fixed number We are using the IA-32. that you get two 38.26 bit fixed point numbers over the stack which you have to multiply while considering the case that one of the numbers maybe less then zero. the architecture only has 32 bit registers so the output has to be in edx:eax, two 32 bit registers.