I'm glad this is being discussed, because for many years inlining functions was considered a performance panacea in C++, with no regard to the size of the object code that was being generated, or the effects inlining was having on instruction cache performance.
The C++ community has brought itself all kinds of complexity and long compile times all in the name of performance which, in my mind, was always pretty suspect.
Comments
I'm glad this is being discussed, because for many years inlining functions was considered a performance panacea in C++, with no regard to the size of the object code that was being generated, or the effects inlining was having on instruction cache performance.
The C++ community has brought itself all kinds of complexity and long compile times all in the name of performance which, in my mind, was always pretty suspect.