Code: Select all

template <typename T>
T fun(Eigen::Matrix<T,Eigen::Dynamic,1> const &x){
   T y;
   y = sin(x(0));
   return y;
}


test_eigen_autodiff_2nd_derivative: /home/denzer/local/include/eigen3/Eigen/src/Core/CwiseBinaryOp.h:131: Eigen::CwiseBinaryOp<BinaryOp, Lhs, Rhs>::CwiseBinaryOp(const Lhs&, const Rhs&, const BinaryOp&) [with BinaryOp = Eigen::internal::scalar_sum_op<double>; Lhs = const Eigen::CwiseUnaryOp<Eigen::internal::scalar_multiple_op<double>, const Eigen::Matrix<double, -1, 1> >; Rhs = const Eigen::CwiseUnaryOp<Eigen::internal::scalar_multiple_op<double>, const Eigen::Matrix<double, -1, 1> >]: Assertion `lhs.rows() == rhs.rows() && lhs.cols() == rhs.cols()' failed.

Code: Select all

template <int order, int num_vars>
struct AutoDiffNdScalar {
    static_assert(order > 0 && order <= 4, "Must have order between 1 and 4");
    typedef typename AutoDiffNdScalar<order - 1, num_vars>::type prev_type;
    typedef Eigen::AutoDiffScalar<Eigen::Matrix<prev_type, num_vars, 1> > type;
};

// Base Case
template<int num_vars>
struct AutoDiffNdScalar<0, num_vars> {
    using type = double;
};
// Alternative: Define AutoDiffNdScalar<1, num_vars>

template <int order, int num_vars>
using AutoDiffNd = typename AutoDiffNdScalar<order, num_vars>::type;

...
    // AutoDiffNd<0, 1> x_bad(1); // Fails as expected
    AutoDiffNd<2, 1> x_taylor(1);


Code: Select all

template<typename _DerType>
class AutoDiffScalar {
...
  protected:
    Scalar m_value;
    DerType m_derivatives;
};


Code: Select all

// Explicitly specify scalar value, rather than implicitly deriving from derivative
template<typename _Scalar, typename _DerType>
class AutoDiffScalar {
...
  protected:
    Scalar m_value;
    DerType m_derivatives;
};

// Keep the existing interface for any prior consumers
// - Most likely invalid C++, but hopefully it gets the point across
template<typename _DerType>
using AutoDiffScalar = template AutoDiffScalar<typename internal::traits<typename internal::remove_all<_DerType>::type>::Scalar, _DerType>;