Hello,

I need to extend the precision of Eigen's complex matrix eigenmode solver -- long double is 16 bytes on my compiler, so this seems sufficient. I see in this thread (https://forum.kde.org/viewtopic.php?f=74&t=85550) that Eigen can be straightforwardly modified to support complex long double matrices.

However, if I correspondingly call (the currently fictitious) eigenmode solver ComplexEigenSolver<MatrixXcld>, will the computations automatically be done in extended precision (long double format), or will I need to modify those methods as well?

Thanks

Just use MatrixXcld everywhere you need it, nothing to change within Eigen!

Thanks for the help, ggael. I'm a novice at C++ and programming, and I'm not too sure about templating.

From what I can tell, all I needed to do is insert the template (as done here: https://forum.kde.org/viewtopic.php?f=74&t=124589)

Code: Select all

typedef Matrix<std::complex<long double>,Dynamic,Dynamic> MatrixXcld;

, and then use MatrixXcld everywhere where I previously would have had MatrixXcd. However, my compiler is now complaining that

Use of overloaded operator '*' is ambiguous (with operand types 'complex<double>' and 'MatrixXcld' (aka 'Matrix<complex<long double>, Dynamic, Dynamic>'))

I assume this means I have to define basic mathematical operation for my new MatrixXcld class as well -- is there a straightforward way to do this?

[Solved]

The std::complex was to blame. It didn't know what to do with multiplying complex<double> with a MatrixXcld (long double). Converting the numbers to complex<long double>() solved the problems!

Indeed, implicit conversion of the nested scalar type is not allowed.