Hi,
To those minding about jitter.
I've taken some measures in a 1Khz interrupt service routine doing about
48 Kflops (multiply/add) per interrupt. With interrupt hard cli the
execution time of the floating point operations, measured with the cpu
clock (TSC), can take from 840 to (even) 970 us depending on the machine
load. Stable 840 for a low load wild oscillations up to the maximum
above under very heavy load, both in cpu and in I/O, mouse difficult to
track in X. (PII 233 Mhz).
By the way are you aware of how many interrupts generates a heavy mouse
activity? A LOT-LOT-LOT.
Due the the asm ("cli") in front of the multiply add loop the culprit
can only be the cache. Or there can be something else?
In my opinion no jitter machines are only DSPs with no cache and
strictly linear address space. With PCs and workstation you have to
accept a compromise.
If you disable the cache then things are more stable but you can be
catched by the raptus to throw the PC out of the window. Try and see
what happen when Linux boots up. That's why the first Celerons were
almost pure s.....
Ciao, Paolo.
--- [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/
