| 1 | /* | 
| 2 | * Copyright (c) 2005 The University of Notre Dame. All Rights Reserved. | 
| 3 | * | 
| 4 | * The University of Notre Dame grants you ("Licensee") a | 
| 5 | * non-exclusive, royalty free, license to use, modify and | 
| 6 | * redistribute this software in source and binary code form, provided | 
| 7 | * that the following conditions are met: | 
| 8 | * | 
| 9 | * 1. Acknowledgement of the program authors must be made in any | 
| 10 | *    publication of scientific results based in part on use of the | 
| 11 | *    program.  An acceptable form of acknowledgement is citation of | 
| 12 | *    the article in which the program was described (Matthew | 
| 13 | *    A. Meineke, Charles F. Vardeman II, Teng Lin, Christopher | 
| 14 | *    J. Fennell and J. Daniel Gezelter, "OOPSE: An Object-Oriented | 
| 15 | *    Parallel Simulation Engine for Molecular Dynamics," | 
| 16 | *    J. Comput. Chem. 26, pp. 252-271 (2005)) | 
| 17 | * | 
| 18 | * 2. Redistributions of source code must retain the above copyright | 
| 19 | *    notice, this list of conditions and the following disclaimer. | 
| 20 | * | 
| 21 | * 3. Redistributions in binary form must reproduce the above copyright | 
| 22 | *    notice, this list of conditions and the following disclaimer in the | 
| 23 | *    documentation and/or other materials provided with the | 
| 24 | *    distribution. | 
| 25 | * | 
| 26 | * This software is provided "AS IS," without a warranty of any | 
| 27 | * kind. All express or implied conditions, representations and | 
| 28 | * warranties, including any implied warranty of merchantability, | 
| 29 | * fitness for a particular purpose or non-infringement, are hereby | 
| 30 | * excluded.  The University of Notre Dame and its licensors shall not | 
| 31 | * be liable for any damages suffered by licensee as a result of | 
| 32 | * using, modifying or distributing the software or its | 
| 33 | * derivatives. In no event will the University of Notre Dame or its | 
| 34 | * licensors be liable for any lost revenue, profit or data, or for | 
| 35 | * direct, indirect, special, consequential, incidental or punitive | 
| 36 | * damages, however caused and regardless of the theory of liability, | 
| 37 | * arising out of the use of or inability to use software, even if the | 
| 38 | * University of Notre Dame has been advised of the possibility of | 
| 39 | * such damages. | 
| 40 | */ | 
| 41 |  | 
| 42 | /** | 
| 43 | * @file DynamicRectMatrix.hpp | 
| 44 | * @author Teng Lin | 
| 45 | * @date 10/11/2004 | 
| 46 | * @version 1.0 | 
| 47 | */ | 
| 48 |  | 
| 49 | #ifndef MATH_DYNAMICRECTMATRIX_HPP | 
| 50 | #define MATH_DYNAMICRECTMATRIX_HPP | 
| 51 | #include <math.h> | 
| 52 | #include <cmath> | 
| 53 | #include "math/DynamicVector.hpp" | 
| 54 |  | 
| 55 | namespace oopse { | 
| 56 |  | 
| 57 | /** | 
| 58 | * @class DynamicRectMatrix DynamicRectMatrix.hpp "math/DynamicRectMatrix.hpp" | 
| 59 | * @brief rectangular matrix class | 
| 60 | */ | 
| 61 | template<typename Real> | 
| 62 | class DynamicRectMatrix { | 
| 63 | public: | 
| 64 | typedef Real ElemType; | 
| 65 | typedef Real* ElemPoinerType; | 
| 66 | typedef DynamicRectMatrix<Real> SelfType; | 
| 67 |  | 
| 68 | /** default constructor */ | 
| 69 | DynamicRectMatrix(){ | 
| 70 | nrow_ = 0; | 
| 71 | ncol_ = 0; | 
| 72 | data_ = NULL; | 
| 73 | } | 
| 74 |  | 
| 75 | DynamicRectMatrix(int nrow, int ncol) { | 
| 76 | allocate(nrow, ncol); | 
| 77 |  | 
| 78 | for (unsigned int i = 0; i < nrow_; i++) | 
| 79 | for (unsigned int j = 0; j < ncol_; j++) | 
| 80 | this->data_[i][j] = 0.0; | 
| 81 | } | 
| 82 |  | 
| 83 | /** Constructs and initializes every element of this matrix to a scalar */ | 
| 84 | DynamicRectMatrix(int nrow, int ncol, Real s) { | 
| 85 | allocate(nrow, ncol); | 
| 86 | for (unsigned int i = 0; i < nrow_; i++) | 
| 87 | for (unsigned int j = 0; j < ncol_; j++) | 
| 88 | this->data_[i][j] = s; | 
| 89 | } | 
| 90 |  | 
| 91 | DynamicRectMatrix(int nrow, int ncol, Real* array) { | 
| 92 | allocate(nrow, ncol); | 
| 93 | for (unsigned int i = 0; i < nrow_; i++) | 
| 94 | for (unsigned int j = 0; j < ncol_; j++) | 
| 95 | this->data_[i][j] = array[i * nrow_ + j]; | 
| 96 | } | 
| 97 |  | 
| 98 | /** copy constructor */ | 
| 99 | DynamicRectMatrix(const SelfType& m) { | 
| 100 | allocate(m.getNRow(), m.getNCol()); | 
| 101 |  | 
| 102 | for (unsigned int i = 0; i < nrow_; i++) | 
| 103 | for (unsigned int j = 0; j < ncol_; j++) | 
| 104 | this->data_[i][j] = m.data_[i][j]; | 
| 105 | } | 
| 106 |  | 
| 107 | /** destructor*/ | 
| 108 | ~DynamicRectMatrix() { deallocate();} | 
| 109 |  | 
| 110 | /** copy assignment operator */ | 
| 111 | DynamicRectMatrix<Real> operator =(const DynamicRectMatrix<Real> m) { | 
| 112 | if (this == &m) | 
| 113 | return *this; | 
| 114 | if (nrow_ != m.getNRow() || ncol_ != m.getNCol()) { | 
| 115 | deallocate(); | 
| 116 | allocate(m.getNRow(), m.getNCol()); | 
| 117 | } | 
| 118 |  | 
| 119 | for (unsigned int i = 0; i < nrow_; i++) | 
| 120 | for (unsigned int j = 0; j < ncol_; j++) | 
| 121 | this->data_[i][j] = m.data_[i][j]; | 
| 122 | return *this; | 
| 123 | } | 
| 124 |  | 
| 125 | /** | 
| 126 | * Return the reference of a single element of this matrix. | 
| 127 | * @return the reference of a single element of this matrix | 
| 128 | * @param i row index | 
| 129 | * @param j Column index | 
| 130 | */ | 
| 131 | Real& operator()(unsigned int i, unsigned int j) { | 
| 132 | return this->data_[i][j]; | 
| 133 | } | 
| 134 |  | 
| 135 | /** | 
| 136 | * Return the value of a single element of this matrix. | 
| 137 | * @return the value of a single element of this matrix | 
| 138 | * @param i row index | 
| 139 | * @param j Column index | 
| 140 | */ | 
| 141 | Real operator()(unsigned int i, unsigned int j) const  { | 
| 142 |  | 
| 143 | return this->data_[i][j]; | 
| 144 | } | 
| 145 |  | 
| 146 | /** | 
| 147 | * Copy the internal data to an array | 
| 148 | * @param array the pointer of destination array | 
| 149 | */ | 
| 150 | void getArray(Real* array) { | 
| 151 | for (unsigned int i = 0; i < nrow_; i++) { | 
| 152 | for (unsigned int j = 0; j < ncol_; j++) { | 
| 153 | array[i * nrow_ + j] = this->data_[i][j]; | 
| 154 | } | 
| 155 | } | 
| 156 | } | 
| 157 |  | 
| 158 | /** | 
| 159 | * Returns a row of  this matrix as a vector. | 
| 160 | * @return a row of  this matrix as a vector | 
| 161 | * @param row the row index | 
| 162 | */ | 
| 163 | DynamicVector<Real> getRow(unsigned int row) { | 
| 164 | DynamicVector<Real> v; | 
| 165 |  | 
| 166 | for (unsigned int i = 0; i < ncol_; i++) | 
| 167 | v[i] = this->data_[row][i]; | 
| 168 |  | 
| 169 | return v; | 
| 170 | } | 
| 171 |  | 
| 172 | /** | 
| 173 | * Sets a row of  this matrix | 
| 174 | * @param row the row index | 
| 175 | * @param v the vector to be set | 
| 176 | */ | 
| 177 | void setRow(unsigned int row, const DynamicVector<Real>& v) { | 
| 178 | assert(v.size() == nrow_); | 
| 179 | for (unsigned int i = 0; i < ncol_; i++) | 
| 180 | this->data_[row][i] = v[i]; | 
| 181 | } | 
| 182 |  | 
| 183 | /** | 
| 184 | * Returns a column of  this matrix as a vector. | 
| 185 | * @return a column of  this matrix as a vector | 
| 186 | * @param col the column index | 
| 187 | */ | 
| 188 | DynamicVector<Real> getColumn(unsigned int col) { | 
| 189 | DynamicVector<Real> v(ncol_); | 
| 190 |  | 
| 191 | for (unsigned int j = 0; j < nrow_; j++) | 
| 192 | v[j] = this->data_[j][col]; | 
| 193 |  | 
| 194 | return v; | 
| 195 | } | 
| 196 |  | 
| 197 | /** | 
| 198 | * Sets a column of  this matrix | 
| 199 | * @param col the column index | 
| 200 | * @param v the vector to be set | 
| 201 | */ | 
| 202 | void setColumn(unsigned int col, const DynamicVector<Real>& v){ | 
| 203 |  | 
| 204 | for (unsigned int j = 0; j < nrow_; j++) | 
| 205 | this->data_[j][col] = v[j]; | 
| 206 | } | 
| 207 |  | 
| 208 | /** | 
| 209 | * swap two rows of this matrix | 
| 210 | * @param i the first row | 
| 211 | * @param j the second row | 
| 212 | */ | 
| 213 | void swapRow(unsigned int i, unsigned int j){ | 
| 214 | assert(i < nrow_ && j < nrow_); | 
| 215 |  | 
| 216 | for (unsigned int k = 0; k < ncol_; k++) | 
| 217 | std::swap(this->data_[i][k], this->data_[j][k]); | 
| 218 | } | 
| 219 |  | 
| 220 | /** | 
| 221 | * swap two Columns of this matrix | 
| 222 | * @param i the first Column | 
| 223 | * @param j the second Column | 
| 224 | */ | 
| 225 | void swapColumn(unsigned int i, unsigned int j){ | 
| 226 | assert(i < ncol_ && j < ncol_); | 
| 227 |  | 
| 228 | for (unsigned int k = 0; k < nrow_; k++) | 
| 229 | std::swap(this->data_[k][i], this->data_[k][j]); | 
| 230 | } | 
| 231 |  | 
| 232 | /** | 
| 233 | * Tests if this matrix is identical to matrix m | 
| 234 | * @return true if this matrix is equal to the matrix m, return false otherwise | 
| 235 | * @m matrix to be compared | 
| 236 | * | 
| 237 | * @todo replace operator == by template function equal | 
| 238 | */ | 
| 239 | bool operator ==(const DynamicRectMatrix<Real> m) { | 
| 240 | assert(nrow_ == m.getNRow() && ncol_ == m.getNCol()); | 
| 241 | for (unsigned int i = 0; i < nrow_; i++) | 
| 242 | for (unsigned int j = 0; j < ncol_; j++) | 
| 243 | if (!equal(this->data_[i][j], m.data_[i][j])) | 
| 244 | return false; | 
| 245 |  | 
| 246 | return true; | 
| 247 | } | 
| 248 |  | 
| 249 | /** | 
| 250 | * Tests if this matrix is not equal to matrix m | 
| 251 | * @return true if this matrix is not equal to the matrix m, return false otherwise | 
| 252 | * @m matrix to be compared | 
| 253 | */ | 
| 254 | bool operator !=(const DynamicRectMatrix<Real> m) { | 
| 255 | return !(*this == m); | 
| 256 | } | 
| 257 |  | 
| 258 | /** Negates the value of this matrix in place. */ | 
| 259 | inline void negate() { | 
| 260 | for (unsigned int i = 0; i < nrow_; i++) | 
| 261 | for (unsigned int j = 0; j < ncol_; j++) | 
| 262 | this->data_[i][j] = -this->data_[i][j]; | 
| 263 | } | 
| 264 |  | 
| 265 | /** | 
| 266 | * Sets the value of this matrix to the negation of matrix m. | 
| 267 | * @param m the source matrix | 
| 268 | */ | 
| 269 | inline void negate(const DynamicRectMatrix<Real> m) { | 
| 270 | for (unsigned int i = 0; i < nrow_; i++) | 
| 271 | for (unsigned int j = 0; j < ncol_; j++) | 
| 272 | this->data_[i][j] = -m.data_[i][j]; | 
| 273 | } | 
| 274 |  | 
| 275 | /** | 
| 276 | * Sets the value of this matrix to the sum of itself and m (*this += m). | 
| 277 | * @param m the other matrix | 
| 278 | */ | 
| 279 | inline void add( const DynamicRectMatrix<Real> m ) { | 
| 280 | assert(nrow_ == m.getNRow() && ncol_ == m.getNCol()); | 
| 281 | for (unsigned int i = 0; i < nrow_; i++) | 
| 282 | for (unsigned int j = 0; j < ncol_; j++) | 
| 283 | this->data_[i][j] += m.data_[i][j]; | 
| 284 | } | 
| 285 |  | 
| 286 | /** | 
| 287 | * Sets the value of this matrix to the sum of m1 and m2 (*this = m1 + m2). | 
| 288 | * @param m1 the first matrix | 
| 289 | * @param m2 the second matrix | 
| 290 | */ | 
| 291 | inline void add( const DynamicRectMatrix<Real> m1, const DynamicRectMatrix<Real> m2 ) { | 
| 292 | assert(m1.getNRow() == m2.getNRow() && m1.getNCol() == m2.getNCol()); | 
| 293 | for (unsigned int i = 0; i < nrow_; i++) | 
| 294 | for (unsigned int j = 0; j < ncol_; j++) | 
| 295 | this->data_[i][j] = m1.data_[i][j] + m2.data_[i][j]; | 
| 296 | } | 
| 297 |  | 
| 298 | /** | 
| 299 | * Sets the value of this matrix to the difference  of itself and m (*this -= m). | 
| 300 | * @param m the other matrix | 
| 301 | */ | 
| 302 | inline void sub( const DynamicRectMatrix<Real> m ) { | 
| 303 | assert(nrow_ == m.getNRow() && ncol_ == m.getNCol()); | 
| 304 | for (unsigned int i = 0; i < nrow_; i++) | 
| 305 | for (unsigned int j = 0; j < ncol_; j++) | 
| 306 | this->data_[i][j] -= m.data_[i][j]; | 
| 307 | } | 
| 308 |  | 
| 309 | /** | 
| 310 | * Sets the value of this matrix to the difference of matrix m1 and m2 (*this = m1 - m2). | 
| 311 | * @param m1 the first matrix | 
| 312 | * @param m2 the second matrix | 
| 313 | */ | 
| 314 | inline void sub( const DynamicRectMatrix<Real> m1, const DynamicRectMatrix<Real> m2){ | 
| 315 | assert(m1.getNRow() == m2.getNRow() && m1.getNCol() == m2.getNCol()); | 
| 316 | for (unsigned int i = 0; i < nrow_; i++) | 
| 317 | for (unsigned int j = 0; j < ncol_; j++) | 
| 318 | this->data_[i][j] = m1.data_[i][j] - m2.data_[i][j]; | 
| 319 | } | 
| 320 |  | 
| 321 | /** | 
| 322 | * Sets the value of this matrix to the scalar multiplication of itself (*this *= s). | 
| 323 | * @param s the scalar value | 
| 324 | */ | 
| 325 | inline void mul( Real s ) { | 
| 326 | for (unsigned int i = 0; i < nrow_; i++) | 
| 327 | for (unsigned int j = 0; j < ncol_; j++) | 
| 328 | this->data_[i][j] *= s; | 
| 329 | } | 
| 330 |  | 
| 331 | /** | 
| 332 | * Sets the value of this matrix to the scalar multiplication of matrix m  (*this = s * m). | 
| 333 | * @param s the scalar value | 
| 334 | * @param m the matrix | 
| 335 | */ | 
| 336 | inline void mul( Real s, const DynamicRectMatrix<Real> m ) { | 
| 337 | assert(nrow_ == m.getNRow() && ncol_ == m.getNCol()); | 
| 338 | for (unsigned int i = 0; i < nrow_; i++) | 
| 339 | for (unsigned int j = 0; j < ncol_; j++) | 
| 340 | this->data_[i][j] = s * m.data_[i][j]; | 
| 341 | } | 
| 342 |  | 
| 343 | /** | 
| 344 | * Sets the value of this matrix to the scalar division of itself  (*this /= s ). | 
| 345 | * @param s the scalar value | 
| 346 | */ | 
| 347 | inline void div( Real s) { | 
| 348 | for (unsigned int i = 0; i < nrow_; i++) | 
| 349 | for (unsigned int j = 0; j < ncol_; j++) | 
| 350 | this->data_[i][j] /= s; | 
| 351 | } | 
| 352 |  | 
| 353 | /** | 
| 354 | * Sets the value of this matrix to the scalar division of matrix m  (*this = m /s). | 
| 355 | * @param s the scalar value | 
| 356 | * @param m the matrix | 
| 357 | */ | 
| 358 | inline void div( Real s, const DynamicRectMatrix<Real> m ) { | 
| 359 | assert(nrow_ == m.getNRow() && ncol_ == m.getNCol()); | 
| 360 | for (unsigned int i = 0; i < nrow_; i++) | 
| 361 | for (unsigned int j = 0; j < ncol_; j++) | 
| 362 | this->data_[i][j] = m.data_[i][j] / s; | 
| 363 | } | 
| 364 |  | 
| 365 | /** | 
| 366 | *  Multiples a scalar into every element of this matrix. | 
| 367 | * @param s the scalar value | 
| 368 | */ | 
| 369 | DynamicRectMatrix<Real> operator *=(const Real s) { | 
| 370 | this->mul(s); | 
| 371 | return *this; | 
| 372 | } | 
| 373 |  | 
| 374 | /** | 
| 375 | *  Divides every element of this matrix by a scalar. | 
| 376 | * @param s the scalar value | 
| 377 | */ | 
| 378 | DynamicRectMatrix<Real> operator /=(const Real s) { | 
| 379 | this->div(s); | 
| 380 | return *this; | 
| 381 | } | 
| 382 |  | 
| 383 | /** | 
| 384 | * Sets the value of this matrix to the sum of the other matrix and itself (*this += m). | 
| 385 | * @param m the other matrix | 
| 386 | */ | 
| 387 | DynamicRectMatrix<Real> operator += (const DynamicRectMatrix<Real> m) { | 
| 388 | assert(nrow_ == m.getNRow() && ncol_ == m.getNCol()); | 
| 389 | add(m); | 
| 390 | return *this; | 
| 391 | } | 
| 392 |  | 
| 393 | /** | 
| 394 | * Sets the value of this matrix to the differerence of itself and the other matrix (*this -= m) | 
| 395 | * @param m the other matrix | 
| 396 | */ | 
| 397 | DynamicRectMatrix<Real> operator -= (const DynamicRectMatrix<Real> m){ | 
| 398 | assert(nrow_ == m.getNRow() && ncol_ == m.getNCol()); | 
| 399 | sub(m); | 
| 400 | return *this; | 
| 401 | } | 
| 402 |  | 
| 403 | /** Return the transpose of this matrix */ | 
| 404 | DynamicRectMatrix<Real> transpose() const{ | 
| 405 | DynamicRectMatrix<Real> result(ncol_,nrow_); | 
| 406 |  | 
| 407 | for (unsigned int i = 0; i < nrow_; i++) | 
| 408 | for (unsigned int j = 0; j < ncol_; j++) | 
| 409 | result(j, i) = this->data_[i][j]; | 
| 410 |  | 
| 411 | return result; | 
| 412 | } | 
| 413 |  | 
| 414 | unsigned int getNRow() const {return nrow_;} | 
| 415 | unsigned int getNCol() const {return ncol_;} | 
| 416 |  | 
| 417 | template<class MatrixType> | 
| 418 | void setSubMatrix(unsigned int beginRow, unsigned int beginCol, const MatrixType& m) { | 
| 419 | assert(beginRow + m.getNRow() -1 <= nrow_); | 
| 420 | assert(beginCol + m.getNCol() -1 <= ncol_); | 
| 421 |  | 
| 422 | for (unsigned int i = 0; i < m.getNRow(); ++i) | 
| 423 | for (unsigned int j = 0; j < m.getNCol(); ++j) | 
| 424 | this->data_[beginRow+i][beginCol+j] = m(i, j); | 
| 425 | } | 
| 426 |  | 
| 427 | template<class MatrixType> | 
| 428 | void getSubMatrix(unsigned int beginRow, unsigned int beginCol, MatrixType& m) { | 
| 429 | assert(beginRow + m.getNRow() -1 <= nrow_); | 
| 430 | assert(beginCol + m.getNCol() - 1 <= ncol_); | 
| 431 |  | 
| 432 | for (unsigned int i = 0; i < m.getNRow(); ++i) | 
| 433 | for (unsigned int j = 0; j < m.getNCol(); ++j) | 
| 434 | m(i, j) = this->data_[beginRow+i][beginCol+j]; | 
| 435 | } | 
| 436 |  | 
| 437 | protected: | 
| 438 | Real** data_; | 
| 439 | unsigned int nrow_; | 
| 440 | unsigned int ncol_; | 
| 441 | private: | 
| 442 | void allocate(int nrow, int ncol) { | 
| 443 | nrow_ = nrow; | 
| 444 | ncol_ = ncol; | 
| 445 | data_ = new Real*[nrow_]; | 
| 446 | for (int i = 0; i < nrow_; ++i) | 
| 447 | data_[i] = new Real[ncol_]; | 
| 448 | } | 
| 449 |  | 
| 450 | void deallocate() { | 
| 451 | for (int i = 0; i < nrow_; ++i) | 
| 452 | delete data_[i]; | 
| 453 | delete []data_; | 
| 454 |  | 
| 455 | nrow_ = 0; | 
| 456 | ncol_ = 0; | 
| 457 | data_ = NULL; | 
| 458 | } | 
| 459 |  | 
| 460 | }; | 
| 461 |  | 
| 462 | /** Negate the value of every element of this matrix. */ | 
| 463 | template<typename Real> | 
| 464 | inline DynamicRectMatrix<Real> operator -(const DynamicRectMatrix<Real> m) { | 
| 465 | DynamicRectMatrix<Real> result(m); | 
| 466 |  | 
| 467 | result.negate(); | 
| 468 |  | 
| 469 | return result; | 
| 470 | } | 
| 471 |  | 
| 472 | /** | 
| 473 | * Return the sum of two matrixes  (m1 + m2). | 
| 474 | * @return the sum of two matrixes | 
| 475 | * @param m1 the first matrix | 
| 476 | * @param m2 the second matrix | 
| 477 | */ | 
| 478 | template<typename Real> | 
| 479 | inline DynamicRectMatrix<Real> operator + (const DynamicRectMatrix<Real> m1,const DynamicRectMatrix<Real> m2) { | 
| 480 |  | 
| 481 | DynamicRectMatrix<Real> result(m1.getNRow(), m1.getNCol()); | 
| 482 |  | 
| 483 | result.add(m1, m2); | 
| 484 |  | 
| 485 | return result; | 
| 486 | } | 
| 487 |  | 
| 488 | /** | 
| 489 | * Return the difference of two matrixes  (m1 - m2). | 
| 490 | * @return the sum of two matrixes | 
| 491 | * @param m1 the first matrix | 
| 492 | * @param m2 the second matrix | 
| 493 | */ | 
| 494 | template<typename Real> | 
| 495 | inline DynamicRectMatrix<Real> operator - (const DynamicRectMatrix<Real> m1, const DynamicRectMatrix<Real> m2) { | 
| 496 | DynamicRectMatrix<Real> result(m1.getNRow(), m1.getNCol()); | 
| 497 |  | 
| 498 | result.sub(m1, m2); | 
| 499 |  | 
| 500 | return result; | 
| 501 | } | 
| 502 |  | 
| 503 | /** | 
| 504 | * Return the multiplication of scalra and  matrix  (m * s). | 
| 505 | * @return the multiplication of a scalra and  a matrix | 
| 506 | * @param m the matrix | 
| 507 | * @param s the scalar | 
| 508 | */ | 
| 509 | template<typename Real> | 
| 510 | inline DynamicRectMatrix<Real> operator *(const DynamicRectMatrix<Real> m, Real s) { | 
| 511 | DynamicRectMatrix<Real> result(m.getNRow(), m.getNCol()); | 
| 512 |  | 
| 513 | result.mul(s, m); | 
| 514 |  | 
| 515 | return result; | 
| 516 | } | 
| 517 |  | 
| 518 | /** | 
| 519 | * Return the multiplication of a scalra and  a matrix  (s * m). | 
| 520 | * @return the multiplication of a scalra and  a matrix | 
| 521 | * @param s the scalar | 
| 522 | * @param m the matrix | 
| 523 | */ | 
| 524 | template<typename Real> | 
| 525 | inline DynamicRectMatrix<Real> operator *(Real s, const DynamicRectMatrix<Real> m) { | 
| 526 | DynamicRectMatrix<Real> result(m.getNRow(), m.getNCol()); | 
| 527 |  | 
| 528 | result.mul(s, m); | 
| 529 |  | 
| 530 | return result; | 
| 531 | } | 
| 532 |  | 
| 533 | /** | 
| 534 | * Return the multiplication of two matrixes  (m1 * m2). | 
| 535 | * @return the multiplication of two matrixes | 
| 536 | * @param m1 the first matrix | 
| 537 | * @param m2 the second matrix | 
| 538 | */ | 
| 539 | template<typename Real> | 
| 540 | inline DynamicRectMatrix<Real> operator *(const DynamicRectMatrix<Real>& m1, const DynamicRectMatrix<Real>& m2) { | 
| 541 | assert(m1.getNCol() == m2.getNRow()); | 
| 542 | unsigned int sameDim = m1.getNCol(); | 
| 543 | int nrow = m1.getNRow(); | 
| 544 | int ncol = m2.getNCol(); | 
| 545 | DynamicRectMatrix<Real> result(nrow, ncol ); | 
| 546 | for (unsigned int i = 0; i < nrow; i++) | 
| 547 | for (unsigned int j = 0; j < ncol; j++) | 
| 548 | for (unsigned int k = 0; k < sameDim; k++) | 
| 549 | result(i, j)  += m1(i, k) * m2(k, j); | 
| 550 |  | 
| 551 | return result; | 
| 552 | } | 
| 553 |  | 
| 554 | /** | 
| 555 | * Return the multiplication of  a matrix and a vector  (m * v). | 
| 556 | * @return the multiplication of a matrix and a vector | 
| 557 | * @param m the matrix | 
| 558 | * @param v the vector | 
| 559 | */ | 
| 560 | template<typename Real> | 
| 561 | inline DynamicVector<Real> operator *(const DynamicRectMatrix<Real> m, const DynamicVector<Real>& v) { | 
| 562 | int nrow = m.getNRow(); | 
| 563 | int ncol = m.getNCol(); | 
| 564 | assert(ncol = v.size()); | 
| 565 | DynamicVector<Real> result(nrow); | 
| 566 |  | 
| 567 | for (unsigned int i = 0; i < nrow ; i++) | 
| 568 | for (unsigned int j = 0; j < ncol ; j++) | 
| 569 | result[i] += m(i, j) * v[j]; | 
| 570 |  | 
| 571 | return result; | 
| 572 | } | 
| 573 |  | 
| 574 | /** | 
| 575 | * Return the scalar division of matrix   (m / s). | 
| 576 | * @return the scalar division of matrix | 
| 577 | * @param m the matrix | 
| 578 | * @param s the scalar | 
| 579 | */ | 
| 580 | template<typename Real> | 
| 581 | inline DynamicRectMatrix<Real> operator /(const DynamicRectMatrix<Real> m, Real s) { | 
| 582 | DynamicRectMatrix<Real> result(m.getNRow(), m.getNCol()); | 
| 583 |  | 
| 584 | result.div(s, m); | 
| 585 |  | 
| 586 | return result; | 
| 587 | } | 
| 588 |  | 
| 589 | /** | 
| 590 | * Write to an output stream | 
| 591 | */ | 
| 592 | template<typename Real> | 
| 593 | std::ostream &operator<< ( std::ostream& o, const DynamicRectMatrix<Real> m) { | 
| 594 | for (unsigned int i = 0; i < m.getNRow() ; i++) { | 
| 595 | o << "("; | 
| 596 | for (unsigned int j = 0; j < m.getNCol() ; j++) { | 
| 597 | o << m(i, j); | 
| 598 | if (j != m.getNCol() -1) | 
| 599 | o << "\t"; | 
| 600 | } | 
| 601 | o << ")" << std::endl; | 
| 602 | } | 
| 603 | return o; | 
| 604 | } | 
| 605 | } | 
| 606 | #endif //MATH_RECTMATRIX_HPP | 
| 607 |  |