Xenomai test results
The last few days have seen several updates to both the Adeos-ipipe patch and Xenomai to fix a few glitches.. Current build is 2.6.19.5 with the 1.0-05 patch for x86_64.Results: The latency tool reports averages in the order of 5-8uSec and worse case less than 20uSec. At first glance, it appears to be on par with RTAI on i386, so a quick & simple module was thrown together to toggle a bit on the parallel port.. The code runs entirely in user space and uses xeno_native to provide scheduling services..
Considering the standard onboard parport is a slow old dog, a stable base period of 5uSec was achieved, or 100KHz (monitored on a 60MHz oscilloscope) - Even under load running X/KDE along with a couple of stress test loops, pretty good.
Just for fun, the same test program was run with a PCI parport card, 3.5uSec (143KHz) appeared to be the limit, and running in pure text mode, 2.5uSec (200KHz) - Still rock solid with no sign of any jitter..
2.2uSec was reached, but jitter rendered the resulting pulse train unusable ecen although the computer remained responsive.
My hats off to the Xenomai/Adeos team for such hard work with impressive results.
[Boot note] RTAI have released Ver.3.5 for i386/x86_64 - The latter using a slightly older adeos-ipipe patch. On the face of it, RTAI should be able to deliver comparable performance, so it should be interesting to see some test results.