****** STRESS EXAMPLE ******

What you find here is a stress test aimed at verifying that the schedulers 
do preempt well under intense load.
It is a simple test that combines the latency calibration task with a fast and
slow task in such a way to have two levels of preemption, nesting on an odd 
number of tasks. 

To verify what is happening you can launch "./display". It will show you the 
min/max/avrg jitters of the latency checking task, highest priority, and the
max jitter of the fast task, next to highest priority, and of and slow task,
slowest priority. Reasonable jitters are a clear indication of preemption.

Note that under UP the tests suck more than 75% of the computing power. Change 
ticking to decrease the load. So be careful to adjust test parameters to your
machine to avoid unfair blames on the scheduler.

We tested for some hours (up to 10), both 8254 and APIC, with the followings: 
	- at least one "ping -f somewhere";
	- a "ping -f localhost";
	- a "top" on alpha screen and one on an X screen;
	- a "while "true"; do ls -lR / >list; done"
	- a "while "true"; cat /proc/interrupts; done"
	- dd if=/dev/hda2 of=/dev/hda3 bs=1000k &
