 LAPACK  3.10.1 LAPACK: Linear Algebra PACKage

## ◆ dlatdf()

 subroutine dlatdf ( integer IJOB, integer N, double precision, dimension( ldz, * ) Z, integer LDZ, double precision, dimension( * ) RHS, double precision RDSUM, double precision RDSCAL, integer, dimension( * ) IPIV, integer, dimension( * ) JPIV )

DLATDF uses the LU factorization of the n-by-n matrix computed by sgetc2 and computes a contribution to the reciprocal Dif-estimate.

Purpose:
``` DLATDF uses the LU factorization of the n-by-n matrix Z computed by
DGETC2 and computes a contribution to the reciprocal Dif-estimate
by solving Z * x = b for x, and choosing the r.h.s. b such that
the norm of x is as large as possible. On entry RHS = b holds the
contribution from earlier solved sub-systems, and on return RHS = x.

The factorization of Z returned by DGETC2 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.```
Parameters
 [in] IJOB ``` IJOB is INTEGER IJOB = 2: First compute an approximative null-vector e of Z using DGECON, 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 chosen as either +1 or -1 (Default).``` [in] N ``` N is INTEGER The number of columns of the matrix Z.``` [in] Z ``` Z is DOUBLE PRECISION array, dimension (LDZ, N) On entry, the LU part of the factorization of the n-by-n matrix Z computed by DGETC2: Z = P * L * U * Q``` [in] LDZ ``` LDZ is INTEGER The leading dimension of the array Z. LDA >= max(1, N).``` [in,out] RHS ``` RHS is DOUBLE PRECISION 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).``` [in,out] RDSUM ``` RDSUM is DOUBLE PRECISION On entry, the sum of squares of computed contributions to the Dif-estimate under computation by DTGSYL, 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 DTGSY2 is called by STGSYL.``` [in,out] RDSCAL ``` RDSCAL is DOUBLE PRECISION 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 DTGSY2 is called by DTGSYL.``` [in] IPIV ``` IPIV is INTEGER array, dimension (N). The pivot indices; for 1 <= i <= N, row i of the matrix has been interchanged with row IPIV(i).``` [in] JPIV ``` JPIV is 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:
This routine is a further developed implementation of algorithm BSOLVE in  using complete pivoting in the LU factorization.
Contributors:
Bo Kagstrom and Peter Poromaa, Department of Computing Science, Umea University, S-901 87 Umea, Sweden.
References:
```   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.

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

Definition at line 169 of file dlatdf.f.

171 *
172 * -- LAPACK auxiliary routine --
173 * -- LAPACK is a software package provided by Univ. of Tennessee, --
174 * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
175 *
176 * .. Scalar Arguments ..
177  INTEGER IJOB, LDZ, N
178  DOUBLE PRECISION RDSCAL, RDSUM
179 * ..
180 * .. Array Arguments ..
181  INTEGER IPIV( * ), JPIV( * )
182  DOUBLE PRECISION RHS( * ), Z( LDZ, * )
183 * ..
184 *
185 * =====================================================================
186 *
187 * .. Parameters ..
188  INTEGER MAXDIM
189  parameter( maxdim = 8 )
190  DOUBLE PRECISION ZERO, ONE
191  parameter( zero = 0.0d+0, one = 1.0d+0 )
192 * ..
193 * .. Local Scalars ..
194  INTEGER I, INFO, J, K
195  DOUBLE PRECISION BM, BP, PMONE, SMINU, SPLUS, TEMP
196 * ..
197 * .. Local Arrays ..
198  INTEGER IWORK( MAXDIM )
199  DOUBLE PRECISION WORK( 4*MAXDIM ), XM( MAXDIM ), XP( MAXDIM )
200 * ..
201 * .. External Subroutines ..
202  EXTERNAL daxpy, dcopy, dgecon, dgesc2, dlassq, dlaswp,
203  \$ dscal
204 * ..
205 * .. External Functions ..
206  DOUBLE PRECISION DASUM, DDOT
207  EXTERNAL dasum, ddot
208 * ..
209 * .. Intrinsic Functions ..
210  INTRINSIC abs, sqrt
211 * ..
212 * .. Executable Statements ..
213 *
214  IF( ijob.NE.2 ) THEN
215 *
216 * Apply permutations IPIV to RHS
217 *
218  CALL dlaswp( 1, rhs, ldz, 1, n-1, ipiv, 1 )
219 *
220 * Solve for L-part choosing RHS either to +1 or -1.
221 *
222  pmone = -one
223 *
224  DO 10 j = 1, n - 1
225  bp = rhs( j ) + one
226  bm = rhs( j ) - one
227  splus = one
228 *
229 * Look-ahead for L-part RHS(1:N-1) = + or -1, SPLUS and
230 * SMIN computed more efficiently than in BSOLVE .
231 *
232  splus = splus + ddot( n-j, z( j+1, j ), 1, z( j+1, j ), 1 )
233  sminu = ddot( n-j, z( j+1, j ), 1, rhs( j+1 ), 1 )
234  splus = splus*rhs( j )
235  IF( splus.GT.sminu ) THEN
236  rhs( j ) = bp
237  ELSE IF( sminu.GT.splus ) THEN
238  rhs( j ) = bm
239  ELSE
240 *
241 * In this case the updating sums are equal and we can
242 * choose RHS(J) +1 or -1. The first time this happens
243 * we choose -1, thereafter +1. This is a simple way to
244 * get good estimates of matrices like Byers well-known
245 * example (see ). (Not done in BSOLVE.)
246 *
247  rhs( j ) = rhs( j ) + pmone
248  pmone = one
249  END IF
250 *
251 * Compute the remaining r.h.s.
252 *
253  temp = -rhs( j )
254  CALL daxpy( n-j, temp, z( j+1, j ), 1, rhs( j+1 ), 1 )
255 *
256  10 CONTINUE
257 *
258 * Solve for U-part, look-ahead for RHS(N) = +-1. This is not done
259 * in BSOLVE and will hopefully give us a better estimate because
260 * any ill-conditioning of the original matrix is transferred to U
261 * and not to L. U(N, N) is an approximation to sigma_min(LU).
262 *
263  CALL dcopy( n-1, rhs, 1, xp, 1 )
264  xp( n ) = rhs( n ) + one
265  rhs( n ) = rhs( n ) - one
266  splus = zero
267  sminu = zero
268  DO 30 i = n, 1, -1
269  temp = one / z( i, i )
270  xp( i ) = xp( i )*temp
271  rhs( i ) = rhs( i )*temp
272  DO 20 k = i + 1, n
273  xp( i ) = xp( i ) - xp( k )*( z( i, k )*temp )
274  rhs( i ) = rhs( i ) - rhs( k )*( z( i, k )*temp )
275  20 CONTINUE
276  splus = splus + abs( xp( i ) )
277  sminu = sminu + abs( rhs( i ) )
278  30 CONTINUE
279  IF( splus.GT.sminu )
280  \$ CALL dcopy( n, xp, 1, rhs, 1 )
281 *
282 * Apply the permutations JPIV to the computed solution (RHS)
283 *
284  CALL dlaswp( 1, rhs, ldz, 1, n-1, jpiv, -1 )
285 *
286 * Compute the sum of squares
287 *
288  CALL dlassq( n, rhs, 1, rdscal, rdsum )
289 *
290  ELSE
291 *
292 * IJOB = 2, Compute approximate nullvector XM of Z
293 *
294  CALL dgecon( 'I', n, z, ldz, one, temp, work, iwork, info )
295  CALL dcopy( n, work( n+1 ), 1, xm, 1 )
296 *
297 * Compute RHS
298 *
299  CALL dlaswp( 1, xm, ldz, 1, n-1, ipiv, -1 )
300  temp = one / sqrt( ddot( n, xm, 1, xm, 1 ) )
301  CALL dscal( n, temp, xm, 1 )
302  CALL dcopy( n, xm, 1, xp, 1 )
303  CALL daxpy( n, one, rhs, 1, xp, 1 )
304  CALL daxpy( n, -one, xm, 1, rhs, 1 )
305  CALL dgesc2( n, z, ldz, rhs, ipiv, jpiv, temp )
306  CALL dgesc2( n, z, ldz, xp, ipiv, jpiv, temp )
307  IF( dasum( n, xp, 1 ).GT.dasum( n, rhs, 1 ) )
308  \$ CALL dcopy( n, xp, 1, rhs, 1 )
309 *
310 * Compute the sum of squares
311 *
312  CALL dlassq( n, rhs, 1, rdscal, rdsum )
313 *
314  END IF
315 *
316  RETURN
317 *
318 * End of DLATDF
319 *
subroutine dlassq(n, x, incx, scl, sumsq)
DLASSQ updates a sum of squares represented in scaled form.
Definition: dlassq.f90:137
subroutine dcopy(N, DX, INCX, DY, INCY)
DCOPY
Definition: dcopy.f:82
double precision function dasum(N, DX, INCX)
DASUM
Definition: dasum.f:71
double precision function ddot(N, DX, INCX, DY, INCY)
DDOT
Definition: ddot.f:82
subroutine dscal(N, DA, DX, INCX)
DSCAL
Definition: dscal.f:79
subroutine daxpy(N, DA, DX, INCX, DY, INCY)
DAXPY
Definition: daxpy.f:89
subroutine dgesc2(N, A, LDA, RHS, IPIV, JPIV, SCALE)
DGESC2 solves a system of linear equations using the LU factorization with complete pivoting computed...
Definition: dgesc2.f:114
subroutine dgecon(NORM, N, A, LDA, ANORM, RCOND, WORK, IWORK, INFO)
DGECON
Definition: dgecon.f:124
subroutine dlaswp(N, A, LDA, K1, K2, IPIV, INCX)
DLASWP performs a series of row interchanges on a general rectangular matrix.
Definition: dlaswp.f:115
Here is the call graph for this function:
Here is the caller graph for this function: