/* Subroutine */ int dorgql_(integer *m, integer *n, integer *k, doublereal * a, integer *lda, doublereal *tau, doublereal *work, integer *lwork, integer *info) { /* -- LAPACK routine (version 2.0) -- Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd., Courant Institute, Argonne National Lab, and Rice University September 30, 1994 Purpose ======= DORGQL generates an M-by-N real matrix Q with orthonormal columns, which is defined as the last N columns of a product of K elementary reflectors of order M Q = H(k) . . . H(2) H(1) as returned by DGEQLF. Arguments ========= M (input) INTEGER The number of rows of the matrix Q. M >= 0. N (input) INTEGER The number of columns of the matrix Q. M >= N >= 0. K (input) INTEGER The number of elementary reflectors whose product defines the matrix Q. N >= K >= 0. A (input/output) DOUBLE PRECISION array, dimension (LDA,N) On entry, the (n-k+i)-th column must contain the vector which defines the elementary reflector H(i), for i = 1,2,...,k, as returned by DGEQLF in the last k columns of its array argument A. On exit, the M-by-N matrix Q. LDA (input) INTEGER The first dimension of the array A. LDA >= max(1,M). TAU (input) DOUBLE PRECISION array, dimension (K) TAU(i) must contain the scalar factor of the elementary reflector H(i), as returned by DGEQLF. WORK (workspace/output) DOUBLE PRECISION array, dimension (LWORK) On exit, if INFO = 0, WORK(1) returns the optimal LWORK. LWORK (input) INTEGER The dimension of the array WORK. LWORK >= max(1,N). For optimum performance LWORK >= N*NB, where NB is the optimal blocksize. INFO (output) INTEGER = 0: successful exit < 0: if INFO = -i, the i-th argument has an illegal value ===================================================================== Test the input arguments Parameter adjustments Function Body */ /* Table of constant values */ static integer c__1 = 1; static integer c_n1 = -1; static integer c__3 = 3; static integer c__2 = 2; /* System generated locals */ integer a_dim1, a_offset, i__1, i__2, i__3, i__4; /* Local variables */ static integer i, j, l, nbmin, iinfo; extern /* Subroutine */ int dorg2l_(integer *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *); static integer ib, nb, kk; extern /* Subroutine */ int dlarfb_(char *, char *, char *, char *, integer *, integer *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *); static integer nx; extern /* Subroutine */ int dlarft_(char *, char *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), xerbla_(char *, integer *); extern integer ilaenv_(integer *, char *, char *, integer *, integer *, integer *, integer *, ftnlen, ftnlen); static integer ldwork, iws; #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 || *n > *m) { *info = -2; } else if (*k < 0 || *k > *n) { *info = -3; } else if (*lda < max(1,*m)) { *info = -5; } else if (*lwork < max(1,*n)) { *info = -8; } if (*info != 0) { i__1 = -(*info); xerbla_("DORGQL", &i__1); return 0; } /* Quick return if possible */ if (*n <= 0) { WORK(1) = 1.; return 0; } /* Determine the block size. */ nb = ilaenv_(&c__1, "DORGQL", " ", m, n, k, &c_n1, 6L, 1L); nbmin = 2; nx = 0; iws = *n; if (nb > 1 && nb < *k) { /* Determine when to cross over from blocked to unblocked code. Computing MAX */ i__1 = 0, i__2 = ilaenv_(&c__3, "DORGQL", " ", m, n, k, &c_n1, 6L, 1L) ; nx = max(i__1,i__2); if (nx < *k) { /* Determine if workspace is large enough for blocked co de. */ ldwork = *n; iws = ldwork * nb; if (*lwork < iws) { /* Not enough workspace to use optimal NB: reduc e NB and determine the minimum value of NB. */ nb = *lwork / ldwork; /* Computing MAX */ i__1 = 2, i__2 = ilaenv_(&c__2, "DORGQL", " ", m, n, k, &c_n1, 6L, 1L); nbmin = max(i__1,i__2); } } } if (nb >= nbmin && nb < *k && nx < *k) { /* Use blocked code after the first block. The last kk columns are handled by the block method. Computing MIN */ i__1 = *k, i__2 = (*k - nx + nb - 1) / nb * nb; kk = min(i__1,i__2); /* Set A(m-kk+1:m,1:n-kk) to zero. */ i__1 = *n - kk; for (j = 1; j <= *n-kk; ++j) { i__2 = *m; for (i = *m - kk + 1; i <= *m; ++i) { A(i,j) = 0.; /* L10: */ } /* L20: */ } } else { kk = 0; } /* Use unblocked code for the first or only block. */ i__1 = *m - kk; i__2 = *n - kk; i__3 = *k - kk; dorg2l_(&i__1, &i__2, &i__3, &A(1,1), lda, &TAU(1), &WORK(1), &iinfo) ; if (kk > 0) { /* Use blocked code */ i__1 = *k; i__2 = nb; for (i = *k - kk + 1; nb < 0 ? i >= *k : i <= *k; i += nb) { /* Computing MIN */ i__3 = nb, i__4 = *k - i + 1; ib = min(i__3,i__4); if (*n - *k + i > 1) { /* Form the triangular factor of the block reflec tor H = H(i+ib-1) . . . H(i+1) H(i) */ i__3 = *m - *k + i + ib - 1; dlarft_("Backward", "Columnwise", &i__3, &ib, &A(1,*n-*k+i), lda, &TAU(i), &WORK(1), &ldwork); /* Apply H to A(1:m-k+i+ib-1,1:n-k+i-1) from the left */ i__3 = *m - *k + i + ib - 1; i__4 = *n - *k + i - 1; dlarfb_("Left", "No transpose", "Backward", "Columnwise", & i__3, &i__4, &ib, &A(1,*n-*k+i), lda, &WORK(1), &ldwork, &A(1,1), lda, &WORK(ib + 1), &ldwork); } /* Apply H to rows 1:m-k+i+ib-1 of current block */ i__3 = *m - *k + i + ib - 1; dorg2l_(&i__3, &ib, &ib, &A(1,*n-*k+i), lda, & TAU(i), &WORK(1), &iinfo); /* Set rows m-k+i+ib:m of current block to zero */ i__3 = *n - *k + i + ib - 1; for (j = *n - *k + i; j <= *n-*k+i+ib-1; ++j) { i__4 = *m; for (l = *m - *k + i + ib; l <= *m; ++l) { A(l,j) = 0.; /* L30: */ } /* L40: */ } /* L50: */ } } WORK(1) = (doublereal) iws; return 0; /* End of DORGQL */ } /* dorgql_ */
/* Subroutine */ int dgerqf_(integer *m, integer *n, doublereal *a, integer * lda, doublereal *tau, doublereal *work, integer *lwork, integer *info) { /* System generated locals */ integer a_dim1, a_offset, i__1, i__2, i__3, i__4; /* Local variables */ integer i__, k, ib, nb, ki, kk, mu, nu, nx, iws, nbmin, iinfo; extern /* Subroutine */ int dgerq2_(integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), dlarfb_(char *, char *, char *, char *, integer *, integer *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *), dlarft_(char *, char *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), xerbla_(char *, integer *); extern integer ilaenv_(integer *, char *, char *, integer *, integer *, integer *, integer *); integer ldwork, lwkopt; logical lquery; /* -- LAPACK computational routine (version 3.4.0) -- */ /* -- LAPACK is a software package provided by Univ. of Tennessee, -- */ /* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- */ /* November 2011 */ /* .. Scalar Arguments .. */ /* .. */ /* .. Array Arguments .. */ /* .. */ /* ===================================================================== */ /* .. 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; --tau; --work; /* Function Body */ *info = 0; lquery = *lwork == -1; if (*m < 0) { *info = -1; } else if (*n < 0) { *info = -2; } else if (*lda < max(1,*m)) { *info = -4; } if (*info == 0) { k = min(*m,*n); if (k == 0) { lwkopt = 1; } else { nb = ilaenv_(&c__1, "DGERQF", " ", m, n, &c_n1, &c_n1); lwkopt = *m * nb; } work[1] = (doublereal) lwkopt; if (*lwork < max(1,*m) && ! lquery) { *info = -7; } } if (*info != 0) { i__1 = -(*info); xerbla_("DGERQF", &i__1); return 0; } else if (lquery) { return 0; } /* Quick return if possible */ if (k == 0) { return 0; } nbmin = 2; nx = 1; iws = *m; if (nb > 1 && nb < k) { /* Determine when to cross over from blocked to unblocked code. */ /* Computing MAX */ i__1 = 0; i__2 = ilaenv_(&c__3, "DGERQF", " ", m, n, &c_n1, &c_n1); // , expr subst nx = max(i__1,i__2); if (nx < k) { /* Determine if workspace is large enough for blocked code. */ ldwork = *m; iws = ldwork * nb; if (*lwork < iws) { /* Not enough workspace to use optimal NB: reduce NB and */ /* determine the minimum value of NB. */ nb = *lwork / ldwork; /* Computing MAX */ i__1 = 2; i__2 = ilaenv_(&c__2, "DGERQF", " ", m, n, &c_n1, & c_n1); // , expr subst nbmin = max(i__1,i__2); } } } if (nb >= nbmin && nb < k && nx < k) { /* Use blocked code initially. */ /* The last kk rows are handled by the block method. */ ki = (k - nx - 1) / nb * nb; /* Computing MIN */ i__1 = k; i__2 = ki + nb; // , expr subst kk = min(i__1,i__2); i__1 = k - kk + 1; i__2 = -nb; for (i__ = k - kk + ki + 1; i__2 < 0 ? i__ >= i__1 : i__ <= i__1; i__ += i__2) { /* Computing MIN */ i__3 = k - i__ + 1; ib = min(i__3,nb); /* Compute the RQ factorization of the current block */ /* A(m-k+i:m-k+i+ib-1,1:n-k+i+ib-1) */ i__3 = *n - k + i__ + ib - 1; dgerq2_(&ib, &i__3, &a[*m - k + i__ + a_dim1], lda, &tau[i__], & work[1], &iinfo); if (*m - k + i__ > 1) { /* Form the triangular factor of the block reflector */ /* H = H(i+ib-1) . . . H(i+1) H(i) */ i__3 = *n - k + i__ + ib - 1; dlarft_("Backward", "Rowwise", &i__3, &ib, &a[*m - k + i__ + a_dim1], lda, &tau[i__], &work[1], &ldwork); /* Apply H to A(1:m-k+i-1,1:n-k+i+ib-1) from the right */ i__3 = *m - k + i__ - 1; i__4 = *n - k + i__ + ib - 1; dlarfb_("Right", "No transpose", "Backward", "Rowwise", &i__3, &i__4, &ib, &a[*m - k + i__ + a_dim1], lda, &work[1], &ldwork, &a[a_offset], lda, &work[ib + 1], &ldwork); } /* L10: */ } mu = *m - k + i__ + nb - 1; nu = *n - k + i__ + nb - 1; } else { mu = *m; nu = *n; } /* Use unblocked code to factor the last or only block */ if (mu > 0 && nu > 0) { dgerq2_(&mu, &nu, &a[a_offset], lda, &tau[1], &work[1], &iinfo); } work[1] = (doublereal) iws; return 0; /* End of DGERQF */ }
int dormrq_(char *side, char *trans, int *m, int *n, int *k, double *a, int *lda, double *tau, double * c__, int *ldc, double *work, int *lwork, int *info) { /* System generated locals */ address a__1[2]; int a_dim1, a_offset, c_dim1, c_offset, i__1, i__2, i__3[2], i__4, i__5; char ch__1[2]; /* Builtin functions */ int s_cat(char *, char **, int *, int *, unsigned long); /* Local variables */ int i__; double t[4160] /* was [65][64] */; int i1, i2, i3, ib, nb, mi, ni, nq, nw, iws; int left; extern int lsame_(char *, char *); int nbmin, iinfo; extern int dormr2_(char *, char *, int *, int *, int *, double *, int *, double *, double *, int *, double *, int *), dlarfb_(char *, char *, char *, char *, int *, int *, int *, double *, int *, double *, int *, double *, int *, double *, int *), dlarft_(char *, char *, int *, int *, double *, int *, double *, double *, int *), xerbla_(char *, int *); extern int ilaenv_(int *, char *, char *, int *, int *, int *, int *); int notran; int ldwork; char transt[1]; int lwkopt; int lquery; /* -- LAPACK routine (version 3.2) -- */ /* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */ /* November 2006 */ /* .. Scalar Arguments .. */ /* .. */ /* .. Array Arguments .. */ /* .. */ /* Purpose */ /* ======= */ /* DORMRQ overwrites the general float M-by-N matrix C with */ /* SIDE = 'L' SIDE = 'R' */ /* TRANS = 'N': Q * C C * Q */ /* TRANS = 'T': Q**T * C C * Q**T */ /* where Q is a float orthogonal matrix defined as the product of k */ /* elementary reflectors */ /* Q = H(1) H(2) . . . H(k) */ /* as returned by DGERQF. Q is of order M if SIDE = 'L' and of order N */ /* if SIDE = 'R'. */ /* Arguments */ /* ========= */ /* SIDE (input) CHARACTER*1 */ /* = 'L': apply Q or Q**T from the Left; */ /* = 'R': apply Q or Q**T from the Right. */ /* TRANS (input) CHARACTER*1 */ /* = 'N': No transpose, apply Q; */ /* = 'T': Transpose, apply Q**T. */ /* M (input) INTEGER */ /* The number of rows of the matrix C. M >= 0. */ /* N (input) INTEGER */ /* The number of columns of the matrix C. N >= 0. */ /* K (input) INTEGER */ /* The number of elementary reflectors whose product defines */ /* the matrix Q. */ /* If SIDE = 'L', M >= K >= 0; */ /* if SIDE = 'R', N >= K >= 0. */ /* A (input) DOUBLE PRECISION array, dimension */ /* (LDA,M) if SIDE = 'L', */ /* (LDA,N) if SIDE = 'R' */ /* The i-th row must contain the vector which defines the */ /* elementary reflector H(i), for i = 1,2,...,k, as returned by */ /* DGERQF in the last k rows of its array argument A. */ /* A is modified by the routine but restored on exit. */ /* LDA (input) INTEGER */ /* The leading dimension of the array A. LDA >= MAX(1,K). */ /* TAU (input) DOUBLE PRECISION array, dimension (K) */ /* TAU(i) must contain the scalar factor of the elementary */ /* reflector H(i), as returned by DGERQF. */ /* C (input/output) DOUBLE PRECISION array, dimension (LDC,N) */ /* On entry, the M-by-N matrix C. */ /* On exit, C is overwritten by Q*C or Q**T*C or C*Q**T or C*Q. */ /* LDC (input) INTEGER */ /* The leading dimension of the array C. LDC >= MAX(1,M). */ /* WORK (workspace/output) DOUBLE PRECISION array, dimension (MAX(1,LWORK)) */ /* On exit, if INFO = 0, WORK(1) returns the optimal LWORK. */ /* LWORK (input) INTEGER */ /* The dimension of the array WORK. */ /* If SIDE = 'L', LWORK >= MAX(1,N); */ /* if SIDE = 'R', LWORK >= MAX(1,M). */ /* For optimum performance LWORK >= N*NB if SIDE = 'L', and */ /* LWORK >= M*NB if SIDE = 'R', where NB is the optimal */ /* blocksize. */ /* If LWORK = -1, then a workspace query is assumed; the routine */ /* only calculates the optimal size of the WORK array, returns */ /* this value as the first entry of the WORK array, and no error */ /* message related to LWORK is issued by XERBLA. */ /* INFO (output) INTEGER */ /* = 0: successful exit */ /* < 0: if INFO = -i, the i-th argument had an illegal value */ /* ===================================================================== */ /* .. Parameters .. */ /* .. */ /* .. Local Scalars .. */ /* .. */ /* .. Local Arrays .. */ /* .. */ /* .. External Functions .. */ /* .. */ /* .. External Subroutines .. */ /* .. */ /* .. Intrinsic Functions .. */ /* .. */ /* .. Executable Statements .. */ /* Test the input arguments */ /* Parameter adjustments */ a_dim1 = *lda; a_offset = 1 + a_dim1; a -= a_offset; --tau; c_dim1 = *ldc; c_offset = 1 + c_dim1; c__ -= c_offset; --work; /* Function Body */ *info = 0; left = lsame_(side, "L"); notran = lsame_(trans, "N"); lquery = *lwork == -1; /* NQ is the order of Q and NW is the minimum dimension of WORK */ if (left) { nq = *m; nw = MAX(1,*n); } else { nq = *n; nw = MAX(1,*m); } if (! left && ! lsame_(side, "R")) { *info = -1; } else if (! notran && ! lsame_(trans, "T")) { *info = -2; } else if (*m < 0) { *info = -3; } else if (*n < 0) { *info = -4; } else if (*k < 0 || *k > nq) { *info = -5; } else if (*lda < MAX(1,*k)) { *info = -7; } else if (*ldc < MAX(1,*m)) { *info = -10; } if (*info == 0) { if (*m == 0 || *n == 0) { lwkopt = 1; } else { /* Determine the block size. NB may be at most NBMAX, where */ /* NBMAX is used to define the local array T. */ /* Computing MIN */ /* Writing concatenation */ i__3[0] = 1, a__1[0] = side; i__3[1] = 1, a__1[1] = trans; s_cat(ch__1, a__1, i__3, &c__2, (unsigned long)2); i__1 = 64, i__2 = ilaenv_(&c__1, "DORMRQ", ch__1, m, n, k, &c_n1); nb = MIN(i__1,i__2); lwkopt = nw * nb; } work[1] = (double) lwkopt; if (*lwork < nw && ! lquery) { *info = -12; } } if (*info != 0) { i__1 = -(*info); xerbla_("DORMRQ", &i__1); return 0; } else if (lquery) { return 0; } /* Quick return if possible */ if (*m == 0 || *n == 0) { return 0; } nbmin = 2; ldwork = nw; if (nb > 1 && nb < *k) { iws = nw * nb; if (*lwork < iws) { nb = *lwork / ldwork; /* Computing MAX */ /* Writing concatenation */ i__3[0] = 1, a__1[0] = side; i__3[1] = 1, a__1[1] = trans; s_cat(ch__1, a__1, i__3, &c__2, (unsigned long)2); i__1 = 2, i__2 = ilaenv_(&c__2, "DORMRQ", ch__1, m, n, k, &c_n1); nbmin = MAX(i__1,i__2); } } else { iws = nw; } if (nb < nbmin || nb >= *k) { /* Use unblocked code */ dormr2_(side, trans, m, n, k, &a[a_offset], lda, &tau[1], &c__[ c_offset], ldc, &work[1], &iinfo); } else { /* Use blocked code */ if (left && ! notran || ! left && notran) { i1 = 1; i2 = *k; i3 = nb; } else { i1 = (*k - 1) / nb * nb + 1; i2 = 1; i3 = -nb; } if (left) { ni = *n; } else { mi = *m; } if (notran) { *(unsigned char *)transt = 'T'; } else { *(unsigned char *)transt = 'N'; } i__1 = i2; i__2 = i3; for (i__ = i1; i__2 < 0 ? i__ >= i__1 : i__ <= i__1; i__ += i__2) { /* Computing MIN */ i__4 = nb, i__5 = *k - i__ + 1; ib = MIN(i__4,i__5); /* Form the triangular factor of the block reflector */ /* H = H(i+ib-1) . . . H(i+1) H(i) */ i__4 = nq - *k + i__ + ib - 1; dlarft_("Backward", "Rowwise", &i__4, &ib, &a[i__ + a_dim1], lda, &tau[i__], t, &c__65); if (left) { /* H or H' is applied to C(1:m-k+i+ib-1,1:n) */ mi = *m - *k + i__ + ib - 1; } else { /* H or H' is applied to C(1:m,1:n-k+i+ib-1) */ ni = *n - *k + i__ + ib - 1; } /* Apply H or H' */ dlarfb_(side, transt, "Backward", "Rowwise", &mi, &ni, &ib, &a[ i__ + a_dim1], lda, t, &c__65, &c__[c_offset], ldc, &work[ 1], &ldwork); /* L10: */ } } work[1] = (double) lwkopt; return 0; /* End of DORMRQ */ } /* dormrq_ */
int dgehrd_(int *n, int *ilo, int *ihi, double *a, int *lda, double *tau, double *work, int *lwork, int *info) { /* System generated locals */ int a_dim1, a_offset, i__1, i__2, i__3, i__4; /* Local variables */ int i__, j; double t[4160] /* was [65][64] */; int ib; double ei; int nb, nh, nx, iws; extern int dgemm_(char *, char *, int *, int *, int *, double *, double *, int *, double *, int *, double *, double *, int *); int nbmin, iinfo; extern int dtrmm_(char *, char *, char *, char *, int *, int *, double *, double *, int *, double *, int *), daxpy_( int *, double *, double *, int *, double *, int *), dgehd2_(int *, int *, int *, double *, int *, double *, double *, int *), dlahr2_( int *, int *, int *, double *, int *, double *, double *, int *, double *, int *), dlarfb_(char *, char *, char *, char *, int *, int *, int *, double *, int *, double *, int *, double *, int *, double *, int *), xerbla_(char *, int *); extern int ilaenv_(int *, char *, char *, int *, int *, int *, int *); int ldwork, lwkopt; int lquery; /* -- LAPACK routine (version 3.2) -- */ /* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */ /* November 2006 */ /* .. Scalar Arguments .. */ /* .. */ /* .. Array Arguments .. */ /* .. */ /* Purpose */ /* ======= */ /* DGEHRD reduces a float 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 DGEBAL; otherwise they should be */ /* set to 1 and N respectively. See Further Details. */ /* 1 <= ILO <= IHI <= N, if N > 0; ILO=1 and IHI=0, if N=0. */ /* A (input/output) DOUBLE PRECISION 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) DOUBLE PRECISION array, dimension (N-1) */ /* The scalar factors of the elementary reflectors (see Further */ /* Details). Elements 1:ILO-1 and IHI:N-1 of TAU are set to */ /* zero. */ /* WORK (workspace/output) DOUBLE PRECISION array, dimension (LWORK) */ /* On exit, if INFO = 0, WORK(1) returns the optimal LWORK. */ /* LWORK (input) INTEGER */ /* The length of the array WORK. LWORK >= MAX(1,N). */ /* For optimum performance LWORK >= N*NB, where NB is the */ /* optimal blocksize. */ /* If LWORK = -1, then a workspace query is assumed; the routine */ /* only calculates the optimal size of the WORK array, returns */ /* this value as the first entry of the WORK array, and no error */ /* message related to LWORK is issued by XERBLA. */ /* INFO (output) INTEGER */ /* = 0: successful exit */ /* < 0: if INFO = -i, the i-th argument had an illegal value. */ /* 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 float scalar, and v is a float 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). */ /* This file is a slight modification of LAPACK-3.0's DGEHRD */ /* subroutine incorporating improvements proposed by Quintana-Orti and */ /* Van de Geijn (2005). */ /* ===================================================================== */ /* .. Parameters .. */ /* .. */ /* .. Local Scalars .. */ /* .. */ /* .. Local Arrays .. */ /* .. */ /* .. External Subroutines .. */ /* .. */ /* .. Intrinsic Functions .. */ /* .. */ /* .. External 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; /* Computing MIN */ i__1 = 64, i__2 = ilaenv_(&c__1, "DGEHRD", " ", n, ilo, ihi, &c_n1); nb = MIN(i__1,i__2); lwkopt = *n * nb; work[1] = (double) lwkopt; lquery = *lwork == -1; 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; } else if (*lwork < MAX(1,*n) && ! lquery) { *info = -8; } if (*info != 0) { i__1 = -(*info); xerbla_("DGEHRD", &i__1); return 0; } else if (lquery) { return 0; } /* Set elements 1:ILO-1 and IHI:N-1 of TAU to zero */ i__1 = *ilo - 1; for (i__ = 1; i__ <= i__1; ++i__) { tau[i__] = 0.; /* L10: */ } i__1 = *n - 1; for (i__ = MAX(1,*ihi); i__ <= i__1; ++i__) { tau[i__] = 0.; /* L20: */ } /* Quick return if possible */ nh = *ihi - *ilo + 1; if (nh <= 1) { work[1] = 1.; return 0; } /* Determine the block size */ /* Computing MIN */ i__1 = 64, i__2 = ilaenv_(&c__1, "DGEHRD", " ", n, ilo, ihi, &c_n1); nb = MIN(i__1,i__2); nbmin = 2; iws = 1; if (nb > 1 && nb < nh) { /* Determine when to cross over from blocked to unblocked code */ /* (last block is always handled by unblocked code) */ /* Computing MAX */ i__1 = nb, i__2 = ilaenv_(&c__3, "DGEHRD", " ", n, ilo, ihi, &c_n1); nx = MAX(i__1,i__2); if (nx < nh) { /* Determine if workspace is large enough for blocked code */ iws = *n * nb; if (*lwork < iws) { /* Not enough workspace to use optimal NB: determine the */ /* minimum value of NB, and reduce NB or force use of */ /* unblocked code */ /* Computing MAX */ i__1 = 2, i__2 = ilaenv_(&c__2, "DGEHRD", " ", n, ilo, ihi, & c_n1); nbmin = MAX(i__1,i__2); if (*lwork >= *n * nbmin) { nb = *lwork / *n; } else { nb = 1; } } } } ldwork = *n; if (nb < nbmin || nb >= nh) { /* Use unblocked code below */ i__ = *ilo; } else { /* Use blocked code */ i__1 = *ihi - 1 - nx; i__2 = nb; for (i__ = *ilo; i__2 < 0 ? i__ >= i__1 : i__ <= i__1; i__ += i__2) { /* Computing MIN */ i__3 = nb, i__4 = *ihi - i__; ib = MIN(i__3,i__4); /* Reduce columns i:i+ib-1 to Hessenberg form, returning the */ /* matrices V and T of the block reflector H = I - V*T*V' */ /* which performs the reduction, and also the matrix Y = A*V*T */ dlahr2_(ihi, &i__, &ib, &a[i__ * a_dim1 + 1], lda, &tau[i__], t, & c__65, &work[1], &ldwork); /* Apply the block reflector H to A(1:ihi,i+ib:ihi) from the */ /* right, computing A := A - Y * V'. V(i+ib,ib-1) must be set */ /* to 1 */ ei = a[i__ + ib + (i__ + ib - 1) * a_dim1]; a[i__ + ib + (i__ + ib - 1) * a_dim1] = 1.; i__3 = *ihi - i__ - ib + 1; dgemm_("No transpose", "Transpose", ihi, &i__3, &ib, &c_b25, & work[1], &ldwork, &a[i__ + ib + i__ * a_dim1], lda, & c_b26, &a[(i__ + ib) * a_dim1 + 1], lda); a[i__ + ib + (i__ + ib - 1) * a_dim1] = ei; /* Apply the block reflector H to A(1:i,i+1:i+ib-1) from the */ /* right */ i__3 = ib - 1; dtrmm_("Right", "Lower", "Transpose", "Unit", &i__, &i__3, &c_b26, &a[i__ + 1 + i__ * a_dim1], lda, &work[1], &ldwork); i__3 = ib - 2; for (j = 0; j <= i__3; ++j) { daxpy_(&i__, &c_b25, &work[ldwork * j + 1], &c__1, &a[(i__ + j + 1) * a_dim1 + 1], &c__1); /* L30: */ } /* Apply the block reflector H to A(i+1:ihi,i+ib:n) from the */ /* left */ i__3 = *ihi - i__; i__4 = *n - i__ - ib + 1; dlarfb_("Left", "Transpose", "Forward", "Columnwise", &i__3, & i__4, &ib, &a[i__ + 1 + i__ * a_dim1], lda, t, &c__65, &a[ i__ + 1 + (i__ + ib) * a_dim1], lda, &work[1], &ldwork); /* L40: */ } } /* Use unblocked code to reduce the rest of the matrix */ dgehd2_(n, &i__, ihi, &a[a_offset], lda, &tau[1], &work[1], &iinfo); work[1] = (double) iws; return 0; /* End of DGEHRD */ } /* dgehrd_ */
/* Subroutine */ int dorglq_(integer *m, integer *n, integer *k, doublereal * a, integer *lda, doublereal *tau, doublereal *work, integer *lwork, integer *info) { /* System generated locals */ integer a_dim1, a_offset, i__1, i__2, i__3; /* Local variables */ integer i__, j, l, ib, nb, ki, kk, nx, iws, nbmin, iinfo; extern /* Subroutine */ int dorgl2_(integer *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), dlarfb_(char *, char *, char *, char *, integer *, integer *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *), dlarft_(char *, char *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), xerbla_(char *, integer *); extern integer ilaenv_(integer *, char *, char *, integer *, integer *, integer *, integer *); integer ldwork, lwkopt; logical lquery; /* -- LAPACK routine (version 3.1) -- */ /* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */ /* November 2006 */ /* .. Scalar Arguments .. */ /* .. */ /* .. Array Arguments .. */ /* .. */ /* Purpose */ /* ======= */ /* DORGLQ generates an M-by-N real matrix Q with orthonormal rows, */ /* which is defined as the first M rows of a product of K elementary */ /* reflectors of order N */ /* Q = H(k) . . . H(2) H(1) */ /* as returned by DGELQF. */ /* Arguments */ /* ========= */ /* M (input) INTEGER */ /* The number of rows of the matrix Q. M >= 0. */ /* N (input) INTEGER */ /* The number of columns of the matrix Q. N >= M. */ /* K (input) INTEGER */ /* The number of elementary reflectors whose product defines the */ /* matrix Q. M >= K >= 0. */ /* A (input/output) DOUBLE PRECISION array, dimension (LDA,N) */ /* On entry, the i-th row must contain the vector which defines */ /* the elementary reflector H(i), for i = 1,2,...,k, as returned */ /* by DGELQF in the first k rows of its array argument A. */ /* On exit, the M-by-N matrix Q. */ /* LDA (input) INTEGER */ /* The first dimension of the array A. LDA >= max(1,M). */ /* TAU (input) DOUBLE PRECISION array, dimension (K) */ /* TAU(i) must contain the scalar factor of the elementary */ /* reflector H(i), as returned by DGELQF. */ /* WORK (workspace/output) DOUBLE PRECISION array, dimension (MAX(1,LWORK)) */ /* On exit, if INFO = 0, WORK(1) returns the optimal LWORK. */ /* LWORK (input) INTEGER */ /* The dimension of the array WORK. LWORK >= max(1,M). */ /* For optimum performance LWORK >= M*NB, where NB is */ /* the optimal blocksize. */ /* If LWORK = -1, then a workspace query is assumed; the routine */ /* only calculates the optimal size of the WORK array, returns */ /* this value as the first entry of the WORK array, and no error */ /* message related to LWORK is issued by XERBLA. */ /* INFO (output) INTEGER */ /* = 0: successful exit */ /* < 0: if INFO = -i, the i-th argument has an illegal value */ /* ===================================================================== */ /* .. 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; --tau; --work; /* Function Body */ *info = 0; nb = ilaenv_(&c__1, "DORGLQ", " ", m, n, k, &c_n1); lwkopt = max(1,*m) * nb; work[1] = (doublereal) lwkopt; lquery = *lwork == -1; if (*m < 0) { *info = -1; } else if (*n < *m) { *info = -2; } else if (*k < 0 || *k > *m) { *info = -3; } else if (*lda < max(1,*m)) { *info = -5; } else if (*lwork < max(1,*m) && ! lquery) { *info = -8; } if (*info != 0) { i__1 = -(*info); xerbla_("DORGLQ", &i__1); return 0; } else if (lquery) { return 0; } /* Quick return if possible */ if (*m <= 0) { work[1] = 1.; return 0; } nbmin = 2; nx = 0; iws = *m; if (nb > 1 && nb < *k) { /* Determine when to cross over from blocked to unblocked code. */ /* Computing MAX */ i__1 = 0, i__2 = ilaenv_(&c__3, "DORGLQ", " ", m, n, k, &c_n1); nx = max(i__1,i__2); if (nx < *k) { /* Determine if workspace is large enough for blocked code. */ ldwork = *m; iws = ldwork * nb; if (*lwork < iws) { /* Not enough workspace to use optimal NB: reduce NB and */ /* determine the minimum value of NB. */ nb = *lwork / ldwork; /* Computing MAX */ i__1 = 2, i__2 = ilaenv_(&c__2, "DORGLQ", " ", m, n, k, &c_n1); nbmin = max(i__1,i__2); } } } if (nb >= nbmin && nb < *k && nx < *k) { /* Use blocked code after the last block. */ /* The first kk rows are handled by the block method. */ ki = (*k - nx - 1) / nb * nb; /* Computing MIN */ i__1 = *k, i__2 = ki + nb; kk = min(i__1,i__2); /* Set A(kk+1:m,1:kk) to zero. */ i__1 = kk; for (j = 1; j <= i__1; ++j) { i__2 = *m; for (i__ = kk + 1; i__ <= i__2; ++i__) { a[i__ + j * a_dim1] = 0.; /* L10: */ } /* L20: */ } } else { kk = 0; } /* Use unblocked code for the last or only block. */ if (kk < *m) { i__1 = *m - kk; i__2 = *n - kk; i__3 = *k - kk; dorgl2_(&i__1, &i__2, &i__3, &a[kk + 1 + (kk + 1) * a_dim1], lda, & tau[kk + 1], &work[1], &iinfo); } if (kk > 0) { /* Use blocked code */ i__1 = -nb; for (i__ = ki + 1; i__1 < 0 ? i__ >= 1 : i__ <= 1; i__ += i__1) { /* Computing MIN */ i__2 = nb, i__3 = *k - i__ + 1; ib = min(i__2,i__3); if (i__ + ib <= *m) { /* Form the triangular factor of the block reflector */ /* H = H(i) H(i+1) . . . H(i+ib-1) */ i__2 = *n - i__ + 1; dlarft_("Forward", "Rowwise", &i__2, &ib, &a[i__ + i__ * a_dim1], lda, &tau[i__], &work[1], &ldwork); /* Apply H' to A(i+ib:m,i:n) from the right */ i__2 = *m - i__ - ib + 1; i__3 = *n - i__ + 1; dlarfb_("Right", "Transpose", "Forward", "Rowwise", &i__2, & i__3, &ib, &a[i__ + i__ * a_dim1], lda, &work[1], & ldwork, &a[i__ + ib + i__ * a_dim1], lda, &work[ib + 1], &ldwork); } /* Apply H' to columns i:n of current block */ i__2 = *n - i__ + 1; dorgl2_(&ib, &i__2, &ib, &a[i__ + i__ * a_dim1], lda, &tau[i__], & work[1], &iinfo); /* Set columns 1:i-1 of current block to zero */ i__2 = i__ - 1; for (j = 1; j <= i__2; ++j) { i__3 = i__ + ib - 1; for (l = i__; l <= i__3; ++l) { a[l + j * a_dim1] = 0.; /* L30: */ } /* L40: */ } /* L50: */ } } work[1] = (doublereal) iws; return 0; /* End of DORGLQ */ } /* dorglq_ */
/* Subroutine */ int dorglq_fla(integer *m, integer *n, integer *k, doublereal * a, integer *lda, doublereal *tau, doublereal *work, integer *lwork, integer *info) { /* System generated locals */ integer a_dim1, a_offset, i__1, i__2, i__3; /* Local variables */ integer i__, j, l, ib, nb, ki, kk, nx, iws, nbmin, iinfo; extern /* Subroutine */ int dorgl2_fla(integer *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), dlarfb_(char *, char *, char *, char *, integer *, integer *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *), dlarft_(char *, char *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), xerbla_(char *, integer *); extern integer ilaenv_(integer *, char *, char *, integer *, integer *, integer *, integer *); integer ldwork, lwkopt; logical lquery; /* -- LAPACK computational routine (version 3.4.0) -- */ /* -- LAPACK is a software package provided by Univ. of Tennessee, -- */ /* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- */ /* November 2011 */ /* .. Scalar Arguments .. */ /* .. */ /* .. Array Arguments .. */ /* .. */ /* ===================================================================== */ /* .. 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; --tau; --work; /* Function Body */ *info = 0; nb = ilaenv_(&c__1, "DORGLQ", " ", m, n, k, &c_n1); lwkopt = max(1,*m) * nb; work[1] = (doublereal) lwkopt; lquery = *lwork == -1; if (*m < 0) { *info = -1; } else if (*n < *m) { *info = -2; } else if (*k < 0 || *k > *m) { *info = -3; } else if (*lda < max(1,*m)) { *info = -5; } else if (*lwork < max(1,*m) && ! lquery) { *info = -8; } if (*info != 0) { i__1 = -(*info); xerbla_("DORGLQ", &i__1); return 0; } else if (lquery) { return 0; } /* Quick return if possible */ if (*m <= 0) { work[1] = 1.; return 0; } nbmin = 2; nx = 0; iws = *m; if (nb > 1 && nb < *k) { /* Determine when to cross over from blocked to unblocked code. */ /* Computing MAX */ i__1 = 0; i__2 = ilaenv_(&c__3, "DORGLQ", " ", m, n, k, &c_n1); // , expr subst nx = max(i__1,i__2); if (nx < *k) { /* Determine if workspace is large enough for blocked code. */ ldwork = *m; iws = ldwork * nb; if (*lwork < iws) { /* Not enough workspace to use optimal NB: reduce NB and */ /* determine the minimum value of NB. */ nb = *lwork / ldwork; /* Computing MAX */ i__1 = 2; i__2 = ilaenv_(&c__2, "DORGLQ", " ", m, n, k, &c_n1); // , expr subst nbmin = max(i__1,i__2); } } } if (nb >= nbmin && nb < *k && nx < *k) { /* Use blocked code after the last block. */ /* The first kk rows are handled by the block method. */ ki = (*k - nx - 1) / nb * nb; /* Computing MIN */ i__1 = *k; i__2 = ki + nb; // , expr subst kk = min(i__1,i__2); /* Set A(kk+1:m,1:kk) to zero. */ i__1 = kk; for (j = 1; j <= i__1; ++j) { i__2 = *m; for (i__ = kk + 1; i__ <= i__2; ++i__) { a[i__ + j * a_dim1] = 0.; /* L10: */ } /* L20: */ } } else { kk = 0; } /* Use unblocked code for the last or only block. */ if (kk < *m) { i__1 = *m - kk; i__2 = *n - kk; i__3 = *k - kk; dorgl2_fla(&i__1, &i__2, &i__3, &a[kk + 1 + (kk + 1) * a_dim1], lda, & tau[kk + 1], &work[1], &iinfo); } if (kk > 0) { /* Use blocked code */ i__1 = -nb; for (i__ = ki + 1; i__1 < 0 ? i__ >= 1 : i__ <= 1; i__ += i__1) { /* Computing MIN */ i__2 = nb; i__3 = *k - i__ + 1; // , expr subst ib = min(i__2,i__3); if (i__ + ib <= *m) { /* Form the triangular factor of the block reflector */ /* H = H(i) H(i+1) . . . H(i+ib-1) */ i__2 = *n - i__ + 1; dlarft_("Forward", "Rowwise", &i__2, &ib, &a[i__ + i__ * a_dim1], lda, &tau[i__], &work[1], &ldwork); /* Apply H**T to A(i+ib:m,i:n) from the right */ i__2 = *m - i__ - ib + 1; i__3 = *n - i__ + 1; dlarfb_("Right", "Transpose", "Forward", "Rowwise", &i__2, & i__3, &ib, &a[i__ + i__ * a_dim1], lda, &work[1], & ldwork, &a[i__ + ib + i__ * a_dim1], lda, &work[ib + 1], &ldwork); } /* Apply H**T to columns i:n of current block */ i__2 = *n - i__ + 1; dorgl2_fla(&ib, &i__2, &ib, &a[i__ + i__ * a_dim1], lda, &tau[i__], & work[1], &iinfo); /* Set columns 1:i-1 of current block to zero */ i__2 = i__ - 1; for (j = 1; j <= i__2; ++j) { i__3 = i__ + ib - 1; for (l = i__; l <= i__3; ++l) { a[l + j * a_dim1] = 0.; /* L30: */ } /* L40: */ } /* L50: */ } } work[1] = (doublereal) iws; return 0; /* End of DORGLQ */ }
int dgerqf_(int *m, int *n, double *a, int * lda, double *tau, double *work, int *lwork, int *info) { /* System generated locals */ int a_dim1, a_offset, i__1, i__2, i__3, i__4; /* Local variables */ int i__, k, ib, nb, ki, kk, mu, nu, nx, iws, nbmin, iinfo; extern int dgerq2_(int *, int *, double *, int *, double *, double *, int *), dlarfb_(char *, char *, char *, char *, int *, int *, int *, double *, int *, double *, int *, double *, int *, double *, int *), dlarft_(char *, char *, int *, int *, double *, int *, double *, double *, int *), xerbla_(char *, int *); extern int ilaenv_(int *, char *, char *, int *, int *, int *, int *); int ldwork, lwkopt; int lquery; /* -- LAPACK routine (version 3.2) -- */ /* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */ /* November 2006 */ /* .. Scalar Arguments .. */ /* .. */ /* .. Array Arguments .. */ /* .. */ /* Purpose */ /* ======= */ /* DGERQF computes an RQ factorization of a float 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) DOUBLE PRECISION 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 MIN(m,n) elementary */ /* reflectors (see Further Details). */ /* LDA (input) INTEGER */ /* The leading dimension of the array A. LDA >= MAX(1,M). */ /* TAU (output) DOUBLE PRECISION array, dimension (MIN(M,N)) */ /* The scalar factors of the elementary reflectors (see Further */ /* Details). */ /* WORK (workspace/output) DOUBLE PRECISION array, dimension (MAX(1,LWORK)) */ /* On exit, if INFO = 0, WORK(1) returns the optimal LWORK. */ /* LWORK (input) INTEGER */ /* The dimension of the array WORK. LWORK >= MAX(1,M). */ /* For optimum performance LWORK >= M*NB, where NB is */ /* the optimal blocksize. */ /* If LWORK = -1, then a workspace query is assumed; the routine */ /* only calculates the optimal size of the WORK array, returns */ /* this value as the first entry of the WORK array, and no error */ /* message related to LWORK is issued by XERBLA. */ /* INFO (output) INTEGER */ /* = 0: successful exit */ /* < 0: if INFO = -i, the i-th argument had an illegal value */ /* 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 float scalar, and v is a float 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). */ /* ===================================================================== */ /* .. 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; --tau; --work; /* Function Body */ *info = 0; lquery = *lwork == -1; if (*m < 0) { *info = -1; } else if (*n < 0) { *info = -2; } else if (*lda < MAX(1,*m)) { *info = -4; } if (*info == 0) { k = MIN(*m,*n); if (k == 0) { lwkopt = 1; } else { nb = ilaenv_(&c__1, "DGERQF", " ", m, n, &c_n1, &c_n1); lwkopt = *m * nb; } work[1] = (double) lwkopt; if (*lwork < MAX(1,*m) && ! lquery) { *info = -7; } } if (*info != 0) { i__1 = -(*info); xerbla_("DGERQF", &i__1); return 0; } else if (lquery) { return 0; } /* Quick return if possible */ if (k == 0) { return 0; } nbmin = 2; nx = 1; iws = *m; if (nb > 1 && nb < k) { /* Determine when to cross over from blocked to unblocked code. */ /* Computing MAX */ i__1 = 0, i__2 = ilaenv_(&c__3, "DGERQF", " ", m, n, &c_n1, &c_n1); nx = MAX(i__1,i__2); if (nx < k) { /* Determine if workspace is large enough for blocked code. */ ldwork = *m; iws = ldwork * nb; if (*lwork < iws) { /* Not enough workspace to use optimal NB: reduce NB and */ /* determine the minimum value of NB. */ nb = *lwork / ldwork; /* Computing MAX */ i__1 = 2, i__2 = ilaenv_(&c__2, "DGERQF", " ", m, n, &c_n1, & c_n1); nbmin = MAX(i__1,i__2); } } } if (nb >= nbmin && nb < k && nx < k) { /* Use blocked code initially. */ /* The last kk rows are handled by the block method. */ ki = (k - nx - 1) / nb * nb; /* Computing MIN */ i__1 = k, i__2 = ki + nb; kk = MIN(i__1,i__2); i__1 = k - kk + 1; i__2 = -nb; for (i__ = k - kk + ki + 1; i__2 < 0 ? i__ >= i__1 : i__ <= i__1; i__ += i__2) { /* Computing MIN */ i__3 = k - i__ + 1; ib = MIN(i__3,nb); /* Compute the RQ factorization of the current block */ /* A(m-k+i:m-k+i+ib-1,1:n-k+i+ib-1) */ i__3 = *n - k + i__ + ib - 1; dgerq2_(&ib, &i__3, &a[*m - k + i__ + a_dim1], lda, &tau[i__], & work[1], &iinfo); if (*m - k + i__ > 1) { /* Form the triangular factor of the block reflector */ /* H = H(i+ib-1) . . . H(i+1) H(i) */ i__3 = *n - k + i__ + ib - 1; dlarft_("Backward", "Rowwise", &i__3, &ib, &a[*m - k + i__ + a_dim1], lda, &tau[i__], &work[1], &ldwork); /* Apply H to A(1:m-k+i-1,1:n-k+i+ib-1) from the right */ i__3 = *m - k + i__ - 1; i__4 = *n - k + i__ + ib - 1; dlarfb_("Right", "No transpose", "Backward", "Rowwise", &i__3, &i__4, &ib, &a[*m - k + i__ + a_dim1], lda, &work[1], &ldwork, &a[a_offset], lda, &work[ib + 1], &ldwork); } /* L10: */ } mu = *m - k + i__ + nb - 1; nu = *n - k + i__ + nb - 1; } else { mu = *m; nu = *n; } /* Use unblocked code to factor the last or only block */ if (mu > 0 && nu > 0) { dgerq2_(&mu, &nu, &a[a_offset], lda, &tau[1], &work[1], &iinfo); } work[1] = (double) iws; return 0; /* End of DGERQF */ } /* dgerqf_ */
/* Subroutine */ int dormrq_(char *side, char *trans, integer *m, integer *n, integer *k, doublereal *a, integer *lda, doublereal *tau, doublereal * c, integer *ldc, doublereal *work, integer *lwork, integer *info) { /* -- LAPACK routine (version 2.0) -- Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd., Courant Institute, Argonne National Lab, and Rice University September 30, 1994 Purpose ======= DORMRQ overwrites the general real M-by-N matrix C with SIDE = 'L' SIDE = 'R' TRANS = 'N': Q * C C * Q TRANS = 'T': Q**T * C C * Q**T where Q is a real orthogonal matrix defined as the product of k elementary reflectors Q = H(1) H(2) . . . H(k) as returned by DGERQF. Q is of order M if SIDE = 'L' and of order N if SIDE = 'R'. Arguments ========= SIDE (input) CHARACTER*1 = 'L': apply Q or Q**T from the Left; = 'R': apply Q or Q**T from the Right. TRANS (input) CHARACTER*1 = 'N': No transpose, apply Q; = 'T': Transpose, apply Q**T. M (input) INTEGER The number of rows of the matrix C. M >= 0. N (input) INTEGER The number of columns of the matrix C. N >= 0. K (input) INTEGER The number of elementary reflectors whose product defines the matrix Q. If SIDE = 'L', M >= K >= 0; if SIDE = 'R', N >= K >= 0. A (input) DOUBLE PRECISION array, dimension (LDA,M) if SIDE = 'L', (LDA,N) if SIDE = 'R' The i-th row must contain the vector which defines the elementary reflector H(i), for i = 1,2,...,k, as returned by DGERQF in the last k rows of its array argument A. A is modified by the routine but restored on exit. LDA (input) INTEGER The leading dimension of the array A. LDA >= max(1,K). TAU (input) DOUBLE PRECISION array, dimension (K) TAU(i) must contain the scalar factor of the elementary reflector H(i), as returned by DGERQF. C (input/output) DOUBLE PRECISION array, dimension (LDC,N) On entry, the M-by-N matrix C. On exit, C is overwritten by Q*C or Q**T*C or C*Q**T or C*Q. LDC (input) INTEGER The leading dimension of the array C. LDC >= max(1,M). WORK (workspace/output) DOUBLE PRECISION array, dimension (LWORK) On exit, if INFO = 0, WORK(1) returns the optimal LWORK. LWORK (input) INTEGER The dimension of the array WORK. If SIDE = 'L', LWORK >= max(1,N); if SIDE = 'R', LWORK >= max(1,M). For optimum performance LWORK >= N*NB if SIDE = 'L', and LWORK >= M*NB if SIDE = 'R', where NB is the optimal blocksize. INFO (output) INTEGER = 0: successful exit < 0: if INFO = -i, the i-th argument had an illegal value ===================================================================== Test the input arguments Parameter adjustments Function Body */ /* Table of constant values */ static integer c__1 = 1; static integer c_n1 = -1; static integer c__2 = 2; static integer c__65 = 65; /* System generated locals */ address a__1[2]; integer a_dim1, a_offset, c_dim1, c_offset, i__1, i__2, i__3[2], i__4, i__5; char ch__1[2]; /* Builtin functions Subroutine */ int s_cat(char *, char **, integer *, integer *, ftnlen); /* Local variables */ static logical left; static integer i; static doublereal t[4160] /* was [65][64] */; extern logical lsame_(char *, char *); static integer nbmin, iinfo, i1, i2, i3; extern /* Subroutine */ int dormr2_(char *, char *, integer *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *, doublereal *, integer *); static integer ib, nb, mi, ni; extern /* Subroutine */ int dlarfb_(char *, char *, char *, char *, integer *, integer *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *); static integer nq, nw; extern /* Subroutine */ int dlarft_(char *, char *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), xerbla_(char *, integer *); extern integer ilaenv_(integer *, char *, char *, integer *, integer *, integer *, integer *, ftnlen, ftnlen); static logical notran; static integer ldwork; static char transt[1]; static integer iws; #define T(I) t[(I)] #define WAS(I) was[(I)] #define TAU(I) tau[(I)-1] #define WORK(I) work[(I)-1] #define A(I,J) a[(I)-1 + ((J)-1)* ( *lda)] #define C(I,J) c[(I)-1 + ((J)-1)* ( *ldc)] *info = 0; left = lsame_(side, "L"); notran = lsame_(trans, "N"); /* NQ is the order of Q and NW is the minimum dimension of WORK */ if (left) { nq = *m; nw = *n; } else { nq = *n; nw = *m; } if (! left && ! lsame_(side, "R")) { *info = -1; } else if (! notran && ! lsame_(trans, "T")) { *info = -2; } else if (*m < 0) { *info = -3; } else if (*n < 0) { *info = -4; } else if (*k < 0 || *k > nq) { *info = -5; } else if (*lda < max(1,*k)) { *info = -7; } else if (*ldc < max(1,*m)) { *info = -10; } else if (*lwork < max(1,nw)) { *info = -12; } if (*info != 0) { i__1 = -(*info); xerbla_("DORMRQ", &i__1); return 0; } /* Quick return if possible */ if (*m == 0 || *n == 0 || *k == 0) { WORK(1) = 1.; return 0; } /* Determine the block size. NB may be at most NBMAX, where NBMAX is used to define the local array T. Computing MIN Writing concatenation */ i__3[0] = 1, a__1[0] = side; i__3[1] = 1, a__1[1] = trans; s_cat(ch__1, a__1, i__3, &c__2, 2L); i__1 = 64, i__2 = ilaenv_(&c__1, "DORMRQ", ch__1, m, n, k, &c_n1, 6L, 2L); nb = min(i__1,i__2); nbmin = 2; ldwork = nw; if (nb > 1 && nb < *k) { iws = nw * nb; if (*lwork < iws) { nb = *lwork / ldwork; /* Computing MAX Writing concatenation */ i__3[0] = 1, a__1[0] = side; i__3[1] = 1, a__1[1] = trans; s_cat(ch__1, a__1, i__3, &c__2, 2L); i__1 = 2, i__2 = ilaenv_(&c__2, "DORMRQ", ch__1, m, n, k, &c_n1, 6L, 2L); nbmin = max(i__1,i__2); } } else { iws = nw; } if (nb < nbmin || nb >= *k) { /* Use unblocked code */ dormr2_(side, trans, m, n, k, &A(1,1), lda, &TAU(1), &C(1,1) , ldc, &WORK(1), &iinfo); } else { /* Use blocked code */ if (left && ! notran || ! left && notran) { i1 = 1; i2 = *k; i3 = nb; } else { i1 = (*k - 1) / nb * nb + 1; i2 = 1; i3 = -nb; } if (left) { ni = *n; } else { mi = *m; } if (notran) { *(unsigned char *)transt = 'T'; } else { *(unsigned char *)transt = 'N'; } i__1 = i2; i__2 = i3; for (i = i1; i3 < 0 ? i >= i2 : i <= i2; i += i3) { /* Computing MIN */ i__4 = nb, i__5 = *k - i + 1; ib = min(i__4,i__5); /* Form the triangular factor of the block reflector H = H(i+ib-1) . . . H(i+1) H(i) */ i__4 = nq - *k + i + ib - 1; dlarft_("Backward", "Rowwise", &i__4, &ib, &A(i,1), lda, & TAU(i), t, &c__65); if (left) { /* H or H' is applied to C(1:m-k+i+ib-1,1:n) */ mi = *m - *k + i + ib - 1; } else { /* H or H' is applied to C(1:m,1:n-k+i+ib-1) */ ni = *n - *k + i + ib - 1; } /* Apply H or H' */ dlarfb_(side, transt, "Backward", "Rowwise", &mi, &ni, &ib, &A(i,1), lda, t, &c__65, &C(1,1), ldc, &WORK(1), & ldwork); /* L10: */ } } WORK(1) = (doublereal) iws; return 0; /* End of DORMRQ */ } /* dormrq_ */
/* Subroutine */ int dgerqf_(integer *m, integer *n, doublereal *a, integer * lda, doublereal *tau, doublereal *work, integer *lwork, integer *info) { /* -- LAPACK routine (version 3.0) -- Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd., Courant Institute, Argonne National Lab, and Rice University June 30, 1999 Purpose ======= DGERQF 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) DOUBLE PRECISION 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 min(m,n) elementary reflectors (see Further Details). LDA (input) INTEGER The leading dimension of the array A. LDA >= max(1,M). TAU (output) DOUBLE PRECISION array, dimension (min(M,N)) The scalar factors of the elementary reflectors (see Further Details). WORK (workspace/output) DOUBLE PRECISION array, dimension (LWORK) On exit, if INFO = 0, WORK(1) returns the optimal LWORK. LWORK (input) INTEGER The dimension of the array WORK. LWORK >= max(1,M). For optimum performance LWORK >= M*NB, where NB is the optimal blocksize. If LWORK = -1, then a workspace query is assumed; the routine only calculates the optimal size of the WORK array, returns this value as the first entry of the WORK array, and no error message related to LWORK is issued by XERBLA. INFO (output) INTEGER = 0: successful exit < 0: if INFO = -i, the i-th argument had an illegal value 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 */ /* Table of constant values */ static integer c__1 = 1; static integer c_n1 = -1; static integer c__3 = 3; static integer c__2 = 2; /* System generated locals */ integer a_dim1, a_offset, i__1, i__2, i__3, i__4; /* Local variables */ static integer i__, k, nbmin, iinfo; extern /* Subroutine */ int dgerq2_(integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *); static integer ib, nb, ki, kk; extern /* Subroutine */ int dlarfb_(char *, char *, char *, char *, integer *, integer *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *); static integer mu, nu, nx; extern /* Subroutine */ int dlarft_(char *, char *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), xerbla_(char *, integer *); extern integer ilaenv_(integer *, char *, char *, integer *, integer *, integer *, integer *, ftnlen, ftnlen); static integer ldwork, lwkopt; static logical lquery; static integer iws; #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; nb = ilaenv_(&c__1, "DGERQF", " ", m, n, &c_n1, &c_n1, (ftnlen)6, (ftnlen) 1); lwkopt = *m * nb; work[1] = (doublereal) lwkopt; lquery = *lwork == -1; if (*m < 0) { *info = -1; } else if (*n < 0) { *info = -2; } else if (*lda < max(1,*m)) { *info = -4; } else if (*lwork < max(1,*m) && ! lquery) { *info = -7; } if (*info != 0) { i__1 = -(*info); xerbla_("DGERQF", &i__1); return 0; } else if (lquery) { return 0; } /* Quick return if possible */ k = min(*m,*n); if (k == 0) { work[1] = 1.; return 0; } nbmin = 2; nx = 1; iws = *m; if (nb > 1 && nb < k) { /* Determine when to cross over from blocked to unblocked code. Computing MAX */ i__1 = 0, i__2 = ilaenv_(&c__3, "DGERQF", " ", m, n, &c_n1, &c_n1, ( ftnlen)6, (ftnlen)1); nx = max(i__1,i__2); if (nx < k) { /* Determine if workspace is large enough for blocked code. */ ldwork = *m; iws = ldwork * nb; if (*lwork < iws) { /* Not enough workspace to use optimal NB: reduce NB and determine the minimum value of NB. */ nb = *lwork / ldwork; /* Computing MAX */ i__1 = 2, i__2 = ilaenv_(&c__2, "DGERQF", " ", m, n, &c_n1, & c_n1, (ftnlen)6, (ftnlen)1); nbmin = max(i__1,i__2); } } } if (nb >= nbmin && nb < k && nx < k) { /* Use blocked code initially. The last kk rows are handled by the block method. */ ki = (k - nx - 1) / nb * nb; /* Computing MIN */ i__1 = k, i__2 = ki + nb; kk = min(i__1,i__2); i__1 = k - kk + 1; i__2 = -nb; for (i__ = k - kk + ki + 1; i__2 < 0 ? i__ >= i__1 : i__ <= i__1; i__ += i__2) { /* Computing MIN */ i__3 = k - i__ + 1; ib = min(i__3,nb); /* Compute the RQ factorization of the current block A(m-k+i:m-k+i+ib-1,1:n-k+i+ib-1) */ i__3 = *n - k + i__ + ib - 1; dgerq2_(&ib, &i__3, &a_ref(*m - k + i__, 1), lda, &tau[i__], & work[1], &iinfo); if (*m - k + i__ > 1) { /* Form the triangular factor of the block reflector H = H(i+ib-1) . . . H(i+1) H(i) */ i__3 = *n - k + i__ + ib - 1; dlarft_("Backward", "Rowwise", &i__3, &ib, &a_ref(*m - k + i__, 1), lda, &tau[i__], &work[1], &ldwork); /* Apply H to A(1:m-k+i-1,1:n-k+i+ib-1) from the right */ i__3 = *m - k + i__ - 1; i__4 = *n - k + i__ + ib - 1; dlarfb_("Right", "No transpose", "Backward", "Rowwise", &i__3, &i__4, &ib, &a_ref(*m - k + i__, 1), lda, &work[1], & ldwork, &a[a_offset], lda, &work[ib + 1], &ldwork); } /* L10: */ } mu = *m - k + i__ + nb - 1; nu = *n - k + i__ + nb - 1; } else { mu = *m; nu = *n; } /* Use unblocked code to factor the last or only block */ if (mu > 0 && nu > 0) { dgerq2_(&mu, &nu, &a[a_offset], lda, &tau[1], &work[1], &iinfo); } work[1] = (doublereal) iws; return 0; /* End of DGERQF */ } /* dgerqf_ */
/* Subroutine */ int dgeqlf_(integer *m, integer *n, doublereal *a, integer * lda, doublereal *tau, doublereal *work, integer *lwork, integer *info) { /* -- LAPACK routine (version 2.0) -- Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd., Courant Institute, Argonne National Lab, and Rice University September 30, 1994 Purpose ======= DGEQLF computes a QL factorization of a real M-by-N matrix A: A = Q * L. 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) DOUBLE PRECISION array, dimension (LDA,N) On entry, the M-by-N matrix A. On exit, if m >= n, the lower triangle of the subarray A(m-n+1:m,1:n) contains the N-by-N lower triangular matrix L; if m <= n, the elements on and below the (n-m)-th superdiagonal contain the M-by-N lower trapezoidal matrix L; 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) DOUBLE PRECISION array, dimension (min(M,N)) The scalar factors of the elementary reflectors (see Further Details). WORK (workspace/output) DOUBLE PRECISION array, dimension (LWORK) On exit, if INFO = 0, WORK(1) returns the optimal LWORK. LWORK (input) INTEGER The dimension of the array WORK. LWORK >= max(1,N). For optimum performance LWORK >= N*NB, where NB is the optimal blocksize. 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(k) . . . H(2) H(1), 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(m-k+i+1:m) = 0 and v(m-k+i) = 1; v(1:m-k+i-1) is stored on exit in A(1:m-k+i-1,n-k+i), and tau in TAU(i). ===================================================================== Test the input arguments Parameter adjustments Function Body */ /* Table of constant values */ static integer c__1 = 1; static integer c_n1 = -1; static integer c__3 = 3; static integer c__2 = 2; /* System generated locals */ integer a_dim1, a_offset, i__1, i__2, i__3, i__4; /* Local variables */ static integer i, k, nbmin, iinfo; extern /* Subroutine */ int dgeql2_(integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *); static integer ib, nb, ki, kk; extern /* Subroutine */ int dlarfb_(char *, char *, char *, char *, integer *, integer *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *); static integer mu, nu, nx; extern /* Subroutine */ int dlarft_(char *, char *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), xerbla_(char *, integer *); extern integer ilaenv_(integer *, char *, char *, integer *, integer *, integer *, integer *, ftnlen, ftnlen); static integer ldwork, iws; #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; } else if (*lwork < max(1,*n)) { *info = -7; } if (*info != 0) { i__1 = -(*info); xerbla_("DGEQLF", &i__1); return 0; } /* Quick return if possible */ k = min(*m,*n); if (k == 0) { WORK(1) = 1.; return 0; } /* Determine the block size. */ nb = ilaenv_(&c__1, "DGEQLF", " ", m, n, &c_n1, &c_n1, 6L, 1L); nbmin = 2; nx = 1; iws = *n; if (nb > 1 && nb < k) { /* Determine when to cross over from blocked to unblocked code. Computing MAX */ i__1 = 0, i__2 = ilaenv_(&c__3, "DGEQLF", " ", m, n, &c_n1, &c_n1, 6L, 1L); nx = max(i__1,i__2); if (nx < k) { /* Determine if workspace is large enough for blocked co de. */ ldwork = *n; iws = ldwork * nb; if (*lwork < iws) { /* Not enough workspace to use optimal NB: reduc e NB and determine the minimum value of NB. */ nb = *lwork / ldwork; /* Computing MAX */ i__1 = 2, i__2 = ilaenv_(&c__2, "DGEQLF", " ", m, n, &c_n1, & c_n1, 6L, 1L); nbmin = max(i__1,i__2); } } } if (nb >= nbmin && nb < k && nx < k) { /* Use blocked code initially. The last kk columns are handled by the block method. */ ki = (k - nx - 1) / nb * nb; /* Computing MIN */ i__1 = k, i__2 = ki + nb; kk = min(i__1,i__2); i__1 = k - kk + 1; i__2 = -nb; for (i = k - kk + ki + 1; -nb < 0 ? i >= k-kk+1 : i <= k-kk+1; i += -nb) { /* Computing MIN */ i__3 = k - i + 1; ib = min(i__3,nb); /* Compute the QL factorization of the current block A(1:m-k+i+ib-1,n-k+i:n-k+i+ib-1) */ i__3 = *m - k + i + ib - 1; dgeql2_(&i__3, &ib, &A(1,*n-k+i), lda, &TAU(i), & WORK(1), &iinfo); if (*n - k + i > 1) { /* Form the triangular factor of the block reflec tor H = H(i+ib-1) . . . H(i+1) H(i) */ i__3 = *m - k + i + ib - 1; dlarft_("Backward", "Columnwise", &i__3, &ib, &A(1,*n-k+i), lda, &TAU(i), &WORK(1), &ldwork); /* Apply H' to A(1:m-k+i+ib-1,1:n-k+i-1) from the left */ i__3 = *m - k + i + ib - 1; i__4 = *n - k + i - 1; dlarfb_("Left", "Transpose", "Backward", "Columnwise", &i__3, &i__4, &ib, &A(1,*n-k+i), lda, &WORK( 1), &ldwork, &A(1,1), lda, &WORK(ib + 1), & ldwork); } /* L10: */ } mu = *m - k + i + nb - 1; nu = *n - k + i + nb - 1; } else { mu = *m; nu = *n; } /* Use unblocked code to factor the last or only block */ if (mu > 0 && nu > 0) { dgeql2_(&mu, &nu, &A(1,1), lda, &TAU(1), &WORK(1), &iinfo); } WORK(1) = (doublereal) iws; return 0; /* End of DGEQLF */ } /* dgeqlf_ */
/* Subroutine */ extern "C" int dorgqr_(integer *m, integer *n, integer *k, doublereal * a, integer *lda, doublereal *tau, doublereal *work, integer *lwork, integer *info) { /* -- LAPACK routine (version 3.0) -- Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd., Courant Institute, Argonne National Lab, and Rice University June 30, 1999 Purpose ======= DORGQR generates an M-by-N real matrix Q with orthonormal columns, which is defined as the first N columns of a product of K elementary reflectors of order M Q = H(1) H(2) . . . H(k) as returned by DGEQRF. Arguments ========= M (input) INTEGER The number of rows of the matrix Q. M >= 0. N (input) INTEGER The number of columns of the matrix Q. M >= N >= 0. K (input) INTEGER The number of elementary reflectors whose product defines the matrix Q. N >= K >= 0. A (input/output) DOUBLE PRECISION array, dimension (LDA,N) On entry, the i-th column must contain the vector which defines the elementary reflector H(i), for i = 1,2,...,k, as returned by DGEQRF in the first k columns of its array argument A. On exit, the M-by-N matrix Q. LDA (input) INTEGER The first dimension of the array A. LDA >= max(1,M). TAU (input) DOUBLE PRECISION array, dimension (K) TAU(i) must contain the scalar factor of the elementary reflector H(i), as returned by DGEQRF. WORK (workspace/output) DOUBLE PRECISION array, dimension (LWORK) On exit, if INFO = 0, WORK(1) returns the optimal LWORK. LWORK (input) INTEGER The dimension of the array WORK. LWORK >= max(1,N). For optimum performance LWORK >= N*NB, where NB is the optimal blocksize. If LWORK = -1, then a workspace query is assumed; the routine only calculates the optimal size of the WORK array, returns this value as the first entry of the WORK array, and no error message related to LWORK is issued by XERBLA. INFO (output) INTEGER = 0: successful exit < 0: if INFO = -i, the i-th argument has an illegal value ===================================================================== Test the input arguments Parameter adjustments */ /* Table of constant values */ static integer c__1 = 1; static integer c_n1 = -1; static integer c__3 = 3; static integer c__2 = 2; /* System generated locals */ integer a_dim1, a_offset, i__1, i__2, i__3; /* Local variables */ static integer i__, j, l, nbmin, iinfo; extern /* Subroutine */ int dorg2r_(integer *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *); static integer ib, nb, ki, kk; extern /* Subroutine */ int dlarfb_(char *, char *, char *, char *, integer *, integer *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *); static integer nx; extern /* Subroutine */ int dlarft_(char *, char *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), xerbla_(char *, integer *); extern integer ilaenv_(integer *, char *, char *, integer *, integer *, integer *, integer *, ftnlen, ftnlen); static integer ldwork, lwkopt; static logical lquery; static integer iws; #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; nb = ilaenv_(&c__1, "DORGQR", " ", m, n, k, &c_n1, (ftnlen)6, (ftnlen)1); lwkopt = max(1,*n) * nb; work[1] = (doublereal) lwkopt; lquery = *lwork == -1; if (*m < 0) { *info = -1; } else if (*n < 0 || *n > *m) { *info = -2; } else if (*k < 0 || *k > *n) { *info = -3; } else if (*lda < max(1,*m)) { *info = -5; } else if (*lwork < max(1,*n) && ! lquery) { *info = -8; } if (*info != 0) { i__1 = -(*info); xerbla_("DORGQR", &i__1); return 0; } else if (lquery) { return 0; } /* Quick return if possible */ if (*n <= 0) { work[1] = 1.; return 0; } nbmin = 2; nx = 0; iws = *n; if (nb > 1 && nb < *k) { /* Determine when to cross over from blocked to unblocked code. Computing MAX */ i__1 = 0, i__2 = ilaenv_(&c__3, "DORGQR", " ", m, n, k, &c_n1, ( ftnlen)6, (ftnlen)1); nx = max(i__1,i__2); if (nx < *k) { /* Determine if workspace is large enough for blocked code. */ ldwork = *n; iws = ldwork * nb; if (*lwork < iws) { /* Not enough workspace to use optimal NB: reduce NB and determine the minimum value of NB. */ nb = *lwork / ldwork; /* Computing MAX */ i__1 = 2, i__2 = ilaenv_(&c__2, "DORGQR", " ", m, n, k, &c_n1, (ftnlen)6, (ftnlen)1); nbmin = max(i__1,i__2); } } } if (nb >= nbmin && nb < *k && nx < *k) { /* Use blocked code after the last block. The first kk columns are handled by the block method. */ ki = (*k - nx - 1) / nb * nb; /* Computing MIN */ i__1 = *k, i__2 = ki + nb; kk = min(i__1,i__2); /* Set A(1:kk,kk+1:n) to zero. */ i__1 = *n; for (j = kk + 1; j <= i__1; ++j) { i__2 = kk; for (i__ = 1; i__ <= i__2; ++i__) { a_ref(i__, j) = 0.; /* L10: */ } /* L20: */ } } else { kk = 0; } /* Use unblocked code for the last or only block. */ if (kk < *n) { i__1 = *m - kk; i__2 = *n - kk; i__3 = *k - kk; dorg2r_(&i__1, &i__2, &i__3, &a_ref(kk + 1, kk + 1), lda, &tau[kk + 1] , &work[1], &iinfo); } if (kk > 0) { /* Use blocked code */ i__1 = -nb; for (i__ = ki + 1; i__1 < 0 ? i__ >= 1 : i__ <= 1; i__ += i__1) { /* Computing MIN */ i__2 = nb, i__3 = *k - i__ + 1; ib = min(i__2,i__3); if (i__ + ib <= *n) { /* Form the triangular factor of the block reflector H = H(i) H(i+1) . . . H(i+ib-1) */ i__2 = *m - i__ + 1; dlarft_("Forward", "Columnwise", &i__2, &ib, &a_ref(i__, i__), lda, &tau[i__], &work[1], &ldwork); /* Apply H to A(i:m,i+ib:n) from the left */ i__2 = *m - i__ + 1; i__3 = *n - i__ - ib + 1; dlarfb_("Left", "No transpose", "Forward", "Columnwise", & i__2, &i__3, &ib, &a_ref(i__, i__), lda, &work[1], & ldwork, &a_ref(i__, i__ + ib), lda, &work[ib + 1], & ldwork); } /* Apply H to rows i:m of current block */ i__2 = *m - i__ + 1; dorg2r_(&i__2, &ib, &ib, &a_ref(i__, i__), lda, &tau[i__], &work[ 1], &iinfo); /* Set rows 1:i-1 of current block to zero */ i__2 = i__ + ib - 1; for (j = i__; j <= i__2; ++j) { i__3 = i__ - 1; for (l = 1; l <= i__3; ++l) { a_ref(l, j) = 0.; /* L30: */ } /* L40: */ } /* L50: */ } } work[1] = (doublereal) iws; return 0; /* End of DORGQR */ } /* dorgqr_ */
/* Subroutine */ int dgehrd_(integer *n, integer *ilo, integer *ihi, doublereal *a, integer *lda, doublereal *tau, doublereal *work, integer *lwork, integer *info) { /* -- LAPACK routine (version 3.0) -- Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd., Courant Institute, Argonne National Lab, and Rice University June 30, 1999 Purpose ======= DGEHRD 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 DGEBAL; otherwise they should be set to 1 and N respectively. See Further Details. 1 <= ILO <= IHI <= N, if N > 0; ILO=1 and IHI=0, if N=0. A (input/output) DOUBLE PRECISION 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) DOUBLE PRECISION array, dimension (N-1) The scalar factors of the elementary reflectors (see Further Details). Elements 1:ILO-1 and IHI:N-1 of TAU are set to zero. WORK (workspace/output) DOUBLE PRECISION array, dimension (LWORK) On exit, if INFO = 0, WORK(1) returns the optimal LWORK. LWORK (input) INTEGER The length of the array WORK. LWORK >= max(1,N). For optimum performance LWORK >= N*NB, where NB is the optimal blocksize. If LWORK = -1, then a workspace query is assumed; the routine only calculates the optimal size of the WORK array, returns this value as the first entry of the WORK array, and no error message related to LWORK is issued by XERBLA. INFO (output) INTEGER = 0: successful exit < 0: if INFO = -i, the i-th argument had an illegal value. 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; static integer c_n1 = -1; static integer c__3 = 3; static integer c__2 = 2; static integer c__65 = 65; static doublereal c_b25 = -1.; static doublereal c_b26 = 1.; /* System generated locals */ integer a_dim1, a_offset, i__1, i__2, i__3, i__4; /* Local variables */ static integer i__; static doublereal t[4160] /* was [65][64] */; extern /* Subroutine */ int dgemm_(char *, char *, integer *, integer *, integer *, doublereal *, doublereal *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *); static integer nbmin, iinfo; extern /* Subroutine */ int dgehd2_(integer *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *); static integer ib; static doublereal ei; static integer nb, nh; extern /* Subroutine */ int dlarfb_(char *, char *, char *, char *, integer *, integer *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *), dlahrd_(integer *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *, doublereal *, integer *); static integer nx; extern /* Subroutine */ int xerbla_(char *, integer *); extern integer ilaenv_(integer *, char *, char *, integer *, integer *, integer *, integer *, ftnlen, ftnlen); static integer ldwork, lwkopt; static logical lquery; static integer iws; #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; /* Computing MIN */ i__1 = 64, i__2 = ilaenv_(&c__1, "DGEHRD", " ", n, ilo, ihi, &c_n1, ( ftnlen)6, (ftnlen)1); nb = min(i__1,i__2); lwkopt = *n * nb; work[1] = (doublereal) lwkopt; lquery = *lwork == -1; 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; } else if (*lwork < max(1,*n) && ! lquery) { *info = -8; } if (*info != 0) { i__1 = -(*info); xerbla_("DGEHRD", &i__1); return 0; } else if (lquery) { return 0; } /* Set elements 1:ILO-1 and IHI:N-1 of TAU to zero */ i__1 = *ilo - 1; for (i__ = 1; i__ <= i__1; ++i__) { tau[i__] = 0.; /* L10: */ } i__1 = *n - 1; for (i__ = max(1,*ihi); i__ <= i__1; ++i__) { tau[i__] = 0.; /* L20: */ } /* Quick return if possible */ nh = *ihi - *ilo + 1; if (nh <= 1) { work[1] = 1.; return 0; } /* Determine the block size. Computing MIN */ i__1 = 64, i__2 = ilaenv_(&c__1, "DGEHRD", " ", n, ilo, ihi, &c_n1, ( ftnlen)6, (ftnlen)1); nb = min(i__1,i__2); nbmin = 2; iws = 1; if (nb > 1 && nb < nh) { /* Determine when to cross over from blocked to unblocked code (last block is always handled by unblocked code). Computing MAX */ i__1 = nb, i__2 = ilaenv_(&c__3, "DGEHRD", " ", n, ilo, ihi, &c_n1, ( ftnlen)6, (ftnlen)1); nx = max(i__1,i__2); if (nx < nh) { /* Determine if workspace is large enough for blocked code. */ iws = *n * nb; if (*lwork < iws) { /* Not enough workspace to use optimal NB: determine the minimum value of NB, and reduce NB or force use of unblocked code. Computing MAX */ i__1 = 2, i__2 = ilaenv_(&c__2, "DGEHRD", " ", n, ilo, ihi, & c_n1, (ftnlen)6, (ftnlen)1); nbmin = max(i__1,i__2); if (*lwork >= *n * nbmin) { nb = *lwork / *n; } else { nb = 1; } } } } ldwork = *n; if (nb < nbmin || nb >= nh) { /* Use unblocked code below */ i__ = *ilo; } else { /* Use blocked code */ i__1 = *ihi - 1 - nx; i__2 = nb; for (i__ = *ilo; i__2 < 0 ? i__ >= i__1 : i__ <= i__1; i__ += i__2) { /* Computing MIN */ i__3 = nb, i__4 = *ihi - i__; ib = min(i__3,i__4); /* Reduce columns i:i+ib-1 to Hessenberg form, returning the matrices V and T of the block reflector H = I - V*T*V' which performs the reduction, and also the matrix Y = A*V*T */ dlahrd_(ihi, &i__, &ib, &a_ref(1, i__), lda, &tau[i__], t, &c__65, &work[1], &ldwork); /* Apply the block reflector H to A(1:ihi,i+ib:ihi) from the right, computing A := A - Y * V'. V(i+ib,ib-1) must be set to 1. */ ei = a_ref(i__ + ib, i__ + ib - 1); a_ref(i__ + ib, i__ + ib - 1) = 1.; i__3 = *ihi - i__ - ib + 1; dgemm_("No transpose", "Transpose", ihi, &i__3, &ib, &c_b25, & work[1], &ldwork, &a_ref(i__ + ib, i__), lda, &c_b26, & a_ref(1, i__ + ib), lda); a_ref(i__ + ib, i__ + ib - 1) = ei; /* Apply the block reflector H to A(i+1:ihi,i+ib:n) from the left */ i__3 = *ihi - i__; i__4 = *n - i__ - ib + 1; dlarfb_("Left", "Transpose", "Forward", "Columnwise", &i__3, & i__4, &ib, &a_ref(i__ + 1, i__), lda, t, &c__65, &a_ref( i__ + 1, i__ + ib), lda, &work[1], &ldwork); /* L30: */ } } /* Use unblocked code to reduce the rest of the matrix */ dgehd2_(n, &i__, ihi, &a[a_offset], lda, &tau[1], &work[1], &iinfo); work[1] = (doublereal) iws; return 0; /* End of DGEHRD */ } /* dgehrd_ */
/* Subroutine */ int dgeqrt_(integer *m, integer *n, integer *nb, doublereal * a, integer *lda, doublereal *t, integer *ldt, doublereal *work, integer *info) { /* System generated locals */ integer a_dim1, a_offset, t_dim1, t_offset, i__1, i__2, i__3, i__4, i__5; /* Local variables */ integer i__, k, ib, iinfo; extern /* Subroutine */ int dlarfb_(char *, char *, char *, char *, integer *, integer *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *), xerbla_(char *, integer *), dgeqrt2_(integer *, integer *, doublereal *, integer *, doublereal *, integer *, integer *), dgeqrt3_(integer * , integer *, doublereal *, integer *, doublereal *, integer *, integer *); /* -- LAPACK computational routine (version 3.5.0) -- */ /* -- LAPACK is a software package provided by Univ. of Tennessee, -- */ /* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- */ /* November 2013 */ /* .. Scalar Arguments .. */ /* .. */ /* .. Array Arguments .. */ /* .. */ /* ===================================================================== */ /* .. */ /* .. Local Scalars .. */ /* .. */ /* .. External Subroutines .. */ /* .. */ /* .. Executable Statements .. */ /* Test the input arguments */ /* Parameter adjustments */ a_dim1 = *lda; a_offset = 1 + a_dim1; a -= a_offset; t_dim1 = *ldt; t_offset = 1 + t_dim1; t -= t_offset; --work; /* Function Body */ *info = 0; if (*m < 0) { *info = -1; } else if (*n < 0) { *info = -2; } else if (*nb < 1 || *nb > min(*m,*n) && min(*m,*n) > 0) { *info = -3; } else if (*lda < max(1,*m)) { *info = -5; } else if (*ldt < *nb) { *info = -7; } if (*info != 0) { i__1 = -(*info); xerbla_("DGEQRT", &i__1); return 0; } /* Quick return if possible */ k = min(*m,*n); if (k == 0) { return 0; } /* Blocked loop of length K */ i__1 = k; i__2 = *nb; for (i__ = 1; i__2 < 0 ? i__ >= i__1 : i__ <= i__1; i__ += i__2) { /* Computing MIN */ i__3 = k - i__ + 1; ib = min(i__3,*nb); /* Compute the QR factorization of the current block A(I:M,I:I+IB-1) */ if (TRUE_) { i__3 = *m - i__ + 1; dgeqrt3_(&i__3, &ib, &a[i__ + i__ * a_dim1], lda, &t[i__ * t_dim1 + 1], ldt, &iinfo); } else { i__3 = *m - i__ + 1; dgeqrt2_(&i__3, &ib, &a[i__ + i__ * a_dim1], lda, &t[i__ * t_dim1 + 1], ldt, &iinfo); } if (i__ + ib <= *n) { /* Update by applying H**T to A(I:M,I+IB:N) from the left */ i__3 = *m - i__ + 1; i__4 = *n - i__ - ib + 1; i__5 = *n - i__ - ib + 1; dlarfb_("L", "T", "F", "C", &i__3, &i__4, &ib, &a[i__ + i__ * a_dim1], lda, &t[i__ * t_dim1 + 1], ldt, &a[i__ + (i__ + ib) * a_dim1], lda, &work[1], &i__5); } } return 0; /* End of DGEQRT */ }
/* Subroutine */ int dgeqrf_(integer *m, integer *n, doublereal *a, integer * lda, doublereal *tau, doublereal *work, integer *lwork, integer *info) { /* System generated locals */ integer a_dim1, a_offset, i__1, i__2, i__3, i__4, i__5, i__6; real r__1; /* Local variables */ integer i__, j, k, ib, nb, nt, nx, iws; extern doublereal sceil_(real *); integer nbmin, iinfo; extern /* Subroutine */ int dgeqr2_(integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), dlarfb_(char *, char *, char *, char *, integer *, integer *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *), dlarft_(char *, char *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), xerbla_(char *, integer *); extern integer ilaenv_(integer *, char *, char *, integer *, integer *, integer *, integer *); integer lbwork, llwork, lwkopt; logical lquery; /* -- LAPACK routine (version 3.1) -- */ /* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */ /* March 2008 */ /* .. Scalar Arguments .. */ /* .. */ /* .. Array Arguments .. */ /* .. */ /* Purpose */ /* ======= */ /* DGEQRF computes a QR factorization of a real M-by-N matrix A: */ /* A = Q * R. */ /* This is the left-looking Level 3 BLAS version of the algorithm. */ /* 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) DOUBLE PRECISION array, dimension (LDA,N) */ /* On entry, the M-by-N matrix A. */ /* On exit, the elements on and above the diagonal of the array */ /* contain the min(M,N)-by-N upper trapezoidal matrix R (R is */ /* upper triangular if m >= n); the elements below the diagonal, */ /* with the array TAU, represent the orthogonal matrix Q as a */ /* product of min(m,n) elementary reflectors (see Further */ /* Details). */ /* LDA (input) INTEGER */ /* The leading dimension of the array A. LDA >= max(1,M). */ /* TAU (output) DOUBLE PRECISION array, dimension (min(M,N)) */ /* The scalar factors of the elementary reflectors (see Further */ /* Details). */ /* WORK (workspace/output) DOUBLE PRECISION array, dimension (MAX(1,LWORK)) */ /* On exit, if INFO = 0, WORK(1) returns the optimal LWORK. */ /* LWORK (input) INTEGER */ /* The dimension of the array WORK. The dimension can be divided into three parts. */ /* 1) The part for the triangular factor T. If the very last T is not bigger */ /* than any of the rest, then this part is NB x ceiling(K/NB), otherwise, */ /* NB x (K-NT), where K = min(M,N) and NT is the dimension of the very last T */ /* 2) The part for the very last T when T is bigger than any of the rest T. */ /* The size of this part is NT x NT, where NT = K - ceiling ((K-NX)/NB) x NB, */ /* where K = min(M,N), NX is calculated by */ /* NX = MAX( 0, ILAENV( 3, 'DGEQRF', ' ', M, N, -1, -1 ) ) */ /* 3) The part for dlarfb is of size max((N-M)*K, (N-M)*NB, K*NB, NB*NB) */ /* So LWORK = part1 + part2 + part3 */ /* If LWORK = -1, then a workspace query is assumed; the routine */ /* only calculates the optimal size of the WORK array, returns */ /* this value as the first entry of the WORK array, and no error */ /* message related to LWORK is issued by XERBLA. */ /* INFO (output) INTEGER */ /* = 0: successful exit */ /* < 0: if INFO = -i, the i-th argument had an illegal value */ /* Further Details */ /* =============== */ /* The matrix Q is represented as a product of elementary reflectors */ /* Q = H(1) H(2) . . . H(k), where k = min(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(1:i-1) = 0 and v(i) = 1; v(i+1:m) is stored on exit in A(i+1:m,i), */ /* and tau in TAU(i). */ /* ===================================================================== */ /* .. Local Scalars .. */ /* .. */ /* .. External Subroutines .. */ /* .. */ /* .. Intrinsic Functions .. */ /* .. */ /* .. External Functions .. */ /* .. */ /* .. Executable Statements .. */ /* Parameter adjustments */ a_dim1 = *lda; a_offset = 1 + a_dim1; a -= a_offset; --tau; --work; /* Function Body */ *info = 0; nbmin = 2; nx = 0; iws = *n; k = min(*m,*n); nb = ilaenv_(&c__1, "DGEQRF", " ", m, n, &c_n1, &c_n1); if (nb > 1 && nb < k) { /* Determine when to cross over from blocked to unblocked code. */ /* Computing MAX */ i__1 = 0, i__2 = ilaenv_(&c__3, "DGEQRF", " ", m, n, &c_n1, &c_n1); nx = max(i__1,i__2); } /* Get NT, the size of the very last T, which is the left-over from in-between K-NX and K to K, eg.: */ /* NB=3 2NB=6 K=10 */ /* | | | */ /* 1--2--3--4--5--6--7--8--9--10 */ /* | \________/ */ /* K-NX=5 NT=4 */ /* So here 4 x 4 is the last T stored in the workspace */ r__1 = (real) (k - nx) / (real) nb; nt = k - sceil_(&r__1) * nb; /* optimal workspace = space for dlarfb + space for normal T's + space for the last T */ /* Computing MAX */ /* Computing MAX */ i__3 = (*n - *m) * k, i__4 = (*n - *m) * nb; /* Computing MAX */ i__5 = k * nb, i__6 = nb * nb; i__1 = max(i__3,i__4), i__2 = max(i__5,i__6); llwork = max(i__1,i__2); r__1 = (real) llwork / (real) nb; llwork = sceil_(&r__1); if (nt > nb) { lbwork = k - nt; /* Optimal workspace for dlarfb = MAX(1,N)*NT */ lwkopt = (lbwork + llwork) * nb; work[1] = (doublereal) (lwkopt + nt * nt); } else { r__1 = (real) k / (real) nb; lbwork = sceil_(&r__1) * nb; lwkopt = (lbwork + llwork - nb) * nb; work[1] = (doublereal) lwkopt; } /* Test the input arguments */ lquery = *lwork == -1; if (*m < 0) { *info = -1; } else if (*n < 0) { *info = -2; } else if (*lda < max(1,*m)) { *info = -4; } else if (*lwork < max(1,*n) && ! lquery) { *info = -7; } if (*info != 0) { i__1 = -(*info); xerbla_("DGEQRF", &i__1); return 0; } else if (lquery) { return 0; } /* Quick return if possible */ if (k == 0) { work[1] = 1.; return 0; } if (nb > 1 && nb < k) { if (nx < k) { /* Determine if workspace is large enough for blocked code. */ if (nt <= nb) { iws = (lbwork + llwork - nb) * nb; } else { iws = (lbwork + llwork) * nb + nt * nt; } if (*lwork < iws) { /* Not enough workspace to use optimal NB: reduce NB and */ /* determine the minimum value of NB. */ if (nt <= nb) { nb = *lwork / (llwork + (lbwork - nb)); } else { nb = (*lwork - nt * nt) / (lbwork + llwork); } /* Computing MAX */ i__1 = 2, i__2 = ilaenv_(&c__2, "DGEQRF", " ", m, n, &c_n1, & c_n1); nbmin = max(i__1,i__2); } } } if (nb >= nbmin && nb < k && nx < k) { /* Use blocked code initially */ i__1 = k - nx; i__2 = nb; for (i__ = 1; i__2 < 0 ? i__ >= i__1 : i__ <= i__1; i__ += i__2) { /* Computing MIN */ i__3 = k - i__ + 1; ib = min(i__3,nb); /* Update the current column using old T's */ i__3 = i__ - nb; i__4 = nb; for (j = 1; i__4 < 0 ? j >= i__3 : j <= i__3; j += i__4) { /* Apply H' to A(J:M,I:I+IB-1) from the left */ i__5 = *m - j + 1; dlarfb_("Left", "Transpose", "Forward", "Columnwise", &i__5, & ib, &nb, &a[j + j * a_dim1], lda, &work[j], &lbwork, & a[j + i__ * a_dim1], lda, &work[lbwork * nb + nt * nt + 1], &ib); /* L20: */ } /* Compute the QR factorization of the current block */ /* A(I:M,I:I+IB-1) */ i__4 = *m - i__ + 1; dgeqr2_(&i__4, &ib, &a[i__ + i__ * a_dim1], lda, &tau[i__], &work[ lbwork * nb + nt * nt + 1], &iinfo); if (i__ + ib <= *n) { /* Form the triangular factor of the block reflector */ /* H = H(i) H(i+1) . . . H(i+ib-1) */ i__4 = *m - i__ + 1; dlarft_("Forward", "Columnwise", &i__4, &ib, &a[i__ + i__ * a_dim1], lda, &tau[i__], &work[i__], &lbwork); } /* L10: */ } } else { i__ = 1; } /* Use unblocked code to factor the last or only block. */ if (i__ <= k) { if (i__ != 1) { i__2 = i__ - nb; i__1 = nb; for (j = 1; i__1 < 0 ? j >= i__2 : j <= i__2; j += i__1) { /* Apply H' to A(J:M,I:K) from the left */ i__4 = *m - j + 1; i__3 = k - i__ + 1; i__5 = k - i__ + 1; dlarfb_("Left", "Transpose", "Forward", "Columnwise", &i__4, & i__3, &nb, &a[j + j * a_dim1], lda, &work[j], &lbwork, &a[j + i__ * a_dim1], lda, &work[lbwork * nb + nt * nt + 1], &i__5); /* L30: */ } i__1 = *m - i__ + 1; i__2 = k - i__ + 1; dgeqr2_(&i__1, &i__2, &a[i__ + i__ * a_dim1], lda, &tau[i__], & work[lbwork * nb + nt * nt + 1], &iinfo); } else { /* Use unblocked code to factor the last or only block. */ i__1 = *m - i__ + 1; i__2 = *n - i__ + 1; dgeqr2_(&i__1, &i__2, &a[i__ + i__ * a_dim1], lda, &tau[i__], & work[1], &iinfo); } } /* Apply update to the column M+1:N when N > M */ if (*m < *n && i__ != 1) { /* Form the last triangular factor of the block reflector */ /* H = H(i) H(i+1) . . . H(i+ib-1) */ if (nt <= nb) { i__1 = *m - i__ + 1; i__2 = k - i__ + 1; dlarft_("Forward", "Columnwise", &i__1, &i__2, &a[i__ + i__ * a_dim1], lda, &tau[i__], &work[i__], &lbwork); } else { i__1 = *m - i__ + 1; i__2 = k - i__ + 1; dlarft_("Forward", "Columnwise", &i__1, &i__2, &a[i__ + i__ * a_dim1], lda, &tau[i__], &work[lbwork * nb + 1], &nt); } /* Apply H' to A(1:M,M+1:N) from the left */ i__1 = k - nx; i__2 = nb; for (j = 1; i__2 < 0 ? j >= i__1 : j <= i__1; j += i__2) { /* Computing MIN */ i__4 = k - j + 1; ib = min(i__4,nb); i__4 = *m - j + 1; i__3 = *n - *m; i__5 = *n - *m; dlarfb_("Left", "Transpose", "Forward", "Columnwise", &i__4, & i__3, &ib, &a[j + j * a_dim1], lda, &work[j], &lbwork, &a[ j + (*m + 1) * a_dim1], lda, &work[lbwork * nb + nt * nt + 1], &i__5); /* L40: */ } if (nt <= nb) { i__2 = *m - j + 1; i__1 = *n - *m; i__4 = k - j + 1; i__3 = *n - *m; dlarfb_("Left", "Transpose", "Forward", "Columnwise", &i__2, & i__1, &i__4, &a[j + j * a_dim1], lda, &work[j], &lbwork, & a[j + (*m + 1) * a_dim1], lda, &work[lbwork * nb + nt * nt + 1], &i__3); } else { i__2 = *m - j + 1; i__1 = *n - *m; i__4 = k - j + 1; i__3 = *n - *m; dlarfb_("Left", "Transpose", "Forward", "Columnwise", &i__2, & i__1, &i__4, &a[j + j * a_dim1], lda, &work[lbwork * nb + 1], &nt, &a[j + (*m + 1) * a_dim1], lda, &work[lbwork * nb + nt * nt + 1], &i__3); } } work[1] = (doublereal) iws; return 0; /* End of DGEQRF */ } /* dgeqrf_ */
/* Subroutine */ int dormqr_(char *side, char *trans, integer *m, integer *n, integer *k, doublereal *a, integer *lda, doublereal *tau, doublereal * c__, integer *ldc, doublereal *work, integer *lwork, integer *info) { /* -- LAPACK routine (version 3.0) -- Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd., Courant Institute, Argonne National Lab, and Rice University June 30, 1999 Purpose ======= DORMQR overwrites the general real M-by-N matrix C with SIDE = 'L' SIDE = 'R' TRANS = 'N': Q * C C * Q TRANS = 'T': Q**T * C C * Q**T where Q is a real orthogonal matrix defined as the product of k elementary reflectors Q = H(1) H(2) . . . H(k) as returned by DGEQRF. Q is of order M if SIDE = 'L' and of order N if SIDE = 'R'. Arguments ========= SIDE (input) CHARACTER*1 = 'L': apply Q or Q**T from the Left; = 'R': apply Q or Q**T from the Right. TRANS (input) CHARACTER*1 = 'N': No transpose, apply Q; = 'T': Transpose, apply Q**T. M (input) INTEGER The number of rows of the matrix C. M >= 0. N (input) INTEGER The number of columns of the matrix C. N >= 0. K (input) INTEGER The number of elementary reflectors whose product defines the matrix Q. If SIDE = 'L', M >= K >= 0; if SIDE = 'R', N >= K >= 0. A (input) DOUBLE PRECISION array, dimension (LDA,K) The i-th column must contain the vector which defines the elementary reflector H(i), for i = 1,2,...,k, as returned by DGEQRF in the first k columns of its array argument A. A is modified by the routine but restored on exit. LDA (input) INTEGER The leading dimension of the array A. If SIDE = 'L', LDA >= max(1,M); if SIDE = 'R', LDA >= max(1,N). TAU (input) DOUBLE PRECISION array, dimension (K) TAU(i) must contain the scalar factor of the elementary reflector H(i), as returned by DGEQRF. C (input/output) DOUBLE PRECISION array, dimension (LDC,N) On entry, the M-by-N matrix C. On exit, C is overwritten by Q*C or Q**T*C or C*Q**T or C*Q. LDC (input) INTEGER The leading dimension of the array C. LDC >= max(1,M). WORK (workspace/output) DOUBLE PRECISION array, dimension (LWORK) On exit, if INFO = 0, WORK(1) returns the optimal LWORK. LWORK (input) INTEGER The dimension of the array WORK. If SIDE = 'L', LWORK >= max(1,N); if SIDE = 'R', LWORK >= max(1,M). For optimum performance LWORK >= N*NB if SIDE = 'L', and LWORK >= M*NB if SIDE = 'R', where NB is the optimal blocksize. If LWORK = -1, then a workspace query is assumed; the routine only calculates the optimal size of the WORK array, returns this value as the first entry of the WORK array, and no error message related to LWORK is issued by XERBLA. INFO (output) INTEGER = 0: successful exit < 0: if INFO = -i, the i-th argument had an illegal value ===================================================================== Test the input arguments Parameter adjustments */ /* Table of constant values */ static integer c__1 = 1; static integer c_n1 = -1; static integer c__2 = 2; static integer c__65 = 65; /* System generated locals */ address a__1[2]; integer a_dim1, a_offset, c_dim1, c_offset, i__1, i__2, i__3[2], i__4, i__5; char ch__1[2]; /* Builtin functions Subroutine */ int s_cat(char *, char **, integer *, integer *, ftnlen); /* Local variables */ static logical left; static integer i__; static doublereal t[4160] /* was [65][64] */; extern logical lsame_(char *, char *); static integer nbmin, iinfo, i1, i2, i3; extern /* Subroutine */ int dorm2r_(char *, char *, integer *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *, doublereal *, integer *); static integer ib, ic, jc, nb, mi, ni; extern /* Subroutine */ int dlarfb_(char *, char *, char *, char *, integer *, integer *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *); static integer nq, nw; extern /* Subroutine */ int dlarft_(char *, char *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), xerbla_(char *, integer *); extern integer ilaenv_(integer *, char *, char *, integer *, integer *, integer *, integer *, ftnlen, ftnlen); static logical notran; static integer ldwork, lwkopt; static logical lquery; static integer iws; #define a_ref(a_1,a_2) a[(a_2)*a_dim1 + a_1] #define c___ref(a_1,a_2) c__[(a_2)*c_dim1 + a_1] a_dim1 = *lda; a_offset = 1 + a_dim1 * 1; a -= a_offset; --tau; c_dim1 = *ldc; c_offset = 1 + c_dim1 * 1; c__ -= c_offset; --work; /* Function Body */ *info = 0; left = lsame_(side, "L"); notran = lsame_(trans, "N"); lquery = *lwork == -1; /* NQ is the order of Q and NW is the minimum dimension of WORK */ if (left) { nq = *m; nw = *n; } else { nq = *n; nw = *m; } if (! left && ! lsame_(side, "R")) { *info = -1; } else if (! notran && ! lsame_(trans, "T")) { *info = -2; } else if (*m < 0) { *info = -3; } else if (*n < 0) { *info = -4; } else if (*k < 0 || *k > nq) { *info = -5; } else if (*lda < max(1,nq)) { *info = -7; } else if (*ldc < max(1,*m)) { *info = -10; } else if (*lwork < max(1,nw) && ! lquery) { *info = -12; } if (*info == 0) { /* Determine the block size. NB may be at most NBMAX, where NBMAX is used to define the local array T. Computing MIN Writing concatenation */ i__3[0] = 1, a__1[0] = side; i__3[1] = 1, a__1[1] = trans; s_cat(ch__1, a__1, i__3, &c__2, (ftnlen)2); i__1 = 64, i__2 = ilaenv_(&c__1, "DORMQR", ch__1, m, n, k, &c_n1, ( ftnlen)6, (ftnlen)2); nb = min(i__1,i__2); lwkopt = max(1,nw) * nb; work[1] = (doublereal) lwkopt; } if (*info != 0) { i__1 = -(*info); xerbla_("DORMQR", &i__1); return 0; } else if (lquery) { return 0; } /* Quick return if possible */ if (*m == 0 || *n == 0 || *k == 0) { work[1] = 1.; return 0; } nbmin = 2; ldwork = nw; if (nb > 1 && nb < *k) { iws = nw * nb; if (*lwork < iws) { nb = *lwork / ldwork; /* Computing MAX Writing concatenation */ i__3[0] = 1, a__1[0] = side; i__3[1] = 1, a__1[1] = trans; s_cat(ch__1, a__1, i__3, &c__2, (ftnlen)2); i__1 = 2, i__2 = ilaenv_(&c__2, "DORMQR", ch__1, m, n, k, &c_n1, ( ftnlen)6, (ftnlen)2); nbmin = max(i__1,i__2); } } else { iws = nw; } if (nb < nbmin || nb >= *k) { /* Use unblocked code */ dorm2r_(side, trans, m, n, k, &a[a_offset], lda, &tau[1], &c__[ c_offset], ldc, &work[1], &iinfo); } else { /* Use blocked code */ if (left && ! notran || ! left && notran) { i1 = 1; i2 = *k; i3 = nb; } else { i1 = (*k - 1) / nb * nb + 1; i2 = 1; i3 = -nb; } if (left) { ni = *n; jc = 1; } else { mi = *m; ic = 1; } i__1 = i2; i__2 = i3; for (i__ = i1; i__2 < 0 ? i__ >= i__1 : i__ <= i__1; i__ += i__2) { /* Computing MIN */ i__4 = nb, i__5 = *k - i__ + 1; ib = min(i__4,i__5); /* Form the triangular factor of the block reflector H = H(i) H(i+1) . . . H(i+ib-1) */ i__4 = nq - i__ + 1; dlarft_("Forward", "Columnwise", &i__4, &ib, &a_ref(i__, i__), lda, &tau[i__], t, &c__65); if (left) { /* H or H' is applied to C(i:m,1:n) */ mi = *m - i__ + 1; ic = i__; } else { /* H or H' is applied to C(1:m,i:n) */ ni = *n - i__ + 1; jc = i__; } /* Apply H or H' */ dlarfb_(side, trans, "Forward", "Columnwise", &mi, &ni, &ib, & a_ref(i__, i__), lda, t, &c__65, &c___ref(ic, jc), ldc, & work[1], &ldwork); /* L10: */ } } work[1] = (doublereal) lwkopt; return 0; /* End of DORMQR */ } /* dormqr_ */
/*< SUBROUTINE DORGQR( M, N, K, A, LDA, TAU, WORK, LWORK, INFO ) >*/ /* Subroutine */ int dorgqr_(integer *m, integer *n, integer *k, doublereal * a, integer *lda, doublereal *tau, doublereal *work, integer *lwork, integer *info) { /* System generated locals */ integer a_dim1, a_offset, i__1, i__2, i__3; /* Local variables */ integer i__, j, l, ib, nb, ki=0, kk, nx, iws, nbmin, iinfo; extern /* Subroutine */ int dorg2r_(integer *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), dlarfb_(char *, char *, char *, char *, integer *, integer *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *, ftnlen, ftnlen, ftnlen, ftnlen), dlarft_(char *, char *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *, ftnlen, ftnlen), xerbla_(char *, integer *, ftnlen); extern integer ilaenv_(integer *, char *, char *, integer *, integer *, integer *, integer *, ftnlen, ftnlen); integer ldwork, lwkopt; 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 .. */ /*< INTEGER INFO, K, LDA, LWORK, M, N >*/ /* .. */ /* .. Array Arguments .. */ /*< DOUBLE PRECISION A( LDA, * ), TAU( * ), WORK( * ) >*/ /* .. */ /* Purpose */ /* ======= */ /* DORGQR generates an M-by-N real matrix Q with orthonormal columns, */ /* which is defined as the first N columns of a product of K elementary */ /* reflectors of order M */ /* Q = H(1) H(2) . . . H(k) */ /* as returned by DGEQRF. */ /* Arguments */ /* ========= */ /* M (input) INTEGER */ /* The number of rows of the matrix Q. M >= 0. */ /* N (input) INTEGER */ /* The number of columns of the matrix Q. M >= N >= 0. */ /* K (input) INTEGER */ /* The number of elementary reflectors whose product defines the */ /* matrix Q. N >= K >= 0. */ /* A (input/output) DOUBLE PRECISION array, dimension (LDA,N) */ /* On entry, the i-th column must contain the vector which */ /* defines the elementary reflector H(i), for i = 1,2,...,k, as */ /* returned by DGEQRF in the first k columns of its array */ /* argument A. */ /* On exit, the M-by-N matrix Q. */ /* LDA (input) INTEGER */ /* The first dimension of the array A. LDA >= max(1,M). */ /* TAU (input) DOUBLE PRECISION array, dimension (K) */ /* TAU(i) must contain the scalar factor of the elementary */ /* reflector H(i), as returned by DGEQRF. */ /* WORK (workspace/output) DOUBLE PRECISION array, dimension (LWORK) */ /* On exit, if INFO = 0, WORK(1) returns the optimal LWORK. */ /* LWORK (input) INTEGER */ /* The dimension of the array WORK. LWORK >= max(1,N). */ /* For optimum performance LWORK >= N*NB, where NB is the */ /* optimal blocksize. */ /* If LWORK = -1, then a workspace query is assumed; the routine */ /* only calculates the optimal size of the WORK array, returns */ /* this value as the first entry of the WORK array, and no error */ /* message related to LWORK is issued by XERBLA. */ /* INFO (output) INTEGER */ /* = 0: successful exit */ /* < 0: if INFO = -i, the i-th argument has an illegal value */ /* ===================================================================== */ /* .. Parameters .. */ /*< DOUBLE PRECISION ZERO >*/ /*< PARAMETER ( ZERO = 0.0D+0 ) >*/ /* .. */ /* .. Local Scalars .. */ /*< LOGICAL LQUERY >*/ /*< >*/ /* .. */ /* .. External Subroutines .. */ /*< EXTERNAL DLARFB, DLARFT, DORG2R, XERBLA >*/ /* .. */ /* .. Intrinsic Functions .. */ /*< INTRINSIC MAX, MIN >*/ /* .. */ /* .. External Functions .. */ /*< INTEGER ILAENV >*/ /*< EXTERNAL ILAENV >*/ /* .. */ /* .. Executable Statements .. */ /* Test the input arguments */ /*< INFO = 0 >*/ /* Parameter adjustments */ a_dim1 = *lda; a_offset = 1 + a_dim1; a -= a_offset; --tau; --work; /* Function Body */ *info = 0; /*< NB = ILAENV( 1, 'DORGQR', ' ', M, N, K, -1 ) >*/ nb = ilaenv_(&c__1, "DORGQR", " ", m, n, k, &c_n1, (ftnlen)6, (ftnlen)1); /*< LWKOPT = MAX( 1, N )*NB >*/ lwkopt = max(1,*n) * nb; /*< WORK( 1 ) = LWKOPT >*/ work[1] = (doublereal) lwkopt; /*< LQUERY = ( LWORK.EQ.-1 ) >*/ lquery = *lwork == -1; /*< IF( M.LT.0 ) THEN >*/ if (*m < 0) { /*< INFO = -1 >*/ *info = -1; /*< ELSE IF( N.LT.0 .OR. N.GT.M ) THEN >*/ } else if (*n < 0 || *n > *m) { /*< INFO = -2 >*/ *info = -2; /*< ELSE IF( K.LT.0 .OR. K.GT.N ) THEN >*/ } else if (*k < 0 || *k > *n) { /*< INFO = -3 >*/ *info = -3; /*< ELSE IF( LDA.LT.MAX( 1, M ) ) THEN >*/ } else if (*lda < max(1,*m)) { /*< INFO = -5 >*/ *info = -5; /*< ELSE IF( LWORK.LT.MAX( 1, N ) .AND. .NOT.LQUERY ) THEN >*/ } else if (*lwork < max(1,*n) && ! lquery) { /*< INFO = -8 >*/ *info = -8; /*< END IF >*/ } /*< IF( INFO.NE.0 ) THEN >*/ if (*info != 0) { /*< CALL XERBLA( 'DORGQR', -INFO ) >*/ i__1 = -(*info); xerbla_("DORGQR", &i__1, (ftnlen)6); /*< RETURN >*/ return 0; /*< ELSE IF( LQUERY ) THEN >*/ } else if (lquery) { /*< RETURN >*/ return 0; /*< END IF >*/ } /* Quick return if possible */ /*< IF( N.LE.0 ) THEN >*/ if (*n <= 0) { /*< WORK( 1 ) = 1 >*/ work[1] = 1.; /*< RETURN >*/ return 0; /*< END IF >*/ } /*< NBMIN = 2 >*/ nbmin = 2; /*< NX = 0 >*/ nx = 0; /*< IWS = N >*/ iws = *n; /*< IF( NB.GT.1 .AND. NB.LT.K ) THEN >*/ if (nb > 1 && nb < *k) { /* Determine when to cross over from blocked to unblocked code. */ /*< NX = MAX( 0, ILAENV( 3, 'DORGQR', ' ', M, N, K, -1 ) ) >*/ /* Computing MAX */ i__1 = 0, i__2 = ilaenv_(&c__3, "DORGQR", " ", m, n, k, &c_n1, ( ftnlen)6, (ftnlen)1); nx = max(i__1,i__2); /*< IF( NX.LT.K ) THEN >*/ if (nx < *k) { /* Determine if workspace is large enough for blocked code. */ /*< LDWORK = N >*/ ldwork = *n; /*< IWS = LDWORK*NB >*/ iws = ldwork * nb; /*< IF( LWORK.LT.IWS ) THEN >*/ if (*lwork < iws) { /* Not enough workspace to use optimal NB: reduce NB and */ /* determine the minimum value of NB. */ /*< NB = LWORK / LDWORK >*/ nb = *lwork / ldwork; /*< NBMIN = MAX( 2, ILAENV( 2, 'DORGQR', ' ', M, N, K, -1 ) ) >*/ /* Computing MAX */ i__1 = 2, i__2 = ilaenv_(&c__2, "DORGQR", " ", m, n, k, &c_n1, (ftnlen)6, (ftnlen)1); nbmin = max(i__1,i__2); /*< END IF >*/ } /*< END IF >*/ } /*< END IF >*/ } /*< IF( NB.GE.NBMIN .AND. NB.LT.K .AND. NX.LT.K ) THEN >*/ if (nb >= nbmin && nb < *k && nx < *k) { /* Use blocked code after the last block. */ /* The first kk columns are handled by the block method. */ /*< KI = ( ( K-NX-1 ) / NB )*NB >*/ ki = (*k - nx - 1) / nb * nb; /*< KK = MIN( K, KI+NB ) >*/ /* Computing MIN */ i__1 = *k, i__2 = ki + nb; kk = min(i__1,i__2); /* Set A(1:kk,kk+1:n) to zero. */ /*< DO 20 J = KK + 1, N >*/ i__1 = *n; for (j = kk + 1; j <= i__1; ++j) { /*< DO 10 I = 1, KK >*/ i__2 = kk; for (i__ = 1; i__ <= i__2; ++i__) { /*< A( I, J ) = ZERO >*/ a[i__ + j * a_dim1] = 0.; /*< 10 CONTINUE >*/ /* L10: */ } /*< 20 CONTINUE >*/ /* L20: */ } /*< ELSE >*/ } else { /*< KK = 0 >*/ kk = 0; /*< END IF >*/ } /* Use unblocked code for the last or only block. */ /*< >*/ if (kk < *n) { i__1 = *m - kk; i__2 = *n - kk; i__3 = *k - kk; dorg2r_(&i__1, &i__2, &i__3, &a[kk + 1 + (kk + 1) * a_dim1], lda, & tau[kk + 1], &work[1], &iinfo); } /*< IF( KK.GT.0 ) THEN >*/ if (kk > 0) { /* Use blocked code */ /*< DO 50 I = KI + 1, 1, -NB >*/ i__1 = -nb; for (i__ = ki + 1; i__1 < 0 ? i__ >= 1 : i__ <= 1; i__ += i__1) { /*< IB = MIN( NB, K-I+1 ) >*/ /* Computing MIN */ i__2 = nb, i__3 = *k - i__ + 1; ib = min(i__2,i__3); /*< IF( I+IB.LE.N ) THEN >*/ if (i__ + ib <= *n) { /* Form the triangular factor of the block reflector */ /* H = H(i) H(i+1) . . . H(i+ib-1) */ /*< >*/ i__2 = *m - i__ + 1; dlarft_("Forward", "Columnwise", &i__2, &ib, &a[i__ + i__ * a_dim1], lda, &tau[i__], &work[1], &ldwork, (ftnlen)7, (ftnlen)10); /* Apply H to A(i:m,i+ib:n) from the left */ /*< >*/ i__2 = *m - i__ + 1; i__3 = *n - i__ - ib + 1; dlarfb_("Left", "No transpose", "Forward", "Columnwise", & i__2, &i__3, &ib, &a[i__ + i__ * a_dim1], lda, &work[ 1], &ldwork, &a[i__ + (i__ + ib) * a_dim1], lda, & work[ib + 1], &ldwork, (ftnlen)4, (ftnlen)12, (ftnlen) 7, (ftnlen)10); /*< END IF >*/ } /* Apply H to rows i:m of current block */ /*< >*/ i__2 = *m - i__ + 1; dorg2r_(&i__2, &ib, &ib, &a[i__ + i__ * a_dim1], lda, &tau[i__], & work[1], &iinfo); /* Set rows 1:i-1 of current block to zero */ /*< DO 40 J = I, I + IB - 1 >*/ i__2 = i__ + ib - 1; for (j = i__; j <= i__2; ++j) { /*< DO 30 L = 1, I - 1 >*/ i__3 = i__ - 1; for (l = 1; l <= i__3; ++l) { /*< A( L, J ) = ZERO >*/ a[l + j * a_dim1] = 0.; /*< 30 CONTINUE >*/ /* L30: */ } /*< 40 CONTINUE >*/ /* L40: */ } /*< 50 CONTINUE >*/ /* L50: */ } /*< END IF >*/ } /*< WORK( 1 ) = IWS >*/ work[1] = (doublereal) iws; /*< RETURN >*/ return 0; /* End of DORGQR */ /*< END >*/ } /* dorgqr_ */
/* Subroutine */ int dgeqrf_(integer *m, integer *n, doublereal *a, integer * lda, doublereal *tau, doublereal *work, integer *lwork, integer *info) { /* System generated locals */ integer a_dim1, a_offset, i__1, i__2, i__3, i__4; /* Local variables */ integer i__, k, ib, nb, nx, iws, nbmin, iinfo; extern /* Subroutine */ int dgeqr2_(integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), dlarfb_(char *, char *, char *, char *, integer *, integer *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *), dlarft_(char *, char *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), xerbla_(char *, integer *); extern integer ilaenv_(integer *, char *, char *, integer *, integer *, integer *, integer *); integer ldwork, lwkopt; logical lquery; /* -- LAPACK routine (version 3.2) -- */ /* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */ /* November 2006 */ /* .. Scalar Arguments .. */ /* .. */ /* .. Array Arguments .. */ /* .. */ /* Purpose */ /* ======= */ /* DGEQRF computes a QR factorization of a real M-by-N matrix A: */ /* A = 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) DOUBLE PRECISION array, dimension (LDA,N) */ /* On entry, the M-by-N matrix A. */ /* On exit, the elements on and above the diagonal of the array */ /* contain the min(M,N)-by-N upper trapezoidal matrix R (R is */ /* upper triangular if m >= n); the elements below the diagonal, */ /* with the array TAU, represent the orthogonal matrix Q as a */ /* product of min(m,n) elementary reflectors (see Further */ /* Details). */ /* LDA (input) INTEGER */ /* The leading dimension of the array A. LDA >= max(1,M). */ /* TAU (output) DOUBLE PRECISION array, dimension (min(M,N)) */ /* The scalar factors of the elementary reflectors (see Further */ /* Details). */ /* WORK (workspace/output) DOUBLE PRECISION array, dimension (MAX(1,LWORK)) */ /* On exit, if INFO = 0, WORK(1) returns the optimal LWORK. */ /* LWORK (input) INTEGER */ /* The dimension of the array WORK. LWORK >= max(1,N). */ /* For optimum performance LWORK >= N*NB, where NB is */ /* the optimal blocksize. */ /* If LWORK = -1, then a workspace query is assumed; the routine */ /* only calculates the optimal size of the WORK array, returns */ /* this value as the first entry of the WORK array, and no error */ /* message related to LWORK is issued by XERBLA. */ /* INFO (output) INTEGER */ /* = 0: successful exit */ /* < 0: if INFO = -i, the i-th argument had an illegal value */ /* 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(1:i-1) = 0 and v(i) = 1; v(i+1:m) is stored on exit in A(i+1:m,i), */ /* and tau in TAU(i). */ /* ===================================================================== */ /* .. 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; --tau; --work; /* Function Body */ *info = 0; nb = ilaenv_(&c__1, "DGEQRF", " ", m, n, &c_n1, &c_n1); lwkopt = *n * nb; work[1] = (doublereal) lwkopt; lquery = *lwork == -1; if (*m < 0) { *info = -1; } else if (*n < 0) { *info = -2; } else if (*lda < max(1,*m)) { *info = -4; } else if (*lwork < max(1,*n) && ! lquery) { *info = -7; } if (*info != 0) { i__1 = -(*info); xerbla_("DGEQRF", &i__1); return 0; } else if (lquery) { return 0; } /* Quick return if possible */ k = min(*m,*n); if (k == 0) { work[1] = 1.; return 0; } nbmin = 2; nx = 0; iws = *n; if (nb > 1 && nb < k) { /* Determine when to cross over from blocked to unblocked code. */ /* Computing MAX */ i__1 = 0, i__2 = ilaenv_(&c__3, "DGEQRF", " ", m, n, &c_n1, &c_n1); nx = max(i__1,i__2); if (nx < k) { /* Determine if workspace is large enough for blocked code. */ ldwork = *n; iws = ldwork * nb; if (*lwork < iws) { /* Not enough workspace to use optimal NB: reduce NB and */ /* determine the minimum value of NB. */ nb = *lwork / ldwork; /* Computing MAX */ i__1 = 2, i__2 = ilaenv_(&c__2, "DGEQRF", " ", m, n, &c_n1, & c_n1); nbmin = max(i__1,i__2); } } } if (nb >= nbmin && nb < k && nx < k) { /* Use blocked code initially */ i__1 = k - nx; i__2 = nb; for (i__ = 1; i__2 < 0 ? i__ >= i__1 : i__ <= i__1; i__ += i__2) { /* Computing MIN */ i__3 = k - i__ + 1; ib = min(i__3,nb); /* Compute the QR factorization of the current block */ /* A(i:m,i:i+ib-1) */ i__3 = *m - i__ + 1; dgeqr2_(&i__3, &ib, &a[i__ + i__ * a_dim1], lda, &tau[i__], &work[ 1], &iinfo); if (i__ + ib <= *n) { /* Form the triangular factor of the block reflector */ /* H = H(i) H(i+1) . . . H(i+ib-1) */ i__3 = *m - i__ + 1; dlarft_("Forward", "Columnwise", &i__3, &ib, &a[i__ + i__ * a_dim1], lda, &tau[i__], &work[1], &ldwork); /* Apply H' to A(i:m,i+ib:n) from the left */ i__3 = *m - i__ + 1; i__4 = *n - i__ - ib + 1; dlarfb_("Left", "Transpose", "Forward", "Columnwise", &i__3, & i__4, &ib, &a[i__ + i__ * a_dim1], lda, &work[1], & ldwork, &a[i__ + (i__ + ib) * a_dim1], lda, &work[ib + 1], &ldwork); } /* L10: */ } } else { i__ = 1; } /* Use unblocked code to factor the last or only block. */ if (i__ <= k) { i__2 = *m - i__ + 1; i__1 = *n - i__ + 1; dgeqr2_(&i__2, &i__1, &a[i__ + i__ * a_dim1], lda, &tau[i__], &work[1] , &iinfo); } work[1] = (doublereal) iws; return 0; /* End of DGEQRF */ } /* dgeqrf_ */
/*< SUBROUTINE DGEQRF( M, N, A, LDA, TAU, WORK, LWORK, INFO ) >*/ /* Subroutine */ int dgeqrf_(integer *m, integer *n, doublereal *a, integer * lda, doublereal *tau, doublereal *work, integer *lwork, integer *info) { /* System generated locals */ integer a_dim1, a_offset, i__1, i__2, i__3, i__4; /* Local variables */ integer i__, k, ib, nb, nx, iws, nbmin, iinfo; extern /* Subroutine */ int dgeqr2_(integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *), dlarfb_(char *, char *, char *, char *, integer *, integer *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *, doublereal *, integer *, ftnlen, ftnlen, ftnlen, ftnlen), dlarft_(char *, char *, integer *, integer *, doublereal *, integer *, doublereal *, doublereal *, integer *, ftnlen, ftnlen), xerbla_(char *, integer *, ftnlen); extern integer ilaenv_(integer *, char *, char *, integer *, integer *, integer *, integer *, ftnlen, ftnlen); integer ldwork, lwkopt; 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 .. */ /*< INTEGER INFO, LDA, LWORK, M, N >*/ /* .. */ /* .. Array Arguments .. */ /*< DOUBLE PRECISION A( LDA, * ), TAU( * ), WORK( * ) >*/ /* .. */ /* Purpose */ /* ======= */ /* DGEQRF computes a QR factorization of a real M-by-N matrix A: */ /* A = 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) DOUBLE PRECISION array, dimension (LDA,N) */ /* On entry, the M-by-N matrix A. */ /* On exit, the elements on and above the diagonal of the array */ /* contain the min(M,N)-by-N upper trapezoidal matrix R (R is */ /* upper triangular if m >= n); the elements below the diagonal, */ /* with the array TAU, represent the orthogonal matrix Q as a */ /* product of min(m,n) elementary reflectors (see Further */ /* Details). */ /* LDA (input) INTEGER */ /* The leading dimension of the array A. LDA >= max(1,M). */ /* TAU (output) DOUBLE PRECISION array, dimension (min(M,N)) */ /* The scalar factors of the elementary reflectors (see Further */ /* Details). */ /* WORK (workspace/output) DOUBLE PRECISION array, dimension (LWORK) */ /* On exit, if INFO = 0, WORK(1) returns the optimal LWORK. */ /* LWORK (input) INTEGER */ /* The dimension of the array WORK. LWORK >= max(1,N). */ /* For optimum performance LWORK >= N*NB, where NB is */ /* the optimal blocksize. */ /* If LWORK = -1, then a workspace query is assumed; the routine */ /* only calculates the optimal size of the WORK array, returns */ /* this value as the first entry of the WORK array, and no error */ /* message related to LWORK is issued by XERBLA. */ /* INFO (output) INTEGER */ /* = 0: successful exit */ /* < 0: if INFO = -i, the i-th argument had an illegal value */ /* 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(1:i-1) = 0 and v(i) = 1; v(i+1:m) is stored on exit in A(i+1:m,i), */ /* and tau in TAU(i). */ /* ===================================================================== */ /* .. Local Scalars .. */ /*< LOGICAL LQUERY >*/ /*< >*/ /* .. */ /* .. External Subroutines .. */ /*< EXTERNAL DGEQR2, DLARFB, DLARFT, XERBLA >*/ /* .. */ /* .. Intrinsic Functions .. */ /*< INTRINSIC MAX, MIN >*/ /* .. */ /* .. External Functions .. */ /*< INTEGER ILAENV >*/ /*< EXTERNAL ILAENV >*/ /* .. */ /* .. Executable Statements .. */ /* Test the input arguments */ /*< INFO = 0 >*/ /* Parameter adjustments */ a_dim1 = *lda; a_offset = 1 + a_dim1; a -= a_offset; --tau; --work; /* Function Body */ *info = 0; /*< NB = ILAENV( 1, 'DGEQRF', ' ', M, N, -1, -1 ) >*/ nb = ilaenv_(&c__1, "DGEQRF", " ", m, n, &c_n1, &c_n1, (ftnlen)6, (ftnlen) 1); /*< LWKOPT = N*NB >*/ lwkopt = *n * nb; /*< WORK( 1 ) = LWKOPT >*/ work[1] = (doublereal) lwkopt; /*< LQUERY = ( LWORK.EQ.-1 ) >*/ lquery = *lwork == -1; /*< IF( M.LT.0 ) THEN >*/ if (*m < 0) { /*< INFO = -1 >*/ *info = -1; /*< ELSE IF( N.LT.0 ) THEN >*/ } else if (*n < 0) { /*< INFO = -2 >*/ *info = -2; /*< ELSE IF( LDA.LT.MAX( 1, M ) ) THEN >*/ } else if (*lda < max(1,*m)) { /*< INFO = -4 >*/ *info = -4; /*< ELSE IF( LWORK.LT.MAX( 1, N ) .AND. .NOT.LQUERY ) THEN >*/ } else if (*lwork < max(1,*n) && ! lquery) { /*< INFO = -7 >*/ *info = -7; /*< END IF >*/ } /*< IF( INFO.NE.0 ) THEN >*/ if (*info != 0) { /*< CALL XERBLA( 'DGEQRF', -INFO ) >*/ i__1 = -(*info); xerbla_("DGEQRF", &i__1, (ftnlen)6); /*< RETURN >*/ return 0; /*< ELSE IF( LQUERY ) THEN >*/ } else if (lquery) { /*< RETURN >*/ return 0; /*< END IF >*/ } /* Quick return if possible */ /*< K = MIN( M, N ) >*/ k = min(*m,*n); /*< IF( K.EQ.0 ) THEN >*/ if (k == 0) { /*< WORK( 1 ) = 1 >*/ work[1] = 1.; /*< RETURN >*/ return 0; /*< END IF >*/ } /*< NBMIN = 2 >*/ nbmin = 2; /*< NX = 0 >*/ nx = 0; /*< IWS = N >*/ iws = *n; /*< IF( NB.GT.1 .AND. NB.LT.K ) THEN >*/ if (nb > 1 && nb < k) { /* Determine when to cross over from blocked to unblocked code. */ /*< NX = MAX( 0, ILAENV( 3, 'DGEQRF', ' ', M, N, -1, -1 ) ) >*/ /* Computing MAX */ i__1 = 0, i__2 = ilaenv_(&c__3, "DGEQRF", " ", m, n, &c_n1, &c_n1, ( ftnlen)6, (ftnlen)1); nx = max(i__1,i__2); /*< IF( NX.LT.K ) THEN >*/ if (nx < k) { /* Determine if workspace is large enough for blocked code. */ /*< LDWORK = N >*/ ldwork = *n; /*< IWS = LDWORK*NB >*/ iws = ldwork * nb; /*< IF( LWORK.LT.IWS ) THEN >*/ if (*lwork < iws) { /* Not enough workspace to use optimal NB: reduce NB and */ /* determine the minimum value of NB. */ /*< NB = LWORK / LDWORK >*/ nb = *lwork / ldwork; /*< >*/ /* Computing MAX */ i__1 = 2, i__2 = ilaenv_(&c__2, "DGEQRF", " ", m, n, &c_n1, & c_n1, (ftnlen)6, (ftnlen)1); nbmin = max(i__1,i__2); /*< END IF >*/ } /*< END IF >*/ } /*< END IF >*/ } /*< IF( NB.GE.NBMIN .AND. NB.LT.K .AND. NX.LT.K ) THEN >*/ if (nb >= nbmin && nb < k && nx < k) { /* Use blocked code initially */ /*< DO 10 I = 1, K - NX, NB >*/ i__1 = k - nx; i__2 = nb; for (i__ = 1; i__2 < 0 ? i__ >= i__1 : i__ <= i__1; i__ += i__2) { /*< IB = MIN( K-I+1, NB ) >*/ /* Computing MIN */ i__3 = k - i__ + 1; ib = min(i__3,nb); /* Compute the QR factorization of the current block */ /* A(i:m,i:i+ib-1) */ /*< >*/ i__3 = *m - i__ + 1; dgeqr2_(&i__3, &ib, &a[i__ + i__ * a_dim1], lda, &tau[i__], &work[ 1], &iinfo); /*< IF( I+IB.LE.N ) THEN >*/ if (i__ + ib <= *n) { /* Form the triangular factor of the block reflector */ /* H = H(i) H(i+1) . . . H(i+ib-1) */ /*< >*/ i__3 = *m - i__ + 1; dlarft_("Forward", "Columnwise", &i__3, &ib, &a[i__ + i__ * a_dim1], lda, &tau[i__], &work[1], &ldwork, (ftnlen)7, (ftnlen)10); /* Apply H' to A(i:m,i+ib:n) from the left */ /*< >*/ i__3 = *m - i__ + 1; i__4 = *n - i__ - ib + 1; dlarfb_("Left", "Transpose", "Forward", "Columnwise", &i__3, & i__4, &ib, &a[i__ + i__ * a_dim1], lda, &work[1], & ldwork, &a[i__ + (i__ + ib) * a_dim1], lda, &work[ib + 1], &ldwork, (ftnlen)4, (ftnlen)9, (ftnlen)7, ( ftnlen)10); /*< END IF >*/ } /*< 10 CONTINUE >*/ /* L10: */ } /*< ELSE >*/ } else { /*< I = 1 >*/ i__ = 1; /*< END IF >*/ } /* Use unblocked code to factor the last or only block. */ /*< >*/ if (i__ <= k) { i__2 = *m - i__ + 1; i__1 = *n - i__ + 1; dgeqr2_(&i__2, &i__1, &a[i__ + i__ * a_dim1], lda, &tau[i__], &work[1] , &iinfo); } /*< WORK( 1 ) = IWS >*/ work[1] = (doublereal) iws; /*< RETURN >*/ return 0; /* End of DGEQRF */ /*< END >*/ } /* dgeqrf_ */