Mercurial > hg > octave-nkf
comparison liboctave/fCmplxSVD.cc @ 7789:82be108cc558
First attempt at single precision tyeps
* * *
corrections to qrupdate single precision routines
* * *
prefer demotion to single over promotion to double
* * *
Add single precision support to log2 function
* * *
Trivial PROJECT file update
* * *
Cache optimized hermitian/transpose methods
* * *
Add tests for tranpose/hermitian and ChangeLog entry for new transpose code
author | David Bateman <dbateman@free.fr> |
---|---|
date | Sun, 27 Apr 2008 22:34:17 +0200 |
parents | |
children | eb63fbe60fab |
comparison
equal
deleted
inserted
replaced
7788:45f5faba05a2 | 7789:82be108cc558 |
---|---|
1 /* | |
2 | |
3 Copyright (C) 1994, 1995, 1996, 1997, 1999, 2002, 2003, 2004, 2005, | |
4 2007 John W. Eaton | |
5 | |
6 This file is part of Octave. | |
7 | |
8 Octave is free software; you can redistribute it and/or modify it | |
9 under the terms of the GNU General Public License as published by the | |
10 Free Software Foundation; either version 3 of the License, or (at your | |
11 option) any later version. | |
12 | |
13 Octave is distributed in the hope that it will be useful, but WITHOUT | |
14 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
15 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License | |
16 for more details. | |
17 | |
18 You should have received a copy of the GNU General Public License | |
19 along with Octave; see the file COPYING. If not, see | |
20 <http://www.gnu.org/licenses/>. | |
21 | |
22 */ | |
23 | |
24 #ifdef HAVE_CONFIG_H | |
25 #include <config.h> | |
26 #endif | |
27 | |
28 #include "fCmplxSVD.h" | |
29 #include "f77-fcn.h" | |
30 #include "lo-error.h" | |
31 | |
32 extern "C" | |
33 { | |
34 F77_RET_T | |
35 F77_FUNC (cgesvd, CGESVD) (F77_CONST_CHAR_ARG_DECL, | |
36 F77_CONST_CHAR_ARG_DECL, | |
37 const octave_idx_type&, const octave_idx_type&, FloatComplex*, | |
38 const octave_idx_type&, float*, FloatComplex*, const octave_idx_type&, | |
39 FloatComplex*, const octave_idx_type&, FloatComplex*, const octave_idx_type&, | |
40 float*, octave_idx_type& | |
41 F77_CHAR_ARG_LEN_DECL | |
42 F77_CHAR_ARG_LEN_DECL); | |
43 } | |
44 | |
45 FloatComplexMatrix | |
46 FloatComplexSVD::left_singular_matrix (void) const | |
47 { | |
48 if (type_computed == SVD::sigma_only) | |
49 { | |
50 (*current_liboctave_error_handler) | |
51 ("FloatComplexSVD: U not computed because type == SVD::sigma_only"); | |
52 return FloatComplexMatrix (); | |
53 } | |
54 else | |
55 return left_sm; | |
56 } | |
57 | |
58 FloatComplexMatrix | |
59 FloatComplexSVD::right_singular_matrix (void) const | |
60 { | |
61 if (type_computed == SVD::sigma_only) | |
62 { | |
63 (*current_liboctave_error_handler) | |
64 ("FloatComplexSVD: V not computed because type == SVD::sigma_only"); | |
65 return FloatComplexMatrix (); | |
66 } | |
67 else | |
68 return right_sm; | |
69 } | |
70 | |
71 octave_idx_type | |
72 FloatComplexSVD::init (const FloatComplexMatrix& a, SVD::type svd_type) | |
73 { | |
74 octave_idx_type info; | |
75 | |
76 octave_idx_type m = a.rows (); | |
77 octave_idx_type n = a.cols (); | |
78 | |
79 FloatComplexMatrix atmp = a; | |
80 FloatComplex *tmp_data = atmp.fortran_vec (); | |
81 | |
82 octave_idx_type min_mn = m < n ? m : n; | |
83 octave_idx_type max_mn = m > n ? m : n; | |
84 | |
85 char jobu = 'A'; | |
86 char jobv = 'A'; | |
87 | |
88 octave_idx_type ncol_u = m; | |
89 octave_idx_type nrow_vt = n; | |
90 octave_idx_type nrow_s = m; | |
91 octave_idx_type ncol_s = n; | |
92 | |
93 switch (svd_type) | |
94 { | |
95 case SVD::economy: | |
96 jobu = jobv = 'S'; | |
97 ncol_u = nrow_vt = nrow_s = ncol_s = min_mn; | |
98 break; | |
99 | |
100 case SVD::sigma_only: | |
101 | |
102 // Note: for this case, both jobu and jobv should be 'N', but | |
103 // there seems to be a bug in dgesvd from Lapack V2.0. To | |
104 // demonstrate the bug, set both jobu and jobv to 'N' and find | |
105 // the singular values of [eye(3), eye(3)]. The result is | |
106 // [-sqrt(2), -sqrt(2), -sqrt(2)]. | |
107 // | |
108 // For Lapack 3.0, this problem seems to be fixed. | |
109 | |
110 jobu = 'N'; | |
111 jobv = 'N'; | |
112 ncol_u = nrow_vt = 1; | |
113 break; | |
114 | |
115 default: | |
116 break; | |
117 } | |
118 | |
119 type_computed = svd_type; | |
120 | |
121 if (! (jobu == 'N' || jobu == 'O')) | |
122 left_sm.resize (m, ncol_u); | |
123 | |
124 FloatComplex *u = left_sm.fortran_vec (); | |
125 | |
126 sigma.resize (nrow_s, ncol_s); | |
127 float *s_vec = sigma.fortran_vec (); | |
128 | |
129 if (! (jobv == 'N' || jobv == 'O')) | |
130 right_sm.resize (nrow_vt, n); | |
131 | |
132 FloatComplex *vt = right_sm.fortran_vec (); | |
133 | |
134 octave_idx_type lrwork = 5*max_mn; | |
135 | |
136 Array<float> rwork (lrwork); | |
137 | |
138 // Ask ZGESVD what the dimension of WORK should be. | |
139 | |
140 octave_idx_type lwork = -1; | |
141 | |
142 Array<FloatComplex> work (1); | |
143 | |
144 F77_XFCN (cgesvd, CGESVD, (F77_CONST_CHAR_ARG2 (&jobu, 1), | |
145 F77_CONST_CHAR_ARG2 (&jobv, 1), | |
146 m, n, tmp_data, m, s_vec, u, m, vt, | |
147 nrow_vt, work.fortran_vec (), lwork, | |
148 rwork.fortran_vec (), info | |
149 F77_CHAR_ARG_LEN (1) | |
150 F77_CHAR_ARG_LEN (1))); | |
151 | |
152 lwork = static_cast<octave_idx_type> (work(0).real ()); | |
153 work.resize (lwork); | |
154 | |
155 F77_XFCN (cgesvd, CGESVD, (F77_CONST_CHAR_ARG2 (&jobu, 1), | |
156 F77_CONST_CHAR_ARG2 (&jobv, 1), | |
157 m, n, tmp_data, m, s_vec, u, m, vt, | |
158 nrow_vt, work.fortran_vec (), lwork, | |
159 rwork.fortran_vec (), info | |
160 F77_CHAR_ARG_LEN (1) | |
161 F77_CHAR_ARG_LEN (1))); | |
162 | |
163 if (! (jobv == 'N' || jobv == 'O')) | |
164 right_sm = right_sm.hermitian (); | |
165 | |
166 return info; | |
167 } | |
168 | |
169 /* | |
170 ;;; Local Variables: *** | |
171 ;;; mode: C++ *** | |
172 ;;; End: *** | |
173 */ |