Please, help us to better know about our user community by answering the following short survey: https://forms.gle/wpyrxWi18ox9Z5ae9 Eigen  3.3.90 (git rev 598e1b6e54ec51c2448b1a10d4354f165f0b083e) Using custom scalar types

By default, Eigen currently supports standard floating-point types (float, double, std::complex<float>, std::complex<double>, long double), as well as all native integer types (e.g., int, unsigned int, short, etc.), and bool. On x86-64 systems, long double permits to locally enforces the use of x87 registers with extended accuracy (in comparison to SSE).

In order to add support for a custom type T you need:

1. make sure the common operator (+,-,*,/,etc.) are supported by the type T
2. add a specialization of struct Eigen::NumTraits<T> (see NumTraits)
3. define the math functions that makes sense for your type. This includes standard ones like sqrt, pow, sin, tan, conj, real, imag, etc, as well as abs2 which is Eigen specific. (see the file Eigen/src/Core/MathFunctions.h)

The math function should be defined in the same namespace than T, or in the std namespace though that second approach is not recommended.

Here is a concrete example adding support for the Adolc's adouble type. Adolc is an automatic differentiation library. The type adouble is basically a real value tracking the values of any number of partial derivatives.

#include <Eigen/Core>
namespace Eigen {
: NumTraits<double> // permits to get the epsilon, dummy_precision, lowest, highest functions
{
enum {
IsComplex = 0,
IsInteger = 0,
IsSigned = 1,
RequireInitialization = 1,
MulCost = 3
};
};
}
}

This other example adds support for the mpq_class type from GMP. It shows in particular how to change the way Eigen picks the best pivot during LU factorization. It selects the coefficient with the highest score, where the score is by default the absolute value of a number, but we can define a different score, for instance to prefer pivots with a more compact representation (this is an example, not a recommendation). Note that the scores should always be non-negative and only zero is allowed to have a score of zero. Also, this can interact badly with thresholds for inexact scalar types.

#include <gmpxx.h>
#include <Eigen/Core>
#include <boost/operators.hpp>
namespace Eigen {
template<> struct NumTraits<mpq_class> : GenericNumTraits<mpq_class>
{
typedef mpq_class Real;
typedef mpq_class NonInteger;
typedef mpq_class Nested;
static inline Real epsilon() { return 0; }
static inline Real dummy_precision() { return 0; }
static inline int digits10() { return 0; }
enum {
IsInteger = 0,
IsSigned = 1,
IsComplex = 0,
RequireInitialization = 1,
MulCost = 100
};
};
namespace internal {
template<> struct scalar_score_coeff_op<mpq_class> {
struct result_type : boost::totally_ordered1<result_type> {
std::size_t len;
result_type(int i = 0) : len(i) {} // Eigen uses Score(0) and Score()
result_type(mpq_class const& q) :
len(mpz_size(q.get_num_mpz_t())+
mpz_size(q.get_den_mpz_t())-1) {}
friend bool operator<(result_type x, result_type y) {
// 0 is the worst possible pivot
if (x.len == 0) return y.len > 0;
if (y.len == 0) return false;
// Prefer a pivot with a small representation
return x.len > y.len;
}
friend bool operator==(result_type x, result_type y) {
// Only used to test if the score is 0
return x.len == y.len;
}
};
result_type operator()(mpq_class const& x) const { return x; }
};
}
}
Eigen
Namespace containing all symbols from the Eigen library.
Definition: Core:134
Eigen::real
const Eigen::CwiseUnaryOp< Eigen::internal::scalar_real_op< typename Derived::Scalar >, const Derived > real(const Eigen::ArrayBase< Derived > &x)
Eigen::abs2
const Eigen::CwiseUnaryOp< Eigen::internal::scalar_abs2_op< typename Derived::Scalar >, const Derived > abs2(const Eigen::ArrayBase< Derived > &x)
Eigen::abs
const Eigen::CwiseUnaryOp< Eigen::internal::scalar_abs_op< typename Derived::Scalar >, const Derived > abs(const Eigen::ArrayBase< Derived > &x)
Eigen::imag
const Eigen::CwiseUnaryOp< Eigen::internal::scalar_imag_op< typename Derived::Scalar >, const Derived > imag(const Eigen::ArrayBase< Derived > &x)
Eigen::conj
const Eigen::CwiseUnaryOp< Eigen::internal::scalar_conjugate_op< typename Derived::Scalar >, const Derived > conj(const Eigen::ArrayBase< Derived > &x)