Quote:Original post by the_edd

Quote:Original post by davetyler

void threaded_function(int shared_variable){  while(1)  {    shared_variable++;    thread_sleep(100); //Implementation defined sleep function taking ms param.  }}void main_function(int shared_variable){  int temp_var;    while(1)  {    temp_var = shared_variable;    thread_sleep(100);  }}

shared_variable isn't shared. Each function gets its own copy. I assume this is a thinko when entering the code?

Quote:Original post by outRider

Quote:Original post by MaulingMonkey

Quote:Original post by Hodgman
On x86, i believe all 32bit (int sized) reads and writes are atomic (they fetch the full 4 bytes in one go).

Nope. Consider an unaligned int straddling page boundaries.

A few more cases: straddling cache lines, straddling bus width boundaries (64-bit value on a system with a 32-bit bus, etc). Although not guaranteed to be atomic, some might still be depending on the memory controller, since it can simply stall the CPU for however many bus cycles are needed to return the result atomically.