xref: /aosp_15_r20/external/eigen/Eigen/src/Core/Dot.h (revision bf2c37156dfe67e5dfebd6d394bad8b2ab5804d4)
1 // This file is part of Eigen, a lightweight C++ template library
2 // for linear algebra.
3 //
4 // Copyright (C) 2006-2008, 2010 Benoit Jacob <[email protected]>
5 //
6 // This Source Code Form is subject to the terms of the Mozilla
7 // Public License v. 2.0. If a copy of the MPL was not distributed
8 // with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
9 
10 #ifndef EIGEN_DOT_H
11 #define EIGEN_DOT_H
12 
13 namespace Eigen {
14 
15 namespace internal {
16 
17 // helper function for dot(). The problem is that if we put that in the body of dot(), then upon calling dot
18 // with mismatched types, the compiler emits errors about failing to instantiate cwiseProduct BEFORE
19 // looking at the static assertions. Thus this is a trick to get better compile errors.
20 template<typename T, typename U,
21 // the NeedToTranspose condition here is taken straight from Assign.h
22          bool NeedToTranspose = T::IsVectorAtCompileTime
23                 && U::IsVectorAtCompileTime
24                 && ((int(T::RowsAtCompileTime) == 1 && int(U::ColsAtCompileTime) == 1)
25                       |  // FIXME | instead of || to please GCC 4.4.0 stupid warning "suggest parentheses around &&".
26                          // revert to || as soon as not needed anymore.
27                     (int(T::ColsAtCompileTime) == 1 && int(U::RowsAtCompileTime) == 1))
28 >
29 struct dot_nocheck
30 {
31   typedef scalar_conj_product_op<typename traits<T>::Scalar,typename traits<U>::Scalar> conj_prod;
32   typedef typename conj_prod::result_type ResScalar;
33   EIGEN_DEVICE_FUNC
34   EIGEN_STRONG_INLINE
rundot_nocheck35   static ResScalar run(const MatrixBase<T>& a, const MatrixBase<U>& b)
36   {
37     return a.template binaryExpr<conj_prod>(b).sum();
38   }
39 };
40 
41 template<typename T, typename U>
42 struct dot_nocheck<T, U, true>
43 {
44   typedef scalar_conj_product_op<typename traits<T>::Scalar,typename traits<U>::Scalar> conj_prod;
45   typedef typename conj_prod::result_type ResScalar;
46   EIGEN_DEVICE_FUNC
47   EIGEN_STRONG_INLINE
48   static ResScalar run(const MatrixBase<T>& a, const MatrixBase<U>& b)
49   {
50     return a.transpose().template binaryExpr<conj_prod>(b).sum();
51   }
52 };
53 
54 } // end namespace internal
55 
56 /** \fn MatrixBase::dot
57   * \returns the dot product of *this with other.
58   *
59   * \only_for_vectors
60   *
61   * \note If the scalar type is complex numbers, then this function returns the hermitian
62   * (sesquilinear) dot product, conjugate-linear in the first variable and linear in the
63   * second variable.
64   *
65   * \sa squaredNorm(), norm()
66   */
67 template<typename Derived>
68 template<typename OtherDerived>
69 EIGEN_DEVICE_FUNC
70 EIGEN_STRONG_INLINE
71 typename ScalarBinaryOpTraits<typename internal::traits<Derived>::Scalar,typename internal::traits<OtherDerived>::Scalar>::ReturnType
72 MatrixBase<Derived>::dot(const MatrixBase<OtherDerived>& other) const
73 {
74   EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
75   EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived)
76   EIGEN_STATIC_ASSERT_SAME_VECTOR_SIZE(Derived,OtherDerived)
77 #if !(defined(EIGEN_NO_STATIC_ASSERT) && defined(EIGEN_NO_DEBUG))
78   typedef internal::scalar_conj_product_op<Scalar,typename OtherDerived::Scalar> func;
79   EIGEN_CHECK_BINARY_COMPATIBILIY(func,Scalar,typename OtherDerived::Scalar);
80 #endif
81 
82   eigen_assert(size() == other.size());
83 
84   return internal::dot_nocheck<Derived,OtherDerived>::run(*this, other);
85 }
86 
87 //---------- implementation of L2 norm and related functions ----------
88 
89 /** \returns, for vectors, the squared \em l2 norm of \c *this, and for matrices the squared Frobenius norm.
90   * In both cases, it consists in the sum of the square of all the matrix entries.
91   * For vectors, this is also equals to the dot product of \c *this with itself.
92   *
93   * \sa dot(), norm(), lpNorm()
94   */
95 template<typename Derived>
96 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename NumTraits<typename internal::traits<Derived>::Scalar>::Real MatrixBase<Derived>::squaredNorm() const
97 {
98   return numext::real((*this).cwiseAbs2().sum());
99 }
100 
101 /** \returns, for vectors, the \em l2 norm of \c *this, and for matrices the Frobenius norm.
102   * In both cases, it consists in the square root of the sum of the square of all the matrix entries.
103   * For vectors, this is also equals to the square root of the dot product of \c *this with itself.
104   *
105   * \sa lpNorm(), dot(), squaredNorm()
106   */
107 template<typename Derived>
108 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename NumTraits<typename internal::traits<Derived>::Scalar>::Real MatrixBase<Derived>::norm() const
109 {
110   return numext::sqrt(squaredNorm());
111 }
112 
113 /** \returns an expression of the quotient of \c *this by its own norm.
114   *
115   * \warning If the input vector is too small (i.e., this->norm()==0),
116   *          then this function returns a copy of the input.
117   *
118   * \only_for_vectors
119   *
120   * \sa norm(), normalize()
121   */
122 template<typename Derived>
123 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::PlainObject
124 MatrixBase<Derived>::normalized() const
125 {
126   typedef typename internal::nested_eval<Derived,2>::type _Nested;
127   _Nested n(derived());
128   RealScalar z = n.squaredNorm();
129   // NOTE: after extensive benchmarking, this conditional does not impact performance, at least on recent x86 CPU
130   if(z>RealScalar(0))
131     return n / numext::sqrt(z);
132   else
133     return n;
134 }
135 
136 /** Normalizes the vector, i.e. divides it by its own norm.
137   *
138   * \only_for_vectors
139   *
140   * \warning If the input vector is too small (i.e., this->norm()==0), then \c *this is left unchanged.
141   *
142   * \sa norm(), normalized()
143   */
144 template<typename Derived>
145 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void MatrixBase<Derived>::normalize()
146 {
147   RealScalar z = squaredNorm();
148   // NOTE: after extensive benchmarking, this conditional does not impact performance, at least on recent x86 CPU
149   if(z>RealScalar(0))
150     derived() /= numext::sqrt(z);
151 }
152 
153 /** \returns an expression of the quotient of \c *this by its own norm while avoiding underflow and overflow.
154   *
155   * \only_for_vectors
156   *
157   * This method is analogue to the normalized() method, but it reduces the risk of
158   * underflow and overflow when computing the norm.
159   *
160   * \warning If the input vector is too small (i.e., this->norm()==0),
161   *          then this function returns a copy of the input.
162   *
163   * \sa stableNorm(), stableNormalize(), normalized()
164   */
165 template<typename Derived>
166 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::PlainObject
167 MatrixBase<Derived>::stableNormalized() const
168 {
169   typedef typename internal::nested_eval<Derived,3>::type _Nested;
170   _Nested n(derived());
171   RealScalar w = n.cwiseAbs().maxCoeff();
172   RealScalar z = (n/w).squaredNorm();
173   if(z>RealScalar(0))
174     return n / (numext::sqrt(z)*w);
175   else
176     return n;
177 }
178 
179 /** Normalizes the vector while avoid underflow and overflow
180   *
181   * \only_for_vectors
182   *
183   * This method is analogue to the normalize() method, but it reduces the risk of
184   * underflow and overflow when computing the norm.
185   *
186   * \warning If the input vector is too small (i.e., this->norm()==0), then \c *this is left unchanged.
187   *
188   * \sa stableNorm(), stableNormalized(), normalize()
189   */
190 template<typename Derived>
191 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void MatrixBase<Derived>::stableNormalize()
192 {
193   RealScalar w = cwiseAbs().maxCoeff();
194   RealScalar z = (derived()/w).squaredNorm();
195   if(z>RealScalar(0))
196     derived() /= numext::sqrt(z)*w;
197 }
198 
199 //---------- implementation of other norms ----------
200 
201 namespace internal {
202 
203 template<typename Derived, int p>
204 struct lpNorm_selector
205 {
206   typedef typename NumTraits<typename traits<Derived>::Scalar>::Real RealScalar;
207   EIGEN_DEVICE_FUNC
208   static inline RealScalar run(const MatrixBase<Derived>& m)
209   {
210     EIGEN_USING_STD(pow)
211     return pow(m.cwiseAbs().array().pow(p).sum(), RealScalar(1)/p);
212   }
213 };
214 
215 template<typename Derived>
216 struct lpNorm_selector<Derived, 1>
217 {
218   EIGEN_DEVICE_FUNC
219   static inline typename NumTraits<typename traits<Derived>::Scalar>::Real run(const MatrixBase<Derived>& m)
220   {
221     return m.cwiseAbs().sum();
222   }
223 };
224 
225 template<typename Derived>
226 struct lpNorm_selector<Derived, 2>
227 {
228   EIGEN_DEVICE_FUNC
229   static inline typename NumTraits<typename traits<Derived>::Scalar>::Real run(const MatrixBase<Derived>& m)
230   {
231     return m.norm();
232   }
233 };
234 
235 template<typename Derived>
236 struct lpNorm_selector<Derived, Infinity>
237 {
238   typedef typename NumTraits<typename traits<Derived>::Scalar>::Real RealScalar;
239   EIGEN_DEVICE_FUNC
240   static inline RealScalar run(const MatrixBase<Derived>& m)
241   {
242     if(Derived::SizeAtCompileTime==0 || (Derived::SizeAtCompileTime==Dynamic && m.size()==0))
243       return RealScalar(0);
244     return m.cwiseAbs().maxCoeff();
245   }
246 };
247 
248 } // end namespace internal
249 
250 /** \returns the \b coefficient-wise \f$ \ell^p \f$ norm of \c *this, that is, returns the p-th root of the sum of the p-th powers of the absolute values
251   *          of the coefficients of \c *this. If \a p is the special value \a Eigen::Infinity, this function returns the \f$ \ell^\infty \f$
252   *          norm, that is the maximum of the absolute values of the coefficients of \c *this.
253   *
254   * In all cases, if \c *this is empty, then the value 0 is returned.
255   *
256   * \note For matrices, this function does not compute the <a href="https://en.wikipedia.org/wiki/Operator_norm">operator-norm</a>. That is, if \c *this is a matrix, then its coefficients are interpreted as a 1D vector. Nonetheless, you can easily compute the 1-norm and \f$\infty\f$-norm matrix operator norms using \link TutorialReductionsVisitorsBroadcastingReductionsNorm partial reductions \endlink.
257   *
258   * \sa norm()
259   */
260 template<typename Derived>
261 template<int p>
262 #ifndef EIGEN_PARSED_BY_DOXYGEN
263 EIGEN_DEVICE_FUNC inline typename NumTraits<typename internal::traits<Derived>::Scalar>::Real
264 #else
265 EIGEN_DEVICE_FUNC MatrixBase<Derived>::RealScalar
266 #endif
267 MatrixBase<Derived>::lpNorm() const
268 {
269   return internal::lpNorm_selector<Derived, p>::run(*this);
270 }
271 
272 //---------- implementation of isOrthogonal / isUnitary ----------
273 
274 /** \returns true if *this is approximately orthogonal to \a other,
275   *          within the precision given by \a prec.
276   *
277   * Example: \include MatrixBase_isOrthogonal.cpp
278   * Output: \verbinclude MatrixBase_isOrthogonal.out
279   */
280 template<typename Derived>
281 template<typename OtherDerived>
282 bool MatrixBase<Derived>::isOrthogonal
283 (const MatrixBase<OtherDerived>& other, const RealScalar& prec) const
284 {
285   typename internal::nested_eval<Derived,2>::type nested(derived());
286   typename internal::nested_eval<OtherDerived,2>::type otherNested(other.derived());
287   return numext::abs2(nested.dot(otherNested)) <= prec * prec * nested.squaredNorm() * otherNested.squaredNorm();
288 }
289 
290 /** \returns true if *this is approximately an unitary matrix,
291   *          within the precision given by \a prec. In the case where the \a Scalar
292   *          type is real numbers, a unitary matrix is an orthogonal matrix, whence the name.
293   *
294   * \note This can be used to check whether a family of vectors forms an orthonormal basis.
295   *       Indeed, \c m.isUnitary() returns true if and only if the columns (equivalently, the rows) of m form an
296   *       orthonormal basis.
297   *
298   * Example: \include MatrixBase_isUnitary.cpp
299   * Output: \verbinclude MatrixBase_isUnitary.out
300   */
301 template<typename Derived>
302 bool MatrixBase<Derived>::isUnitary(const RealScalar& prec) const
303 {
304   typename internal::nested_eval<Derived,1>::type self(derived());
305   for(Index i = 0; i < cols(); ++i)
306   {
307     if(!internal::isApprox(self.col(i).squaredNorm(), static_cast<RealScalar>(1), prec))
308       return false;
309     for(Index j = 0; j < i; ++j)
310       if(!internal::isMuchSmallerThan(self.col(i).dot(self.col(j)), static_cast<Scalar>(1), prec))
311         return false;
312   }
313   return true;
314 }
315 
316 } // end namespace Eigen
317 
318 #endif // EIGEN_DOT_H
319