Mercurial > octave-libgccjit
comparison liboctave/fCRowVector.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 // RowVector manipulations. | |
2 /* | |
3 | |
4 Copyright (C) 1994, 1995, 1996, 1997, 1999, 2000, 2001, 2002, 2003, | |
5 2004, 2005, 2006, 2007 John W. Eaton | |
6 | |
7 This file is part of Octave. | |
8 | |
9 Octave is free software; you can redistribute it and/or modify it | |
10 under the terms of the GNU General Public License as published by the | |
11 Free Software Foundation; either version 3 of the License, or (at your | |
12 option) any later version. | |
13 | |
14 Octave is distributed in the hope that it will be useful, but WITHOUT | |
15 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
16 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License | |
17 for more details. | |
18 | |
19 You should have received a copy of the GNU General Public License | |
20 along with Octave; see the file COPYING. If not, see | |
21 <http://www.gnu.org/licenses/>. | |
22 | |
23 */ | |
24 | |
25 #ifdef HAVE_CONFIG_H | |
26 #include <config.h> | |
27 #endif | |
28 | |
29 #include <iostream> | |
30 | |
31 #include "Array-util.h" | |
32 #include "f77-fcn.h" | |
33 #include "functor.h" | |
34 #include "lo-error.h" | |
35 #include "mx-base.h" | |
36 #include "mx-inlines.cc" | |
37 #include "oct-cmplx.h" | |
38 | |
39 // Fortran functions we call. | |
40 | |
41 extern "C" | |
42 { | |
43 F77_RET_T | |
44 F77_FUNC (cgemv, CGEMV) (F77_CONST_CHAR_ARG_DECL, | |
45 const octave_idx_type&, const octave_idx_type&, const FloatComplex&, | |
46 const FloatComplex*, const octave_idx_type&, const FloatComplex*, | |
47 const octave_idx_type&, const FloatComplex&, FloatComplex*, const octave_idx_type& | |
48 F77_CHAR_ARG_LEN_DECL); | |
49 | |
50 F77_RET_T | |
51 F77_FUNC (xcdotu, XCDOTU) (const octave_idx_type&, const FloatComplex*, const octave_idx_type&, | |
52 const FloatComplex*, const octave_idx_type&, FloatComplex&); | |
53 } | |
54 | |
55 // FloatComplex Row Vector class | |
56 | |
57 FloatComplexRowVector::FloatComplexRowVector (const FloatRowVector& a) | |
58 : MArray<FloatComplex> (a.length ()) | |
59 { | |
60 for (octave_idx_type i = 0; i < length (); i++) | |
61 elem (i) = a.elem (i); | |
62 } | |
63 | |
64 bool | |
65 FloatComplexRowVector::operator == (const FloatComplexRowVector& a) const | |
66 { | |
67 octave_idx_type len = length (); | |
68 if (len != a.length ()) | |
69 return 0; | |
70 return mx_inline_equal (data (), a.data (), len); | |
71 } | |
72 | |
73 bool | |
74 FloatComplexRowVector::operator != (const FloatComplexRowVector& a) const | |
75 { | |
76 return !(*this == a); | |
77 } | |
78 | |
79 // destructive insert/delete/reorder operations | |
80 | |
81 FloatComplexRowVector& | |
82 FloatComplexRowVector::insert (const FloatRowVector& a, octave_idx_type c) | |
83 { | |
84 octave_idx_type a_len = a.length (); | |
85 | |
86 if (c < 0 || c + a_len > length ()) | |
87 { | |
88 (*current_liboctave_error_handler) ("range error for insert"); | |
89 return *this; | |
90 } | |
91 | |
92 if (a_len > 0) | |
93 { | |
94 make_unique (); | |
95 | |
96 for (octave_idx_type i = 0; i < a_len; i++) | |
97 xelem (c+i) = a.elem (i); | |
98 } | |
99 | |
100 return *this; | |
101 } | |
102 | |
103 FloatComplexRowVector& | |
104 FloatComplexRowVector::insert (const FloatComplexRowVector& a, octave_idx_type c) | |
105 { | |
106 octave_idx_type a_len = a.length (); | |
107 | |
108 if (c < 0 || c + a_len > length ()) | |
109 { | |
110 (*current_liboctave_error_handler) ("range error for insert"); | |
