• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1namespace Eigen {
2
3/** \page Eigen2ToEigen3 Porting from Eigen2 to Eigen3
4
5This page lists the most important API changes between Eigen2 and Eigen3,
6and gives tips to help porting your application from Eigen2 to Eigen3.
7
8\eigenAutoToc
9
10\section CompatibilitySupport Eigen2 compatibility support
11
12Up to version 3.2 %Eigen provides <a href="http://eigen.tuxfamily.org/dox-3.2/Eigen2SupportModes.html">Eigen2 support modes</a>. These are removed now, because they were barely used anymore and became hard to maintain after internal re-designs.
13You can still use them by first <a href="http://eigen.tuxfamily.org/dox-3.2/Eigen2ToEigen3.html">porting your code to Eigen 3.2</a>.
14
15\section Using The USING_PART_OF_NAMESPACE_EIGEN macro
16
17The USING_PART_OF_NAMESPACE_EIGEN macro has been removed. In Eigen 3, just do:
18\code
19using namespace Eigen;
20\endcode
21
22\section ComplexDot Dot products over complex numbers
23
24This is the single trickiest change between Eigen 2 and Eigen 3. It only affects code using \c std::complex numbers as scalar type.
25
26Eigen 2's dot product was linear in the first variable. Eigen 3's dot product is linear in the second variable. In other words, the Eigen 2 code \code x.dot(y) \endcode is equivalent to the Eigen 3 code \code y.dot(x) \endcode In yet other words, dot products are complex-conjugated in Eigen 3 compared to Eigen 2. The switch to the new convention was commanded by common usage, especially with the notation \f$ x^Ty \f$ for dot products of column-vectors.
27
28\section VectorBlocks Vector blocks
29
30<table class="manual">
31<tr><th>Eigen 2</th><th>Eigen 3</th></th>
32<tr><td>\code
33vector.start(length)
34vector.start<length>()
35vector.end(length)
36vector.end<length>()
37\endcode</td><td>\code
38vector.head(length)
39vector.head<length>()
40vector.tail(length)
41vector.tail<length>()
42\endcode</td></tr>
43</table>
44
45
46\section Corners Matrix Corners
47
48<table class="manual">
49<tr><th>Eigen 2</th><th>Eigen 3</th></th>
50<tr><td>\code
51matrix.corner(TopLeft,r,c)
52matrix.corner(TopRight,r,c)
53matrix.corner(BottomLeft,r,c)
54matrix.corner(BottomRight,r,c)
55matrix.corner<r,c>(TopLeft)
56matrix.corner<r,c>(TopRight)
57matrix.corner<r,c>(BottomLeft)
58matrix.corner<r,c>(BottomRight)
59\endcode</td><td>\code
60matrix.topLeftCorner(r,c)
61matrix.topRightCorner(r,c)
62matrix.bottomLeftCorner(r,c)
63matrix.bottomRightCorner(r,c)
64matrix.topLeftCorner<r,c>()
65matrix.topRightCorner<r,c>()
66matrix.bottomLeftCorner<r,c>()
67matrix.bottomRightCorner<r,c>()
68\endcode</td>
69</tr>
70</table>
71
72Notice that Eigen3 also provides these new convenience methods: topRows(), bottomRows(), leftCols(), rightCols(). See in class DenseBase.
73
74\section CoefficientWiseOperations Coefficient wise operations
75
76In Eigen2, coefficient wise operations which have no proper mathematical definition (as a coefficient wise product)
77were achieved using the .cwise() prefix, e.g.:
78\code a.cwise() * b \endcode
79In Eigen3 this .cwise() prefix has been superseded by a new kind of matrix type called
80Array for which all operations are performed coefficient wise. You can easily view a matrix as an array and vice versa using
81the MatrixBase::array() and ArrayBase::matrix() functions respectively. Here is an example:
82\code
83Vector4f a, b, c;
84c = a.array() * b.array();
85\endcode
86Note that the .array() function is not at all a synonym of the deprecated .cwise() prefix.
87While the .cwise() prefix changed the behavior of the following operator, the array() function performs
88a permanent conversion to the array world. Therefore, for binary operations such as the coefficient wise product,
89both sides must be converted to an \em array as in the above example. On the other hand, when you
90concatenate multiple coefficient wise operations you only have to do the conversion once, e.g.:
91\code
92Vector4f a, b, c;
93c = a.array().abs().pow(3) * b.array().abs().sin();
94\endcode
95With Eigen2 you would have written:
96\code
97c = (a.cwise().abs().cwise().pow(3)).cwise() * (b.cwise().abs().cwise().sin());
98\endcode
99
100\section PartAndExtract Triangular and self-adjoint matrices
101
102In Eigen 2 you had to play with the part, extract, and marked functions to deal with triangular and selfadjoint matrices. In Eigen 3, all these functions have been removed in favor of the concept of \em views:
103
104<table class="manual">
105<tr><th>Eigen 2</th><th>Eigen 3</th></tr>
106<tr><td>\code
107A.part<UpperTriangular>();
108A.part<StrictlyLowerTriangular>(); \endcode</td>
109<td>\code
110A.triangularView<Upper>()
111A.triangularView<StrictlyLower>()\endcode</td></tr>
112<tr><td>\code
113A.extract<UpperTriangular>();
114A.extract<StrictlyLowerTriangular>();\endcode</td>
115<td>\code
116A.triangularView<Upper>()
117A.triangularView<StrictlyLower>()\endcode</td></tr>
118<tr><td>\code
119A.marked<UpperTriangular>();
120A.marked<StrictlyLowerTriangular>();\endcode</td>
121<td>\code
122A.triangularView<Upper>()
123A.triangularView<StrictlyLower>()\endcode</td></tr>
124<tr><td colspan="2"></td></tr>
125<tr><td>\code
126A.part<SelfAdfjoint|UpperTriangular>();
127A.extract<SelfAdfjoint|LowerTriangular>();\endcode</td>
128<td>\code
129A.selfadjointView<Upper>()
130A.selfadjointView<Lower>()\endcode</td></tr>
131<tr><td colspan="2"></td></tr>
132<tr><td>\code
133UpperTriangular
134LowerTriangular
135UnitUpperTriangular
136UnitLowerTriangular
137StrictlyUpperTriangular
138StrictlyLowerTriangular
139\endcode</td><td>\code
140Upper
141Lower
142UnitUpper
143UnitLower
144StrictlyUpper
145StrictlyLower
146\endcode</td>
147</tr>
148</table>
149
150\sa class TriangularView, class SelfAdjointView
151
152\section TriangularSolveInPlace Triangular in-place solving
153
154<table class="manual">
155<tr><th>Eigen 2</th><th>Eigen 3</th></tr>
156<tr><td>\code A.triangularSolveInPlace<XxxTriangular>(Y);\endcode</td><td>\code A.triangularView<Xxx>().solveInPlace(Y);\endcode</td></tr>
157</table>
158
159
160\section Decompositions Matrix decompositions
161
162Some of Eigen 2's matrix decompositions have been renamed in Eigen 3, while some others have been removed and are replaced by other decompositions in Eigen 3.
163
164<table class="manual">
165  <tr>
166    <th>Eigen 2</th>
167    <th>Eigen 3</th>
168    <th>Notes</th>
169  </tr>
170  <tr>
171    <td>LU</td>
172    <td>FullPivLU</td>
173    <td class="alt">See also the new PartialPivLU, it's much faster</td>
174  </tr>
175  <tr>
176    <td>QR</td>
177    <td>HouseholderQR</td>
178    <td class="alt">See also the new ColPivHouseholderQR, it's more reliable</td>
179  </tr>
180  <tr>
181    <td>SVD</td>
182    <td>JacobiSVD</td>
183    <td class="alt">We currently don't have a bidiagonalizing SVD; of course this is planned.</td>
184  </tr>
185  <tr>
186    <td>EigenSolver and friends</td>
187    <td>\code #include<Eigen/Eigenvalues> \endcode </td>
188    <td class="alt">Moved to separate module</td>
189  </tr>
190</table>
191
192\section LinearSolvers Linear solvers
193
194<table class="manual">
195<tr><th>Eigen 2</th><th>Eigen 3</th><th>Notes</th></tr>
196<tr><td>\code A.lu();\endcode</td>
197<td>\code A.fullPivLu();\endcode</td>
198<td class="alt">Now A.lu() returns a PartialPivLU</td></tr>
199<tr><td>\code A.lu().solve(B,&X);\endcode</td>
200<td>\code X = A.lu().solve(B);
201 X = A.fullPivLu().solve(B);\endcode</td>
202<td class="alt">The returned by value is fully optimized</td></tr>
203<tr><td>\code A.llt().solve(B,&X);\endcode</td>
204<td>\code X = A.llt().solve(B);
205 X = A.selfadjointView<Lower>.llt().solve(B);
206 X = A.selfadjointView<Upper>.llt().solve(B);\endcode</td>
207<td class="alt">The returned by value is fully optimized and \n
208the selfadjointView API allows you to select the \n
209triangular part to work on (default is lower part)</td></tr>
210<tr><td>\code A.llt().solveInPlace(B);\endcode</td>
211<td>\code B = A.llt().solve(B);
212 B = A.selfadjointView<Lower>.llt().solve(B);
213 B = A.selfadjointView<Upper>.llt().solve(B);\endcode</td>
214<td class="alt">In place solving</td></tr>
215<tr><td>\code A.ldlt().solve(B,&X);\endcode</td>
216<td>\code X = A.ldlt().solve(B);
217 X = A.selfadjointView<Lower>.ldlt().solve(B);
218 X = A.selfadjointView<Upper>.ldlt().solve(B);\endcode</td>
219<td class="alt">The returned by value is fully optimized and \n
220the selfadjointView API allows you to select the \n
221triangular part to work on</td></tr>
222</table>
223
224\section GeometryModule Changes in the Geometry module
225
226The Geometry module is the one that changed the most. If you rely heavily on it, it's probably a good idea to use the <a href="http://eigen.tuxfamily.org/dox-3.2/Eigen2SupportModes.html">"Eigen 2 support modes"</a> to perform your migration.
227
228\section Transform The Transform class
229
230In Eigen 2, the Transform class didn't really know whether it was a projective or affine transformation. In Eigen 3, it takes a new \a Mode template parameter, which indicates whether it's \a Projective or \a Affine transform. There is no default value.
231
232The Transform3f (etc) typedefs are no more. In Eigen 3, the Transform typedefs explicitly refer to the \a Projective and \a Affine modes:
233
234<table class="manual">
235<tr><th>Eigen 2</th><th>Eigen 3</th><th>Notes</th></tr>
236<tr>
237  <td> Transform3f </td>
238  <td> Affine3f or Projective3f </td>
239  <td> Of course 3f is just an example here </td>
240</tr>
241</table>
242
243
244\section LazyVsNoalias Lazy evaluation and noalias
245
246In Eigen all operations are performed in a lazy fashion except the matrix products which are always evaluated into a temporary by default.
247In Eigen2, lazy evaluation could be enforced by tagging a product using the .lazy() function. However, in complex expressions it was not
248easy to determine where to put the lazy() function. In Eigen3, the lazy() feature has been superseded by the MatrixBase::noalias() function
249which can be used on the left hand side of an assignment when no aliasing can occur. Here is an example:
250\code
251MatrixXf a, b, c;
252...
253c.noalias() += 2 * a.transpose() * b;
254\endcode
255However, the noalias mechanism does not cover all the features of the old .lazy(). Indeed, in some extremely rare cases,
256it might be useful to explicit request for a lay product, i.e., for a product which will be evaluated one coefficient at once, on request,
257just like any other expressions. To this end you can use the MatrixBase::lazyProduct() function, however we strongly discourage you to
258use it unless you are sure of what you are doing, i.e., you have rigourosly measured a speed improvement.
259
260\section AlignMacros Alignment-related macros
261
262The EIGEN_ALIGN_128 macro has been renamed to EIGEN_ALIGN16. Don't be surprised, it's just that we switched to counting in bytes ;-)
263
264The \link TopicPreprocessorDirectivesPerformance EIGEN_DONT_ALIGN \endlink option still exists in Eigen 3, but it has a new cousin: \link TopicPreprocessorDirectivesPerformance  EIGEN_DONT_ALIGN_STATICALLY.\endlink It allows to get rid of all static alignment issues while keeping alignment of dynamic-size heap-allocated arrays. Vectorization of statically allocated arrays is still preserved (unless you define \link TopicPreprocessorDirectivesPerformance EIGEN_UNALIGNED_VECTORIZE \endlink =0), at the cost of unaligned memory stores.
265
266\section AlignedMap Aligned Map objects
267
268A common issue with Eigen 2 was that when mapping an array with Map, there was no way to tell Eigen that your array was aligned. There was a ForceAligned option but it didn't mean that; it was just confusing and has been removed.
269
270New in Eigen3 is the #Aligned option. See the documentation of class Map. Use it like this:
271\code
272Map<Vector4f, Aligned> myMappedVector(some_aligned_array);
273\endcode
274There also are related convenience static methods, which actually are the preferred way as they take care of such things as constness:
275\code
276result = Vector4f::MapAligned(some_aligned_array);
277\endcode
278
279\section StdContainers STL Containers
280
281In Eigen2, <tt>\#include\<Eigen/StdVector\></tt> tweaked std::vector to automatically align elements. The problem was that that was quite invasive. In Eigen3, we only override standard behavior if you use Eigen::aligned_allocator<T> as your allocator type. So for example, if you use std::vector<Matrix4f>, you need to do the following change (note that aligned_allocator is under namespace Eigen):
282
283<table class="manual">
284<tr><th>Eigen 2</th><th>Eigen 3</th></tr>
285<tr>
286  <td> \code std::vector<Matrix4f> \endcode </td>
287  <td> \code std::vector<Matrix4f, aligned_allocator<Matrix4f> > \endcode </td>
288</tr>
289</table>
290
291\section eiPrefix Internal ei_ prefix
292
293In Eigen2, global internal functions and structures were prefixed by \c ei_. In Eigen3, they all have been moved into the more explicit \c internal namespace. So, e.g., \c ei_sqrt(x) now becomes \c internal::sqrt(x). Of course it is not recommended to rely on Eigen's internal features.
294
295
296
297*/
298
299}
300