Hello David,
[...]
> Are you using FP or some more "advanced" math functions?
[...]
Nope. At the moment its just bit shifting and int calculations :). But I
will get some FP to do when I have
AD/DA cards in my system, which will be in the next project.
[...]
> A third possibility: It seems to me that it isn't X, but rather the
> applications that affect you RT task. (Otherwise running X on NT - or any
> other alternative to a local X server - probably wouldn't have the same
effect
> as running a local server. It *could* be the library and not the server,
> but...) That is, it might be *any* form of system stress that affects the
RT
> task.
[...]
If you run graphical X apps on an X Client, you get some network traffic
which depends on window size/
color depth/some other stuff I don't know. If I have a basic xterm, almost
nothing happens. When running KDevelop
fullscreen (without doing anything, no scrolling, compiling, etc ... ), I
get an additional 70-80 us. Its MUCH worse
when I'm scrolling through source code :)
[...]
> My guess would be that the fluctuations in execution time are the result
of
> cache misses caused by the applications flushing RT data out of the cache.
The
> Celeron has 128 kB of full core speed cache, and a 66 MHz system bus, so
there
> can be quite some difference between accessing hot and cold data. Are you
using
> lots of data, look-up tables or anything like that?
[...]
Cache misses could also be part of the problem ... my module is now 800k in
memory, depeding on how
large I set my data structures. But only a small part of it is accessed with
my current configuration, so I think
the biggest part is the network traffic.
Marco
-- [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/
