Considering that the code is mostly x86 assembly, the gains from such optimization are quite unlikely.