Huge performance difference btw double and long double

Question

Regarding performance of double vs long double:

On Macbook, Core i7 2017, MacOs 11, Xcode 12.2, C++17, Clang, with and without optimization