minus_op and so on about fixed_point HOT 1 CLOSED

Thanks for the observations. I aim to provide zero-cost abstractions but put most of my effort into ensuring the right API design so any feedback on performance is very valuable.

Your example, cls is a more complex than a fundamental type but it may still be pretty easy to optimize for compared to more complicated types such as, say, Boost.Multiprecision. Also, when I add the -O3 flag, the compiler eats up all of the code in your example so it's hard to determine what really strains the optimizer.

Also, I have not really explored using uvalues (xvalues? I loose track), i.e. (Rhs&& rhs).

Regarding the technique of passing a copy as the first parameter, that's a good approach. But unfortunately, I am maintaining support for C++11 which does not allow statements which mutate state (or indeed anything that isn't a return statement). So the lhs += rhs; generates a compiler error.

from fixed_point.