C++ class function in assembly

Question

Hello Community I am look at C++ assembly, I have compiled a benchmark from the PARSEC suite and I am having difficulty knowing how do they name the class attribute functions in assembly language. for example if I have a class with some functions to manipulate it, in cpp we call them like test.increment(); After some investigation I found out that this function is

atomic_load_acq_ptr

represented as:

_ZL19atomic_load_acq_intPVj

in assembly, or at least this is what I have found out.

Let me know if I am wrong! Is there some fixed rule for the mapping? or are they random? Thanks

Answer 1

It's called name mangling, is necessary because of overloads and templates and such (i.e. the plain chars-and-numbers name isn't enough to identify a chunk of code unambiguously; embedding spaces or <> or :: in names usually isn't legal; copying the additional information in uncondensed, human-readable form would be wasteful), and it therefore depends on types, arity, etc.

The exact scheme can vary, but usually each compiler is self-consistent for a relatively long time (sometimes even several compilers can settle for one way).

Answer 2

That's called name mangling.. It is compiler dependant. No standard way, sorry :)

Answer 3

Yes there's a standard method for creating these symbols known as name mangling.

Answer 4

C++ allows function overloading, this means that one can have two functions with the same name but different parameters. Since your binary formats do not understand type this is a proble. The way that this is worked around is to use a scheme called name mangling. This adds a whole function of type information to the name used in the source file and ensures one calls the correct overload.

The extra letters etc that are added are governed by the particular Application Binary Interface (ABI) being used. Different compilers (and sometimes even different versions) may use different ABIs.