + (m22 - 1) * 3;
if (bmp22) {
if (k == *ktop - 1) {
slaqr1_(&c__2, &h__[k + 1 + (k + 1) * h_dim1], ldh, &sr[(
m22 << 1) - 1], &si[(m22 << 1) - 1], &sr[m22 * 2],
&si[m22 * 2], &v[m22 * v_dim1 + 1]);
beta = v[m22 * v_dim1 + 1];
slarfg_(&c__2, &beta, &v[m22 * v_dim1 + 2], &c__1, &v[m22
* v_dim1 + 1]);
} else {
beta = h__[k + 1 + k * h_dim1];
v[m22 * v_dim1 + 2] = h__[k + 2 + k * h_dim1];
slarfg_(&c__2, &beta, &v[m22 * v_dim1 + 2], &c__1, &v[m22
* v_dim1 + 1]);
h__[k + 1 + k * h_dim1] = beta;
h__[k + 2 + k * h_dim1] = 0.f;
}
}
/* ==== Multiply H by reflections from the left ==== */
if (accum) {
jbot = min(ndcol,*kbot);
} else if (*wantt) {
jbot = *n;
} else {
jbot = *kbot;
}
i__4 = jbot;
for (j = max(*ktop,krcol); j <= i__4; ++j) {
/* Computing MIN */
i__5 = mbot, i__6 = (j - krcol + 2) / 3;
mend = min(i__5,i__6);
i__5 = mend;
for (m = mtop; m <= i__5; ++m) {
k = krcol + (m - 1) * 3;
refsum = v[m * v_dim1 + 1] * (h__[k + 1 + j * h_dim1] + v[
m * v_dim1 + 2] * h__[k + 2 + j * h_dim1] + v[m *
v_dim1 + 3] * h__[k + 3 + j * h_dim1]);
h__[k + 1 + j * h_dim1] -= refsum;
h__[k + 2 + j * h_dim1] -= refsum * v[m * v_dim1 + 2];
h__[k + 3 + j * h_dim1] -= refsum * v[m * v_dim1 + 3];
/* L30: */
}
/* L40: */
}
if (bmp22) {
k = krcol + (m22 - 1) * 3;
/* Computing MAX */
i__4 = k + 1;
i__5 = jbot;
for (j = max(i__4,*ktop); j <= i__5; ++j) {
refsum = v[m22 * v_dim1 + 1] * (h__[k + 1 + j * h_dim1] +
v[m22 * v_dim1 + 2] * h__[k + 2 + j * h_dim1]);
h__[k + 1 + j * h_dim1] -= refsum;
h__[k + 2 + j * h_dim1] -= refsum * v[m22 * v_dim1 + 2];
/* L50: */
}
}
/* ==== Multiply H by reflections from the right. */
/* . Delay filling in the last row until the */
/* . vigilant deflation check is complete. ==== */
if (accum) {
jtop = max(*ktop,incol);
} else if (*wantt) {
jtop = 1;
} else {
jtop = *ktop;
}
i__5 = mbot;
for (m = mtop; m <= i__5; ++m) {
if (v[m * v_dim1 + 1] != 0.f) {
k = krcol + (m - 1) * 3;
/* Computing MIN */
i__6 = *kbot, i__7 = k + 3;
i__4 = min(i__6,i__7);
for (j = jtop; j <= i__4; ++j) {
refsum = v[m * v_dim1 + 1] * (h__[j + (k + 1) *
h_dim1] + v[m * v_dim1 + 2] * h__[j + (k + 2)
* h_dim1] + v[m * v_dim1 + 3] * h__[j + (k +
3) * h_dim1]);
h__[j + (k + 1) * h_dim1] -= refsum;
h__[j + (k + 2) * h_dim1] -= refsum * v[m * v_dim1 +
2];
h__[j + (k + 3) * h_dim1] -= refsum * v[m * v_dim1 +
3];
/* L60: */
}
if (accum) {
/* ==== Accumulate U. (If necessary, update Z later */
/* . with with an efficient matrix-matrix */
/* . multiply.) ==== */
kms = k - incol;
/* Computing MAX */
i__4 = 1, i__6 = *ktop - incol;
i__7 = kdu;
for (j = max(i__4,i__6); j <= i__7; ++j) {
refsum = v[m * v_dim1 + 1] * (u[j + (kms + 1) *
u_dim1] + v[m * v_dim1 + 2] * u[j + (kms
+ 2) * u_dim1] + v[m * v_dim1 + 3] * u[j
+ (kms + 3) * u_dim1]);
u[j + (kms + 1) * u_dim1] -= refsum;
u[j + (kms + 2) * u_dim1] -= refsum * v[m *
v_dim1 + 2];
u[j + (kms + 3) * u_dim1] -= refsum * v[m *
v_dim1 + 3];
/* L70: */
}
} else if (*wantz) {
/* ==== U is not accumulated, so update Z */
/* . now by multiplying by reflections */
/* . from the right. ==== */
i__7 = *ihiz;
for (j = *iloz; j <= i__7; ++j) {
refsum = v[m * v_dim1 + 1] * (z__[j + (k + 1) *
z_dim1] + v[m * v_dim1 + 2] * z__[j + (k
+ 2) * z_dim1] + v[m * v_dim1 + 3] * z__[
j + (k + 3) * z_dim1]);
z__[j + (k + 1) * z_dim1] -= refsum;
z__[j + (k + 2) * z_dim1] -= refsum * v[m *
v_dim1 + 2];
z__[j + (k + 3) * z_dim1] -= refsum * v[m *
v_dim1 + 3];
/* L80: */
}
}
}
/* L90: */
}
/* ==== Special case: 2-by-2 reflection (if needed) ==== */
k = krcol + (m22 - 1) * 3;
if (bmp22 && v[m22 * v_dim1 + 1] != 0.f) {
/* Computing MIN */
i__7 = *kbot, i__4 = k + 3;
i__5 = min(i__7,i__4);
for (j = jtop; j <= i__5; ++j) {
refsum = v[m22 * v_dim1 + 1] * (h__[j + (k + 1) * h_dim1]
+ v[m22 * v_dim1 + 2] * h__[j + (k + 2) * h_dim1])
;
h__[j + (k + 1) * h_dim1] -= refsum;
h__[j + (k + 2) * h_dim1] -= refsum * v[m22 * v_dim1 + 2];
/* L100: */
}
if (accum) {
kms = k - incol;
/* Computing MAX */
i__5 = 1, i__7 = *ktop - incol;
i__4 = kdu;
for (j = max(i__5,i__7); j <= i__4; ++j) {
refsum = v[m22 * v_dim1 + 1] * (u[j + (kms + 1) *
u_dim1] + v[m22 * v_dim1 + 2] * u[j + (kms +
2) * u_dim1]);
u[j + (kms + 1) * u_dim1] -= refsum;
u[j + (kms + 2) * u_dim1] -= refsum * v[m22 * v_dim1
+ 2];
/* L110: */
}
} else if (*wantz) {
i__4 = *ihiz;
for (j = *iloz; j <= i__4; ++j) {
refsum = v[m22 * v_dim1 + 1] * (z__[j + (k + 1) *
z_dim1] + v[m22 * v_dim1 + 2] * z__[j + (k +
2) * z_dim1]);
z__[j + (k + 1) * z_dim1] -= refsum;
z__[j + (k + 2) * z_dim1] -= refsum * v[m22 * v_dim1
+ 2];
/* L120: */
}
}
}
/* ==== Vigilant deflation check ==== */
mstart = mtop;
if (krcol + (mstart - 1) * 3 < *ktop) {
++mstart;
}
mend = mbot;
if (bmp22) {
++mend;
}
if (krcol == *kbot - 2) {
++mend;
}
i__4 = mend;
for (m = mstart; m <= i__4; ++m) {
/* Computing MIN */
i__5 = *kbot - 1, i__7 = krcol + (m - 1) * 3;
k = min(i__5,i__7);
/* ==== The following convergence test requires that */
/* . the tradition small-compared-to-nearby-diagonals */
/* . criterion and the Ahues & Tisseur (LAWN 122, 1997) */
/* . criteria both be satisfied. The latter improves */
/* . accuracy in some examples. Falling back on an */
/* . alternate convergence criterion when TST1 or TST2 */
/* . is zero (as done here) is traditional but probably */
/* . unnecessary. ==== */
if (h__[k + 1 + k * h_dim1] != 0.f) {
tst1 = (r__1 = h__[k + k * h_dim1], dabs(r__1)) + (r__2 =
h__[k + 1 + (k + 1) * h_dim1], dabs(r__2));
if (tst1 == 0.f) {
if (k >= *ktop + 1) {
tst1 += (r__1 = h__[k + (k - 1) * h_dim1], dabs(
r__1));
}
if (k >= *ktop + 2) {
tst1 += (r__1 = h__[k + (k - 2) * h_dim1], dabs(
r__1));
}
if (k >= *ktop + 3) {
tst1 += (r__1 = h__[k + (k - 3) * h_dim1], dabs(
r__1));
}
if (k <= *kbot - 2) {
tst1 += (r__1 = h__[k + 2 + (k + 1) * h_dim1],
dabs(r__1));
}
if (k <= *kbot - 3) {
tst1 += (r__1 = h__[k + 3 + (k + 1) * h_dim1],
dabs(r__1));
}
if (k <= *kbot - 4) {
tst1 += (r__1 = h__[k + 4 + (k + 1) * h_dim1],
dabs(r__1));
}
}
/* Computing MAX */
r__2 = smlnum, r__3 = ulp * tst1;
if ((r__1 = h__[k + 1 + k * h_dim1], dabs(r__1)) <= dmax(
r__2,r__3)) {
/* Computing MAX */
r__3 = (r__1 = h__[k + 1 + k * h_dim1], dabs(r__1)),
r__4 = (r__2 = h__[k + (k + 1) * h_dim1],
dabs(r__2));
h12 = dmax(r__3,r__4);
/* Computing MIN */
r__3 = (r__1 = h__[k + 1 + k * h_dim1], dabs(r__1)),
r__4 = (r__2 = h__[k + (k + 1) * h_dim1],
dabs(r__2));
h21 = dmin(r__3,r__4);
/* Computing MAX */
r__3 = (r__1 = h__[k + 1 + (k + 1) * h_dim1], dabs(
r__1)), r__4 = (r__2 = h__[k + k * h_dim1] -
h__[k + 1 + (k + 1) * h_dim1], dabs(r__2));
h11 = dmax(r__3,r__4);
/* Computing MIN */
r__3 = (r__1 = h__[k + 1 + (k + 1) * h_dim1], dabs(
r__1)), r__4 = (r__2 = h__[k + k * h_dim1] -
h__[k + 1 + (k + 1) * h_dim1], dabs(r__2));
h22 = dmin(r__3,r__4);
scl = h11 + h12;
tst2 = h22 * (h11 / scl);
/* Computing MAX */
r__1 = smlnum, r__2 = ulp * tst2;
if (tst2 == 0.f || h21 * (h12 / scl) <= dmax(r__1,
r__2)) {
h__[k + 1 + k * h_dim1] = 0.f;
}
}
}
/* L130: */
}
/* ==== Fill in the last row of each bulge. ==== */
/* Computing MIN */
i__4 = nbmps, i__5 = (*kbot - krcol - 1) / 3;
mend = min(i__4,i__5);
i__4 = mend;
for (m = mtop; m <= i__4; ++m) {
k = krcol + (m - 1) * 3;
refsum = v[m * v_dim1 + 1] * v[m * v_dim1 + 3] * h__[k + 4 + (
k + 3) * h_dim1];
h__[k + 4 + (k + 1) * h_dim1] = -refsum;
h__[k + 4 + (k + 2) * h_dim1] = -refsum * v[m * v_dim1 + 2];
h__[k + 4 + (k + 3) * h_dim1] -= refsum * v[m * v_dim1 + 3];
/* L140: */
}
/* ==== End of near-the-diagonal bulge chase. ==== */
/* L150: */
}
/* ==== Use U (if accumulated) to update far-from-diagonal */
/* . entries in H. If required, use U to update Z as */
/* . well. ==== */
if (accum) {
if (*wantt) {
jtop = 1;
jbot = *n;
} else {
jtop = *ktop;
jbot = *kbot;
}
if (! blk22 || incol < *ktop || ndcol > *kbot || ns <= 2) {
/* ==== Updates not exploiting the 2-by-2 block */
/* . structure of U. K1 and NU keep track of */
/* . the location and size of U in the special */
/* . cases of introducing bulges and chasing */
/* . bulges off the bottom. In these special */
/* . cases and in case the number of shifts */
/* . is NS = 2, there is no 2-by-2 block */
/* . structure to exploit. ==== */
/* Computing MAX */
i__3 = 1, i__4 = *ktop - incol;
k1 = max(i__3,i__4);
/* Computing MAX */
i__3 = 0, i__4 = ndcol - *kbot;
nu = kdu - max(i__3,i__4) - k1 + 1;
/* ==== Horizontal Multiply ==== */
i__3 = jbot;
i__4 = *nh;
for (jcol = min(ndcol,*kbot) + 1; i__4 < 0 ? jcol >= i__3 :
jcol <= i__3; jcol += i__4) {
/* Computing MIN */
i__5 = *nh, i__7 = jbot - jcol + 1;
jlen = min(i__5,i__7);
sgemm_("C", "N", &nu, &jlen, &nu, &c_b8, &u[k1 + k1 *
u_dim1], ldu, &h__[incol + k1 + jcol * h_dim1],
ldh, &c_b7, &wh[wh_offset], ldwh);
slacpy_("ALL", &nu, &jlen, &wh[wh_offset], ldwh, &h__[
incol + k1 + jcol * h_dim1], ldh);
/* L160: */
}
/* ==== Vertical multiply ==== */
i__4 = max(*ktop,incol) - 1;
i__3 = *nv;
for (jrow = jtop; i__3 < 0 ? jrow >= i__4 : jrow <= i__4;
jrow += i__3) {
/* Computing MIN */
i__5 = *nv, i__7 = max(*ktop,incol) - jrow;
jlen = min(i__5,i__7);
sgemm_("N", "N", &jlen, &nu, &nu, &c_b8, &h__[jrow + (
incol + k1) * h_dim1], ldh, &u[k1 + k1 * u_dim1],
ldu, &c_b7, &wv[wv_offset], ldwv);
slacpy_("ALL", &jlen, &nu, &wv[wv_offset], ldwv, &h__[
jrow + (incol + k1) * h_dim1], ldh);
/* L170: */
}
/* ==== Z multiply (also vertical) ==== */
if (*wantz) {
i__3 = *ihiz;
i__4 = *nv;
for (jrow = *iloz; i__4 < 0 ? jrow >= i__3 : jrow <= i__3;
jrow += i__4) {
/* Computing MIN */
i__5 = *nv, i__7 = *ihiz - jrow + 1;
jlen = min(i__5,i__7);
sgemm_("N", "N", &jlen, &nu, &nu, &c_b8, &z__[jrow + (
incol + k1) * z_dim1], ldz, &u[k1 + k1 *
u_dim1], ldu, &c_b7, &wv[wv_offset], ldwv);
slacpy_("ALL", &jlen, &nu, &wv[wv_offset], ldwv, &z__[
jrow + (incol + k1) * z_dim1], ldz)
;
/* L180: */
}
}
} else {
/* ==== Updates exploiting U's 2-by-2 block structure. */
/* . (I2, I4, J2, J4 are the last rows and columns */
/* . of the blocks.) ==== */
i2 = (kdu + 1) / 2;
i4 = kdu;
j2 = i4 - i2;
j4 = kdu;
/* ==== KZS and KNZ deal with the band of zeros */
/* . along the diagonal of one of the triangular */
/* . blocks. ==== */
kzs = j4 - j2 - (ns + 1);
knz = ns + 1;
/* ==== Horizontal multiply ==== */
i__4 = jbot;
i__3 = *nh;
for (jcol = min(ndcol,*kbot) + 1; i__3 < 0 ? jcol >= i__4 :
jcol <= i__4; jcol += i__3) {
/* Computing MIN */
i__5 = *nh, i__7 = jbot - jcol + 1;
jlen = min(i__5,i__7);
/* ==== Copy bottom of H to top+KZS of scratch ==== */
/* (The first KZS rows get multiplied by zero.) ==== */
slacpy_("ALL", &knz, &jlen, &h__[incol + 1 + j2 + jcol *
h_dim1], ldh, &wh[kzs + 1 + wh_dim1], ldwh);
/* ==== Multiply by U21' ==== */
slaset_("ALL", &kzs, &jlen, &c_b7, &c_b7, &wh[wh_offset],
ldwh);
strmm_("L", "U", "C", "N", &knz, &jlen, &c_b8, &u[j2 + 1
+ (kzs + 1) * u_dim1], ldu, &wh[kzs + 1 + wh_dim1]
, ldwh);
/* ==== Multiply top of H by U11' ==== */
sgemm_("C", "N", &i2, &jlen, &j2, &c_b8, &u[u_offset],
ldu, &h__[incol + 1 + jcol * h_dim1], ldh, &c_b8,
&wh[wh_offset], ldwh);
/* ==== Copy top of H to bottom of WH ==== */
slacpy_("ALL", &j2, &jlen, &h__[incol + 1 + jcol * h_dim1]
, ldh, &wh[i2 + 1 + wh_dim1], ldwh);
/* ==== Multiply by U21' ==== */
strmm_("L", "L", "C", "N", &j2, &jlen, &c_b8, &u[(i2 + 1)
* u_dim1 + 1], ldu, &wh[i2 + 1 + wh_dim1], ldwh);
/* ==== Multiply by U22 ==== */
i__5 = i4 - i2;
i__7 = j4 - j2;
sgemm_("C", "N", &i__5, &jlen, &i__7, &c_b8, &u[j2 + 1 + (
i2 + 1) * u_dim1], ldu, &h__[incol + 1 + j2 +
jcol * h_dim1], ldh, &c_b8, &wh[i2 + 1 + wh_dim1],
ldwh);
/* ==== Copy it back ==== */
slacpy_("ALL", &kdu, &jlen, &wh[wh_offset], ldwh, &h__[
incol + 1 + jcol * h_dim1], ldh);
/* L190: */
}
/* ==== Vertical multiply ==== */
i__3 = max(incol,*ktop) - 1;
i__4 = *nv;
for (jrow = jtop; i__4 < 0 ? jrow >= i__3 : jrow <= i__3;
jrow += i__4) {
/* Computing MIN */
i__5 = *nv, i__7 = max(incol,*ktop) - jrow;
jlen = min(i__5,i__7);
/* ==== Copy right of H to scratch (the first KZS */
/* . columns get multiplied by zero) ==== */
slacpy_("ALL", &jlen, &knz, &h__[jrow + (incol + 1 + j2) *
h_dim1], ldh, &wv[(kzs + 1) * wv_dim1 + 1], ldwv);
/* ==== Multiply by U21 ==== */
slaset_("ALL", &jlen, &kzs, &c_b7, &c_b7, &wv[wv_offset],
ldwv);
strmm_("R", "U", "N", "N", &jlen, &knz, &c_b8, &u[j2 + 1
+ (kzs + 1) * u_dim1], ldu, &wv[(kzs + 1) *
wv_dim1 + 1], ldwv);
/* ==== Multiply by U11 ==== */
sgemm_("N", "N", &jlen, &i2, &j2, &c_b8, &h__[jrow + (
incol + 1) * h_dim1], ldh, &u[u_offset], ldu, &
c_b8, &wv[wv_offset], ldwv);
/* ==== Copy left of H to right of scratch ==== */
slacpy_("ALL", &jlen, &j2, &h__[jrow + (incol + 1) *
h_dim1], ldh, &wv[(i2 + 1) * wv_dim1 + 1], ldwv);
/* ==== Multiply by U21 ==== */
i__5 = i4 - i2;
strmm_("R", "L", "N", "N", &jlen, &i__5, &c_b8, &u[(i2 +
1) * u_dim1 + 1], ldu, &wv[(i2 + 1) * wv_dim1 + 1]
, ldwv);
/* ==== Multiply by U22 ==== */
i__5 = i4 - i2;
i__7 = j4 - j2;
sgemm_("N", "N", &jlen, &i__5, &i__7, &c_b8, &h__[jrow + (
incol + 1 + j2) * h_dim1], ldh, &u[j2 + 1 + (i2 +
1) * u_dim1], ldu, &c_b8, &wv[(i2 + 1) * wv_dim1
+ 1], ldwv);
/* ==== Copy it back ==== */
slacpy_("ALL", &jlen, &kdu, &wv[wv_offset], ldwv, &h__[
jrow + (incol + 1) * h_dim1], ldh);
/* L200: */
}
/* ==== Multiply Z (also vertical) ==== */
if (*wantz) {
i__4 = *ihiz;
i__3 = *nv;
for (jrow = *iloz; i__3 < 0 ? jrow >= i__4 : jrow <= i__4;
jrow += i__3) {
/* Computing MIN */
i__5 = *nv, i__7 = *ihiz - jrow + 1;
jlen = min(i__5,i__7);
/* ==== Copy right of Z to left of scratch (first */
/* . KZS columns get multiplied by zero) ==== */
slacpy_("ALL", &jlen, &knz, &z__[jrow + (incol + 1 +
j2) * z_dim1], ldz, &wv[(kzs + 1) * wv_dim1 +
1], ldwv);
/* ==== Multiply by U12 ==== */
slaset_("ALL", &jlen, &kzs, &c_b7, &c_b7, &wv[
wv_offset], ldwv);
strmm_("R", "U", "N", "N", &jlen, &knz, &c_b8, &u[j2
+ 1 + (kzs + 1) * u_dim1], ldu, &wv[(kzs + 1)
* wv_dim1 + 1], ldwv);
/* ==== Multiply by U11 ==== */
sgemm_("N", "N", &jlen, &i2, &j2, &c_b8, &z__[jrow + (
incol + 1) * z_dim1], ldz, &u[u_offset], ldu,
&c_b8, &wv[wv_offset], ldwv);
/* ==== Copy left of Z to right of scratch ==== */
slacpy_("ALL", &jlen, &j2, &z__[jrow + (incol + 1) *
z_dim1], ldz, &wv[(i2 + 1) * wv_dim1 + 1],
ldwv);
/* ==== Multiply by U21 ==== */
i__5 = i4 - i2;
strmm_("R", "L", "N", "N", &jlen, &i__5, &c_b8, &u[(
i2 + 1) * u_dim1 + 1], ldu, &wv[(i2 + 1) *
wv_dim1 + 1], ldwv);
/* ==== Multiply by U22 ==== */
i__5 = i4 - i2;
i__7 = j4 - j2;
sgemm_("N", "N", &jlen, &i__5, &i__7, &c_b8, &z__[
jrow + (incol + 1 + j2) * z_dim1], ldz, &u[j2
+ 1 + (i2 + 1) * u_dim1], ldu, &c_b8, &wv[(i2
+ 1) * wv_dim1 + 1], ldwv);
/* ==== Copy the result back to Z ==== */
slacpy_("ALL", &jlen, &kdu, &wv[wv_offset], ldwv, &
z__[jrow + (incol + 1) * z_dim1], ldz);
/* L210: */
}
}
}
}
/* L220: */
}
/* ==== End of SLAQR5 ==== */
return 0;
} /* slaqr5_ */
/* Subroutine */ int sdrvgg_(integer *nsizes, integer *nn, integer *ntypes,
logical *dotype, integer *iseed, real *thresh, real *thrshn, integer *
nounit, real *a, integer *lda, real *b, real *s, real *t, real *s2,
real *t2, real *q, integer *ldq, real *z__, real *alphr1, real *
alphi1, real *beta1, real *alphr2, real *alphi2, real *beta2, real *
vl, real *vr, real *work, integer *lwork, real *result, integer *info)
{
/* Initialized data */
static integer kclass[26] = { 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,2,2,2,2,2,2,2,
2,2,2,3 };
static integer kbmagn[26] = { 1,1,1,1,1,1,1,1,3,2,3,2,2,3,1,1,1,1,1,1,1,3,
2,3,2,1 };
static integer ktrian[26] = { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,
1,1,1,1 };
static integer iasign[26] = { 0,0,0,0,0,0,2,0,2,2,0,0,2,2,2,0,2,0,0,0,2,2,
2,2,2,0 };
static integer ibsign[26] = { 0,0,0,0,0,0,0,2,0,0,2,2,0,0,2,0,2,0,0,0,0,0,
0,0,0,0 };
static integer kz1[6] = { 0,1,2,1,3,3 };
static integer kz2[6] = { 0,0,1,2,1,1 };
static integer kadd[6] = { 0,0,0,0,3,2 };
static integer katype[26] = { 0,1,0,1,2,3,4,1,4,4,1,1,4,4,4,2,4,5,8,7,9,4,
4,4,4,0 };
static integer kbtype[26] = { 0,0,1,1,2,-3,1,4,1,1,4,4,1,1,-4,2,-4,8,8,8,
8,8,8,8,8,0 };
static integer kazero[26] = { 1,1,1,1,1,1,2,1,2,2,1,1,2,2,3,1,3,5,5,5,5,3,
3,3,3,1 };
static integer kbzero[26] = { 1,1,1,1,1,1,1,2,1,1,2,2,1,1,4,1,4,6,6,6,6,4,
4,4,4,1 };
static integer kamagn[26] = { 1,1,1,1,1,1,1,1,2,3,2,3,2,3,1,1,1,1,1,1,1,2,
3,3,2,1 };
/* Format strings */
static char fmt_9999[] = "(\002 SDRVGG: \002,a,\002 returned INFO=\002,i"
"6,\002.\002,/9x,\002N=\002,i6,\002, JTYPE=\002,i6,\002, ISEED="
"(\002,3(i5,\002,\002),i5,\002)\002)";
static char fmt_9997[] = "(\002 SDRVGG: SGET53 returned INFO=\002,i1,"
"\002 for eigenvalue \002,i6,\002.\002,/9x,\002N=\002,i6,\002, JT"
"YPE=\002,i6,\002, ISEED=(\002,3(i5,\002,\002),i5,\002)\002)";
static char fmt_9996[] = "(\002 SDRVGG: S not in Schur form at eigenvalu"
"e \002,i6,\002.\002,/9x,\002N=\002,i6,\002, JTYPE=\002,i6,\002, "
"ISEED=(\002,3(i5,\002,\002),i5,\002)\002)";
static char fmt_9998[] = "(\002 SDRVGG: \002,a,\002 Eigenvectors from"
" \002,a,\002 incorrectly \002,\002normalized.\002,/\002 Bits of "
"error=\002,0p,g10.3,\002,\002,9x,\002N=\002,i6,\002, JTYPE=\002,"
"i6,\002, ISEED=(\002,3(i5,\002,\002),i5,\002)\002)";
static char fmt_9995[] = "(/1x,a3,\002 -- Real Generalized eigenvalue pr"
"oblem driver\002)";
static char fmt_9994[] = "(\002 Matrix types (see SDRVGG for details):"
" \002)";
static char fmt_9993[] = "(\002 Special Matrices:\002,23x,\002(J'=transp"
"osed Jordan block)\002,/\002 1=(0,0) 2=(I,0) 3=(0,I) 4=(I,I"
") 5=(J',J') \002,\0026=(diag(J',I), diag(I,J'))\002,/\002 Diag"
"onal Matrices: ( \002,\002D=diag(0,1,2,...) )\002,/\002 7=(D,"
"I) 9=(large*D, small*I\002,\002) 11=(large*I, small*D) 13=(l"
"arge*D, large*I)\002,/\002 8=(I,D) 10=(small*D, large*I) 12="
"(small*I, large*D) \002,\002 14=(small*D, small*I)\002,/\002 15"
"=(D, reversed D)\002)";
static char fmt_9992[] = "(\002 Matrices Rotated by Random \002,a,\002 M"
"atrices U, V:\002,/\002 16=Transposed Jordan Blocks "
" 19=geometric \002,\002alpha, beta=0,1\002,/\002 17=arithm. alp"
"ha&beta \002,\002 20=arithmetic alpha, beta=0,"
"1\002,/\002 18=clustered \002,\002alpha, beta=0,1 21"
"=random alpha, beta=0,1\002,/\002 Large & Small Matrices:\002,"
"/\002 22=(large, small) \002,\00223=(small,large) 24=(smal"
"l,small) 25=(large,large)\002,/\002 26=random O(1) matrices"
".\002)";
static char fmt_9991[] = "(/\002 Tests performed: (S is Schur, T is tri"
"angular, \002,\002Q and Z are \002,a,\002,\002,/20x,\002l and r "
"are the appropriate left and right\002,/19x,\002eigenvectors, re"
"sp., a is alpha, b is beta, and\002,/19x,a,\002 means \002,a,"
"\002.)\002,/\002 1 = | A - Q S Z\002,a,\002 | / ( |A| n ulp ) "
" 2 = | B - Q T Z\002,a,\002 | / ( |B| n ulp )\002,/\002 3 = | "
"I - QQ\002,a,\002 | / ( n ulp ) 4 = | I - ZZ\002,a"
",\002 | / ( n ulp )\002,/\002 5 = difference between (alpha,beta"
") and diagonals of\002,\002 (S,T)\002,/\002 6 = max | ( b A - a "
"B )\002,a,\002 l | / const. 7 = max | ( b A - a B ) r | / cons"
"t.\002,/1x)";
static char fmt_9990[] = "(\002 Matrix order=\002,i5,\002, type=\002,i2"
",\002, seed=\002,4(i4,\002,\002),\002 result \002,i3,\002 is\002"
",0p,f8.2)";
static char fmt_9989[] = "(\002 Matrix order=\002,i5,\002, type=\002,i2"
",\002, seed=\002,4(i4,\002,\002),\002 result \002,i3,\002 is\002"
",1p,e10.3)";
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, q_dim1, q_offset, s_dim1,
s_offset, s2_dim1, s2_offset, t_dim1, t_offset, t2_dim1,
t2_offset, vl_dim1, vl_offset, vr_dim1, vr_offset, z_dim1,
z_offset, i__1, i__2, i__3, i__4;
real r__1, r__2, r__3, r__4, r__5, r__6, r__7, r__8, r__9, r__10;
/* Builtin functions */
double r_sign(real *, real *);
integer s_wsfe(cilist *), do_fio(integer *, char *, ftnlen), e_wsfe(void);
/* Local variables */
integer j, n, i1, n1, jc, nb, in, jr, ns, nbz;
real ulp;
integer iadd, nmax;
real temp1, temp2;
logical badnn;
real dumma[4];
integer iinfo;
real rmagn[4];
extern /* Subroutine */ int sgegs_(char *, char *, integer *, real *,
integer *, real *, integer *, real *, real *, real *, real *,
integer *, real *, integer *, real *, integer *, integer *), sget51_(integer *, integer *, real *, integer *,
real *, integer *, real *, integer *, real *, integer *, real *,
real *), sget52_(logical *, integer *, real *, integer *, real *,
integer *, real *, integer *, real *, real *, real *, real *,
real *), sgegv_(char *, char *, integer *, real *, integer *,
real *, integer *, real *, real *, real *, real *, integer *,
real *, integer *, real *, integer *, integer *),
sget53_(real *, integer *, real *, integer *, real *, real *,
real *, real *, integer *);
integer nmats, jsize, nerrs, jtype, ntest;
extern /* Subroutine */ int slatm4_(integer *, integer *, integer *,
integer *, integer *, real *, real *, real *, integer *, integer *
, real *, integer *);
logical ilabad;
extern /* Subroutine */ int sorm2r_(char *, char *, integer *, integer *,
integer *, real *, integer *, real *, real *, integer *, real *,
integer *), slabad_(real *, real *);
extern doublereal slamch_(char *);
real safmin;
integer ioldsd[4];
real safmax;
extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
integer *, integer *);
extern /* Subroutine */ int slarfg_(integer *, real *, real *, integer *,
real *);
extern doublereal slarnd_(integer *, integer *);
extern /* Subroutine */ int alasvm_(char *, integer *, integer *, integer
*, integer *), xerbla_(char *, integer *),
