Visual Servoing Platform  version 3.2.0 under development (2019-01-22)
vpGEMM.h
1 /****************************************************************************
2  *
3  * ViSP, open source Visual Servoing Platform software.
4  * Copyright (C) 2005 - 2019 by Inria. All rights reserved.
5  *
6  * This software is free software; you can redistribute it and/or modify
7  * it under the terms of the GNU General Public License as published by
8  * the Free Software Foundation; either version 2 of the License, or
9  * (at your option) any later version.
10  * See the file LICENSE.txt at the root directory of this source
11  * distribution for additional information about the GNU GPL.
12  *
13  * For using ViSP with software that can not be combined with the GNU
14  * GPL, please contact Inria about acquiring a ViSP Professional
15  * Edition License.
16  *
17  * See http://visp.inria.fr for more information.
18  *
19  * This software was developed at:
20  * Inria Rennes - Bretagne Atlantique
21  * Campus Universitaire de Beaulieu
22  * 35042 Rennes Cedex
23  * France
24  *
25  * If you have questions regarding the use of this file, please contact
26  * Inria at visp@inria.fr
27  *
28  * This file is provided AS IS with NO WARRANTY OF ANY KIND, INCLUDING THE
29  * WARRANTY OF DESIGN, MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
30  *
31  * Description:
32  * Matrix generalized multiplication.
33  *
34  * Authors:
35  * Laneurit Jean
36  *
37  *****************************************************************************/
38 
39 #ifndef _vpGEMM_h_
40 #define _vpGEMM_h_
41 
42 #include <visp3/core/vpArray2D.h>
43 #include <visp3/core/vpException.h>
44 
45 const vpArray2D<double> null(0, 0);
46 
57 typedef enum {
58  VP_GEMM_A_T = 1,
59  VP_GEMM_B_T = 2,
60  VP_GEMM_C_T = 4,
61 } vpGEMMmethod;
62 
63 template <unsigned int>
64 inline void GEMMsize(const vpArray2D<double> & /*A*/, const vpArray2D<double> & /*B*/, unsigned int & /*Arows*/,
65  unsigned int & /*Acols*/, unsigned int & /*Brows*/, unsigned int & /*Bcols*/)
66 {
67 }
68 
69 template <>
70 void inline GEMMsize<0>(const vpArray2D<double> &A, const vpArray2D<double> &B, unsigned int &Arows,
71  unsigned int &Acols, unsigned int &Brows, unsigned int &Bcols)
72 {
73  Arows = A.getRows();
74  Acols = A.getCols();
75  Brows = B.getRows();
76  Bcols = B.getCols();
77 }
78 
79 template <>
80 inline void GEMMsize<1>(const vpArray2D<double> &A, const vpArray2D<double> &B, unsigned int &Arows,
81  unsigned int &Acols, unsigned int &Brows, unsigned int &Bcols)
82 {
83  Arows = A.getCols();
84  Acols = A.getRows();
85  Brows = B.getRows();
86  Bcols = B.getCols();
87 }
88 template <>
89 inline void GEMMsize<2>(const vpArray2D<double> &A, const vpArray2D<double> &B, unsigned int &Arows,
90  unsigned int &Acols, unsigned int &Brows, unsigned int &Bcols)
91 {
92  Arows = A.getRows();
93  Acols = A.getCols();
94  Brows = B.getCols();
95  Bcols = B.getRows();
96 }
97 template <>
98 inline void GEMMsize<3>(const vpArray2D<double> &A, const vpArray2D<double> &B, unsigned int &Arows,
99  unsigned int &Acols, unsigned int &Brows, unsigned int &Bcols)
100 {
101  Arows = A.getCols();
102  Acols = A.getRows();
103  Brows = B.getCols();
104  Bcols = B.getRows();
105 }
106 
107 template <>
108 inline void GEMMsize<4>(const vpArray2D<double> &A, const vpArray2D<double> &B, unsigned int &Arows,
109  unsigned int &Acols, unsigned int &Brows, unsigned int &Bcols)
110 {
111  Arows = A.getRows();
112  Acols = A.getCols();
113  Brows = B.getRows();
114  Bcols = B.getCols();
115 }
116 
117 template <>
118 inline void GEMMsize<5>(const vpArray2D<double> &A, const vpArray2D<double> &B, unsigned int &Arows,
119  unsigned int &Acols, unsigned int &Brows, unsigned int &Bcols)
120 {
121  Arows = A.getCols();
122  Acols = A.getRows();
123  Brows = B.getRows();
124  Bcols = B.getCols();
125 }
126 
127 template <>
128 inline void GEMMsize<6>(const vpArray2D<double> &A, const vpArray2D<double> &B, unsigned int &Arows,
129  unsigned int &Acols, unsigned int &Brows, unsigned int &Bcols)
130 {
131  Arows = A.getRows();
132  Acols = A.getCols();
133  Brows = B.getCols();
134  Bcols = B.getRows();
135 }
136 
137 template <>
138 inline void GEMMsize<7>(const vpArray2D<double> &A, const vpArray2D<double> &B, unsigned int &Arows,
139  unsigned int &Acols, unsigned int &Brows, unsigned int &Bcols)
140 {
141  Arows = A.getCols();
142  Acols = A.getRows();
143  Brows = B.getCols();
144  Bcols = B.getRows();
145 }
146 
147 template <unsigned int>
148 inline void GEMM1(const unsigned int & /*Arows*/, const unsigned int & /*Brows*/, const unsigned int & /*Bcols*/,
149  const vpArray2D<double> & /*A*/, const vpArray2D<double> & /*B*/, const double & /*alpha*/,
150  vpArray2D<double> & /*D*/)
151 {
152 }
153 
154 template <>
155 inline void GEMM1<0>(const unsigned int &Arows, const unsigned int &Brows, const unsigned int &Bcols,
156  const vpArray2D<double> &A, const vpArray2D<double> &B, const double &alpha, vpArray2D<double> &D)
157 {
158  for (unsigned int r = 0; r < Arows; r++)
159  for (unsigned int c = 0; c < Bcols; c++) {
160  double sum = 0;
161  for (unsigned int n = 0; n < Brows; n++)
162  sum += A[r][n] * B[n][c] * alpha;
163  D[r][c] = sum;
164  }
165 }
166 
167 template <>
168 inline void GEMM1<1>(const unsigned int &Arows, const unsigned int &Brows, const unsigned int &Bcols,
169  const vpArray2D<double> &A, const vpArray2D<double> &B, const double &alpha, vpArray2D<double> &D)
170 {
171  for (unsigned int r = 0; r < Arows; r++)
172  for (unsigned int c = 0; c < Bcols; c++) {
173  double sum = 0;
174  for (unsigned int n = 0; n < Brows; n++)
175  sum += A[n][r] * B[n][c] * alpha;
176  D[r][c] = sum;
177  }
178 }
179 
180 template <>
181 inline void GEMM1<2>(const unsigned int &Arows, const unsigned int &Brows, const unsigned int &Bcols,
182  const vpArray2D<double> &A, const vpArray2D<double> &B, const double &alpha, vpArray2D<double> &D)
183 {
184  for (unsigned int r = 0; r < Arows; r++)
185  for (unsigned int c = 0; c < Bcols; c++) {
186  double sum = 0;
187  for (unsigned int n = 0; n < Brows; n++)
188  sum += A[r][n] * B[c][n] * alpha;
189  D[r][c] = sum;
190  }
191 }
192 
193 template <>
194 inline void GEMM1<3>(const unsigned int &Arows, const unsigned int &Brows, const unsigned int &Bcols,
195  const vpArray2D<double> &A, const vpArray2D<double> &B, const double &alpha, vpArray2D<double> &D)
196 {
197  for (unsigned int r = 0; r < Arows; r++)
198  for (unsigned int c = 0; c < Bcols; c++) {
199  double sum = 0;
200  for (unsigned int n = 0; n < Brows; n++)
201  sum += A[n][r] * B[c][n] * alpha;
202  D[r][c] = sum;
203  }
204 }
205 
206 template <unsigned int>
207 inline void GEMM2(const unsigned int & /*Arows*/, const unsigned int & /*Brows*/, const unsigned int & /*Bcols*/,
208  const vpArray2D<double> & /*A*/, const vpArray2D<double> & /*B*/, const double & /*alpha*/,
209  const vpArray2D<double> & /*C*/, const double & /*beta*/, vpArray2D<double> & /*D*/)
210 {
211 }
212 
