Question

I was wondering if local variables in assembly are faster than the global variables that we use. The context for this is that I am learning some 2d animation using the win32 api, from a book. The author uses a function to initialize(Creation,registration, showing and updating of the window) the main window for the program. I wrote that function in asm(just to practice some asm). So, I was wondering if there is any performance advantage involved, since in the asm function i used, the WNDCLASSEX structure was created locally(in stack). I know that the local variables in assembly are supposed to be faster, but having gone through the disassembly for another program(entirely in cpp), I noticed that the compiler creates the WNDCLASSEX locally as well. This confused me about the topic. So i want to know if there is any difference in performance between the asm code and C++ code.

Devjeet

Answer 1

The top of the stack is touched by a lot of code. That means the the top of the stack is usually in CPU cache. Accessing this would be faster than accessing other memory areas (from .bss, etc).

But for a function like CreateWindow which is called only a few times per program this doesn’t really matter. The difference is less than a a few hundred CPU cycles. For other parts of the code the difference might be more noticeable. But an important thing to note is that if you’re doing the same thing repeatedly with the same piece of data that data too will end up in CPU cache and thus the performance difference would be negated.