On Mon, Sep 06, 2004 at 10:38:41PM +0300, David Harel wrote:
> Below pthread code example.
> for (i = 0; i < ITERATIONS; i++)
> {
> pthread_mutex_lock(&mutex);
> pthread_mutex_unlock(&mutex);
> j = myglobal;
> j++;
> myglobal = j;
> printf("%s%d", arg, myglobal);
> }
Don't do that. Very buggy code. Unlock the mutex *after* you change the global.
On the other hand, using a local variable doesn't buy you anything; just myglobal++ is
fine, but *when you're holding the mutex*.
Also, if you want to ensure there's actually any system-level IO in each exectution,
flush with fflush(stdout) after each printf.
> My question.
> What is the scheduling policy? is it not that whenever a thread does IO
> it looses CPU?
Depends on so many things, like where the output is going (terminal, file, pipe,
socket), what the file
system/network stack are, do they have delayed writes, etc. etc. A thread loses CPU
when it has to wait for a
physical IO operation to complete. Not every printf() call and not even any write()
call results in a physical
IO operation.
Just don't depend on that.
--
avva
"There's nothing simply good, nor ill alone" -- John Donne
=================================================================
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]
