> Keeping everything the same, if you can rewrite your code so that it generates fewer instructions, it should be faster.

Not necessarily. Sometimes those additional instructions allow you to vectorize loops through eliminating the need for one loop iteration to depend on the previous one. For example: https://stackoverflow.com/questions/72306573/why-does-this-c...

> "Let us consider the following C++ code:"

But, C code is shown. I do not get why people tend to avoid using C name.

If it's compiled with a C++ compiler as C++ then it's C++ code. In this example the meaning is the same but some "C code" will produce different results depending on whether you compile it as C or C++.

It's compiled with a C++ compiler ('_Z2f2i' in the disassembly) and the function name is 'f.cpp', so I suppose it's in the 'C is a subset of C++' part of C.

Still that's C code. Code is an input to the compiler. So, I would not call the code differently just because later a C++ comiler is used. In C++ code, we expect C++ specific features be used.