Latency Testing and Calibration
===============================

This directory contains a test module to verify architectural latency and 
jitter, up to the scheduling of a task. When used with the hard timer in
oneshot mode it can serve as a fine tuning of the average real-time 
scheduling latency.
In oneshot mode the idea is to measure the difference in time between the 
expected switch time and the time when a task is actually called by the 
scheduler. In periodic mode instead the variations of the task period are
used as measures of the scheduling latency/jitter. The reason for such a
choice are:
- there will be always a drift because the timer interrupt is based on a time 
baseline that is different from the one used to carry out measures calculations;
in case of a very good calibration such a drift can be made very small but 
users will rarely do so and thus they'll tend to be misled by the results;
- as far as there is no loss of timer interrupts there will be no drift in
periodic mode, unless the hardware goes ashtray, so period variations are 
valid measures.
The execution is fully parametrisable at insmod as it is possible to set up:
- timer mode
- timer period
- averaging time
- per average time or overall worst cases
- use of the FPU (with some calculations)
