Subj : Re: stupid threading question ....
To   : comp.programming.threads
From : David Schwartz
Date : Mon Aug 15 2005 11:33 am


"Daniel Rayment" <dwrayment@rogers.com> wrote in message 
news:TYWdnbU705-2pp3eRVn-pg@rogers.com...

> but how bout on a 2 or more processor machine?  in the unlikely event that 
> 2 threads want to call the same function at the same time.   (Once again 
> we are not talking about accessing data types or sharing global variable 
> or anything that "obviously is bad").  ssuppose the function that gets 
> called at the same time is

    Why would there be a problem? The function isn't changing, so two 
threads are just reading the same unchanging information.

> function int doit(int &a,  &b)
> {
>   int c, d;
>
>   c = a + b;
>   a = b *c;
>   d = c - a;
>
>  if (a<b) return d;
>  else return c
>
> }
>
> all variables are local  or parameter reference.
>
>
> thread::code()
> {
>  static int (*func)(int &, int &)fn = doit;
>
> while (1)
> {
>  WaitForSingleObject( whatever, INFINITE);
>
>  int a, b;
>  (*fn)(a,b);
> }
>
> }
>
> and of course under murphys law  both wake at same time and call the 
> "doit"  function  supllyied as a pointer to func.

    So what do you think will go wrong? As you said, the variables are 
local. Even with one thread, there will need to be one copy for each 
invocation of the function. Consider:

int fact(int i)
{
 int j=i-1;
 if(i==00) return 1;
 return i * fact(j);
}

    Clearly there has to be one 'j' for each invocation of this function, 
right?

    DS

.