1 namespace Eigen { 2 3 /** \eigenManualPage TopicStructHavingEigenMembers Structures Having Eigen Members 4 5 \eigenAutoToc 6 7 \section summary Executive Summary 8 9 If you define a structure having members of \ref TopicFixedSizeVectorizable "fixed-size vectorizable Eigen types", you must overload its "operator new" so that it generates 16-bytes-aligned pointers. Fortunately, Eigen provides you with a macro EIGEN_MAKE_ALIGNED_OPERATOR_NEW that does that for you. 10 11 \section what What kind of code needs to be changed? 12 13 The kind of code that needs to be changed is this: 14 15 \code 16 class Foo 17 { 18 ... 19 Eigen::Vector2d v; 20 ... 21 }; 22 23 ... 24 25 Foo *foo = new Foo; 26 \endcode 27 28 In other words: you have a class that has as a member a \ref TopicFixedSizeVectorizable "fixed-size vectorizable Eigen object", and then you dynamically create an object of that class. 29 30 \section how How should such code be modified? 31 32 Very easy, you just need to put a EIGEN_MAKE_ALIGNED_OPERATOR_NEW macro in a public part of your class, like this: 33 34 \code 35 class Foo 36 { 37 ... 38 Eigen::Vector2d v; 39 ... 40 public: 41 EIGEN_MAKE_ALIGNED_OPERATOR_NEW 42 }; 43 44 ... 45 46 Foo *foo = new Foo; 47 \endcode 48 49 This macro makes "new Foo" always return an aligned pointer. 50 51 If this approach is too intrusive, see also the \ref othersolutions. 52 53 \section why Why is this needed? 54 55 OK let's say that your code looks like this: 56 57 \code 58 class Foo 59 { 60 ... 61 Eigen::Vector2d v; 62 ... 63 }; 64 65 ... 66 67 Foo *foo = new Foo; 68 \endcode 69 70 A Eigen::Vector2d consists of 2 doubles, which is 128 bits. Which is exactly the size of a SSE packet, which makes it possible to use SSE for all sorts of operations on this vector. But SSE instructions (at least the ones that Eigen uses, which are the fast ones) require 128-bit alignment. Otherwise you get a segmentation fault. 71 72 For this reason, Eigen takes care by itself to require 128-bit alignment for Eigen::Vector2d, by doing two things: 73 \li Eigen requires 128-bit alignment for the Eigen::Vector2d's array (of 2 doubles). With GCC, this is done with a __attribute__ ((aligned(16))). 74 \li Eigen overloads the "operator new" of Eigen::Vector2d so it will always return 128-bit aligned pointers. 75 76 Thus, normally, you don't have to worry about anything, Eigen handles alignment for you... 77 78 ... except in one case. When you have a class Foo like above, and you dynamically allocate a new Foo as above, then, since Foo doesn't have aligned "operator new", the returned pointer foo is not necessarily 128-bit aligned. 79 80 The alignment attribute of the member v is then relative to the start of the class, foo. If the foo pointer wasn't aligned, then foo->v won't be aligned either! 81 82 The solution is to let class Foo have an aligned "operator new", as we showed in the previous section. 83 84 \section movetotop Should I then put all the members of Eigen types at the beginning of my class? 85 86 That's not required. Since Eigen takes care of declaring 128-bit alignment, all members that need it are automatically 128-bit aligned relatively to the class. So code like this works fine: 87 88 \code 89 class Foo 90 { 91 double x; 92 Eigen::Vector2d v; 93 public: 94 EIGEN_MAKE_ALIGNED_OPERATOR_NEW 95 }; 96 \endcode 97 98 \section dynamicsize What about dynamic-size matrices and vectors? 99 100 Dynamic-size matrices and vectors, such as Eigen::VectorXd, allocate dynamically their own array of coefficients, so they take care of requiring absolute alignment automatically. So they don't cause this issue. The issue discussed here is only with \ref TopicFixedSizeVectorizable "fixed-size vectorizable matrices and vectors". 101 102 \section bugineigen So is this a bug in Eigen? 103 104 No, it's not our bug. It's more like an inherent problem of the C++98 language specification, and seems to be taken care of in the upcoming language revision: <a href="http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2341.pdf">see this document</a>. 105 106 \section conditional What if I want to do this conditionnally (depending on template parameters) ? 107 108 For this situation, we offer the macro EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF(NeedsToAlign). It will generate aligned operators like EIGEN_MAKE_ALIGNED_OPERATOR_NEW if NeedsToAlign is true. It will generate operators with the default alignment if NeedsToAlign is false. 109 110 Example: 111 112 \code 113 template<int n> class Foo 114 { 115 typedef Eigen::Matrix<float,n,1> Vector; 116 enum { NeedsToAlign = (sizeof(Vector)%16)==0 }; 117 ... 118 Vector v; 119 ... 120 public: 121 EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF(NeedsToAlign) 122 }; 123 124 ... 125 126 Foo<4> *foo4 = new Foo<4>; // foo4 is guaranteed to be 128bit-aligned 127 Foo<3> *foo3 = new Foo<3>; // foo3 has only the system default alignment guarantee 128 \endcode 129 130 131 \section othersolutions Other solutions 132 133 In case putting the EIGEN_MAKE_ALIGNED_OPERATOR_NEW macro everywhere is too intrusive, there exists at least two other solutions. 134 135 \subsection othersolutions1 Disabling alignment 136 137 The first is to disable alignment requirement for the fixed size members: 138 \code 139 class Foo 140 { 141 ... 142 Eigen::Matrix<double,2,1,Eigen::DontAlign> v; 143 ... 144 }; 145 \endcode 146 This has for effect to disable vectorization when using \c v. 147 If a function of Foo uses it several times, then it still possible to re-enable vectorization by copying it into an aligned temporary vector: 148 \code 149 void Foo::bar() 150 { 151 Eigen::Vector2d av(v); 152 // use av instead of v 153 ... 154 // if av changed, then do: 155 v = av; 156 } 157 \endcode 158 159 \subsection othersolutions2 Private structure 160 161 The second consist in storing the fixed-size objects into a private struct which will be dynamically allocated at the construction time of the main object: 162 163 \code 164 struct Foo_d 165 { 166 EIGEN_MAKE_ALIGNED_OPERATOR_NEW 167 Vector2d v; 168 ... 169 }; 170 171 172 struct Foo { 173 Foo() { init_d(); } 174 ~Foo() { delete d; } 175 void bar() 176 { 177 // use d->v instead of v 178 ... 179 } 180 private: 181 void init_d() { d = new Foo_d; } 182 Foo_d* d; 183 }; 184 \endcode 185 186 The clear advantage here is that the class Foo remains unchanged regarding alignment issues. The drawback is that a heap allocation will be required whatsoever. 187 188 */ 189 190 } 191
