On Fri, Aug 20, 2004 at 08:15:22PM +0200, gilles wrote: > Marc Kleine-Budde wrote: > > Conclusion: As ''slow'' as the the last tests, but we can guarantee a > > upper limit (about 115 us) but that's not the point.
> I think we already mentioned it, but the figures you are comparing are : > RTAI/LXRT, sampling period: 100us, jitter 40us > RTAI/fusion, sampling period: 1ms, jitter 100us Yes - I remember that :) I've made these tests and graphes for my thesis talk, and now I want to share them with you.... > What you are observing here is very likely the effect of Linux using the > processor for a longer time and kicking RTAI out of the cache. It is a > well known side effect of raising the sampling period. Just run LXRT > with the same load and a 1ms period, and you should observe a raise in > latency. I tried to (quickly) port the programm to LXRT, but I've had no success (yet). I'll measure again with my skin, sampling period decresed to 100 us, on monday. (I have no access to the machine during the weekend) > I have just run testsuite/latency/latency, using the RTAI native > user-space skin, and observed a maximum jitter of 21 us on a Duron 950 > MHz with the same load as yours. > P.S: it is hard to issue any further comments without the sources of > your skin. I'll have to talk to Jan, if and when we're going to release the code... thanks - Marc -- #!/bin/sh set - `type $0` 'tr "[a-zA-Z]" "[n-za-mN-ZA-M]"';while [ "$2" != "" ];do \ shift;done; echo 'frq -a -rc '`echo "$0"| $1 `'>$UBZR/.`rpub signature|'`\ echo $1|$1`'`;rpub "Jr ner fvtangher bs obet. Erfvfgnapr vf shgvyr!"'|$1|sh