slacpy_(char *, integer *, integer *, real *, integer *, real *,
integer *), slaset_(char *, integer *, integer *, real *,
real *, real *, integer *);
real ulpinv;
integer lwkopt, mtypes, ntestt;
/* Fortran I/O blocks */
static cilist io___42 = { 0, 0, 0, fmt_9999, 0 };
static cilist io___43 = { 0, 0, 0, fmt_9999, 0 };
static cilist io___47 = { 0, 0, 0, fmt_9997, 0 };
static cilist io___48 = { 0, 0, 0, fmt_9996, 0 };
static cilist io___49 = { 0, 0, 0, fmt_9999, 0 };
static cilist io___51 = { 0, 0, 0, fmt_9998, 0 };
static cilist io___52 = { 0, 0, 0, fmt_9998, 0 };
static cilist io___53 = { 0, 0, 0, fmt_9996, 0 };
static cilist io___54 = { 0, 0, 0, fmt_9995, 0 };
static cilist io___55 = { 0, 0, 0, fmt_9994, 0 };
static cilist io___56 = { 0, 0, 0, fmt_9993, 0 };
static cilist io___57 = { 0, 0, 0, fmt_9992, 0 };
static cilist io___58 = { 0, 0, 0, fmt_9991, 0 };
static cilist io___59 = { 0, 0, 0, fmt_9990, 0 };
static cilist io___60 = { 0, 0, 0, fmt_9989, 0 };
/* -- LAPACK test routine (version 3.1) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* SDRVGG checks the nonsymmetric generalized eigenvalue driver */
/* routines. */
/* T T T */
/* SGEGS factors A and B as Q S Z and Q T Z , where means */
/* transpose, T is upper triangular, S is in generalized Schur form */
/* (block upper triangular, with 1x1 and 2x2 blocks on the diagonal, */
/* the 2x2 blocks corresponding to complex conjugate pairs of */
/* generalized eigenvalues), and Q and Z are orthogonal. It also */
/* computes the generalized eigenvalues (alpha(1),beta(1)), ..., */
/* (alpha(n),beta(n)), where alpha(j)=S(j,j) and beta(j)=P(j,j) -- */
/* thus, w(j) = alpha(j)/beta(j) is a root of the generalized */
/* eigenvalue problem */
/* det( A - w(j) B ) = 0 */
/* and m(j) = beta(j)/alpha(j) is a root of the essentially equivalent */
/* problem */
/* det( m(j) A - B ) = 0 */
/* SGEGV computes the generalized eigenvalues (alpha(1),beta(1)), ..., */
/* (alpha(n),beta(n)), the matrix L whose columns contain the */
/* generalized left eigenvectors l, and the matrix R whose columns */
/* contain the generalized right eigenvectors r for the pair (A,B). */
/* When SDRVGG is called, a number of matrix "sizes" ("n's") and a */
/* number of matrix "types" are specified. For each size ("n") */
/* and each type of matrix, one matrix will be generated and used */
/* to test the nonsymmetric eigenroutines. For each matrix, 7 */
/* tests will be performed and compared with the threshhold THRESH: */
/* Results from SGEGS: */
/* T */
/* (1) | A - Q S Z | / ( |A| n ulp ) */
/* T */
/* (2) | B - Q T Z | / ( |B| n ulp ) */
/* T */
/* (3) | I - QQ | / ( n ulp ) */
/* T */
/* (4) | I - ZZ | / ( n ulp ) */
/* (5) maximum over j of D(j) where: */
/* if alpha(j) is real: */
/* |alpha(j) - S(j,j)| |beta(j) - T(j,j)| */
/* D(j) = ------------------------ + ----------------------- */
/* max(|alpha(j)|,|S(j,j)|) max(|beta(j)|,|T(j,j)|) */
/* if alpha(j) is complex: */
/* | det( s S - w T ) | */
/* D(j) = --------------------------------------------------- */
/* ulp max( s norm(S), |w| norm(T) )*norm( s S - w T ) */
/* and S and T are here the 2 x 2 diagonal blocks of S and T */
/* corresponding to the j-th eigenvalue. */
/* Results from SGEGV: */
/* (6) max over all left eigenvalue/-vector pairs (beta/alpha,l) of */
/* | l**H * (beta A - alpha B) | / ( ulp max( |beta A|, |alpha B| ) ) */
/* where l**H is the conjugate tranpose of l. */
/* (7) max over all right eigenvalue/-vector pairs (beta/alpha,r) of */
/* | (beta A - alpha B) r | / ( ulp max( |beta A|, |alpha B| ) ) */
/* Test Matrices */
/* ---- -------- */
/* The sizes of the test matrices are specified by an array */
/* NN(1:NSIZES); the value of each element NN(j) specifies one size. */
/* The "types" are specified by a logical array DOTYPE( 1:NTYPES ); if */
/* DOTYPE(j) is .TRUE., then matrix type "j" will be generated. */
/* Currently, the list of possible types is: */
/* (1) ( 0, 0 ) (a pair of zero matrices) */
/* (2) ( I, 0 ) (an identity and a zero matrix) */
/* (3) ( 0, I ) (an identity and a zero matrix) */
/* (4) ( I, I ) (a pair of identity matrices) */
/* t t */
/* (5) ( J , J ) (a pair of transposed Jordan blocks) */
/* t ( I 0 ) */
/* (6) ( X, Y ) where X = ( J 0 ) and Y = ( t ) */
/* ( 0 I ) ( 0 J ) */
/* and I is a k x k identity and J a (k+1)x(k+1) */
/* Jordan block; k=(N-1)/2 */
/* (7) ( D, I ) where D is diag( 0, 1,..., N-1 ) (a diagonal */
/* matrix with those diagonal entries.) */
/* (8) ( I, D ) */
/* (9) ( big*D, small*I ) where "big" is near overflow and small=1/big */
/* (10) ( small*D, big*I ) */
/* (11) ( big*I, small*D ) */
/* (12) ( small*I, big*D ) */
/* (13) ( big*D, big*I ) */
/* (14) ( small*D, small*I ) */
/* (15) ( D1, D2 ) where D1 is diag( 0, 0, 1, ..., N-3, 0 ) and */
/* D2 is diag( 0, N-3, N-4,..., 1, 0, 0 ) */
/* t t */
/* (16) Q ( J , J ) Z where Q and Z are random orthogonal matrices. */
/* (17) Q ( T1, T2 ) Z where T1 and T2 are upper triangular matrices */
/* with random O(1) entries above the diagonal */
/* and diagonal entries diag(T1) = */
/* ( 0, 0, 1, ..., N-3, 0 ) and diag(T2) = */
/* ( 0, N-3, N-4,..., 1, 0, 0 ) */
/* (18) Q ( T1, T2 ) Z diag(T1) = ( 0, 0, 1, 1, s, ..., s, 0 ) */
/* diag(T2) = ( 0, 1, 0, 1,..., 1, 0 ) */
/* s = machine precision. */
/* (19) Q ( T1, T2 ) Z diag(T1)=( 0,0,1,1, 1-d, ..., 1-(N-5)*d=s, 0 ) */
/* diag(T2) = ( 0, 1, 0, 1, ..., 1, 0 ) */
/* N-5 */
/* (20) Q ( T1, T2 ) Z diag(T1)=( 0, 0, 1, 1, a, ..., a =s, 0 ) */
/* diag(T2) = ( 0, 1, 0, 1, ..., 1, 0, 0 ) */
/* (21) Q ( T1, T2 ) Z diag(T1)=( 0, 0, 1, r1, r2, ..., r(N-4), 0 ) */
/* diag(T2) = ( 0, 1, 0, 1, ..., 1, 0, 0 ) */
/* where r1,..., r(N-4) are random. */
/* (22) Q ( big*T1, small*T2 ) Z diag(T1) = ( 0, 0, 1, ..., N-3, 0 ) */
/* diag(T2) = ( 0, 1, ..., 1, 0, 0 ) */
/* (23) Q ( small*T1, big*T2 ) Z diag(T1) = ( 0, 0, 1, ..., N-3, 0 ) */
/* diag(T2) = ( 0, 1, ..., 1, 0, 0 ) */
/* (24) Q ( small*T1, small*T2 ) Z diag(T1) = ( 0, 0, 1, ..., N-3, 0 ) */
/* diag(T2) = ( 0, 1, ..., 1, 0, 0 ) */
/* (25) Q ( big*T1, big*T2 ) Z diag(T1) = ( 0, 0, 1, ..., N-3, 0 ) */
/* diag(T2) = ( 0, 1, ..., 1, 0, 0 ) */
/* (26) Q ( T1, T2 ) Z where T1 and T2 are random upper-triangular */
/* matrices. */
/* Arguments */
/* ========= */
/* NSIZES (input) INTEGER */
/* The number of sizes of matrices to use. If it is zero, */
/* SDRVGG does nothing. It must be at least zero. */
/* NN (input) INTEGER array, dimension (NSIZES) */
/* An array containing the sizes to be used for the matrices. */
/* Zero values will be skipped. The values must be at least */
/* zero. */
/* NTYPES (input) INTEGER */
/* The number of elements in DOTYPE. If it is zero, SDRVGG */
/* does nothing. It must be at least zero. If it is MAXTYP+1 */
/* and NSIZES is 1, then an additional type, MAXTYP+1 is */
/* defined, which is to use whatever matrix is in A. This */
/* is only useful if DOTYPE(1:MAXTYP) is .FALSE. and */
/* DOTYPE(MAXTYP+1) is .TRUE. . */
/* DOTYPE (input) LOGICAL array, dimension (NTYPES) */
/* If DOTYPE(j) is .TRUE., then for each size in NN a */
/* matrix of that size and of type j will be generated. */
/* If NTYPES is smaller than the maximum number of types */
/* defined (PARAMETER MAXTYP), then types NTYPES+1 through */
/* MAXTYP will not be generated. If NTYPES is larger */
/* than MAXTYP, DOTYPE(MAXTYP+1) through DOTYPE(NTYPES) */
/* will be ignored. */
/* ISEED (input/output) INTEGER array, dimension (4) */
/* On entry ISEED specifies the seed of the random number */
/* generator. The array elements should be between 0 and 4095; */
/* if not they will be reduced mod 4096. Also, ISEED(4) must */
/* be odd. The random number generator uses a linear */
/* congruential sequence limited to small integers, and so */
/* should produce machine independent random numbers. The */
/* values of ISEED are changed on exit, and can be used in the */
/* next call to SDRVGG to continue the same random number */
/* sequence. */
/* THRESH (input) REAL */
/* A test will count as "failed" if the "error", computed as */
/* described above, exceeds THRESH. Note that the error is */
/* scaled to be O(1), so THRESH should be a reasonably small */
/* multiple of 1, e.g., 10 or 100. In particular, it should */
/* not depend on the precision (single vs. double) or the size */
/* of the matrix. It must be at least zero. */
/* THRSHN (input) REAL */
/* Threshhold for reporting eigenvector normalization error. */
/* If the normalization of any eigenvector differs from 1 by */
/* more than THRSHN*ulp, then a special error message will be */
/* printed. (This is handled separately from the other tests, */
/* since only a compiler or programming error should cause an */
/* error message, at least if THRSHN is at least 5--10.) */
/* NOUNIT (input) INTEGER */
/* The FORTRAN unit number for printing out error messages */
/* (e.g., if a routine returns IINFO not equal to 0.) */
/* A (input/workspace) REAL array, dimension */
/* (LDA, max(NN)) */
/* Used to hold the original A matrix. Used as input only */
/* if NTYPES=MAXTYP+1, DOTYPE(1:MAXTYP)=.FALSE., and */
/* DOTYPE(MAXTYP+1)=.TRUE. */
/* LDA (input) INTEGER */
/* The leading dimension of A, B, S, T, S2, and T2. */
/* It must be at least 1 and at least max( NN ). */
/* B (input/workspace) REAL array, dimension */
/* (LDA, max(NN)) */
/* Used to hold the original B matrix. Used as input only */
/* if NTYPES=MAXTYP+1, DOTYPE(1:MAXTYP)=.FALSE., and */
/* DOTYPE(MAXTYP+1)=.TRUE. */
/* S (workspace) REAL array, dimension (LDA, max(NN)) */
/* The Schur form matrix computed from A by SGEGS. On exit, S */
/* contains the Schur form matrix corresponding to the matrix */
/* in A. */
/* T (workspace) REAL array, dimension (LDA, max(NN)) */
/* The upper triangular matrix computed from B by SGEGS. */
/* S2 (workspace) REAL array, dimension (LDA, max(NN)) */
/* The matrix computed from A by SGEGV. This will be the */
/* Schur form of some matrix related to A, but will not, in */
/* general, be the same as S. */
/* T2 (workspace) REAL array, dimension (LDA, max(NN)) */
/* The matrix computed from B by SGEGV. This will be the */
/* Schur form of some matrix related to B, but will not, in */
/* general, be the same as T. */
/* Q (workspace) REAL array, dimension (LDQ, max(NN)) */
/* The (left) orthogonal matrix computed by SGEGS. */
/* LDQ (input) INTEGER */
/* The leading dimension of Q, Z, VL, and VR. It must */
/* be at least 1 and at least max( NN ). */
/* Z (workspace) REAL array of */
/* dimension( LDQ, max(NN) ) */
/* The (right) orthogonal matrix computed by SGEGS. */
/* ALPHR1 (workspace) REAL array, dimension (max(NN)) */
/* ALPHI1 (workspace) REAL array, dimension (max(NN)) */
/* BETA1 (workspace) REAL array, dimension (max(NN)) */
/* The generalized eigenvalues of (A,B) computed by SGEGS. */
/* ( ALPHR1(k)+ALPHI1(k)*i ) / BETA1(k) is the k-th */
/* generalized eigenvalue of the matrices in A and B. */
/* ALPHR2 (workspace) REAL array, dimension (max(NN)) */
/* ALPHI2 (workspace) REAL array, dimension (max(NN)) */
/* BETA2 (workspace) REAL array, dimension (max(NN)) */
/* The generalized eigenvalues of (A,B) computed by SGEGV. */
/* ( ALPHR2(k)+ALPHI2(k)*i ) / BETA2(k) is the k-th */
/* generalized eigenvalue of the matrices in A and B. */
/* VL (workspace) REAL array, dimension (LDQ, max(NN)) */
/* The (block lower triangular) left eigenvector matrix for */
/* the matrices in A and B. (See STGEVC for the format.) */
/* VR (workspace) REAL array, dimension (LDQ, max(NN)) */
/* The (block upper triangular) right eigenvector matrix for */
/* the matrices in A and B. (See STGEVC for the format.) */
/* WORK (workspace) REAL array, dimension (LWORK) */
/* LWORK (input) INTEGER */
/* The number of entries in WORK. This must be at least */
/* 2*N + MAX( 6*N, N*(NB+1), (k+1)*(2*k+N+1) ), where */
/* "k" is the sum of the blocksize and number-of-shifts for */
/* SHGEQZ, and NB is the greatest of the blocksizes for */
/* SGEQRF, SORMQR, and SORGQR. (The blocksizes and the */
/* number-of-shifts are retrieved through calls to ILAENV.) */
/* RESULT (output) REAL array, dimension (15) */
/* The values computed by the tests described above. */
/* The values are currently limited to 1/ulp, to avoid */
/* overflow. */
/* INFO (output) INTEGER */
/* = 0: successful exit */
/* < 0: if INFO = -i, the i-th argument had an illegal value. */
/* > 0: A routine returned an error code. INFO is the */
/* absolute value of the INFO value returned. */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. Local Arrays .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Data statements .. */
/* Parameter adjustments */
--nn;
--dotype;
--iseed;
t2_dim1 = *lda;
t2_offset = 1 + t2_dim1;
t2 -= t2_offset;
s2_dim1 = *lda;
s2_offset = 1 + s2_dim1;
s2 -= s2_offset;
t_dim1 = *lda;
t_offset = 1 + t_dim1;
t -= t_offset;
s_dim1 = *lda;
s_offset = 1 + s_dim1;
s -= s_offset;
b_dim1 = *lda;
b_offset = 1 + b_dim1;
b -= b_offset;
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
vr_dim1 = *ldq;
vr_offset = 1 + vr_dim1;
vr -= vr_offset;
vl_dim1 = *ldq;
vl_offset = 1 + vl_dim1;
vl -= vl_offset;
z_dim1 = *ldq;
z_offset = 1 + z_dim1;
z__ -= z_offset;
q_dim1 = *ldq;
q_offset = 1 + q_dim1;
q -= q_offset;
--alphr1;
--alphi1;
--beta1;
--alphr2;
--alphi2;
--beta2;
--work;
--result;
/* Function Body */
/* .. */
/* .. Executable Statements .. */
/* Check for errors */
*info = 0;
badnn = FALSE_;
nmax = 1;
i__1 = *nsizes;
for (j = 1; j <= i__1; ++j) {
/* Computing MAX */
i__2 = nmax, i__3 = nn[j];
nmax = max(i__2,i__3);
if (nn[j] < 0) {
badnn = TRUE_;
}
/* L10: */
}
/* Maximum blocksize and shift -- we assume that blocksize and number */
/* of shifts are monotone increasing functions of N. */
/* Computing MAX */
i__1 = 1, i__2 = ilaenv_(&c__1, "SGEQRF", " ", &nmax, &nmax, &c_n1, &c_n1), i__1 = max(i__1,i__2), i__2 = ilaenv_(&
c__1, "SORMQR", "LT", &nmax, &nmax, &nmax, &c_n1), i__1 = max(i__1,i__2), i__2 = ilaenv_(&c__1, "SORGQR",
" ", &nmax, &nmax, &nmax, &c_n1);
nb = max(i__1,i__2);
nbz = ilaenv_(&c__1, "SHGEQZ", "SII", &nmax, &c__1, &nmax, &c__0);
ns = ilaenv_(&c__4, "SHGEQZ", "SII", &nmax, &c__1, &nmax, &c__0);
i1 = nbz + ns;
/* Computing MAX */
i__1 = nmax * 6, i__2 = nmax * (nb + 1), i__1 = max(i__1,i__2), i__2 = ((
i1 << 1) + nmax + 1) * (i1 + 1);
lwkopt = (nmax << 1) + max(i__1,i__2);
/* Check for errors */
if (*nsizes < 0) {
*info = -1;
} else if (badnn) {
*info = -2;
} else if (*ntypes < 0) {
*info = -3;
} else if (*thresh < 0.f) {
*info = -6;
} else if (*lda <= 1 || *lda < nmax) {
*info = -10;
} else if (*ldq <= 1 || *ldq < nmax) {
*info = -19;
} else if (lwkopt > *lwork) {
*info = -30;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("SDRVGG", &i__1);
return 0;
}
/* Quick return if possible */
if (*nsizes == 0 || *ntypes == 0) {
return 0;
}
safmin = slamch_("Safe minimum");
ulp = slamch_("Epsilon") * slamch_("Base");
safmin /= ulp;
safmax = 1.f / safmin;
slabad_(&safmin, &safmax);
ulpinv = 1.f / ulp;
/* The values RMAGN(2:3) depend on N, see below. */
rmagn[0] = 0.f;
rmagn[1] = 1.f;
/* Loop over sizes, types */
ntestt = 0;
nerrs = 0;
nmats = 0;
i__1 = *nsizes;
for (jsize = 1; jsize <= i__1; ++jsize) {
n = nn[jsize];
n1 = max(1,n);
rmagn[2] = safmax * ulp / (real) n1;
rmagn[3] = safmin * ulpinv * n1;
if (*nsizes != 1) {
mtypes = min(26,*ntypes);
} else {
mtypes = min(27,*ntypes);
}
i__2 = mtypes;
for (jtype = 1; jtype <= i__2; ++jtype) {
if (! dotype[jtype]) {
goto L160;
}
++nmats;
ntest = 0;
/* Save ISEED in case of an error. */
for (j = 1; j <= 4; ++j) {
ioldsd[j - 1] = iseed[j];
/* L20: */
}
/* Initialize RESULT */
for (j = 1; j <= 15; ++j) {
result[j] = 0.f;
/* L30: */
}
/* Compute A and B */
/* Description of control parameters: */
/* KCLASS: =1 means w/o rotation, =2 means w/ rotation, */
/* =3 means random. */
/* KATYPE: the "type" to be passed to SLATM4 for computing A. */
/* KAZERO: the pattern of zeros on the diagonal for A: */
/* =1: ( xxx ), =2: (0, xxx ) =3: ( 0, 0, xxx, 0 ), */
/* =4: ( 0, xxx, 0, 0 ), =5: ( 0, 0, 1, xxx, 0 ), */
/* =6: ( 0, 1, 0, xxx, 0 ). (xxx means a string of */
/* non-zero entries.) */
/* KAMAGN: the magnitude of the matrix: =0: zero, =1: O(1), */
/* =2: large, =3: small. */
/* IASIGN: 1 if the diagonal elements of A are to be */
/* multiplied by a random magnitude 1 number, =2 if */
/* randomly chosen diagonal blocks are to be rotated */
/* to form 2x2 blocks. */
/* KBTYPE, KBZERO, KBMAGN, IBSIGN: the same, but for B. */
/* KTRIAN: =0: don't fill in the upper triangle, =1: do. */
/* KZ1, KZ2, KADD: used to implement KAZERO and KBZERO. */
/* RMAGN: used to implement KAMAGN and KBMAGN. */
if (mtypes > 26) {
goto L110;
}
iinfo = 0;
if (kclass[jtype - 1] < 3) {
/* Generate A (w/o rotation) */
if ((i__3 = katype[jtype - 1], abs(i__3)) == 3) {
in = ((n - 1) / 2 << 1) + 1;
if (in != n) {
slaset_("Full", &n, &n, &c_b36, &c_b36, &a[a_offset],
lda);
}
} else {
in = n;
}
slatm4_(&katype[jtype - 1], &in, &kz1[kazero[jtype - 1] - 1],
&kz2[kazero[jtype - 1] - 1], &iasign[jtype - 1], &
rmagn[kamagn[jtype - 1]], &ulp, &rmagn[ktrian[jtype -
1] * kamagn[jtype - 1]], &c__2, &iseed[1], &a[
a_offset], lda);
iadd = kadd[kazero[jtype - 1] - 1];
if (iadd > 0 && iadd <= n) {
a[iadd + iadd * a_dim1] = 1.f;
}
/* Generate B (w/o rotation) */
if ((i__3 = kbtype[jtype - 1], abs(i__3)) == 3) {
in = ((n - 1) / 2 << 1) + 1;
if (in != n) {
slaset_("Full", &n, &n, &c_b36, &c_b36, &b[b_offset],
lda);
}
} else {
in = n;
}
slatm4_(&kbtype[jtype - 1], &in, &kz1[kbzero[jtype - 1] - 1],
&kz2[kbzero[jtype - 1] - 1], &ibsign[jtype - 1], &
rmagn[kbmagn[jtype - 1]], &c_b42, &rmagn[ktrian[jtype
- 1] * kbmagn[jtype - 1]], &c__2, &iseed[1], &b[
b_offset], lda);
iadd = kadd[kbzero[jtype - 1] - 1];
if (iadd != 0 && iadd <= n) {
b[iadd + iadd * b_dim1] = 1.f;
}
if (kclass[jtype - 1] == 2 && n > 0) {
/* Include rotations */
/* Generate Q, Z as Householder transformations times */
/* a diagonal matrix. */
i__3 = n - 1;
for (jc = 1; jc <= i__3; ++jc) {
i__4 = n;
for (jr = jc; jr <= i__4; ++jr) {
q[jr + jc * q_dim1] = slarnd_(&c__3, &iseed[1]);
z__[jr + jc * z_dim1] = slarnd_(&c__3, &iseed[1]);
/* L40: */
}
i__4 = n + 1 - jc;
slarfg_(&i__4, &q[jc + jc * q_dim1], &q[jc + 1 + jc *
q_dim1], &c__1, &work[jc]);
work[(n << 1) + jc] = r_sign(&c_b42, &q[jc + jc *
q_dim1]);
q[jc + jc * q_dim1] = 1.f;
i__4 = n + 1 - jc;
slarfg_(&i__4, &z__[jc + jc * z_dim1], &z__[jc + 1 +
jc * z_dim1], &c__1, &work[n + jc]);
work[n * 3 + jc] = r_sign(&c_b42, &z__[jc + jc *
z_dim1]);
z__[jc + jc * z_dim1] = 1.f;
/* L50: */
}
q[n + n * q_dim1] = 1.f;
work[n] = 0.f;
r__1 = slarnd_(&c__2, &iseed[1]);
work[n * 3] = r_sign(&c_b42, &r__1);
z__[n + n * z_dim1] = 1.f;
work[n * 2] = 0.f;
r__1 = slarnd_(&c__2, &iseed[1]);
work[n * 4] = r_sign(&c_b42, &r__1);
/* Apply the diagonal matrices */
i__3 = n;
for (jc = 1; jc <= i__3; ++jc) {
i__4 = n;
for (jr = 1; jr <= i__4; ++jr) {
a[jr + jc * a_dim1] = work[(n << 1) + jr] * work[
n * 3 + jc] * a[jr + jc * a_dim1];
b[jr + jc * b_dim1] = work[(n << 1) + jr] * work[
n * 3 + jc] * b[jr + jc * b_dim1];
/* L60: */
}
/* L70: */
}
i__3 = n - 1;
sorm2r_("L", "N", &n, &n, &i__3, &q[q_offset], ldq, &work[
1], &a[a_offset], lda, &work[(n << 1) + 1], &
iinfo);
if (iinfo != 0) {
goto L100;
}
i__3 = n - 1;
sorm2r_("R", "T", &n, &n, &i__3, &z__[z_offset], ldq, &
work[n + 1], &a[a_offset], lda, &work[(n << 1) +
1], &iinfo);
if (iinfo != 0) {
goto L100;
}
i__3 = n - 1;
sorm2r_("L", "N", &n, &n, &i__3, &q[q_offset], ldq, &work[
1], &b[b_offset], lda, &work[(n << 1) + 1], &
iinfo);
if (iinfo != 0) {
goto L100;
}
i__3 = n - 1;
sorm2r_("R", "T", &n, &n, &i__3, &z__[z_offset], ldq, &
work[n + 1], &b[b_offset], lda, &work[(n << 1) +
1], &iinfo);
if (iinfo != 0) {
goto L100;
}
}
} else {
/* Random matrices */
i__3 = n;
for (jc = 1; jc <= i__3; ++jc) {
i__4 = n;
for (jr = 1; jr <= i__4; ++jr) {
a[jr + jc * a_dim1] = rmagn[kamagn[jtype - 1]] *
slarnd_(&c__2, &iseed[1]);
b[jr + jc * b_dim1] = rmagn[kbmagn[jtype - 1]] *
slarnd_(&c__2, &iseed[1]);
/* L80: */
}
/* L90: */
}
}
L100:
if (iinfo != 0) {
io___42.ciunit = *nounit;
s_wsfe(&io___42);
do_fio(&c__1, "Generator", (ftnlen)9);
do_fio(&c__1, (char *)&iinfo, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&n, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&jtype, (ftnlen)sizeof(integer));
do_fio(&c__4, (char *)&ioldsd[0], (ftnlen)sizeof(integer));
e_wsfe();
*info = abs(iinfo);
return 0;
}
L110:
/* Call SGEGS to compute H, T, Q, Z, alpha, and beta. */
slacpy_(" ", &n, &n, &a[a_offset], lda, &s[s_offset], lda);
slacpy_(" ", &n, &n, &b[b_offset], lda, &t[t_offset], lda);
ntest = 1;
result[1] = ulpinv;
sgegs_("V", "V", &n, &s[s_offset], lda, &t[t_offset], lda, &
alphr1[1], &alphi1[1], &beta1[1], &q[q_offset], ldq, &z__[
z_offset], ldq, &work[1], lwork, &iinfo);
if (iinfo != 0) {
io___43.ciunit = *nounit;
s_wsfe(&io___43);
do_fio(&c__1, "SGEGS", (ftnlen)5);
do_fio(&c__1, (char *)&iinfo, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&n, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&jtype, (ftnlen)sizeof(integer));
do_fio(&c__4, (char *)&ioldsd[0], (ftnlen)sizeof(integer));
e_wsfe();
*info = abs(iinfo);
goto L140;
}
ntest = 4;
/* Do tests 1--4 */
sget51_(&c__1, &n, &a[a_offset], lda, &s[s_offset], lda, &q[
q_offset], ldq, &z__[z_offset], ldq, &work[1], &result[1])
;
sget51_(&c__1, &n, &b[b_offset], lda, &t[t_offset], lda, &q[
q_offset], ldq, &z__[z_offset], ldq, &work[1], &result[2])
;
sget51_(&c__3, &n, &b[b_offset], lda, &t[t_offset], lda, &q[
q_offset], ldq, &q[q_offset], ldq, &work[1], &result[3]);
sget51_(&c__3, &n, &b[b_offset], lda, &t[t_offset], lda, &z__[
z_offset], ldq, &z__[z_offset], ldq, &work[1], &result[4])
;
/* Do test 5: compare eigenvalues with diagonals. */
/* Also check Schur form of A. */
temp1 = 0.f;
i__3 = n;
for (j = 1; j <= i__3; ++j) {
ilabad = FALSE_;
if (alphi1[j] == 0.f) {
/* Computing MAX */
r__7 = safmin, r__8 = (r__2 = alphr1[j], dabs(r__2)),
r__7 = max(r__7,r__8), r__8 = (r__3 = s[j + j *
s_dim1], dabs(r__3));
/* Computing MAX */
r__9 = safmin, r__10 = (r__5 = beta1[j], dabs(r__5)),
r__9 = max(r__9,r__10), r__10 = (r__6 = t[j + j *
t_dim1], dabs(r__6));
temp2 = ((r__1 = alphr1[j] - s[j + j * s_dim1], dabs(r__1)
) / dmax(r__7,r__8) + (r__4 = beta1[j] - t[j + j *
t_dim1], dabs(r__4)) / dmax(r__9,r__10)) / ulp;
if (j < n) {
if (s[j + 1 + j * s_dim1] != 0.f) {
ilabad = TRUE_;
}
}
if (j > 1) {
if (s[j + (j - 1) * s_dim1] != 0.f) {
ilabad = TRUE_;
}
}
} else {
if (alphi1[j] > 0.f) {
i1 = j;
} else {
i1 = j - 1;
}
if (i1 <= 0 || i1 >= n) {
ilabad = TRUE_;
} else if (i1 < n - 1) {
if (s[i1 + 2 + (i1 + 1) * s_dim1] != 0.f) {
ilabad = TRUE_;
}
} else if (i1 > 1) {
if (s[i1 + (i1 - 1) * s_dim1] != 0.f) {
ilabad = TRUE_;
}
}
if (! ilabad) {
sget53_(&s[i1 + i1 * s_dim1], lda, &t[i1 + i1 *
t_dim1], lda, &beta1[j], &alphr1[j], &alphi1[
j], &temp2, &iinfo);
if (iinfo >= 3) {
io___47.ciunit = *nounit;
s_wsfe(&io___47);
do_fio(&c__1, (char *)&iinfo, (ftnlen)sizeof(
integer));
do_fio(&c__1, (char *)&j, (ftnlen)sizeof(integer))
;
do_fio(&c__1, (char *)&n, (ftnlen)sizeof(integer))
;
do_fio(&c__1, (char *)&jtype, (ftnlen)sizeof(
integer));
do_fio(&c__4, (char *)&ioldsd[0], (ftnlen)sizeof(
integer));
e_wsfe();
*info = abs(iinfo);
}
} else {
temp2 = ulpinv;
}
}
temp1 = dmax(temp1,temp2);
if (ilabad) {
io___48.ciunit = *nounit;
s_wsfe(&io___48);
do_fio(&c__1, (char *)&j, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&n, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&jtype, (ftnlen)sizeof(integer));
do_fio(&c__4, (char *)&ioldsd[0], (ftnlen)sizeof(integer))
;
e_wsfe();
}
/* L120: */
}
result[5] = temp1;
/* Call SGEGV to compute S2, T2, VL, and VR, do tests. */
/* Eigenvalues and Eigenvectors */
slacpy_(" ", &n, &n, &a[a_offset], lda, &s2[s2_offset], lda);
slacpy_(" ", &n, &n, &b[b_offset], lda, &t2[t2_offset], lda);
ntest = 6;
result[6] = ulpinv;
sgegv_("V", "V", &n, &s2[s2_offset], lda, &t2[t2_offset], lda, &
alphr2[1], &alphi2[1], &beta2[1], &vl[vl_offset], ldq, &
vr[vr_offset], ldq, &work[1], lwork, &iinfo);
if (iinfo != 0) {
io___49.ciunit = *nounit;
s_wsfe(&io___49);
do_fio(&c__1, "SGEGV", (ftnlen)5);
do_fio(&c__1, (char *)&iinfo, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&n, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&jtype, (ftnlen)sizeof(integer));
do_fio(&c__4, (char *)&ioldsd[0], (ftnlen)sizeof(integer));
e_wsfe();
*info = abs(iinfo);
goto L140;
}
ntest = 7;
/* Do Tests 6 and 7 */
sget52_(&c_true, &n, &a[a_offset], lda, &b[b_offset], lda, &vl[
vl_offset], ldq, &alphr2[1], &alphi2[1], &beta2[1], &work[
1], dumma);
result[6] = dumma[0];
if (dumma[1] > *thrshn) {
io___51.ciunit = *nounit;
s_wsfe(&io___51);
do_fio(&c__1, "Left", (ftnlen)4);
do_fio(&c__1, "SGEGV", (ftnlen)5);
do_fio(&c__1, (char *)&dumma[1], (ftnlen)sizeof(real));
do_fio(&c__1, (char *)&n, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&jtype, (ftnlen)sizeof(integer));
do_fio(&c__4, (char *)&ioldsd[0], (ftnlen)sizeof(integer));
e_wsfe();
}
sget52_(&c_false, &n, &a[a_offset], lda, &b[b_offset], lda, &vr[
vr_offset], ldq, &alphr2[1], &alphi2[1], &beta2[1], &work[
1], dumma);
result[7] = dumma[0];
if (dumma[1] > *thresh) {
io___52.ciunit = *nounit;
s_wsfe(&io___52);
do_fio(&c__1, "Right", (ftnlen)5);
do_fio(&c__1, "SGEGV", (ftnlen)5);
do_fio(&c__1, (char *)&dumma[1], (ftnlen)sizeof(real));
do_fio(&c__1, (char *)&n, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&jtype, (ftnlen)sizeof(integer));
do_fio(&c__4, (char *)&ioldsd[0], (ftnlen)sizeof(integer));
e_wsfe();
}
/* Check form of Complex eigenvalues. */
i__3 = n;
for (j = 1; j <= i__3; ++j) {
ilabad = FALSE_;
if (alphi2[j] > 0.f) {
if (j == n) {
ilabad = TRUE_;
} else if (alphi2[j + 1] >= 0.f) {
ilabad = TRUE_;
}
} else if (alphi2[j] < 0.f) {
if (j == 1) {
ilabad = TRUE_;
} else if (alphi2[j - 1] <= 0.f) {
ilabad = TRUE_;
}
}
if (ilabad) {
io___53.ciunit = *nounit;
s_wsfe(&io___53);
do_fio(&c__1, (char *)&j, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&n, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&jtype, (ftnlen)sizeof(integer));
do_fio(&c__4, (char *)&ioldsd[0], (ftnlen)sizeof(integer))
;
e_wsfe();
}
/* L130: */
}
/* End of Loop -- Check for RESULT(j) > THRESH */
L140:
ntestt += ntest;
/* Print out tests which fail. */
i__3 = ntest;
for (jr = 1; jr <= i__3; ++jr) {
if (result[jr] >= *thresh) {
/* If this is the first test to fail, */
/* print a header to the data file. */
if (nerrs == 0) {
io___54.ciunit = *nounit;
s_wsfe(&io___54);
do_fio(&c__1, "SGG", (ftnlen)3);
e_wsfe();
/* Matrix types */
io___55.ciunit = *nounit;
s_wsfe(&io___55);
e_wsfe();
io___56.ciunit = *nounit;
s_wsfe(&io___56);
e_wsfe();
io___57.ciunit = *nounit;
s_wsfe(&io___57);
do_fio(&c__1, "Orthogonal", (ftnlen)10);
e_wsfe();
/* Tests performed */
io___58.ciunit = *nounit;
s_wsfe(&io___58);
do_fio(&c__1, "orthogonal", (ftnlen)10);
do_fio(&c__1, "'", (ftnlen)1);
do_fio(&c__1, "transpose", (ftnlen)9);
for (j = 1; j <= 5; ++j) {
do_fio(&c__1, "'", (ftnlen)1);
}
e_wsfe();
}
++nerrs;
if (result[jr] < 1e4f) {
io___59.ciunit = *nounit;
s_wsfe(&io___59);
do_fio(&c__1, (char *)&n, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&jtype, (ftnlen)sizeof(integer))
;
do_fio(&c__4, (char *)&ioldsd[0], (ftnlen)sizeof(
integer));
do_fio(&c__1, (char *)&jr, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&result[jr], (ftnlen)sizeof(
real));
e_wsfe();
} else {
io___60.ciunit = *nounit;
s_wsfe(&io___60);
do_fio(&c__1, (char *)&n, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&jtype, (ftnlen)sizeof(integer))
;
do_fio(&c__4, (char *)&ioldsd[0], (ftnlen)sizeof(
integer));
do_fio(&c__1, (char *)&jr, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&result[jr], (ftnlen)sizeof(
real));
e_wsfe();
}
}
/* L150: */
}
L160:
;
}
/* L170: */
}
/* Summary */
alasvm_("SGG", nounit, &nerrs, &ntestt, &c__0);
return 0;
/* End of SDRVGG */
} /* sdrvgg_ */
/* Subroutine */ int slaqps_(integer *m, integer *n, integer *offset, integer
*nb, integer *kb, real *a, integer *lda, integer *jpvt, real *tau,
real *vn1, real *vn2, real *auxv, real *f, integer *ldf)
{
/* -- LAPACK auxiliary 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
=======
SLAQPS computes a step of QR factorization with column pivoting
of a real M-by-N matrix A by using Blas-3. It tries to factorize
NB columns from A starting from the row OFFSET+1, and updates all
of the matrix with Blas-3 xGEMM.
In some cases, due to catastrophic cancellations, it cannot
factorize NB columns. Hence, the actual number of factorized
columns is returned in KB.
Block A(1:OFFSET,1:N) is accordingly pivoted, but not factorized.
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
OFFSET (input) INTEGER
The number of rows of A that have been factorized in
previous steps.
NB (input) INTEGER
The number of columns to factorize.
KB (output) INTEGER
The number of columns actually factorized.
A (input/output) REAL array, dimension (LDA,N)
On entry, the M-by-N matrix A.
On exit, block A(OFFSET+1:M,1:KB) is the triangular
factor obtained and block A(1:OFFSET,1:N) has been
accordingly pivoted, but no factorized.
The rest of the matrix, block A(OFFSET+1:M,KB+1:N) has
been updated.
LDA (input) INTEGER
The leading dimension of the array A. LDA >= max(1,M).
JPVT (input/output) INTEGER array, dimension (N)
JPVT(I) = K <==> Column K of the full matrix A has been
permuted into position I in AP.
TAU (output) REAL array, dimension (KB)
The scalar factors of the elementary reflectors.
VN1 (input/output) REAL array, dimension (N)
The vector with the partial column norms.
VN2 (input/output) REAL array, dimension (N)
The vector with the exact column norms.
AUXV (input/output) REAL array, dimension (NB)
Auxiliar vector.
F (input/output) REAL array, dimension (LDF,NB)
Matrix F' = L*Y'*A.
LDF (input) INTEGER
The leading dimension of the array F. LDF >= max(1,N).
Further Details
===============
Based on contributions by
G. Quintana-Orti, Depto. de Informatica, Universidad Jaime I, Spain
X. Sun, Computer Science Dept., Duke University, USA
=====================================================================
Parameter adjustments */
/* Table of constant values */
static integer c__1 = 1;
static real c_b7 = -1.f;
static real c_b8 = 1.f;
static real c_b15 = 0.f;
/* System generated locals */
integer a_dim1, a_offset, f_dim1, f_offset, i__1, i__2;
real r__1, r__2;
/* Builtin functions */
double sqrt(doublereal);
integer i_nint(real *);
/* Local variables */
static real temp, temp2;
extern doublereal snrm2_(integer *, real *, integer *);
static integer j, k;
extern /* Subroutine */ int sgemm_(char *, char *, integer *, integer *,
integer *, real *, real *, integer *, real *, integer *, real *,
real *, integer *);
static integer itemp;
extern /* Subroutine */ int sgemv_(char *, integer *, integer *, real *,
real *, integer *, real *, integer *, real *, real *, integer *), sswap_(integer *, real *, integer *, real *, integer *);
static integer rk;
extern /* Subroutine */ int slarfg_(integer *, real *, real *, integer *,
real *);
static integer lsticc;
extern integer isamax_(integer *, real *, integer *);
static integer lastrk;
static real akk;
static integer pvt;
#define a_ref(a_1,a_2) a[(a_2)*a_dim1 + a_1]
#define f_ref(a_1,a_2) f[(a_2)*f_dim1 + a_1]
a_dim1 = *lda;
a_offset = 1 + a_dim1 * 1;
a -= a_offset;
--jpvt;
--tau;
--vn1;
--vn2;
--auxv;
f_dim1 = *ldf;
f_offset = 1 + f_dim1 * 1;
f -= f_offset;
/* Function Body
Computing MIN */
i__1 = *m, i__2 = *n + *offset;
lastrk = min(i__1,i__2);
lsticc = 0;
k = 0;
/* Beginning of while loop. */
L10:
if (k < *nb && lsticc == 0) {
++k;
rk = *offset + k;
/* Determine ith pivot column and swap if necessary */
i__1 = *n - k + 1;
pvt = k - 1 + isamax_(&i__1, &vn1[k], &c__1);
if (pvt != k) {
sswap_(m, &a_ref(1, pvt), &c__1, &a_ref(1, k), &c__1);
i__1 = k - 1;
sswap_(&i__1, &f_ref(pvt, 1), ldf, &f_ref(k, 1), ldf);
itemp = jpvt[pvt];
jpvt[pvt] = jpvt[k];
jpvt[k] = itemp;
vn1[pvt] = vn1[k];
vn2[pvt] = vn2[k];
}
/* Apply previous Householder reflectors to column K:
A(RK:M,K) := A(RK:M,K) - A(RK:M,1:K-1)*F(K,1:K-1)'. */
if (k > 1) {
i__1 = *m - rk + 1;
i__2 = k - 1;
sgemv_("No transpose", &i__1, &i__2, &c_b7, &a_ref(rk, 1), lda, &
f_ref(k, 1), ldf, &c_b8, &a_ref(rk, k), &c__1)
;
}
/* Generate elementary reflector H(k). */
if (rk < *m) {
i__1 = *m - rk + 1;
slarfg_(&i__1, &a_ref(rk, k), &a_ref(rk + 1, k), &c__1, &tau[k]);
} else {
slarfg_(&c__1, &a_ref(rk, k), &a_ref(rk, k), &c__1, &tau[k]);
}
akk = a_ref(rk, k);
a_ref(rk, k) = 1.f;
/* Compute Kth column of F:
Compute F(K+1:N,K) := tau(K)*A(RK:M,K+1:N)'*A(RK:M,K). */
if (k < *n) {
i__1 = *m - rk + 1;
i__2 = *n - k;
sgemv_("Transpose", &i__1, &i__2, &tau[k], &a_ref(rk, k + 1), lda,
&a_ref(rk, k), &c__1, &c_b15, &f_ref(k + 1, k), &c__1);
}
/* Padding F(1:K,K) with zeros. */
i__1 = k;
for (j = 1; j <= i__1; ++j) {
f_ref(j, k) = 0.f;
/* L20: */
}
/* Incremental updating of F:
F(1:N,K) := F(1:N,K) - tau(K)*F(1:N,1:K-1)*A(RK:M,1:K-1)'
*A(RK:M,K). */
if (k > 1) {
i__1 = *m - rk + 1;
i__2 = k - 1;
r__1 = -tau[k];
sgemv_("Transpose", &i__1, &i__2, &r__1, &a_ref(rk, 1), lda, &
a_ref(rk, k), &c__1, &c_b15, &auxv[1], &c__1);
i__1 = k - 1;
sgemv_("No transpose", n, &i__1, &c_b8, &f_ref(1, 1), ldf, &auxv[
1], &c__1, &c_b8, &f_ref(1, k), &c__1);
}
/* Update the current row of A:
A(RK,K+1:N) := A(RK,K+1:N) - A(RK,1:K)*F(K+1:N,1:K)'. */
if (k < *n) {
i__1 = *n - k;
sgemv_("No transpose", &i__1, &k, &c_b7, &f_ref(k + 1, 1), ldf, &
a_ref(rk, 1), lda, &c_b8, &a_ref(rk, k + 1), lda);
}
/* Update partial column norms. */
if (rk < lastrk) {
i__1 = *n;
for (j = k + 1; j <= i__1; ++j) {
if (vn1[j] != 0.f) {
temp = (r__1 = a_ref(rk, j), dabs(r__1)) / vn1[j];
/* Computing MAX */
r__1 = 0.f, r__2 = (temp + 1.f) * (1.f - temp);
temp = dmax(r__1,r__2);
/* Computing 2nd power */
r__1 = vn1[j] / vn2[j];
temp2 = temp * .05f * (r__1 * r__1) + 1.f;
if (temp2 == 1.f) {
vn2[j] = (real) lsticc;
lsticc = j;
} else {
vn1[j] *= sqrt(temp);
}
}
/* L30: */
}
}
a_ref(rk, k) = akk;
/* End of while loop. */
goto L10;
}
*kb = k;
rk = *offset + *kb;
/* Apply the block reflector to the rest of the matrix:
A(OFFSET+KB+1:M,KB+1:N) := A(OFFSET+KB+1:M,KB+1:N) -
A(OFFSET+KB+1:M,1:KB)*F(KB+1:N,1:KB)'.
Computing MIN */
i__1 = *n, i__2 = *m - *offset;
if (*kb < min(i__1,i__2)) {
i__1 = *m - rk;
i__2 = *n - *kb;
sgemm_("No transpose", "Transpose", &i__1, &i__2, kb, &c_b7, &a_ref(
rk + 1, 1), lda, &f_ref(*kb + 1, 1), ldf, &c_b8, &a_ref(rk +
1, *kb + 1), lda);
}
/* Recomputation of difficult columns. */
L40:
if (lsticc > 0) {
itemp = i_nint(&vn2[lsticc]);
i__1 = *m - rk;
vn1[lsticc] = snrm2_(&i__1, &a_ref(rk + 1, lsticc), &c__1);
vn2[lsticc] = vn1[lsticc];
lsticc = itemp;
goto L40;
}
return 0;
/* End of SLAQPS */
} /* slaqps_ */