213 template <>
214 inline void GEMM2<0>(const unsigned int &Arows, const unsigned int &Brows, const unsigned int &Bcols,
215  const vpArray2D<double> &A, const vpArray2D<double> &B, const double &alpha,
216  const vpArray2D<double> &C, const double &beta, vpArray2D<double> &D)
217 {
218  for (unsigned int r = 0; r < Arows; r++)
219  for (unsigned int c = 0; c < Bcols; c++) {
220  double sum = 0;
221  for (unsigned int n = 0; n < Brows; n++)
222  sum += A[r][n] * B[n][c] * alpha;
223  D[r][c] = sum + C[r][c] * beta;
224  }
225 }
226 
227 template <>
228 inline void GEMM2<1>(const unsigned int &Arows, const unsigned int &Brows, const unsigned int &Bcols,
229  const vpArray2D<double> &A, const vpArray2D<double> &B, const double &alpha,
230  const vpArray2D<double> &C, const double &beta, vpArray2D<double> &D)
231 {
232  for (unsigned int r = 0; r < Arows; r++)
233  for (unsigned int c = 0; c < Bcols; c++) {
234  double sum = 0;
235  for (unsigned int n = 0; n < Brows; n++)
236  sum += A[n][r] * B[n][c] * alpha;
237  D[r][c] = sum + C[r][c] * beta;
238  }
239 }
240 
241 template <>
242 inline void GEMM2<2>(const unsigned int &Arows, const unsigned int &Brows, const unsigned int &Bcols,
243  const vpArray2D<double> &A, const vpArray2D<double> &B, const double &alpha,
244  const vpArray2D<double> &C, const double &beta, vpArray2D<double> &D)
245 {
246  for (unsigned int r = 0; r < Arows; r++)
247  for (unsigned int c = 0; c < Bcols; c++) {
248  double sum = 0;
249  for (unsigned int n = 0; n < Brows; n++)
250  sum += A[r][n] * B[c][n] * alpha;
251  D[r][c] = sum + C[r][c] * beta;
252  }
253 }
254 
255 template <>
256 inline void GEMM2<3>(const unsigned int &Arows, const unsigned int &Brows, const unsigned int &Bcols,
257  const vpArray2D<double> &A, const vpArray2D<double> &B, const double &alpha,
258  const vpArray2D<double> &C, const double &beta, vpArray2D<double> &D)
259 {
260  for (unsigned int r = 0; r < Arows; r++)
261  for (unsigned int c = 0; c < Bcols; c++) {
262  double sum = 0;
263  for (unsigned int n = 0; n < Brows; n++)
264  sum += A[n][r] * B[c][n] * alpha;
265  D[r][c] = sum + C[r][c] * beta;
266  }
267 }
268 
269 template <>
270 inline void GEMM2<4>(const unsigned int &Arows, const unsigned int &Brows, const unsigned int &Bcols,
271  const vpArray2D<double> &A, const vpArray2D<double> &B, const double &alpha,
272  const vpArray2D<double> &C, const double &beta, vpArray2D<double> &D)
273 {
274  for (unsigned int r = 0; r < Arows; r++)
275  for (unsigned int c = 0; c < Bcols; c++) {
276  double sum = 0;
277  for (unsigned int n = 0; n < Brows; n++)
278  sum += A[r][n] * B[n][c] * alpha;
279  D[r][c] = sum + C[c][r] * beta;
280  }
281 }
282 
283 template <>
284 inline void GEMM2<5>(const unsigned int &Arows, const unsigned int &Brows, const unsigned int &Bcols,
285  const vpArray2D<double> &A, const vpArray2D<double> &B, const double &alpha,
286  const vpArray2D<double> &C, const double &beta, vpArray2D<double> &D)
287 {
288  for (unsigned int r = 0; r < Arows; r++)
289  for (unsigned int c = 0; c < Bcols; c++) {
290  double sum = 0;
291  for (unsigned int n = 0; n < Brows; n++)
292  sum += A[n][r] * B[n][c] * alpha;
293  D[r][c] = sum + C[c][r] * beta;
294  }
295 }
296 
297 template <>
298 inline void GEMM2<6>(const unsigned int &Arows, const unsigned int &Brows, const unsigned int &Bcols,
299  const vpArray2D<double> &A, const vpArray2D<double> &B, const double &alpha,
300  const vpArray2D<double> &C, const double &beta, vpArray2D<double> &D)
301 {
302  for (unsigned int r = 0; r < Arows; r++)
303  for (unsigned int c = 0; c < Bcols; c++) {
304  double sum = 0;
305  for (unsigned int n = 0; n < Brows; n++)
306  sum += A[r][n] * B[c][n] * alpha;
307  D[r][c] = sum + C[c][r] * beta;
308  }
309 }
310 
311 template <>
312 inline void GEMM2<7>(const unsigned int &Arows, const unsigned int &Brows, const unsigned int &Bcols,
313  const vpArray2D<double> &A, const vpArray2D<double> &B, const double &alpha,
314  const vpArray2D<double> &C, const double &beta, vpArray2D<double> &D)
315 {
316  for (unsigned int r = 0; r < Arows; r++)
317  for (unsigned int c = 0; c < Bcols; c++) {
318  double sum = 0;
319  for (unsigned int n = 0; n < Brows; n++)
320  sum += A[n][r] * B[c][n] * alpha;
321  D[r][c] = sum + C[c][r] * beta;
322  }
323 }
324 
325 template <unsigned int T>
326 inline void vpTGEMM(const vpArray2D<double> &A, const vpArray2D<double> &B, const double &alpha,
327  const vpArray2D<double> &C, const double &beta, vpArray2D<double> &D)
328 {
329  unsigned int Arows;
330  unsigned int Acols;
331  unsigned int Brows;
332  unsigned int Bcols;
333 
334  GEMMsize<T>(A, B, Arows, Acols, Brows, Bcols);
335 
336  try {
337  if ((Arows != D.getRows()) || (Bcols != D.getCols()))
338  D.resize(Arows, Bcols);
339  } catch (...) {
340  throw;
341  }
342 
343  if (Acols != Brows) {
344  throw(vpException(vpException::dimensionError, "In vpGEMM, cannot multiply (%dx%d) matrix by (%dx%d) matrix", Arows,
345  Acols, Brows, Bcols));
346  }
347 
348  if (C.getRows() != 0 && C.getCols() != 0) {
349  if ((Arows != C.getRows()) || (Bcols != C.getCols())) {
350  throw(vpException(vpException::dimensionError, "In vpGEMM, cannot add resulting (%dx%d) matrix to (%dx%d) matrix",
351  Arows, Bcols, C.getRows(), C.getCols()));
352  }
353 
354  GEMM2<T>(Arows, Brows, Bcols, A, B, alpha, C, beta, D);
355  } else {
356  GEMM1<T>(Arows, Brows, Bcols, A, B, alpha, D);
357  }
358 }
359 
393 inline void vpGEMM(const vpArray2D<double> &A, const vpArray2D<double> &B, const double &alpha,
394  const vpArray2D<double> &C, const double &beta, vpArray2D<double> &D, const unsigned int &ops = 0)
395 {
396  switch (ops) {
397  case 0:
398  vpTGEMM<0>(A, B, alpha, C, beta, D);
399  break;
400  case 1:
401  vpTGEMM<1>(A, B, alpha, C, beta, D);
402  break;
403  case 2:
404  vpTGEMM<2>(A, B, alpha, C, beta, D);
405  break;
406  case 3:
407  vpTGEMM<3>(A, B, alpha, C, beta, D);
408  break;
409  case 4:
410  vpTGEMM<4>(A, B, alpha, C, beta, D);
411  break;
412  case 5:
413  vpTGEMM<5>(A, B, alpha, C, beta, D);
414  break;
415  case 6:
416  vpTGEMM<6>(A, B, alpha, C, beta, D);
417  break;
418  case 7:
419  vpTGEMM<7>(A, B, alpha, C, beta, D);
420  break;
421  default:
422  throw(vpException(vpException::functionNotImplementedError, "Operation on vpGEMM not implemented"));
423  break;
424  }
425 }
426 
427 #endif
void resize(const unsigned int nrows, const unsigned int ncols, const bool flagNullify=true, const bool recopy_=true)
Definition: vpArray2D.h:171
error that can be emited by ViSP classes.
Definition: vpException.h:71
unsigned int getCols() const
Definition: vpArray2D.h:146
unsigned int getRows() const
Definition: vpArray2D.h:156
vpGEMMmethod
Definition: vpGEMM.h:57
void vpGEMM(const vpArray2D< double > &A, const vpArray2D< double > &B, const double &alpha, const vpArray2D< double > &C, const double &beta, vpArray2D< double > &D, const unsigned int &ops=0)
Definition: vpGEMM.h:393
Function not implemented.
Definition: vpException.h:90