When you are profiling code that's been generated by an optimizing compiler, you often have to manually reverse-translate from the compiler generated assembler back into the C/C++ source to understand the profile results. The reason you have to do it manually is because one line of C or C++ can easily be compiled into a dozen or more instructions which may not even be adjacent to each other. I don't write much assembler but I find myself reading assembler at least once a week.