111 return *this; | |
112 } | |
113 | |
114 if (a_len > 0) | |
115 { | |
116 make_unique (); | |
117 | |
118 for (octave_idx_type i = 0; i < a_len; i++) | |
119 xelem (c+i) = a.elem (i); | |
120 } | |
121 | |
122 return *this; | |
123 } | |
124 | |
125 FloatComplexRowVector& | |
126 FloatComplexRowVector::fill (float val) | |
127 { | |
128 octave_idx_type len = length (); | |
129 | |
130 if (len > 0) | |
131 { | |
132 make_unique (); | |
133 | |
134 for (octave_idx_type i = 0; i < len; i++) | |
135 xelem (i) = val; | |
136 } | |
137 | |
138 return *this; | |
139 } | |
140 | |
141 FloatComplexRowVector& | |
142 FloatComplexRowVector::fill (const FloatComplex& val) | |
143 { | |
144 octave_idx_type len = length (); | |
145 | |
146 if (len > 0) | |
147 { | |
148 make_unique (); | |
149 | |
150 for (octave_idx_type i = 0; i < len; i++) | |
151 xelem (i) = val; | |
152 } | |
153 | |
154 return *this; | |
155 } | |
156 | |
157 FloatComplexRowVector& | |
158 FloatComplexRowVector::fill (float val, octave_idx_type c1, octave_idx_type c2) | |
159 { | |
160 octave_idx_type len = length (); | |
161 | |
162 if (c1 < 0 || c2 < 0 || c1 >= len || c2 >= len) | |
163 { | |
164 (*current_liboctave_error_handler) ("range error for fill"); | |
165 return *this; | |
166 } | |
167 | |
168 if (c1 > c2) { octave_idx_type tmp = c1; c1 = c2; c2 = tmp; } | |
169 | |
170 if (c2 >= c1) | |
171 { | |
172 make_unique (); | |
173 | |
174 for (octave_idx_type i = c1; i <= c2; i++) | |
175 xelem (i) = val; | |
176 } | |
177 | |
178 return *this; | |
179 } | |
180 | |
181 FloatComplexRowVector& | |
182 FloatComplexRowVector::fill (const FloatComplex& val, octave_idx_type c1, octave_idx_type c2) | |
183 { | |
184 octave_idx_type len = length (); | |
185 | |
186 if (c1 < 0 || c2 < 0 || c1 >= len || c2 >= len) | |
187 { | |
188 (*current_liboctave_error_handler) ("range error for fill"); | |
189 return *this; | |
190 } | |
191 | |
192 if (c1 > c2) { octave_idx_type tmp = c1; c1 = c2; c2 = tmp; } | |
193 | |
194 if (c2 >= c1) | |
195 { | |
196 make_unique (); | |
197 | |
198 for (octave_idx_type i = c1; i <= c2; i++) | |
199 xelem (i) = val; | |
200 } | |
201 | |
202 return *this; | |
203 } | |
204 | |
205 FloatComplexRowVector | |
206 FloatComplexRowVector::append (const FloatRowVector& a) const | |
207 { | |
208 octave_idx_type len = length (); | |
209 octave_idx_type nc_insert = len; | |
210 FloatComplexRowVector retval (len + a.length ()); | |
211 retval.insert (*this, 0); | |
212 retval.insert (a, nc_insert); | |
213 return retval; | |
214 } | |
215 | |
216 FloatComplexRowVector | |
217 FloatComplexRowVector::append (const FloatComplexRowVector& a) const | |
218 { | |
219 octave_idx_type len = length (); | |
220 octave_idx_type nc_insert = len; | |
221 FloatComplexRowVector retval (len + a.length ()); | |
222 retval.insert (*this, 0); | |
223 retval.insert (a, nc_insert); | |
224 return retval; | |
225 } | |
226 | |
227 FloatComplexColumnVector | |
228 FloatComplexRowVector::hermitian (void) const | |
229 { | |
230 return MArray<FloatComplex>::hermitian (std::conj); | |
231 } | |
232 | |
233 FloatComplexColumnVector | |
234 FloatComplexRowVector::transpose (void) const | |
235 { | |
236 return MArray<FloatComplex>::transpose (); | |
237 } | |
238 | |
239 FloatComplexRowVector | |
240 conj (const FloatComplexRowVector& a) | |
241 { | |
242 octave_idx_type a_len = a.length (); | |
243 FloatComplexRowVector retval; | |
244 if (a_len > 0) | |
245 retval = FloatComplexRowVector (mx_inline_conj_dup (a.data (), a_len), a_len); | |
246 return retval; | |
247 } | |
248 | |
249 // resize is the destructive equivalent for this one | |
250 | |
251 FloatComplexRowVector | |
252 FloatComplexRowVector::extract (octave_idx_type c1, octave_idx_type c2) const | |
253 { | |
254 if (c1 > c2) { octave_idx_type tmp = c1; c1 = c2; c2 = tmp; } | |
255 | |
256 octave_idx_type new_c = c2 - c1 + 1; | |
257 | |
258 FloatComplexRowVector result (new_c); | |
259 | |
260 for (octave_idx_type i = 0; i < new_c; i++) | |
261 result.elem (i) = elem (c1+i); | |
262 | |
263 return result; | |
264 } | |
265 | |
266 FloatComplexRowVector | |
267 FloatComplexRowVector::extract_n (octave_idx_type r1, octave_idx_type n) const | |
268 { | |
269 FloatComplexRowVector result (n); | |
270 | |
271 for (octave_idx_type i = 0; i < n; i++) | |
272 result.elem (i) = elem (r1+i); | |
273 | |
274 return result; | |
275 } | |
276 | |
277 // row vector by row vector -> row vector operations | |
278 | |
279 FloatComplexRowVector& | |
280 FloatComplexRowVector::operator += (const FloatRowVector& a) | |
281 { | |
282 octave_idx_type len = length (); | |
283 | |
284 octave_idx_type a_len = a.length (); | |
285 | |
286 if (len != a_len) | |
287 { | |
288 gripe_nonconformant ("operator +=", len, a_len); | |
289 return *this; | |
290 } | |
291 | |
292 if (len == 0) | |
293 return *this; | |
294 | |
295 FloatComplex *d = fortran_vec (); // Ensures only one reference to my privates! | |
296 | |
297 mx_inline_add2 (d, a.data (), len); | |
298 return *this; | |
299 } | |
300 | |
301 FloatComplexRowVector& | |
302 FloatComplexRowVector::operator -= (const FloatRowVector& a) | |
303 { | |
304 octave_idx_type len = length (); | |
305 | |
306 octave_idx_type a_len = a.length (); | |
307 | |
308 if (len != a_len) | |
309 { | |
310 gripe_nonconformant ("operator -=", len, a_len); | |
311 return *this; | |
312 } | |
313 | |
314 if (len == 0) | |
315 return *this; | |
316 | |
317 FloatComplex *d = fortran_vec (); // Ensures only one reference to my privates! | |
318 | |
319 mx_inline_subtract2 (d, a.data (), len); | |
320 return *this; | |
321 } | |
322 | |
323 // row vector by matrix -> row vector | |
324 | |
325 FloatComplexRowVector | |
326 operator * (const FloatComplexRowVector& v, const FloatComplexMatrix& a) | |
327 { | |
328 FloatComplexRowVector retval; | |
329 | |
330 octave_idx_type len = v.length (); | |
331 | |
332 octave_idx_type a_nr = a.rows (); | |
333 octave_idx_type a_nc = a.cols (); | |
334 | |
335 if (a_nr != len) | |
336 gripe_nonconformant ("operator *", 1, len, a_nr, a_nc); | |
337 else | |
338 { | |
339 if (len == 0) | |
340 retval.resize (a_nc, 0.0); | |
341 else | |
342 { | |
343 // Transpose A to form A'*x == (x'*A)' | |
344 | |
345 octave_idx_type ld = a_nr; | |
346 | |
347 retval.resize (a_nc); | |
348 FloatComplex *y = retval.fortran_vec (); | |
349 | |
350 F77_XFCN (cgemv, CGEMV, (F77_CONST_CHAR_ARG2 ("T", 1), | |
351 a_nr, a_nc, 1.0, a.data (), | |
352 ld, v.data (), 1, 0.0, y, 1 | |
353 F77_CHAR_ARG_LEN (1))); | |
354 } | |
355 } | |
356 | |
357 return retval; | |
358 } | |
359 | |
360 FloatComplexRowVector | |
361 operator * (const FloatRowVector& v, const FloatComplexMatrix& a) | |
362 { | |
363 FloatComplexRowVector tmp (v); | |
364 return tmp * a; | |
365 } | |
366 | |
367 // other operations | |
368 | |
369 FloatRowVector | |
370 FloatComplexRowVector::map (dmapper fcn) const | |
371 { | |
372 return MArray<FloatComplex>::map<float> (func_ptr (fcn)); | |
373 } | |
374 | |
375 FloatComplexRowVector | |
376 FloatComplexRowVector::map (cmapper fcn) const | |
377 { | |
378 return MArray<FloatComplex>::map<FloatComplex> (func_ptr (fcn)); | |
379 } | |
380 | |
381 FloatComplex | |
382 FloatComplexRowVector::min (void) const | |
383 { | |
384 octave_idx_type len = length (); | |
385 if (len == 0) | |
386 return FloatComplex (0.0); | |
387 | |
388 FloatComplex res = elem (0); | |
389 float absres = std::abs (res); | |
390 | |
391 for (octave_idx_type i = 1; i < len; i++) | |
392 if (std::abs (elem (i)) < absres) | |
393 { | |
394 res = elem (i); | |
395 absres = std::abs (res); | |
396 } | |
397 | |
398 return res; | |
399 } | |
400 | |
401 FloatComplex | |
402 FloatComplexRowVector::max (void) const | |
403 { | |
404 octave_idx_type len = length (); | |
405 if (len == 0) | |
406 return FloatComplex (0.0); | |
407 | |
408 FloatComplex res = elem (0); | |
409 float absres = std::abs (res); | |
410 | |
411 for (octave_idx_type i = 1; i < len; i++) | |
412 if (std::abs (elem (i)) > absres) | |
413 { | |
414 res = elem (i); | |
415 absres = std::abs (res); | |
416 } | |
417 | |
418 return res; | |
419 } | |
420 | |
421 // i/o | |
422 | |
423 std::ostream& | |
424 operator << (std::ostream& os, const FloatComplexRowVector& a) | |
425 { | |
426 // int field_width = os.precision () + 7; | |
427 for (octave_idx_type i = 0; i < a.length (); i++) | |
428 os << " " /* setw (field_width) */ << a.elem (i); | |
429 return os; | |
430 } | |
431 | |
432 std::istream& | |
433 operator >> (std::istream& is, FloatComplexRowVector& a) | |
434 { | |
435 octave_idx_type len = a.length(); | |
436 | |
437 if (len < 1) | |
438 is.clear (std::ios::badbit); | |
439 else | |
440 { | |
441 FloatComplex tmp; | |
442 for (octave_idx_type i = 0; i < len; i++) | |
443 { | |
444 is >> tmp; | |
445 if (is) | |
446 a.elem (i) = tmp; | |
447 else | |
448 break; | |
449 } | |
450 } | |
451 return is; | |
452 } | |
453 | |
454 // row vector by column vector -> scalar | |
455 | |
456 // row vector by column vector -> scalar | |
457 | |
458 FloatComplex | |
459 operator * (const FloatComplexRowVector& v, const FloatColumnVector& a) | |
460 { | |
461 FloatComplexColumnVector tmp (a); | |
462 return v * tmp; | |
463 } | |
464 | |
465 FloatComplex | |
466 operator * (const FloatComplexRowVector& v, const FloatComplexColumnVector& a) | |
467 { | |
468 FloatComplex retval (0.0, 0.0); | |
469 | |
470 octave_idx_type len = v.length (); | |
471 | |
472 octave_idx_type a_len = a.length (); | |
473 | |
474 if (len != a_len) | |
475 gripe_nonconformant ("operator *", len, a_len); | |
476 else if (len != 0) | |
477 F77_FUNC (xcdotu, XCDOTU) (len, v.data (), 1, a.data (), 1, retval); | |
478 | |
479 return retval; | |
480 } | |
481 | |
482 // other operations | |
483 | |
484 FloatComplexRowVector | |
485 linspace (const FloatComplex& x1, const FloatComplex& x2, octave_idx_type n) | |
486 { | |
487 FloatComplexRowVector retval; | |
488 | |
489 if (n > 0) | |
490 { | |
491 retval.resize (n); | |
492 FloatComplex delta = (x2 - x1) / static_cast<float> (n - 1.0); | |
493 retval.elem (0) = x1; | |
494 for (octave_idx_type i = 1; i < n-1; i++) | |
495 retval.elem (i) = x1 + static_cast<float> (1.0) * i * delta; | |
496 retval.elem (n-1) = x2; | |
497 } | |
498 else | |
499 { | |
500 retval.resize (1); | |
501 retval.elem (0) = x2; | |
502 } | |
503 | |
504 return retval; | |
505 } | |
506 | |
507 /* | |
508 ;;; Local Variables: *** | |
509 ;;; mode: C++ *** | |
510 ;;; End: *** | |
511 */ |