>>>>> "blux" == blux <[EMAIL PROTECTED]> writes:
blux> I tested polling instead of irq to compare reaction times (rtl
blux> v1.1, P200 : about 1 to 4 micros with polling on lpt port and
blux> 2,5 to 3 micros with irq -> anybody can give me his results to
blux> compare) Two surprises : polling is less regular than irqs
blux> . (can give sources for critic)...
That's not a surprise. A simple comparison between the two runs like
this:
1. Advantages of interrupts (possible ones; present if implementation
is good enough)
a. Lower latency
b. Lower jitter
2. Advantages of polling (ditto)
a. Lower bound on response time under high load
b. Easy to guarantee liveness (i.e., ensure that everything that must
run will run)
c. Easier to implement
d. Easier to analyze
e. Easier to make correct (avoid deadlock bugs, critical section bugs,
resource conflict bugs, &c.)
2(b) is a particularly important one. For example, in networking
devices I would want to use polling on general principles because of
issues like that. There are many examples of devices that malfunction
under load because they use misdesigned priority scheduling systems
and/or interrupts. Conversely, anything that uses polling and a
single priority schedule tends to be extremely robust no matter how
you overload it.
paul
--- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
----
For more information on Real-Time Linux see:
http://www.rtlinux.org/~rtlinux/
