On Mar 24, 2006, at 8:26 AM, Robert Ammerman wrote:

> By definition, "hard" real time systems must meet certain timing
> requirements, or they "fail".
>
> Both the "one big loop" and he "multi thread" techniques can be
> used to ensure this timing, but in both cases it requires
> careful analysis to ensure timing requirements are met.
>
Yes.  "Hard" real time is very hard indeed.  And rarely actually
necessary.  Hard RTOSes usually give up a great deal of performance
to get that guarantee; the MINIMUM time it might take to accomplish
something is typically very much lower than the guaranteed time.

I remember circa 1989, when BBN proudly announced their new
"Butterfly" multiprocessing RTOS could switch a network packet
in a guaranteed 1.2ms or so.  Since we at cisco had just finished
HW/SW that routed an honest 10,000+pps (100us per packet), I was
not very impressed. I'm more impressed NOW, since I'm older and
wiser and SW routing rates are up over 1e6pps and "we" still can't
"guarantee" 1.2ms for any particular packet...  (however, it was
more important to do the 1e4pps than to meet a 1.2ms guarantee,
at least in the eyes of the market.)

BillW

-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist