Jonathan Blow

[1] Well, the compiler usually keeps a track of the values stored at registers, so they can share the value of the registers through different procedures. An example will guide you better. Suppose that you have this block of C code:

[2] void doit(int* vari) { CallAWindowsProcedure(vari) ; } void main() { int avar =56; doit(&avar) ; } So what happens next if we compile with -O3, is that the compiler will look if the whole program can be optimized, if it doesn't(because there's a call to an extern..

[3]..symbol) , then the compiler will check if avar could be optimized to an immediate value in asm(imm32), but it does not because the same as before, the external symbol. So it would probably just store the value 56 at rcx, rdx, or some register, and pass the address of the...

[4].... register at other register(following the windows calling conventions) and then call the external symbol, right?

[5] Now suppose we add the following to the main procedure of before

[6] void main() { //it's c++ but whatever int avar = 56; std::thread t ([avar] () CallOtherWindowsProcedure(&avar) ; }) ; ... } Obviously, the compiler would simply store avar at the data section and reference to it in the main and t threads, but it wouldn't be needed

Since we want a copy of avar on the thread t, we could simply do this instead of allocating 4 bytes at the data section: main: mov eax, 56 ... t_anonymousmethod001: mov eax, 56 ...