Revisit ffmin/ffmax?

[manas96]

In Listing 22, I assume the ff in functions ffmin and ffmax stands for floatfloatmin and floatfloatmax. However, these functions use doubles.

[hollasch]

In the C++ standard library, the f prefix generally stands for floating point, and double is the standard default. See https://en.cppreference.com/w/cpp/numeric/math/fmin.

These functions were defined before we based our code on C++11 (for std::shared_ptr), among other things. We're generally trying to limit the amount of C++11 we use for simplicity, but we might consider just adopting std::fmin and std::fmax.

However, fmin/fmax support additional float-specific behavior, such as handling of NaN values (only returning the non-NaN value), and some implementations return ±0 properly (std::fmin(+0,-0) may guarantee return of -0, for example).

In our simple implementation, we really just emulate std::min and std::max. It might be better just to rename these functions min and max, relying on specific overloads. image_texture for example could use max(int,int).

Generally, not sure if it's worth the effort to refactor these, but thought I'd write down the considerations here.