simplified and documented the type specifications in lambda_lanczos_util.hpp

[jewettaij]

I was able to simplify the code for real_t.

(I also made the code for ConjugateProduct shorter.)

This is not an important pull request. I will not be offended if you reject it.

Why?

Most LAMMPS programmers are not familiar with C++11. When I add your code to LAMMPS, I need to make your code short and easy to understand. I am sending you the version of lambda_lanczos_util.hpp that I am using in LAMMPS. I don't know if it is helpful to you.
But understanding your code has taught me new C++ features!

Cheers
Andrew

[jewettaij]

Here is one more way to simplify this file:

1. You can remove lines 68-69:

template <typename T>
double norm(const vector<double>&);

PLEASE IGNORE EVERYTHING BELOW THIS LINE (doesn't work)

2) Perhaps we can get simplify ConjugateProduct.

Perhaps instead we can replace ConjugateProduct::prod() with something simpler:

template

T ConjugateProd(T a, T b) { return std::conj(a)*b; }

I think this is possible because std::conj() is defined for all ordinary numeric types (float, double, long double), as well as complex types. So I think it is probably safe to use ConjugateProd(a,b) on ordinary numbers as well as complex types. (I have not tested simplification #2 yet.)

I did not include either of these two changes in my pull request.
Cheers

Andrew

[jewettaij]

Sorry.

Nevermind.

I just realized that this file contains a mistake.

I'll close this pull request and send another pull request (with some of the same changes).