I've been testing interrupt latencies with RTLinux-3 on the 200MHz Pentium
machine. I have an external logic that sends an interrupt to the PC and
latches a veto which blocks further interrupts. The PC outputs a pulse
which clears the veto thus allowing further interrupts. I measure, on a
scope, the time difference between the input signal and the output signal
thereby giving me an pretty decent view of what the latency looks like.
I have the following numbers:
Case 1: the IRQ handler sends the output pulse - 6us.
Case 2: the IRQ handler dispatches a rt thread which sends the output pulse
- 10us.
Case 3: the IRQ handler stuffs an RT fifo with an int; a user space process
does a blocking read on the fifo and sends the output pulse - 40us
(sometimes 50).
In the case of #3, I'm wondering if a faster processor would do any better.
Does anyone know? I'm debating whether to do my task as a rt thread or as a
user space process. My task would take up to 1ms to complete. Obviously,
the rt task is faster, but would lack user space fault protection. If it is
possible to bring #3 down to about 10us with a faster processor, then it may
be worth it.
Thanks! Ken
-- [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/
