 LAPACK  3.10.1 LAPACK: Linear Algebra PACKage

## ◆ dpprfs()

 subroutine dpprfs ( character UPLO, integer N, integer NRHS, double precision, dimension( * ) AP, double precision, dimension( * ) AFP, double precision, dimension( ldb, * ) B, integer LDB, double precision, dimension( ldx, * ) X, integer LDX, double precision, dimension( * ) FERR, double precision, dimension( * ) BERR, double precision, dimension( * ) WORK, integer, dimension( * ) IWORK, integer INFO )

DPPRFS

Purpose:
``` DPPRFS improves the computed solution to a system of linear
equations when the coefficient matrix is symmetric positive definite
and packed, and provides error bounds and backward error estimates
for the solution.```
Parameters
 [in] UPLO ``` UPLO is CHARACTER*1 = 'U': Upper triangle of A is stored; = 'L': Lower triangle of A is stored.``` [in] N ``` N is INTEGER The order of the matrix A. N >= 0.``` [in] NRHS ``` NRHS is INTEGER The number of right hand sides, i.e., the number of columns of the matrices B and X. NRHS >= 0.``` [in] AP ``` AP is DOUBLE PRECISION array, dimension (N*(N+1)/2) The upper or lower triangle of the symmetric matrix A, packed columnwise in a linear array. The j-th column of A is stored in the array AP as follows: if UPLO = 'U', AP(i + (j-1)*j/2) = A(i,j) for 1<=i<=j; if UPLO = 'L', AP(i + (j-1)*(2n-j)/2) = A(i,j) for j<=i<=n.``` [in] AFP ``` AFP is DOUBLE PRECISION array, dimension (N*(N+1)/2) The triangular factor U or L from the Cholesky factorization A = U**T*U or A = L*L**T, as computed by DPPTRF/ZPPTRF, packed columnwise in a linear array in the same format as A (see AP).``` [in] B ``` B is DOUBLE PRECISION array, dimension (LDB,NRHS) The right hand side matrix B.``` [in] LDB ``` LDB is INTEGER The leading dimension of the array B. LDB >= max(1,N).``` [in,out] X ``` X is DOUBLE PRECISION array, dimension (LDX,NRHS) On entry, the solution matrix X, as computed by DPPTRS. On exit, the improved solution matrix X.``` [in] LDX ``` LDX is INTEGER The leading dimension of the array X. LDX >= max(1,N).``` [out] FERR ``` FERR is DOUBLE PRECISION array, dimension (NRHS) The estimated forward error bound for each solution vector X(j) (the j-th column of the solution matrix X). If XTRUE is the true solution corresponding to X(j), FERR(j) is an estimated upper bound for the magnitude of the largest element in (X(j) - XTRUE) divided by the magnitude of the largest element in X(j). The estimate is as reliable as the estimate for RCOND, and is almost always a slight overestimate of the true error.``` [out] BERR ``` BERR is DOUBLE PRECISION array, dimension (NRHS) The componentwise relative backward error of each solution vector X(j) (i.e., the smallest relative change in any element of A or B that makes X(j) an exact solution).``` [out] WORK ` WORK is DOUBLE PRECISION array, dimension (3*N)` [out] IWORK ` IWORK is INTEGER array, dimension (N)` [out] INFO ``` INFO is INTEGER = 0: successful exit < 0: if INFO = -i, the i-th argument had an illegal value```
Internal Parameters:
`  ITMAX is the maximum number of steps of iterative refinement.`

Definition at line 169 of file dpprfs.f.

171 *
172 * -- LAPACK computational 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  CHARACTER UPLO
178  INTEGER INFO, LDB, LDX, N, NRHS
179 * ..
180 * .. Array Arguments ..
181  INTEGER IWORK( * )
182  DOUBLE PRECISION AFP( * ), AP( * ), B( LDB, * ), BERR( * ),
183  \$ FERR( * ), WORK( * ), X( LDX, * )
184 * ..
185 *
186 * =====================================================================
187 *
188 * .. Parameters ..
189  INTEGER ITMAX
190  parameter( itmax = 5 )
191  DOUBLE PRECISION ZERO
192  parameter( zero = 0.0d+0 )
193  DOUBLE PRECISION ONE
194  parameter( one = 1.0d+0 )
195  DOUBLE PRECISION TWO
196  parameter( two = 2.0d+0 )
197  DOUBLE PRECISION THREE
198  parameter( three = 3.0d+0 )
199 * ..
200 * .. Local Scalars ..
201  LOGICAL UPPER
202  INTEGER COUNT, I, IK, J, K, KASE, KK, NZ
203  DOUBLE PRECISION EPS, LSTRES, S, SAFE1, SAFE2, SAFMIN, XK
204 * ..
205 * .. Local Arrays ..
206  INTEGER ISAVE( 3 )
207 * ..
208 * .. External Subroutines ..
209  EXTERNAL daxpy, dcopy, dlacn2, dpptrs, dspmv, xerbla
210 * ..
211 * .. Intrinsic Functions ..
212  INTRINSIC abs, max
213 * ..
214 * .. External Functions ..
215  LOGICAL LSAME
216  DOUBLE PRECISION DLAMCH
217  EXTERNAL lsame, dlamch
218 * ..
219 * .. Executable Statements ..
220 *
221 * Test the input parameters.
222 *
223  info = 0
224  upper = lsame( uplo, 'U' )
225  IF( .NOT.upper .AND. .NOT.lsame( uplo, 'L' ) ) THEN
226  info = -1
227  ELSE IF( n.LT.0 ) THEN
228  info = -2
229  ELSE IF( nrhs.LT.0 ) THEN
230  info = -3
231  ELSE IF( ldb.LT.max( 1, n ) ) THEN
232  info = -7
233  ELSE IF( ldx.LT.max( 1, n ) ) THEN
234  info = -9
235  END IF
236  IF( info.NE.0 ) THEN
237  CALL xerbla( 'DPPRFS', -info )
238  RETURN
239  END IF
240 *
241 * Quick return if possible
242 *
243  IF( n.EQ.0 .OR. nrhs.EQ.0 ) THEN
244  DO 10 j = 1, nrhs
245  ferr( j ) = zero
246  berr( j ) = zero
247  10 CONTINUE
248  RETURN
249  END IF
250 *
251 * NZ = maximum number of nonzero elements in each row of A, plus 1
252 *
253  nz = n + 1
254  eps = dlamch( 'Epsilon' )
255  safmin = dlamch( 'Safe minimum' )
256  safe1 = nz*safmin
257  safe2 = safe1 / eps
258 *
259 * Do for each right hand side
260 *
261  DO 140 j = 1, nrhs
262 *
263  count = 1
264  lstres = three
265  20 CONTINUE
266 *
267 * Loop until stopping criterion is satisfied.
268 *
269 * Compute residual R = B - A * X
270 *
271  CALL dcopy( n, b( 1, j ), 1, work( n+1 ), 1 )
272  CALL dspmv( uplo, n, -one, ap, x( 1, j ), 1, one, work( n+1 ),
273  \$ 1 )
274 *
275 * Compute componentwise relative backward error from formula
276 *
277 * max(i) ( abs(R(i)) / ( abs(A)*abs(X) + abs(B) )(i) )
278 *
279 * where abs(Z) is the componentwise absolute value of the matrix
280 * or vector Z. If the i-th component of the denominator is less
281 * than SAFE2, then SAFE1 is added to the i-th components of the
282 * numerator and denominator before dividing.
283 *
284  DO 30 i = 1, n
285  work( i ) = abs( b( i, j ) )
286  30 CONTINUE
287 *
288 * Compute abs(A)*abs(X) + abs(B).
289 *
290  kk = 1
291  IF( upper ) THEN
292  DO 50 k = 1, n
293  s = zero
294  xk = abs( x( k, j ) )
295  ik = kk
296  DO 40 i = 1, k - 1
297  work( i ) = work( i ) + abs( ap( ik ) )*xk
298  s = s + abs( ap( ik ) )*abs( x( i, j ) )
299  ik = ik + 1
300  40 CONTINUE
301  work( k ) = work( k ) + abs( ap( kk+k-1 ) )*xk + s
302  kk = kk + k
303  50 CONTINUE
304  ELSE
305  DO 70 k = 1, n
306  s = zero
307  xk = abs( x( k, j ) )
308  work( k ) = work( k ) + abs( ap( kk ) )*xk
309  ik = kk + 1
310  DO 60 i = k + 1, n
311  work( i ) = work( i ) + abs( ap( ik ) )*xk
312  s = s + abs( ap( ik ) )*abs( x( i, j ) )
313  ik = ik + 1
314  60 CONTINUE
315  work( k ) = work( k ) + s
316  kk = kk + ( n-k+1 )
317  70 CONTINUE
318  END IF
319  s = zero
320  DO 80 i = 1, n
321  IF( work( i ).GT.safe2 ) THEN
322  s = max( s, abs( work( n+i ) ) / work( i ) )
323  ELSE
324  s = max( s, ( abs( work( n+i ) )+safe1 ) /
325  \$ ( work( i )+safe1 ) )
326  END IF
327  80 CONTINUE
328  berr( j ) = s
329 *
330 * Test stopping criterion. Continue iterating if
331 * 1) The residual BERR(J) is larger than machine epsilon, and
332 * 2) BERR(J) decreased by at least a factor of 2 during the
333 * last iteration, and
334 * 3) At most ITMAX iterations tried.
335 *
336  IF( berr( j ).GT.eps .AND. two*berr( j ).LE.lstres .AND.
337  \$ count.LE.itmax ) THEN
338 *
339 * Update solution and try again.
340 *
341  CALL dpptrs( uplo, n, 1, afp, work( n+1 ), n, info )
342  CALL daxpy( n, one, work( n+1 ), 1, x( 1, j ), 1 )
343  lstres = berr( j )
344  count = count + 1
345  GO TO 20
346  END IF
347 *
348 * Bound error from formula
349 *
350 * norm(X - XTRUE) / norm(X) .le. FERR =
351 * norm( abs(inv(A))*
352 * ( abs(R) + NZ*EPS*( abs(A)*abs(X)+abs(B) ))) / norm(X)
353 *
354 * where
355 * norm(Z) is the magnitude of the largest component of Z
356 * inv(A) is the inverse of A
357 * abs(Z) is the componentwise absolute value of the matrix or
358 * vector Z
359 * NZ is the maximum number of nonzeros in any row of A, plus 1
360 * EPS is machine epsilon
361 *
362 * The i-th component of abs(R)+NZ*EPS*(abs(A)*abs(X)+abs(B))
363 * is incremented by SAFE1 if the i-th component of
364 * abs(A)*abs(X) + abs(B) is less than SAFE2.
365 *
366 * Use DLACN2 to estimate the infinity-norm of the matrix
367 * inv(A) * diag(W),
368 * where W = abs(R) + NZ*EPS*( abs(A)*abs(X)+abs(B) )))
369 *
370  DO 90 i = 1, n
371  IF( work( i ).GT.safe2 ) THEN
372  work( i ) = abs( work( n+i ) ) + nz*eps*work( i )
373  ELSE
374  work( i ) = abs( work( n+i ) ) + nz*eps*work( i ) + safe1
375  END IF
376  90 CONTINUE
377 *
378  kase = 0
379  100 CONTINUE
380  CALL dlacn2( n, work( 2*n+1 ), work( n+1 ), iwork, ferr( j ),
381  \$ kase, isave )
382  IF( kase.NE.0 ) THEN
383  IF( kase.EQ.1 ) THEN
384 *
385 * Multiply by diag(W)*inv(A**T).
386 *
387  CALL dpptrs( uplo, n, 1, afp, work( n+1 ), n, info )
388  DO 110 i = 1, n
389  work( n+i ) = work( i )*work( n+i )
390  110 CONTINUE
391  ELSE IF( kase.EQ.2 ) THEN
392 *
393 * Multiply by inv(A)*diag(W).
394 *
395  DO 120 i = 1, n
396  work( n+i ) = work( i )*work( n+i )
397  120 CONTINUE
398  CALL dpptrs( uplo, n, 1, afp, work( n+1 ), n, info )
399  END IF
400  GO TO 100
401  END IF
402 *
403 * Normalize error.
404 *
405  lstres = zero
406  DO 130 i = 1, n
407  lstres = max( lstres, abs( x( i, j ) ) )
408  130 CONTINUE
409  IF( lstres.NE.zero )
410  \$ ferr( j ) = ferr( j ) / lstres
411 *
412  140 CONTINUE
413 *
414  RETURN
415 *
416 * End of DPPRFS
417 *
double precision function dlamch(CMACH)
DLAMCH
Definition: dlamch.f:69
subroutine xerbla(SRNAME, INFO)
XERBLA
Definition: xerbla.f:60
logical function lsame(CA, CB)
LSAME
Definition: lsame.f:53
subroutine dcopy(N, DX, INCX, DY, INCY)
DCOPY
Definition: dcopy.f:82
subroutine daxpy(N, DA, DX, INCX, DY, INCY)
DAXPY
Definition: daxpy.f:89
subroutine dspmv(UPLO, N, ALPHA, AP, X, INCX, BETA, Y, INCY)
DSPMV
Definition: dspmv.f:147
subroutine dlacn2(N, V, X, ISGN, EST, KASE, ISAVE)
DLACN2 estimates the 1-norm of a square matrix, using reverse communication for evaluating matrix-vec...
Definition: dlacn2.f:136
subroutine dpptrs(UPLO, N, NRHS, AP, B, LDB, INFO)
DPPTRS
Definition: dpptrs.f:108
Here is the call graph for this function:
Here is the caller graph for this function: