Annotation of rpl/lapack/lapack/dsyrfsx.f, revision 1.15
1.5 bertrand 1: *> \brief \b DSYRFSX
2: *
3: * =========== DOCUMENTATION ===========
4: *
1.12 bertrand 5: * Online html documentation available at
6: * http://www.netlib.org/lapack/explore-html/
1.5 bertrand 7: *
8: *> \htmlonly
1.12 bertrand 9: *> Download DSYRFSX + dependencies
10: *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/dsyrfsx.f">
11: *> [TGZ]</a>
12: *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/dsyrfsx.f">
13: *> [ZIP]</a>
14: *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/dsyrfsx.f">
1.5 bertrand 15: *> [TXT]</a>
1.12 bertrand 16: *> \endhtmlonly
1.5 bertrand 17: *
18: * Definition:
19: * ===========
20: *
21: * SUBROUTINE DSYRFSX( UPLO, EQUED, N, NRHS, A, LDA, AF, LDAF, IPIV,
22: * S, B, LDB, X, LDX, RCOND, BERR, N_ERR_BNDS,
23: * ERR_BNDS_NORM, ERR_BNDS_COMP, NPARAMS, PARAMS,
24: * WORK, IWORK, INFO )
1.12 bertrand 25: *
1.5 bertrand 26: * .. Scalar Arguments ..
27: * CHARACTER UPLO, EQUED
28: * INTEGER INFO, LDA, LDAF, LDB, LDX, N, NRHS, NPARAMS,
29: * $ N_ERR_BNDS
30: * DOUBLE PRECISION RCOND
31: * ..
32: * .. Array Arguments ..
33: * INTEGER IPIV( * ), IWORK( * )
34: * DOUBLE PRECISION A( LDA, * ), AF( LDAF, * ), B( LDB, * ),
35: * $ X( LDX, * ), WORK( * )
36: * DOUBLE PRECISION S( * ), PARAMS( * ), BERR( * ),
37: * $ ERR_BNDS_NORM( NRHS, * ),
38: * $ ERR_BNDS_COMP( NRHS, * )
39: * ..
1.12 bertrand 40: *
1.5 bertrand 41: *
42: *> \par Purpose:
43: * =============
44: *>
45: *> \verbatim
46: *>
47: *> DSYRFSX improves the computed solution to a system of linear
48: *> equations when the coefficient matrix is symmetric indefinite, and
49: *> provides error bounds and backward error estimates for the
50: *> solution. In addition to normwise error bound, the code provides
51: *> maximum componentwise error bound if possible. See comments for
52: *> ERR_BNDS_NORM and ERR_BNDS_COMP for details of the error bounds.
53: *>
54: *> The original system of linear equations may have been equilibrated
55: *> before calling this routine, as described by arguments EQUED and S
56: *> below. In this case, the solution and error bounds returned are
57: *> for the original unequilibrated system.
58: *> \endverbatim
59: *
60: * Arguments:
61: * ==========
62: *
63: *> \verbatim
64: *> Some optional parameters are bundled in the PARAMS array. These
65: *> settings determine how refinement is performed, but often the
66: *> defaults are acceptable. If the defaults are acceptable, users
67: *> can pass NPARAMS = 0 which prevents the source code from accessing
68: *> the PARAMS argument.
69: *> \endverbatim
70: *>
71: *> \param[in] UPLO
72: *> \verbatim
73: *> UPLO is CHARACTER*1
74: *> = 'U': Upper triangle of A is stored;
75: *> = 'L': Lower triangle of A is stored.
76: *> \endverbatim
77: *>
78: *> \param[in] EQUED
79: *> \verbatim
80: *> EQUED is CHARACTER*1
81: *> Specifies the form of equilibration that was done to A
82: *> before calling this routine. This is needed to compute
83: *> the solution and error bounds correctly.
84: *> = 'N': No equilibration
85: *> = 'Y': Both row and column equilibration, i.e., A has been
86: *> replaced by diag(S) * A * diag(S).
87: *> The right hand side B has been changed accordingly.
88: *> \endverbatim
89: *>
90: *> \param[in] N
91: *> \verbatim
92: *> N is INTEGER
93: *> The order of the matrix A. N >= 0.
94: *> \endverbatim
95: *>
96: *> \param[in] NRHS
97: *> \verbatim
98: *> NRHS is INTEGER
99: *> The number of right hand sides, i.e., the number of columns
100: *> of the matrices B and X. NRHS >= 0.
101: *> \endverbatim
102: *>
103: *> \param[in] A
104: *> \verbatim
105: *> A is DOUBLE PRECISION array, dimension (LDA,N)
106: *> The symmetric matrix A. If UPLO = 'U', the leading N-by-N
107: *> upper triangular part of A contains the upper triangular
108: *> part of the matrix A, and the strictly lower triangular
109: *> part of A is not referenced. If UPLO = 'L', the leading
110: *> N-by-N lower triangular part of A contains the lower
111: *> triangular part of the matrix A, and the strictly upper
112: *> triangular part of A is not referenced.
113: *> \endverbatim
114: *>
115: *> \param[in] LDA
116: *> \verbatim
117: *> LDA is INTEGER
118: *> The leading dimension of the array A. LDA >= max(1,N).
119: *> \endverbatim
120: *>
121: *> \param[in] AF
122: *> \verbatim
123: *> AF is DOUBLE PRECISION array, dimension (LDAF,N)
124: *> The factored form of the matrix A. AF contains the block
125: *> diagonal matrix D and the multipliers used to obtain the
126: *> factor U or L from the factorization A = U*D*U**T or A =
127: *> L*D*L**T as computed by DSYTRF.
128: *> \endverbatim
129: *>
130: *> \param[in] LDAF
131: *> \verbatim
132: *> LDAF is INTEGER
133: *> The leading dimension of the array AF. LDAF >= max(1,N).
134: *> \endverbatim
135: *>
136: *> \param[in] IPIV
137: *> \verbatim
138: *> IPIV is INTEGER array, dimension (N)
139: *> Details of the interchanges and the block structure of D
140: *> as determined by DSYTRF.
141: *> \endverbatim
142: *>
143: *> \param[in,out] S
144: *> \verbatim
1.7 bertrand 145: *> S is DOUBLE PRECISION array, dimension (N)
1.5 bertrand 146: *> The scale factors for A. If EQUED = 'Y', A is multiplied on
147: *> the left and right by diag(S). S is an input argument if FACT =
148: *> 'F'; otherwise, S is an output argument. If FACT = 'F' and EQUED
149: *> = 'Y', each element of S must be positive. If S is output, each
150: *> element of S is a power of the radix. If S is input, each element
151: *> of S should be a power of the radix to ensure a reliable solution
152: *> and error estimates. Scaling by powers of the radix does not cause
153: *> rounding errors unless the result underflows or overflows.
154: *> Rounding errors during scaling lead to refining with a matrix that
155: *> is not equivalent to the input matrix, producing error estimates
156: *> that may not be reliable.
157: *> \endverbatim
158: *>
159: *> \param[in] B
160: *> \verbatim
161: *> B is DOUBLE PRECISION array, dimension (LDB,NRHS)
162: *> The right hand side matrix B.
163: *> \endverbatim
164: *>
165: *> \param[in] LDB
166: *> \verbatim
167: *> LDB is INTEGER
168: *> The leading dimension of the array B. LDB >= max(1,N).
169: *> \endverbatim
170: *>
171: *> \param[in,out] X
172: *> \verbatim
173: *> X is DOUBLE PRECISION array, dimension (LDX,NRHS)
174: *> On entry, the solution matrix X, as computed by DGETRS.
175: *> On exit, the improved solution matrix X.
176: *> \endverbatim
177: *>
178: *> \param[in] LDX
179: *> \verbatim
180: *> LDX is INTEGER
181: *> The leading dimension of the array X. LDX >= max(1,N).
182: *> \endverbatim
183: *>
184: *> \param[out] RCOND
185: *> \verbatim
186: *> RCOND is DOUBLE PRECISION
187: *> Reciprocal scaled condition number. This is an estimate of the
188: *> reciprocal Skeel condition number of the matrix A after
189: *> equilibration (if done). If this is less than the machine
190: *> precision (in particular, if it is zero), the matrix is singular
191: *> to working precision. Note that the error may still be small even
192: *> if this number is very small and the matrix appears ill-
193: *> conditioned.
194: *> \endverbatim
195: *>
196: *> \param[out] BERR
197: *> \verbatim
198: *> BERR is DOUBLE PRECISION array, dimension (NRHS)
199: *> Componentwise relative backward error. This is the
200: *> componentwise relative backward error of each solution vector X(j)
201: *> (i.e., the smallest relative change in any element of A or B that
202: *> makes X(j) an exact solution).
203: *> \endverbatim
204: *>
205: *> \param[in] N_ERR_BNDS
206: *> \verbatim
207: *> N_ERR_BNDS is INTEGER
208: *> Number of error bounds to return for each right hand side
209: *> and each type (normwise or componentwise). See ERR_BNDS_NORM and
210: *> ERR_BNDS_COMP below.
211: *> \endverbatim
212: *>
213: *> \param[out] ERR_BNDS_NORM
214: *> \verbatim
215: *> ERR_BNDS_NORM is DOUBLE PRECISION array, dimension (NRHS, N_ERR_BNDS)
216: *> For each right-hand side, this array contains information about
217: *> various error bounds and condition numbers corresponding to the
218: *> normwise relative error, which is defined as follows:
219: *>
220: *> Normwise relative error in the ith solution vector:
221: *> max_j (abs(XTRUE(j,i) - X(j,i)))
222: *> ------------------------------
223: *> max_j abs(X(j,i))
224: *>
225: *> The array is indexed by the type of error information as described
226: *> below. There currently are up to three pieces of information
227: *> returned.
228: *>
229: *> The first index in ERR_BNDS_NORM(i,:) corresponds to the ith
230: *> right-hand side.
231: *>
232: *> The second index in ERR_BNDS_NORM(:,err) contains the following
233: *> three fields:
234: *> err = 1 "Trust/don't trust" boolean. Trust the answer if the
235: *> reciprocal condition number is less than the threshold
236: *> sqrt(n) * dlamch('Epsilon').
237: *>
238: *> err = 2 "Guaranteed" error bound: The estimated forward error,
239: *> almost certainly within a factor of 10 of the true error
240: *> so long as the next entry is greater than the threshold
241: *> sqrt(n) * dlamch('Epsilon'). This error bound should only
242: *> be trusted if the previous boolean is true.
243: *>
244: *> err = 3 Reciprocal condition number: Estimated normwise
245: *> reciprocal condition number. Compared with the threshold
246: *> sqrt(n) * dlamch('Epsilon') to determine if the error
247: *> estimate is "guaranteed". These reciprocal condition
248: *> numbers are 1 / (norm(Z^{-1},inf) * norm(Z,inf)) for some
249: *> appropriately scaled matrix Z.
250: *> Let Z = S*A, where S scales each row by a power of the
251: *> radix so all absolute row sums of Z are approximately 1.
252: *>
253: *> See Lapack Working Note 165 for further details and extra
254: *> cautions.
255: *> \endverbatim
256: *>
257: *> \param[out] ERR_BNDS_COMP
258: *> \verbatim
259: *> ERR_BNDS_COMP is DOUBLE PRECISION array, dimension (NRHS, N_ERR_BNDS)
260: *> For each right-hand side, this array contains information about
261: *> various error bounds and condition numbers corresponding to the
262: *> componentwise relative error, which is defined as follows:
263: *>
264: *> Componentwise relative error in the ith solution vector:
265: *> abs(XTRUE(j,i) - X(j,i))
266: *> max_j ----------------------
267: *> abs(X(j,i))
268: *>
269: *> The array is indexed by the right-hand side i (on which the
270: *> componentwise relative error depends), and the type of error
271: *> information as described below. There currently are up to three
272: *> pieces of information returned for each right-hand side. If
273: *> componentwise accuracy is not requested (PARAMS(3) = 0.0), then
1.15 ! bertrand 274: *> ERR_BNDS_COMP is not accessed. If N_ERR_BNDS < 3, then at most
1.5 bertrand 275: *> the first (:,N_ERR_BNDS) entries are returned.
276: *>
277: *> The first index in ERR_BNDS_COMP(i,:) corresponds to the ith
278: *> right-hand side.
279: *>
280: *> The second index in ERR_BNDS_COMP(:,err) contains the following
281: *> three fields:
282: *> err = 1 "Trust/don't trust" boolean. Trust the answer if the
283: *> reciprocal condition number is less than the threshold
284: *> sqrt(n) * dlamch('Epsilon').
285: *>
286: *> err = 2 "Guaranteed" error bound: The estimated forward error,
287: *> almost certainly within a factor of 10 of the true error
288: *> so long as the next entry is greater than the threshold
289: *> sqrt(n) * dlamch('Epsilon'). This error bound should only
290: *> be trusted if the previous boolean is true.
291: *>
292: *> err = 3 Reciprocal condition number: Estimated componentwise
293: *> reciprocal condition number. Compared with the threshold
294: *> sqrt(n) * dlamch('Epsilon') to determine if the error
295: *> estimate is "guaranteed". These reciprocal condition
296: *> numbers are 1 / (norm(Z^{-1},inf) * norm(Z,inf)) for some
297: *> appropriately scaled matrix Z.
298: *> Let Z = S*(A*diag(x)), where x is the solution for the
299: *> current right-hand side and S scales each row of
300: *> A*diag(x) by a power of the radix so all absolute row
301: *> sums of Z are approximately 1.
302: *>
303: *> See Lapack Working Note 165 for further details and extra
304: *> cautions.
305: *> \endverbatim
306: *>
307: *> \param[in] NPARAMS
308: *> \verbatim
309: *> NPARAMS is INTEGER
1.15 ! bertrand 310: *> Specifies the number of parameters set in PARAMS. If <= 0, the
1.5 bertrand 311: *> PARAMS array is never referenced and default values are used.
312: *> \endverbatim
313: *>
314: *> \param[in,out] PARAMS
315: *> \verbatim
1.7 bertrand 316: *> PARAMS is DOUBLE PRECISION array, dimension (NPARAMS)
1.15 ! bertrand 317: *> Specifies algorithm parameters. If an entry is < 0.0, then
1.5 bertrand 318: *> that entry will be filled with default value used for that
319: *> parameter. Only positions up to NPARAMS are accessed; defaults
320: *> are used for higher-numbered parameters.
321: *>
322: *> PARAMS(LA_LINRX_ITREF_I = 1) : Whether to perform iterative
323: *> refinement or not.
324: *> Default: 1.0D+0
1.15 ! bertrand 325: *> = 0.0: No refinement is performed, and no error bounds are
1.5 bertrand 326: *> computed.
1.15 ! bertrand 327: *> = 1.0: Use the double-precision refinement algorithm,
1.5 bertrand 328: *> possibly with doubled-single computations if the
329: *> compilation environment does not support DOUBLE
330: *> PRECISION.
331: *> (other values are reserved for future use)
332: *>
333: *> PARAMS(LA_LINRX_ITHRESH_I = 2) : Maximum number of residual
334: *> computations allowed for refinement.
335: *> Default: 10
336: *> Aggressive: Set to 100 to permit convergence using approximate
337: *> factorizations or factorizations other than LU. If
338: *> the factorization uses a technique other than
339: *> Gaussian elimination, the guarantees in
340: *> err_bnds_norm and err_bnds_comp may no longer be
341: *> trustworthy.
342: *>
343: *> PARAMS(LA_LINRX_CWISE_I = 3) : Flag determining if the code
344: *> will attempt to find a solution with small componentwise
345: *> relative error in the double-precision algorithm. Positive
346: *> is true, 0.0 is false.
347: *> Default: 1.0 (attempt componentwise convergence)
348: *> \endverbatim
349: *>
350: *> \param[out] WORK
351: *> \verbatim
352: *> WORK is DOUBLE PRECISION array, dimension (4*N)
353: *> \endverbatim
354: *>
355: *> \param[out] IWORK
356: *> \verbatim
357: *> IWORK is INTEGER array, dimension (N)
358: *> \endverbatim
359: *>
360: *> \param[out] INFO
361: *> \verbatim
362: *> INFO is INTEGER
363: *> = 0: Successful exit. The solution to every right-hand side is
364: *> guaranteed.
365: *> < 0: If INFO = -i, the i-th argument had an illegal value
366: *> > 0 and <= N: U(INFO,INFO) is exactly zero. The factorization
367: *> has been completed, but the factor U is exactly singular, so
368: *> the solution and error bounds could not be computed. RCOND = 0
369: *> is returned.
370: *> = N+J: The solution corresponding to the Jth right-hand side is
371: *> not guaranteed. The solutions corresponding to other right-
372: *> hand sides K with K > J may not be guaranteed as well, but
373: *> only the first such right-hand side is reported. If a small
374: *> componentwise error is not requested (PARAMS(3) = 0.0) then
375: *> the Jth right-hand side is the first with a normwise error
376: *> bound that is not guaranteed (the smallest J such
377: *> that ERR_BNDS_NORM(J,1) = 0.0). By default (PARAMS(3) = 1.0)
378: *> the Jth right-hand side is the first with either a normwise or
379: *> componentwise error bound that is not guaranteed (the smallest
380: *> J such that either ERR_BNDS_NORM(J,1) = 0.0 or
381: *> ERR_BNDS_COMP(J,1) = 0.0). See the definition of
382: *> ERR_BNDS_NORM(:,1) and ERR_BNDS_COMP(:,1). To get information
383: *> about all of the right-hand sides check ERR_BNDS_NORM or
384: *> ERR_BNDS_COMP.
385: *> \endverbatim
386: *
387: * Authors:
388: * ========
389: *
1.12 bertrand 390: *> \author Univ. of Tennessee
391: *> \author Univ. of California Berkeley
392: *> \author Univ. of Colorado Denver
393: *> \author NAG Ltd.
1.5 bertrand 394: *
1.7 bertrand 395: *> \date April 2012
1.5 bertrand 396: *
397: *> \ingroup doubleSYcomputational
398: *
399: * =====================================================================
1.1 bertrand 400: SUBROUTINE DSYRFSX( UPLO, EQUED, N, NRHS, A, LDA, AF, LDAF, IPIV,
401: $ S, B, LDB, X, LDX, RCOND, BERR, N_ERR_BNDS,
402: $ ERR_BNDS_NORM, ERR_BNDS_COMP, NPARAMS, PARAMS,
403: $ WORK, IWORK, INFO )
404: *
1.12 bertrand 405: * -- LAPACK computational routine (version 3.7.0) --
1.5 bertrand 406: * -- LAPACK is a software package provided by Univ. of Tennessee, --
407: * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
1.7 bertrand 408: * April 2012
1.1 bertrand 409: *
410: * .. Scalar Arguments ..
411: CHARACTER UPLO, EQUED
412: INTEGER INFO, LDA, LDAF, LDB, LDX, N, NRHS, NPARAMS,
413: $ N_ERR_BNDS
414: DOUBLE PRECISION RCOND
415: * ..
416: * .. Array Arguments ..
417: INTEGER IPIV( * ), IWORK( * )
418: DOUBLE PRECISION A( LDA, * ), AF( LDAF, * ), B( LDB, * ),
419: $ X( LDX, * ), WORK( * )
420: DOUBLE PRECISION S( * ), PARAMS( * ), BERR( * ),
421: $ ERR_BNDS_NORM( NRHS, * ),
422: $ ERR_BNDS_COMP( NRHS, * )
423: * ..
424: *
1.5 bertrand 425: * ==================================================================
1.1 bertrand 426: *
427: * .. Parameters ..
428: DOUBLE PRECISION ZERO, ONE
429: PARAMETER ( ZERO = 0.0D+0, ONE = 1.0D+0 )
430: DOUBLE PRECISION ITREF_DEFAULT, ITHRESH_DEFAULT
431: DOUBLE PRECISION COMPONENTWISE_DEFAULT, RTHRESH_DEFAULT
432: DOUBLE PRECISION DZTHRESH_DEFAULT
433: PARAMETER ( ITREF_DEFAULT = 1.0D+0 )
434: PARAMETER ( ITHRESH_DEFAULT = 10.0D+0 )
435: PARAMETER ( COMPONENTWISE_DEFAULT = 1.0D+0 )
436: PARAMETER ( RTHRESH_DEFAULT = 0.5D+0 )
437: PARAMETER ( DZTHRESH_DEFAULT = 0.25D+0 )
438: INTEGER LA_LINRX_ITREF_I, LA_LINRX_ITHRESH_I,
439: $ LA_LINRX_CWISE_I
440: PARAMETER ( LA_LINRX_ITREF_I = 1,
441: $ LA_LINRX_ITHRESH_I = 2 )
442: PARAMETER ( LA_LINRX_CWISE_I = 3 )
443: INTEGER LA_LINRX_TRUST_I, LA_LINRX_ERR_I,
444: $ LA_LINRX_RCOND_I
445: PARAMETER ( LA_LINRX_TRUST_I = 1, LA_LINRX_ERR_I = 2 )
446: PARAMETER ( LA_LINRX_RCOND_I = 3 )
447: * ..
448: * .. Local Scalars ..
449: CHARACTER(1) NORM
450: LOGICAL RCEQU
451: INTEGER J, PREC_TYPE, REF_TYPE, N_NORMS
452: DOUBLE PRECISION ANORM, RCOND_TMP
453: DOUBLE PRECISION ILLRCOND_THRESH, ERR_LBND, CWISE_WRONG
454: LOGICAL IGNORE_CWISE
455: INTEGER ITHRESH
456: DOUBLE PRECISION RTHRESH, UNSTABLE_THRESH
457: * ..
458: * .. External Subroutines ..
459: EXTERNAL XERBLA, DSYCON, DLA_SYRFSX_EXTENDED
460: * ..
461: * .. Intrinsic Functions ..
462: INTRINSIC MAX, SQRT
463: * ..
464: * .. External Functions ..
1.12 bertrand 465: EXTERNAL LSAME, ILAPREC
1.1 bertrand 466: EXTERNAL DLAMCH, DLANSY, DLA_SYRCOND
467: DOUBLE PRECISION DLAMCH, DLANSY, DLA_SYRCOND
468: LOGICAL LSAME
1.12 bertrand 469: INTEGER ILAPREC
1.1 bertrand 470: * ..
471: * .. Executable Statements ..
472: *
473: * Check the input parameters.
474: *
475: INFO = 0
476: REF_TYPE = INT( ITREF_DEFAULT )
477: IF ( NPARAMS .GE. LA_LINRX_ITREF_I ) THEN
478: IF ( PARAMS( LA_LINRX_ITREF_I ) .LT. 0.0D+0 ) THEN
479: PARAMS( LA_LINRX_ITREF_I ) = ITREF_DEFAULT
480: ELSE
481: REF_TYPE = PARAMS( LA_LINRX_ITREF_I )
482: END IF
483: END IF
484: *
485: * Set default parameters.
486: *
487: ILLRCOND_THRESH = DBLE( N )*DLAMCH( 'Epsilon' )
488: ITHRESH = INT( ITHRESH_DEFAULT )
489: RTHRESH = RTHRESH_DEFAULT
490: UNSTABLE_THRESH = DZTHRESH_DEFAULT
491: IGNORE_CWISE = COMPONENTWISE_DEFAULT .EQ. 0.0D+0
492: *
493: IF ( NPARAMS.GE.LA_LINRX_ITHRESH_I ) THEN
494: IF ( PARAMS( LA_LINRX_ITHRESH_I ).LT.0.0D+0 ) THEN
495: PARAMS( LA_LINRX_ITHRESH_I ) = ITHRESH
496: ELSE
497: ITHRESH = INT( PARAMS( LA_LINRX_ITHRESH_I ) )
498: END IF
499: END IF
500: IF ( NPARAMS.GE.LA_LINRX_CWISE_I ) THEN
501: IF ( PARAMS( LA_LINRX_CWISE_I ).LT.0.0D+0 ) THEN
502: IF ( IGNORE_CWISE ) THEN
503: PARAMS( LA_LINRX_CWISE_I ) = 0.0D+0
504: ELSE
505: PARAMS( LA_LINRX_CWISE_I ) = 1.0D+0
506: END IF
507: ELSE
508: IGNORE_CWISE = PARAMS( LA_LINRX_CWISE_I ) .EQ. 0.0D+0
509: END IF
510: END IF
511: IF ( REF_TYPE .EQ. 0 .OR. N_ERR_BNDS .EQ. 0 ) THEN
512: N_NORMS = 0
513: ELSE IF ( IGNORE_CWISE ) THEN
514: N_NORMS = 1
515: ELSE
516: N_NORMS = 2
517: END IF
518: *
519: RCEQU = LSAME( EQUED, 'Y' )
520: *
521: * Test input parameters.
522: *
523: IF ( .NOT.LSAME( UPLO, 'U' ) .AND. .NOT.LSAME( UPLO, 'L' ) ) THEN
524: INFO = -1
525: ELSE IF( .NOT.RCEQU .AND. .NOT.LSAME( EQUED, 'N' ) ) THEN
526: INFO = -2
527: ELSE IF( N.LT.0 ) THEN
528: INFO = -3
529: ELSE IF( NRHS.LT.0 ) THEN
530: INFO = -4
531: ELSE IF( LDA.LT.MAX( 1, N ) ) THEN
532: INFO = -6
533: ELSE IF( LDAF.LT.MAX( 1, N ) ) THEN
534: INFO = -8
535: ELSE IF( LDB.LT.MAX( 1, N ) ) THEN
1.7 bertrand 536: INFO = -12
1.1 bertrand 537: ELSE IF( LDX.LT.MAX( 1, N ) ) THEN
1.7 bertrand 538: INFO = -14
1.1 bertrand 539: END IF
540: IF( INFO.NE.0 ) THEN
541: CALL XERBLA( 'DSYRFSX', -INFO )
542: RETURN
543: END IF
544: *
545: * Quick return if possible.
546: *
547: IF( N.EQ.0 .OR. NRHS.EQ.0 ) THEN
548: RCOND = 1.0D+0
549: DO J = 1, NRHS
550: BERR( J ) = 0.0D+0
551: IF ( N_ERR_BNDS .GE. 1 ) THEN
552: ERR_BNDS_NORM( J, LA_LINRX_TRUST_I ) = 1.0D+0
553: ERR_BNDS_COMP( J, LA_LINRX_TRUST_I ) = 1.0D+0
554: END IF
555: IF ( N_ERR_BNDS .GE. 2 ) THEN
556: ERR_BNDS_NORM( J, LA_LINRX_ERR_I ) = 0.0D+0
557: ERR_BNDS_COMP( J, LA_LINRX_ERR_I ) = 0.0D+0
558: END IF
559: IF ( N_ERR_BNDS .GE. 3 ) THEN
560: ERR_BNDS_NORM( J, LA_LINRX_RCOND_I ) = 1.0D+0
561: ERR_BNDS_COMP( J, LA_LINRX_RCOND_I ) = 1.0D+0
562: END IF
563: END DO
564: RETURN
565: END IF
566: *
567: * Default to failure.
568: *
569: RCOND = 0.0D+0
570: DO J = 1, NRHS
571: BERR( J ) = 1.0D+0
572: IF ( N_ERR_BNDS .GE. 1 ) THEN
573: ERR_BNDS_NORM( J, LA_LINRX_TRUST_I ) = 1.0D+0
574: ERR_BNDS_COMP( J, LA_LINRX_TRUST_I ) = 1.0D+0
575: END IF
576: IF ( N_ERR_BNDS .GE. 2 ) THEN
577: ERR_BNDS_NORM( J, LA_LINRX_ERR_I ) = 1.0D+0
578: ERR_BNDS_COMP( J, LA_LINRX_ERR_I ) = 1.0D+0
579: END IF
580: IF ( N_ERR_BNDS .GE. 3 ) THEN
581: ERR_BNDS_NORM( J, LA_LINRX_RCOND_I ) = 0.0D+0
582: ERR_BNDS_COMP( J, LA_LINRX_RCOND_I ) = 0.0D+0
583: END IF
584: END DO
585: *
586: * Compute the norm of A and the reciprocal of the condition
587: * number of A.
588: *
589: NORM = 'I'
590: ANORM = DLANSY( NORM, UPLO, N, A, LDA, WORK )
591: CALL DSYCON( UPLO, N, AF, LDAF, IPIV, ANORM, RCOND, WORK,
592: $ IWORK, INFO )
593: *
594: * Perform refinement on each right-hand side
595: *
596: IF ( REF_TYPE .NE. 0 ) THEN
597:
598: PREC_TYPE = ILAPREC( 'E' )
599:
600: CALL DLA_SYRFSX_EXTENDED( PREC_TYPE, UPLO, N,
601: $ NRHS, A, LDA, AF, LDAF, IPIV, RCEQU, S, B,
602: $ LDB, X, LDX, BERR, N_NORMS, ERR_BNDS_NORM, ERR_BNDS_COMP,
603: $ WORK( N+1 ), WORK( 1 ), WORK( 2*N+1 ), WORK( 1 ), RCOND,
604: $ ITHRESH, RTHRESH, UNSTABLE_THRESH, IGNORE_CWISE,
605: $ INFO )
606: END IF
607:
608: ERR_LBND = MAX( 10.0D+0, SQRT( DBLE( N ) ) )*DLAMCH( 'Epsilon' )
609: IF (N_ERR_BNDS .GE. 1 .AND. N_NORMS .GE. 1) THEN
610: *
611: * Compute scaled normwise condition number cond(A*C).
612: *
613: IF ( RCEQU ) THEN
614: RCOND_TMP = DLA_SYRCOND( UPLO, N, A, LDA, AF, LDAF, IPIV,
615: $ -1, S, INFO, WORK, IWORK )
616: ELSE
617: RCOND_TMP = DLA_SYRCOND( UPLO, N, A, LDA, AF, LDAF, IPIV,
618: $ 0, S, INFO, WORK, IWORK )
619: END IF
620: DO J = 1, NRHS
621: *
622: * Cap the error at 1.0.
623: *
624: IF (N_ERR_BNDS .GE. LA_LINRX_ERR_I
625: $ .AND. ERR_BNDS_NORM( J, LA_LINRX_ERR_I ) .GT. 1.0D+0)
626: $ ERR_BNDS_NORM( J, LA_LINRX_ERR_I ) = 1.0D+0
627: *
628: * Threshold the error (see LAWN).
629: *
630: IF ( RCOND_TMP .LT. ILLRCOND_THRESH ) THEN
631: ERR_BNDS_NORM( J, LA_LINRX_ERR_I ) = 1.0D+0
632: ERR_BNDS_NORM( J, LA_LINRX_TRUST_I ) = 0.0D+0
633: IF ( INFO .LE. N ) INFO = N + J
634: ELSE IF (ERR_BNDS_NORM( J, LA_LINRX_ERR_I ) .LT. ERR_LBND)
635: $ THEN
636: ERR_BNDS_NORM( J, LA_LINRX_ERR_I ) = ERR_LBND
637: ERR_BNDS_NORM( J, LA_LINRX_TRUST_I ) = 1.0D+0
638: END IF
639: *
640: * Save the condition number.
641: *
642: IF (N_ERR_BNDS .GE. LA_LINRX_RCOND_I) THEN
643: ERR_BNDS_NORM( J, LA_LINRX_RCOND_I ) = RCOND_TMP
644: END IF
645: END DO
646: END IF
647:
648: IF ( N_ERR_BNDS .GE. 1 .AND. N_NORMS .GE. 2 ) THEN
649: *
650: * Compute componentwise condition number cond(A*diag(Y(:,J))) for
651: * each right-hand side using the current solution as an estimate of
652: * the true solution. If the componentwise error estimate is too
653: * large, then the solution is a lousy estimate of truth and the
654: * estimated RCOND may be too optimistic. To avoid misleading users,
655: * the inverse condition number is set to 0.0 when the estimated
656: * cwise error is at least CWISE_WRONG.
657: *
658: CWISE_WRONG = SQRT( DLAMCH( 'Epsilon' ) )
659: DO J = 1, NRHS
660: IF ( ERR_BNDS_COMP( J, LA_LINRX_ERR_I ) .LT. CWISE_WRONG )
661: $ THEN
662: RCOND_TMP = DLA_SYRCOND( UPLO, N, A, LDA, AF, LDAF, IPIV,
663: $ 1, X(1,J), INFO, WORK, IWORK )
664: ELSE
665: RCOND_TMP = 0.0D+0
666: END IF
667: *
668: * Cap the error at 1.0.
669: *
670: IF ( N_ERR_BNDS .GE. LA_LINRX_ERR_I
671: $ .AND. ERR_BNDS_COMP( J, LA_LINRX_ERR_I ) .GT. 1.0D+0 )
672: $ ERR_BNDS_COMP( J, LA_LINRX_ERR_I ) = 1.0D+0
673: *
674: * Threshold the error (see LAWN).
675: *
676: IF ( RCOND_TMP .LT. ILLRCOND_THRESH ) THEN
677: ERR_BNDS_COMP( J, LA_LINRX_ERR_I ) = 1.0D+0
678: ERR_BNDS_COMP( J, LA_LINRX_TRUST_I ) = 0.0D+0
1.7 bertrand 679: IF ( .NOT. IGNORE_CWISE
1.1 bertrand 680: $ .AND. INFO.LT.N + J ) INFO = N + J
681: ELSE IF ( ERR_BNDS_COMP( J, LA_LINRX_ERR_I )
682: $ .LT. ERR_LBND ) THEN
683: ERR_BNDS_COMP( J, LA_LINRX_ERR_I ) = ERR_LBND
684: ERR_BNDS_COMP( J, LA_LINRX_TRUST_I ) = 1.0D+0
685: END IF
686: *
687: * Save the condition number.
688: *
689: IF ( N_ERR_BNDS .GE. LA_LINRX_RCOND_I ) THEN
690: ERR_BNDS_COMP( J, LA_LINRX_RCOND_I ) = RCOND_TMP
691: END IF
692:
693: END DO
694: END IF
695: *
696: RETURN
697: *
698: * End of DSYRFSX
699: *
700: END
CVSweb interface <joel.bertrand@systella.fr>
![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to dsyrfsx.f CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [local] / rpl / lapack / lapack