#include "blaswrap.h"
#include "f2c.h"

/* Subroutine */ int clatdf_(integer *ijob, integer *n, complex *z__, integer 
	*ldz, complex *rhs, real *rdsum, real *rdscal, integer *ipiv, integer 
	*jpiv)
{
/*  -- LAPACK auxiliary routine (version 3.1) --   
       Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd..   
       November 2006   


    Purpose   
    =======   

    CLATDF computes the contribution to the reciprocal Dif-estimate   
    by solving for x in Z * x = b, where b is chosen such that the norm   
    of x is as large as possible. It is assumed that LU decomposition   
    of Z has been computed by CGETC2. On entry RHS = f holds the   
    contribution from earlier solved sub-systems, and on return RHS = x.   

    The factorization of Z returned by CGETC2 has the form   
    Z = P * L * U * Q, where P and Q are permutation matrices. L is lower   
    triangular with unit diagonal elements and U is upper triangular.   

    Arguments   
    =========   

    IJOB    (input) INTEGER   
            IJOB = 2: First compute an approximative null-vector e   
                of Z using CGECON, e is normalized and solve for   
                Zx = +-e - f with the sign giving the greater value of   
                2-norm(x).  About 5 times as expensive as Default.   
            IJOB .ne. 2: Local look ahead strategy where   
                all entries of the r.h.s. b is choosen as either +1 or   
                -1.  Default.   

    N       (input) INTEGER   
            The number of columns of the matrix Z.   

    Z       (input) REAL array, dimension (LDZ, N)   
            On entry, the LU part of the factorization of the n-by-n   
            matrix Z computed by CGETC2:  Z = P * L * U * Q   

    LDZ     (input) INTEGER   
            The leading dimension of the array Z.  LDA >= max(1, N).   

    RHS     (input/output) REAL array, dimension (N).   
            On entry, RHS contains contributions from other subsystems.   
            On exit, RHS contains the solution of the subsystem with   
            entries according to the value of IJOB (see above).   

    RDSUM   (input/output) REAL   
            On entry, the sum of squares of computed contributions to   
            the Dif-estimate under computation by CTGSYL, where the   
            scaling factor RDSCAL (see below) has been factored out.   
            On exit, the corresponding sum of squares updated with the   
            contributions from the current sub-system.   
            If TRANS = 'T' RDSUM is not touched.   
            NOTE: RDSUM only makes sense when CTGSY2 is called by CTGSYL.   

    RDSCAL  (input/output) REAL   
            On entry, scaling factor used to prevent overflow in RDSUM.   
            On exit, RDSCAL is updated w.r.t. the current contributions   
            in RDSUM.   
            If TRANS = 'T', RDSCAL is not touched.   
            NOTE: RDSCAL only makes sense when CTGSY2 is called by   
            CTGSYL.   

    IPIV    (input) INTEGER array, dimension (N).   
            The pivot indices; for 1 <= i <= N, row i of the   
            matrix has been interchanged with row IPIV(i).   

    JPIV    (input) INTEGER array, dimension (N).   
            The pivot indices; for 1 <= j <= N, column j of the   
            matrix has been interchanged with column JPIV(j).   

    Further Details   
    ===============   

    Based on contributions by   
       Bo Kagstrom and Peter Poromaa, Department of Computing Science,   
       Umea University, S-901 87 Umea, Sweden.   

    This routine is a further developed implementation of algorithm   
    BSOLVE in [1] using complete pivoting in the LU factorization.   

     [1]   Bo Kagstrom and Lars Westin,   
           Generalized Schur Methods with Condition Estimators for   
           Solving the Generalized Sylvester Equation, IEEE Transactions   
           on Automatic Control, Vol. 34, No. 7, July 1989, pp 745-751.   

     [2]   Peter Poromaa,   
           On Efficient and Robust Estimators for the Separation   
           between two Regular Matrix Pairs with Applications in   
           Condition Estimation. Report UMINF-95.05, Department of   
           Computing Science, Umea University, S-901 87 Umea, Sweden,   
           1995.   

    =====================================================================   


       Parameter adjustments */
    /* Table of constant values */
    static complex c_b1 = {1.f,0.f};
    static integer c__1 = 1;
    static integer c_n1 = -1;
    static real c_b24 = 1.f;
    
    /* System generated locals */
    integer z_dim1, z_offset, i__1, i__2, i__3, i__4, i__5;
    complex q__1, q__2, q__3;
    /* Builtin functions */
    void c_div(complex *, complex *, complex *);
    double c_abs(complex *);
    void c_sqrt(complex *, complex *);
    /* Local variables */
    static integer i__, j, k;
    static complex bm, bp, xm[2], xp[2];
    static integer info;
    static complex temp, work[8];
    extern /* Subroutine */ int cscal_(integer *, complex *, complex *, 
	    integer *);
    static real scale;
    extern /* Complex */ VOID cdotc_(complex *, integer *, complex *, integer 
	    *, complex *, integer *);
    extern /* Subroutine */ int ccopy_(integer *, complex *, integer *, 
	    complex *, integer *);
    static complex pmone;
    extern /* Subroutine */ int caxpy_(integer *, complex *, complex *, 
	    integer *, complex *, integer *);
    static real rtemp, sminu, rwork[2], splus;
    extern /* Subroutine */ int cgesc2_(integer *, complex *, integer *, 
	    complex *, integer *, integer *, real *), cgecon_(char *, integer 
	    *, complex *, integer *, real *, real *, complex *, real *, 
	    integer *), classq_(integer *, complex *, integer *, real 
	    *, real *), claswp_(integer *, complex *, integer *, integer *, 
	    integer *, integer *, integer *);
    extern doublereal scasum_(integer *, complex *, integer *);


    z_dim1 = *ldz;
    z_offset = 1 + z_dim1;
    z__ -= z_offset;
    --rhs;
    --ipiv;
    --jpiv;

    /* Function Body */
    if (*ijob != 2) {

/*        Apply permutations IPIV to RHS */

	i__1 = *n - 1;
	claswp_(&c__1, &rhs[1], ldz, &c__1, &i__1, &ipiv[1], &c__1);

/*        Solve for L-part choosing RHS either to +1 or -1. */

	q__1.r = -1.f, q__1.i = -0.f;
	pmone.r = q__1.r, pmone.i = q__1.i;
	i__1 = *n - 1;
	for (j = 1; j <= i__1; ++j) {
	    i__2 = j;
	    q__1.r = rhs[i__2].r + 1.f, q__1.i = rhs[i__2].i + 0.f;
	    bp.r = q__1.r, bp.i = q__1.i;
	    i__2 = j;
	    q__1.r = rhs[i__2].r - 1.f, q__1.i = rhs[i__2].i - 0.f;
	    bm.r = q__1.r, bm.i = q__1.i;
	    splus = 1.f;

/*           Lockahead for L- part RHS(1:N-1) = +-1   
             SPLUS and SMIN computed more efficiently than in BSOLVE[1]. */

	    i__2 = *n - j;
	    cdotc_(&q__1, &i__2, &z__[j + 1 + j * z_dim1], &c__1, &z__[j + 1 
		    + j * z_dim1], &c__1);
	    splus += q__1.r;
	    i__2 = *n - j;
	    cdotc_(&q__1, &i__2, &z__[j + 1 + j * z_dim1], &c__1, &rhs[j + 1], 
		     &c__1);
	    sminu = q__1.r;
	    i__2 = j;
	    splus *= rhs[i__2].r;
	    if (splus > sminu) {
		i__2 = j;
		rhs[i__2].r = bp.r, rhs[i__2].i = bp.i;
	    } else if (sminu > splus) {
		i__2 = j;
		rhs[i__2].r = bm.r, rhs[i__2].i = bm.i;
	    } else {

/*              In this case the updating sums are equal and we can   
                choose RHS(J) +1 or -1. The first time this happens we   
                choose -1, thereafter +1. This is a simple way to get   
                good estimates of matrices like Byers well-known example   
                (see [1]). (Not done in BSOLVE.) */

		i__2 = j;
		i__3 = j;
		q__1.r = rhs[i__3].r + pmone.r, q__1.i = rhs[i__3].i + 
			pmone.i;
		rhs[i__2].r = q__1.r, rhs[i__2].i = q__1.i;
		pmone.r = 1.f, pmone.i = 0.f;
	    }

/*           Compute the remaining r.h.s. */

	    i__2 = j;
	    q__1.r = -rhs[i__2].r, q__1.i = -rhs[i__2].i;
	    temp.r = q__1.r, temp.i = q__1.i;
	    i__2 = *n - j;
	    caxpy_(&i__2, &temp, &z__[j + 1 + j * z_dim1], &c__1, &rhs[j + 1], 
		     &c__1);
/* L10: */
	}

/*        Solve for U- part, lockahead for RHS(N) = +-1. This is not done   
          In BSOLVE and will hopefully give us a better estimate because   
          any ill-conditioning of the original matrix is transfered to U   
          and not to L. U(N, N) is an approximation to sigma_min(LU). */

	i__1 = *n - 1;
	ccopy_(&i__1, &rhs[1], &c__1, work, &c__1);
	i__1 = *n - 1;
	i__2 = *n;
	q__1.r = rhs[i__2].r + 1.f, q__1.i = rhs[i__2].i + 0.f;
	work[i__1].r = q__1.r, work[i__1].i = q__1.i;
	i__1 = *n;
	i__2 = *n;
	q__1.r = rhs[i__2].r - 1.f, q__1.i = rhs[i__2].i - 0.f;
	rhs[i__1].r = q__1.r, rhs[i__1].i = q__1.i;
	splus = 0.f;
	sminu = 0.f;
	for (i__ = *n; i__ >= 1; --i__) {
	    c_div(&q__1, &c_b1, &z__[i__ + i__ * z_dim1]);
	    temp.r = q__1.r, temp.i = q__1.i;
	    i__1 = i__ - 1;
	    i__2 = i__ - 1;
	    q__1.r = work[i__2].r * temp.r - work[i__2].i * temp.i, q__1.i = 
		    work[i__2].r * temp.i + work[i__2].i * temp.r;
	    work[i__1].r = q__1.r, work[i__1].i = q__1.i;
	    i__1 = i__;
	    i__2 = i__;
	    q__1.r = rhs[i__2].r * temp.r - rhs[i__2].i * temp.i, q__1.i = 
		    rhs[i__2].r * temp.i + rhs[i__2].i * temp.r;
	    rhs[i__1].r = q__1.r, rhs[i__1].i = q__1.i;
	    i__1 = *n;
	    for (k = i__ + 1; k <= i__1; ++k) {
		i__2 = i__ - 1;
		i__3 = i__ - 1;
		i__4 = k - 1;
		i__5 = i__ + k * z_dim1;
		q__3.r = z__[i__5].r * temp.r - z__[i__5].i * temp.i, q__3.i =
			 z__[i__5].r * temp.i + z__[i__5].i * temp.r;
		q__2.r = work[i__4].r * q__3.r - work[i__4].i * q__3.i, 
			q__2.i = work[i__4].r * q__3.i + work[i__4].i * 
			q__3.r;
		q__1.r = work[i__3].r - q__2.r, q__1.i = work[i__3].i - 
			q__2.i;
		work[i__2].r = q__1.r, work[i__2].i = q__1.i;
		i__2 = i__;
		i__3 = i__;
		i__4 = k;
		i__5 = i__ + k * z_dim1;
		q__3.r = z__[i__5].r * temp.r - z__[i__5].i * temp.i, q__3.i =
			 z__[i__5].r * temp.i + z__[i__5].i * temp.r;
		q__2.r = rhs[i__4].r * q__3.r - rhs[i__4].i * q__3.i, q__2.i =
			 rhs[i__4].r * q__3.i + rhs[i__4].i * q__3.r;
		q__1.r = rhs[i__3].r - q__2.r, q__1.i = rhs[i__3].i - q__2.i;
		rhs[i__2].r = q__1.r, rhs[i__2].i = q__1.i;
/* L20: */
	    }
	    splus += c_abs(&work[i__ - 1]);
	    sminu += c_abs(&rhs[i__]);
/* L30: */
	}
	if (splus > sminu) {
	    ccopy_(n, work, &c__1, &rhs[1], &c__1);
	}

/*        Apply the permutations JPIV to the computed solution (RHS) */

	i__1 = *n - 1;
	claswp_(&c__1, &rhs[1], ldz, &c__1, &i__1, &jpiv[1], &c_n1);

/*        Compute the sum of squares */

	classq_(n, &rhs[1], &c__1, rdscal, rdsum);
	return 0;
    }

/*     ENTRY IJOB = 2   

       Compute approximate nullvector XM of Z */

    cgecon_("I", n, &z__[z_offset], ldz, &c_b24, &rtemp, work, rwork, &info);
    ccopy_(n, &work[*n], &c__1, xm, &c__1);

/*     Compute RHS */

    i__1 = *n - 1;
    claswp_(&c__1, xm, ldz, &c__1, &i__1, &ipiv[1], &c_n1);
    cdotc_(&q__3, n, xm, &c__1, xm, &c__1);
    c_sqrt(&q__2, &q__3);
    c_div(&q__1, &c_b1, &q__2);
    temp.r = q__1.r, temp.i = q__1.i;
    cscal_(n, &temp, xm, &c__1);
    ccopy_(n, xm, &c__1, xp, &c__1);
    caxpy_(n, &c_b1, &rhs[1], &c__1, xp, &c__1);
    q__1.r = -1.f, q__1.i = -0.f;
    caxpy_(n, &q__1, xm, &c__1, &rhs[1], &c__1);
    cgesc2_(n, &z__[z_offset], ldz, &rhs[1], &ipiv[1], &jpiv[1], &scale);
    cgesc2_(n, &z__[z_offset], ldz, xp, &ipiv[1], &jpiv[1], &scale);
    if (scasum_(n, xp, &c__1) > scasum_(n, &rhs[1], &c__1)) {
	ccopy_(n, xp, &c__1, &rhs[1], &c__1);
    }

/*     Compute the sum of squares */

    classq_(n, &rhs[1], &c__1, rdscal, rdsum);
    return 0;

/*     End of CLATDF */

} /* clatdf_ */