LAPACK  3.6.1 LAPACK: Linear Algebra PACKage
 subroutine cpstrf ( character UPLO, integer N, complex, dimension( lda, * ) A, integer LDA, integer, dimension( n ) PIV, integer RANK, real TOL, real, dimension( 2*n ) WORK, integer INFO )

CPSTRF computes the Cholesky factorization with complete pivoting of complex Hermitian positive semidefinite matrix.

Purpose:
``` CPSTRF computes the Cholesky factorization with complete
pivoting of a complex Hermitian positive semidefinite matrix A.

The factorization has the form
P**T * A * P = U**H * U ,  if UPLO = 'U',
P**T * A * P = L  * L**H,  if UPLO = 'L',
where U is an upper triangular matrix and L is lower triangular, and
P is stored as vector PIV.

This algorithm does not attempt to check that A is positive
semidefinite. This version of the algorithm calls level 3 BLAS.```
Parameters
 [in] UPLO ``` UPLO is CHARACTER*1 Specifies whether the upper or lower triangular part of the symmetric matrix A is stored. = 'U': Upper triangular = 'L': Lower triangular``` [in] N ``` N is INTEGER The order of the matrix A. N >= 0.``` [in,out] A ``` A is COMPLEX array, dimension (LDA,N) On entry, the symmetric matrix A. If UPLO = 'U', the leading n by n upper triangular part of A contains the upper triangular part of the matrix A, and the strictly lower triangular part of A is not referenced. If UPLO = 'L', the leading n by n lower triangular part of A contains the lower triangular part of the matrix A, and the strictly upper triangular part of A is not referenced. On exit, if INFO = 0, the factor U or L from the Cholesky factorization as above.``` [in] LDA ``` LDA is INTEGER The leading dimension of the array A. LDA >= max(1,N).``` [out] PIV ``` PIV is INTEGER array, dimension (N) PIV is such that the nonzero entries are P( PIV(K), K ) = 1.``` [out] RANK ``` RANK is INTEGER The rank of A given by the number of steps the algorithm completed.``` [in] TOL ``` TOL is REAL User defined tolerance. If TOL < 0, then N*U*MAX( A(K,K) ) will be used. The algorithm terminates at the (K-1)st step if the pivot <= TOL.``` [out] WORK ``` WORK is REAL array, dimension (2*N) Work space.``` [out] INFO ``` INFO is INTEGER < 0: If INFO = -K, the K-th argument had an illegal value, = 0: algorithm completed successfully, and > 0: the matrix A is either rank deficient with computed rank as returned in RANK, or is not positive semidefinite. See Section 7 of LAPACK Working Note #161 for further information.```
Date
November 2015

Definition at line 144 of file cpstrf.f.

144 *
145 * -- LAPACK computational routine (version 3.6.0) --
146 * -- LAPACK is a software package provided by Univ. of Tennessee, --
147 * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
148 * November 2015
149 *
150 * .. Scalar Arguments ..
151  REAL tol
152  INTEGER info, lda, n, rank
153  CHARACTER uplo
154 * ..
155 * .. Array Arguments ..
156  COMPLEX a( lda, * )
157  REAL work( 2*n )
158  INTEGER piv( n )
159 * ..
160 *
161 * =====================================================================
162 *
163 * .. Parameters ..
164  REAL one, zero
165  parameter ( one = 1.0e+0, zero = 0.0e+0 )
166  COMPLEX cone
167  parameter ( cone = ( 1.0e+0, 0.0e+0 ) )
168 * ..
169 * .. Local Scalars ..
170  COMPLEX ctemp
171  REAL ajj, sstop, stemp
172  INTEGER i, itemp, j, jb, k, nb, pvt
173  LOGICAL upper
174 * ..
175 * .. External Functions ..
176  REAL slamch
177  INTEGER ilaenv
178  LOGICAL lsame, sisnan
179  EXTERNAL slamch, ilaenv, lsame, sisnan
180 * ..
181 * .. External Subroutines ..
182  EXTERNAL cgemv, cherk, clacgv, cpstf2, csscal, cswap,
183  \$ xerbla
184 * ..
185 * .. Intrinsic Functions ..
186  INTRINSIC conjg, max, min, REAL, sqrt, maxloc
187 * ..
188 * .. Executable Statements ..
189 *
190 * Test the input parameters.
191 *
192  info = 0
193  upper = lsame( uplo, 'U' )
194  IF( .NOT.upper .AND. .NOT.lsame( uplo, 'L' ) ) THEN
195  info = -1
196  ELSE IF( n.LT.0 ) THEN
197  info = -2
198  ELSE IF( lda.LT.max( 1, n ) ) THEN
199  info = -4
200  END IF
201  IF( info.NE.0 ) THEN
202  CALL xerbla( 'CPSTRF', -info )
203  RETURN
204  END IF
205 *
206 * Quick return if possible
207 *
208  IF( n.EQ.0 )
209  \$ RETURN
210 *
211 * Get block size
212 *
213  nb = ilaenv( 1, 'CPOTRF', uplo, n, -1, -1, -1 )
214  IF( nb.LE.1 .OR. nb.GE.n ) THEN
215 *
216 * Use unblocked code
217 *
218  CALL cpstf2( uplo, n, a( 1, 1 ), lda, piv, rank, tol, work,
219  \$ info )
220  GO TO 230
221 *
222  ELSE
223 *
224 * Initialize PIV
225 *
226  DO 100 i = 1, n
227  piv( i ) = i
228  100 CONTINUE
229 *
230 * Compute stopping value
231 *
232  DO 110 i = 1, n
233  work( i ) = REAL( A( I, I ) )
234  110 CONTINUE
235  pvt = maxloc( work( 1:n ), 1 )
236  ajj = REAL( A( PVT, PVT ) )
237  IF( ajj.LE.zero.OR.sisnan( ajj ) ) THEN
238  rank = 0
239  info = 1
240  GO TO 230
241  END IF
242 *
243 * Compute stopping value if not supplied
244 *
245  IF( tol.LT.zero ) THEN
246  sstop = n * slamch( 'Epsilon' ) * ajj
247  ELSE
248  sstop = tol
249  END IF
250 *
251 *
252  IF( upper ) THEN
253 *
254 * Compute the Cholesky factorization P**T * A * P = U**H * U
255 *
256  DO 160 k = 1, n, nb
257 *
258 * Account for last block not being NB wide
259 *
260  jb = min( nb, n-k+1 )
261 *
262 * Set relevant part of first half of WORK to zero,
263 * holds dot products
264 *
265  DO 120 i = k, n
266  work( i ) = 0
267  120 CONTINUE
268 *
269  DO 150 j = k, k + jb - 1
270 *
271 * Find pivot, test for exit, else swap rows and columns
272 * Update dot products, compute possible pivots which are
273 * stored in the second half of WORK
274 *
275  DO 130 i = j, n
276 *
277  IF( j.GT.k ) THEN
278  work( i ) = work( i ) +
279  \$ REAL( CONJG( A( J-1, I ) )*
280  \$ a( j-1, i ) )
281  END IF
282  work( n+i ) = REAL( A( I, I ) ) - work( i )
283 *
284  130 CONTINUE
285 *
286  IF( j.GT.1 ) THEN
287  itemp = maxloc( work( (n+j):(2*n) ), 1 )
288  pvt = itemp + j - 1
289  ajj = work( n+pvt )
290  IF( ajj.LE.sstop.OR.sisnan( ajj ) ) THEN
291  a( j, j ) = ajj
292  GO TO 220
293  END IF
294  END IF
295 *
296  IF( j.NE.pvt ) THEN
297 *
298 * Pivot OK, so can now swap pivot rows and columns
299 *
300  a( pvt, pvt ) = a( j, j )
301  CALL cswap( j-1, a( 1, j ), 1, a( 1, pvt ), 1 )
302  IF( pvt.LT.n )
303  \$ CALL cswap( n-pvt, a( j, pvt+1 ), lda,
304  \$ a( pvt, pvt+1 ), lda )
305  DO 140 i = j + 1, pvt - 1
306  ctemp = conjg( a( j, i ) )
307  a( j, i ) = conjg( a( i, pvt ) )
308  a( i, pvt ) = ctemp
309  140 CONTINUE
310  a( j, pvt ) = conjg( a( j, pvt ) )
311 *
312 * Swap dot products and PIV
313 *
314  stemp = work( j )
315  work( j ) = work( pvt )
316  work( pvt ) = stemp
317  itemp = piv( pvt )
318  piv( pvt ) = piv( j )
319  piv( j ) = itemp
320  END IF
321 *
322  ajj = sqrt( ajj )
323  a( j, j ) = ajj
324 *
325 * Compute elements J+1:N of row J.
326 *
327  IF( j.LT.n ) THEN
328  CALL clacgv( j-1, a( 1, j ), 1 )
329  CALL cgemv( 'Trans', j-k, n-j, -cone, a( k, j+1 ),
330  \$ lda, a( k, j ), 1, cone, a( j, j+1 ),
331  \$ lda )
332  CALL clacgv( j-1, a( 1, j ), 1 )
333  CALL csscal( n-j, one / ajj, a( j, j+1 ), lda )
334  END IF
335 *
336  150 CONTINUE
337 *
338 * Update trailing matrix, J already incremented
339 *
340  IF( k+jb.LE.n ) THEN
341  CALL cherk( 'Upper', 'Conj Trans', n-j+1, jb, -one,
342  \$ a( k, j ), lda, one, a( j, j ), lda )
343  END IF
344 *
345  160 CONTINUE
346 *
347  ELSE
348 *
349 * Compute the Cholesky factorization P**T * A * P = L * L**H
350 *
351  DO 210 k = 1, n, nb
352 *
353 * Account for last block not being NB wide
354 *
355  jb = min( nb, n-k+1 )
356 *
357 * Set relevant part of first half of WORK to zero,
358 * holds dot products
359 *
360  DO 170 i = k, n
361  work( i ) = 0
362  170 CONTINUE
363 *
364  DO 200 j = k, k + jb - 1
365 *
366 * Find pivot, test for exit, else swap rows and columns
367 * Update dot products, compute possible pivots which are
368 * stored in the second half of WORK
369 *
370  DO 180 i = j, n
371 *
372  IF( j.GT.k ) THEN
373  work( i ) = work( i ) +
374  \$ REAL( CONJG( A( I, J-1 ) )*
375  \$ a( i, j-1 ) )
376  END IF
377  work( n+i ) = REAL( A( I, I ) ) - work( i )
378 *
379  180 CONTINUE
380 *
381  IF( j.GT.1 ) THEN
382  itemp = maxloc( work( (n+j):(2*n) ), 1 )
383  pvt = itemp + j - 1
384  ajj = work( n+pvt )
385  IF( ajj.LE.sstop.OR.sisnan( ajj ) ) THEN
386  a( j, j ) = ajj
387  GO TO 220
388  END IF
389  END IF
390 *
391  IF( j.NE.pvt ) THEN
392 *
393 * Pivot OK, so can now swap pivot rows and columns
394 *
395  a( pvt, pvt ) = a( j, j )
396  CALL cswap( j-1, a( j, 1 ), lda, a( pvt, 1 ), lda )
397  IF( pvt.LT.n )
398  \$ CALL cswap( n-pvt, a( pvt+1, j ), 1,
399  \$ a( pvt+1, pvt ), 1 )
400  DO 190 i = j + 1, pvt - 1
401  ctemp = conjg( a( i, j ) )
402  a( i, j ) = conjg( a( pvt, i ) )
403  a( pvt, i ) = ctemp
404  190 CONTINUE
405  a( pvt, j ) = conjg( a( pvt, j ) )
406 *
407 * Swap dot products and PIV
408 *
409  stemp = work( j )
410  work( j ) = work( pvt )
411  work( pvt ) = stemp
412  itemp = piv( pvt )
413  piv( pvt ) = piv( j )
414  piv( j ) = itemp
415  END IF
416 *
417  ajj = sqrt( ajj )
418  a( j, j ) = ajj
419 *
420 * Compute elements J+1:N of column J.
421 *
422  IF( j.LT.n ) THEN
423  CALL clacgv( j-1, a( j, 1 ), lda )
424  CALL cgemv( 'No Trans', n-j, j-k, -cone,
425  \$ a( j+1, k ), lda, a( j, k ), lda, cone,
426  \$ a( j+1, j ), 1 )
427  CALL clacgv( j-1, a( j, 1 ), lda )
428  CALL csscal( n-j, one / ajj, a( j+1, j ), 1 )
429  END IF
430 *
431  200 CONTINUE
432 *
433 * Update trailing matrix, J already incremented
434 *
435  IF( k+jb.LE.n ) THEN
436  CALL cherk( 'Lower', 'No Trans', n-j+1, jb, -one,
437  \$ a( j, k ), lda, one, a( j, j ), lda )
438  END IF
439 *
440  210 CONTINUE
441 *
442  END IF
443  END IF
444 *
445 * Ran to completion, A has full rank
446 *
447  rank = n
448 *
449  GO TO 230
450  220 CONTINUE
451 *
452 * Rank is the number of steps completed. Set INFO = 1 to signal
453 * that the factorization cannot be used to solve a system.
454 *
455  rank = j - 1
456  info = 1
457 *
458  230 CONTINUE
459  RETURN
460 *
461 * End of CPSTRF
462 *
subroutine cpstf2(UPLO, N, A, LDA, PIV, RANK, TOL, WORK, INFO)
CPSTF2 computes the Cholesky factorization with complete pivoting of complex Hermitian positive semid...
Definition: cpstf2.f:144
logical function sisnan(SIN)
SISNAN tests input for NaN.
Definition: sisnan.f:61
subroutine cherk(UPLO, TRANS, N, K, ALPHA, A, LDA, BETA, C, LDC)
CHERK
Definition: cherk.f:175
subroutine xerbla(SRNAME, INFO)
XERBLA
Definition: xerbla.f:62
subroutine cgemv(TRANS, M, N, ALPHA, A, LDA, X, INCX, BETA, Y, INCY)
CGEMV
Definition: cgemv.f:160
integer function ilaenv(ISPEC, NAME, OPTS, N1, N2, N3, N4)
Definition: tstiee.f:83
subroutine clacgv(N, X, INCX)
CLACGV conjugates a complex vector.
Definition: clacgv.f:76
subroutine cswap(N, CX, INCX, CY, INCY)
CSWAP
Definition: cswap.f:52
real function slamch(CMACH)
SLAMCH
Definition: slamch.f:69
logical function lsame(CA, CB)
LSAME
Definition: lsame.f:55
subroutine csscal(N, SA, CX, INCX)
CSSCAL
Definition: csscal.f:54

Here is the call graph for this function:

Here is the caller graph for this function: