1 *> \brief \b ZHETF2 computes the factorization of a complex Hermitian matrix, using the diagonal pivoting method (unblocked algorithm, calling Level 2 BLAS).
3 * =========== DOCUMENTATION ===========
5 * Online html documentation available at
6 * http://www.netlib.org/lapack/explore-html/
9 *> Download ZHETF2 + dependencies
10 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/zhetf2.f">
12 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/zhetf2.f">
14 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/zhetf2.f">
21 * SUBROUTINE ZHETF2( UPLO, N, A, LDA, IPIV, INFO )
23 * .. Scalar Arguments ..
25 * INTEGER INFO, LDA, N
27 * .. Array Arguments ..
29 * COMPLEX*16 A( LDA, * )
38 *> ZHETF2 computes the factorization of a complex Hermitian matrix A
39 *> using the Bunch-Kaufman diagonal pivoting method:
41 *> A = U*D*U**H or A = L*D*L**H
43 *> where U (or L) is a product of permutation and unit upper (lower)
44 *> triangular matrices, U**H is the conjugate transpose of U, and D is
45 *> Hermitian and block diagonal with 1-by-1 and 2-by-2 diagonal blocks.
47 *> This is the unblocked version of the algorithm, calling Level 2 BLAS.
55 *> UPLO is CHARACTER*1
56 *> Specifies whether the upper or lower triangular part of the
57 *> Hermitian matrix A is stored:
58 *> = 'U': Upper triangular
59 *> = 'L': Lower triangular
65 *> The order of the matrix A. N >= 0.
70 *> A is COMPLEX*16 array, dimension (LDA,N)
71 *> On entry, the Hermitian matrix A. If UPLO = 'U', the leading
72 *> n-by-n upper triangular part of A contains the upper
73 *> triangular part of the matrix A, and the strictly lower
74 *> triangular part of A is not referenced. If UPLO = 'L', the
75 *> leading n-by-n lower triangular part of A contains the lower
76 *> triangular part of the matrix A, and the strictly upper
77 *> triangular part of A is not referenced.
79 *> On exit, the block diagonal matrix D and the multipliers used
80 *> to obtain the factor U or L (see below for further details).
86 *> The leading dimension of the array A. LDA >= max(1,N).
91 *> IPIV is INTEGER array, dimension (N)
92 *> Details of the interchanges and the block structure of D.
95 *> If IPIV(k) > 0, then rows and columns k and IPIV(k) were
96 *> interchanged and D(k,k) is a 1-by-1 diagonal block.
98 *> If IPIV(k) = IPIV(k-1) < 0, then rows and columns
99 *> k-1 and -IPIV(k) were interchanged and D(k-1:k,k-1:k)
100 *> is a 2-by-2 diagonal block.
103 *> If IPIV(k) > 0, then rows and columns k and IPIV(k) were
104 *> interchanged and D(k,k) is a 1-by-1 diagonal block.
106 *> If IPIV(k) = IPIV(k+1) < 0, then rows and columns
107 *> k+1 and -IPIV(k) were interchanged and D(k:k+1,k:k+1)
108 *> is a 2-by-2 diagonal block.
114 *> = 0: successful exit
115 *> < 0: if INFO = -k, the k-th argument had an illegal value
116 *> > 0: if INFO = k, D(k,k) is exactly zero. The factorization
117 *> has been completed, but the block diagonal matrix D is
118 *> exactly singular, and division by zero will occur if it
119 *> is used to solve a system of equations.
125 *> \author Univ. of Tennessee
126 *> \author Univ. of California Berkeley
127 *> \author Univ. of Colorado Denver
130 *> \date November 2013
132 *> \ingroup complex16HEcomputational
134 *> \par Further Details:
135 * =====================
139 *> If UPLO = 'U', then A = U*D*U**H, where
140 *> U = P(n)*U(n)* ... *P(k)U(k)* ...,
141 *> i.e., U is a product of terms P(k)*U(k), where k decreases from n to
142 *> 1 in steps of 1 or 2, and D is a block diagonal matrix with 1-by-1
143 *> and 2-by-2 diagonal blocks D(k). P(k) is a permutation matrix as
144 *> defined by IPIV(k), and U(k) is a unit upper triangular matrix, such
145 *> that if the diagonal block D(k) is of order s (s = 1 or 2), then
148 *> U(k) = ( 0 I 0 ) s
152 *> If s = 1, D(k) overwrites A(k,k), and v overwrites A(1:k-1,k).
153 *> If s = 2, the upper triangle of D(k) overwrites A(k-1,k-1), A(k-1,k),
154 *> and A(k,k), and v overwrites A(1:k-2,k-1:k).
156 *> If UPLO = 'L', then A = L*D*L**H, where
157 *> L = P(1)*L(1)* ... *P(k)*L(k)* ...,
158 *> i.e., L is a product of terms P(k)*L(k), where k increases from 1 to
159 *> n in steps of 1 or 2, and D is a block diagonal matrix with 1-by-1
160 *> and 2-by-2 diagonal blocks D(k). P(k) is a permutation matrix as
161 *> defined by IPIV(k), and L(k) is a unit lower triangular matrix, such
162 *> that if the diagonal block D(k) is of order s (s = 1 or 2), then
165 *> L(k) = ( 0 I 0 ) s
169 *> If s = 1, D(k) overwrites A(k,k), and v overwrites A(k+1:n,k).
170 *> If s = 2, the lower triangle of D(k) overwrites A(k,k), A(k+1,k),
171 *> and A(k+1,k+1), and v overwrites A(k+2:n,k:k+1).
174 *> \par Contributors:
178 *> 09-29-06 - patch from
179 *> Bobby Cheng, MathWorks
181 *> Replace l.210 and l.393
182 *> IF( MAX( ABSAKK, COLMAX ).EQ.ZERO ) THEN
184 *> IF( (MAX( ABSAKK, COLMAX ).EQ.ZERO) .OR. DISNAN(ABSAKK) ) THEN
186 *> 01-01-96 - Based on modifications by
187 *> J. Lewis, Boeing Computer Services Company
188 *> A. Petitet, Computer Science Dept., Univ. of Tenn., Knoxville, USA
191 * =====================================================================
192 SUBROUTINE ZHETF2( UPLO, N, A, LDA, IPIV, INFO )
194 * -- LAPACK computational routine (version 3.5.0) --
195 * -- LAPACK is a software package provided by Univ. of Tennessee, --
196 * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
199 * .. Scalar Arguments ..
203 * .. Array Arguments ..
205 COMPLEX*16 A( LDA, * )
208 * =====================================================================
211 DOUBLE PRECISION ZERO, ONE
212 PARAMETER ( ZERO = 0.0D+0, ONE = 1.0D+0 )
213 DOUBLE PRECISION EIGHT, SEVTEN
214 PARAMETER ( EIGHT = 8.0D+0, SEVTEN = 17.0D+0 )
216 * .. Local Scalars ..
218 INTEGER I, IMAX, J, JMAX, K, KK, KP, KSTEP
219 DOUBLE PRECISION ABSAKK, ALPHA, COLMAX, D, D11, D22, R1, ROWMAX,
221 COMPLEX*16 D12, D21, T, WK, WKM1, WKP1, ZDUM
223 * .. External Functions ..
224 LOGICAL LSAME, DISNAN
226 DOUBLE PRECISION DLAPY2
227 EXTERNAL LSAME, IZAMAX, DLAPY2, DISNAN
229 * .. External Subroutines ..
230 EXTERNAL XERBLA, ZDSCAL, ZHER, ZSWAP
232 * .. Intrinsic Functions ..
233 INTRINSIC ABS, DBLE, DCMPLX, DCONJG, DIMAG, MAX, SQRT
235 * .. Statement Functions ..
236 DOUBLE PRECISION CABS1
238 * .. Statement Function definitions ..
239 CABS1( ZDUM ) = ABS( DBLE( ZDUM ) ) + ABS( DIMAG( ZDUM ) )
241 * .. Executable Statements ..
243 * Test the input parameters.
246 UPPER = LSAME( UPLO, 'U' )
247 IF( .NOT.UPPER .AND. .NOT.LSAME( UPLO, 'L' ) ) THEN
249 ELSE IF( N.LT.0 ) THEN
251 ELSE IF( LDA.LT.MAX( 1, N ) ) THEN
255 CALL XERBLA( 'ZHETF2', -INFO )
259 * Initialize ALPHA for use in choosing pivot block size.
261 ALPHA = ( ONE+SQRT( SEVTEN ) ) / EIGHT
265 * Factorize A as U*D*U**H using the upper triangle of A
267 * K is the main loop index, decreasing from N to 1 in steps of
273 * If K < 1, exit from loop
279 * Determine rows and columns to be interchanged and whether
280 * a 1-by-1 or 2-by-2 pivot block will be used
282 ABSAKK = ABS( DBLE( A( K, K ) ) )
284 * IMAX is the row-index of the largest off-diagonal element in
285 * column K, and COLMAX is its absolute value.
286 * Determine both COLMAX and IMAX.
289 IMAX = IZAMAX( K-1, A( 1, K ), 1 )
290 COLMAX = CABS1( A( IMAX, K ) )
295 IF( (MAX( ABSAKK, COLMAX ).EQ.ZERO) .OR. DISNAN(ABSAKK) ) THEN
297 * Column K is zero or underflow, or contains a NaN:
298 * set INFO and continue
303 A( K, K ) = DBLE( A( K, K ) )
306 * ============================================================
308 * Test for interchange
310 IF( ABSAKK.GE.ALPHA*COLMAX ) THEN
312 * no interchange, use 1-by-1 pivot block
317 * JMAX is the column-index of the largest off-diagonal
318 * element in row IMAX, and ROWMAX is its absolute value.
319 * Determine only ROWMAX.
321 JMAX = IMAX + IZAMAX( K-IMAX, A( IMAX, IMAX+1 ), LDA )
322 ROWMAX = CABS1( A( IMAX, JMAX ) )
324 JMAX = IZAMAX( IMAX-1, A( 1, IMAX ), 1 )
325 ROWMAX = MAX( ROWMAX, CABS1( A( JMAX, IMAX ) ) )
328 IF( ABSAKK.GE.ALPHA*COLMAX*( COLMAX / ROWMAX ) ) THEN
330 * no interchange, use 1-by-1 pivot block
334 ELSE IF( ABS( DBLE( A( IMAX, IMAX ) ) ).GE.ALPHA*ROWMAX )
337 * interchange rows and columns K and IMAX, use 1-by-1
343 * interchange rows and columns K-1 and IMAX, use 2-by-2
352 * ============================================================
357 * Interchange rows and columns KK and KP in the leading
358 * submatrix A(1:k,1:k)
360 CALL ZSWAP( KP-1, A( 1, KK ), 1, A( 1, KP ), 1 )
361 DO 20 J = KP + 1, KK - 1
362 T = DCONJG( A( J, KK ) )
363 A( J, KK ) = DCONJG( A( KP, J ) )
366 A( KP, KK ) = DCONJG( A( KP, KK ) )
367 R1 = DBLE( A( KK, KK ) )
368 A( KK, KK ) = DBLE( A( KP, KP ) )
370 IF( KSTEP.EQ.2 ) THEN
371 A( K, K ) = DBLE( A( K, K ) )
373 A( K-1, K ) = A( KP, K )
377 A( K, K ) = DBLE( A( K, K ) )
379 $ A( K-1, K-1 ) = DBLE( A( K-1, K-1 ) )
382 * Update the leading submatrix
384 IF( KSTEP.EQ.1 ) THEN
386 * 1-by-1 pivot block D(k): column k now holds
390 * where U(k) is the k-th column of U
392 * Perform a rank-1 update of A(1:k-1,1:k-1) as
394 * A := A - U(k)*D(k)*U(k)**H = A - W(k)*1/D(k)*W(k)**H
396 R1 = ONE / DBLE( A( K, K ) )
397 CALL ZHER( UPLO, K-1, -R1, A( 1, K ), 1, A, LDA )
399 * Store U(k) in column k
401 CALL ZDSCAL( K-1, R1, A( 1, K ), 1 )
404 * 2-by-2 pivot block D(k): columns k and k-1 now hold
406 * ( W(k-1) W(k) ) = ( U(k-1) U(k) )*D(k)
408 * where U(k) and U(k-1) are the k-th and (k-1)-th columns
411 * Perform a rank-2 update of A(1:k-2,1:k-2) as
413 * A := A - ( U(k-1) U(k) )*D(k)*( U(k-1) U(k) )**H
414 * = A - ( W(k-1) W(k) )*inv(D(k))*( W(k-1) W(k) )**H
418 D = DLAPY2( DBLE( A( K-1, K ) ),
419 $ DIMAG( A( K-1, K ) ) )
420 D22 = DBLE( A( K-1, K-1 ) ) / D
421 D11 = DBLE( A( K, K ) ) / D
422 TT = ONE / ( D11*D22-ONE )
423 D12 = A( K-1, K ) / D
426 DO 40 J = K - 2, 1, -1
427 WKM1 = D*( D11*A( J, K-1 )-DCONJG( D12 )*
429 WK = D*( D22*A( J, K )-D12*A( J, K-1 ) )
431 A( I, J ) = A( I, J ) - A( I, K )*DCONJG( WK ) -
432 $ A( I, K-1 )*DCONJG( WKM1 )
436 A( J, J ) = DCMPLX( DBLE( A( J, J ) ), 0.0D+0 )
444 * Store details of the interchanges in IPIV
446 IF( KSTEP.EQ.1 ) THEN
453 * Decrease K and return to the start of the main loop
460 * Factorize A as L*D*L**H using the lower triangle of A
462 * K is the main loop index, increasing from 1 to N in steps of
468 * If K > N, exit from loop
474 * Determine rows and columns to be interchanged and whether
475 * a 1-by-1 or 2-by-2 pivot block will be used
477 ABSAKK = ABS( DBLE( A( K, K ) ) )
479 * IMAX is the row-index of the largest off-diagonal element in
480 * column K, and COLMAX is its absolute value.
481 * Determine both COLMAX and IMAX.
484 IMAX = K + IZAMAX( N-K, A( K+1, K ), 1 )
485 COLMAX = CABS1( A( IMAX, K ) )
490 IF( (MAX( ABSAKK, COLMAX ).EQ.ZERO) .OR. DISNAN(ABSAKK) ) THEN
492 * Column K is zero or underflow, or contains a NaN:
493 * set INFO and continue
498 A( K, K ) = DBLE( A( K, K ) )
501 * ============================================================
503 * Test for interchange
505 IF( ABSAKK.GE.ALPHA*COLMAX ) THEN
507 * no interchange, use 1-by-1 pivot block
512 * JMAX is the column-index of the largest off-diagonal
513 * element in row IMAX, and ROWMAX is its absolute value.
514 * Determine only ROWMAX.
516 JMAX = K - 1 + IZAMAX( IMAX-K, A( IMAX, K ), LDA )
517 ROWMAX = CABS1( A( IMAX, JMAX ) )
519 JMAX = IMAX + IZAMAX( N-IMAX, A( IMAX+1, IMAX ), 1 )
520 ROWMAX = MAX( ROWMAX, CABS1( A( JMAX, IMAX ) ) )
523 IF( ABSAKK.GE.ALPHA*COLMAX*( COLMAX / ROWMAX ) ) THEN
525 * no interchange, use 1-by-1 pivot block
529 ELSE IF( ABS( DBLE( A( IMAX, IMAX ) ) ).GE.ALPHA*ROWMAX )
532 * interchange rows and columns K and IMAX, use 1-by-1
538 * interchange rows and columns K+1 and IMAX, use 2-by-2
547 * ============================================================
552 * Interchange rows and columns KK and KP in the trailing
553 * submatrix A(k:n,k:n)
556 $ CALL ZSWAP( N-KP, A( KP+1, KK ), 1, A( KP+1, KP ), 1 )
557 DO 60 J = KK + 1, KP - 1
558 T = DCONJG( A( J, KK ) )
559 A( J, KK ) = DCONJG( A( KP, J ) )
562 A( KP, KK ) = DCONJG( A( KP, KK ) )
563 R1 = DBLE( A( KK, KK ) )
564 A( KK, KK ) = DBLE( A( KP, KP ) )
566 IF( KSTEP.EQ.2 ) THEN
567 A( K, K ) = DBLE( A( K, K ) )
569 A( K+1, K ) = A( KP, K )
573 A( K, K ) = DBLE( A( K, K ) )
575 $ A( K+1, K+1 ) = DBLE( A( K+1, K+1 ) )
578 * Update the trailing submatrix
580 IF( KSTEP.EQ.1 ) THEN
582 * 1-by-1 pivot block D(k): column k now holds
586 * where L(k) is the k-th column of L
590 * Perform a rank-1 update of A(k+1:n,k+1:n) as
592 * A := A - L(k)*D(k)*L(k)**H = A - W(k)*(1/D(k))*W(k)**H
594 R1 = ONE / DBLE( A( K, K ) )
595 CALL ZHER( UPLO, N-K, -R1, A( K+1, K ), 1,
596 $ A( K+1, K+1 ), LDA )
598 * Store L(k) in column K
600 CALL ZDSCAL( N-K, R1, A( K+1, K ), 1 )
604 * 2-by-2 pivot block D(k)
608 * Perform a rank-2 update of A(k+2:n,k+2:n) as
610 * A := A - ( L(k) L(k+1) )*D(k)*( L(k) L(k+1) )**H
611 * = A - ( W(k) W(k+1) )*inv(D(k))*( W(k) W(k+1) )**H
613 * where L(k) and L(k+1) are the k-th and (k+1)-th
616 D = DLAPY2( DBLE( A( K+1, K ) ),
617 $ DIMAG( A( K+1, K ) ) )
618 D11 = DBLE( A( K+1, K+1 ) ) / D
619 D22 = DBLE( A( K, K ) ) / D
620 TT = ONE / ( D11*D22-ONE )
621 D21 = A( K+1, K ) / D
625 WK = D*( D11*A( J, K )-D21*A( J, K+1 ) )
626 WKP1 = D*( D22*A( J, K+1 )-DCONJG( D21 )*
629 A( I, J ) = A( I, J ) - A( I, K )*DCONJG( WK ) -
630 $ A( I, K+1 )*DCONJG( WKP1 )
634 A( J, J ) = DCMPLX( DBLE( A( J, J ) ), 0.0D+0 )
640 * Store details of the interchanges in IPIV
642 IF( KSTEP.EQ.1 ) THEN
649 * Increase K and return to the start of the main loop