Am 11.11.2012 um 18:43 schrieb Kent A. Reed: > On 11/11/2012 8:53 AM, Michael Haberler wrote: >> I've worked on Xenomai userthreads build and it is now ready for early >> testers >> >> <...> >> >> This works on both the 2.6.38.8 and 3.2.21 xenomai kernels, recompilation >> not needed. > > Michael, I meant to ask. Do you see any significant difference in the > two kernels? I've been using your 3.2.21 kernel only.
I dont have hard figures yet (would need to run longer and record results), but my gut feeling is the 2.6.38.8 kernel is a bit lower on latency than 3.2.21 with xenomai-user I cant tell any differences between the two; as John reported, and I verified, there are issues with 2.6.38.8 and xenomai-kernel and probably more than one.. I dont see any of those on 3.2.21 > >> This build wiggles parport pins on my atom; there are still isolated >> realtime delays, and reporting of those needs improvement. >> >> The parport driver at this point uses the portable but slower ppdev ioctl >> method; work is underway to enable direct inport/outport I/O if supported by >> the architecture. >> >> This is now sufficiently mature that development of all RTOS support can >> happen in a single branch. >> >> for the daring: >> http://git.mah.priv.at/gitweb/emc2-dev.git/shortlog/refs/heads/rtos-integration-preview1 > > > I built your latest and greatest on my quadcore AMD box a few minutes > ago: Ubuntu 10.04LTS, 3.2.21-xenomai+ kernel. I also built it on my > Toshiba NB-305 netbook with dual-core atom N450: Linux Mint 13, > 3.2.21-xenomai+ kernel. > > Very preliminary indications are that with xenomai-user threads, I also > get better jitter numbers, both lower and more stable, than with > xenomai-kernel on my quadcore. I still see the blowup when I cat > /proc/xenomai/stat but that may be just an ASUS+Athlon II thing because > it does not occur on my atom-equipped netbook (which, however, has > crappy jitter for other reasons). I haven't had time to try any of your > xenomai work on my ASUS atom motherboard. > Curiously, I get better jitter results on the quadcore if I run both > latency-test threads on cpu 0, e.g., a core that is exposed to the Linux > scheduler, than on isolated cpus 2 and 3. On the other hand, > latency-test running entirely on cpu 0 is somewhat more sensitive to me > starting other processes, like glxgears, but the effect seems to be > restricted to a few microseconds accumulated the first time I start up > glxgears (maybe due to fetching glxgears from the disk?---I should try > again with a solid-state disk). these results puzzles me as well. one thing I cant make sense of yet is: isolcpus=1 on an atom bind base-thread in latency-test to cpu1 (the isolated one) and servo-thread to cpu0 (where linux runs too) and I get consistently lower latencies on servo-thread than on fast-thread making sense? none I can see.. > Like I said, this is very preliminary. At the least, it shows that > everything builds reliably from your repo in various environments. thanks for the stubborn testing, I really appreciate the extra pair of eyeballs - Michael > > I admit I was skeptical about the user-thread work but this is beginning > to look like a winner. After 30 minutes, my quadcore is still showing > max jitter less than 15us (1ms slow thread) 20us (25us fast thread) > with both threads running on cpu 0. During the same time, I've been > running several copies of glxgears, surfing the web, and running find > over the whole filesystem. RTAI performance on isolated cores is better > yet, but we all know the long-term problem with RTAI. > > Regards, > Kent > > > > > > > > ------------------------------------------------------------------------------ > Everyone hates slow websites. So do we. > Make your web apps faster with AppDynamics > Download AppDynamics Lite for free today: > http://p.sf.net/sfu/appdyn_d2d_nov > _______________________________________________ > Emc-developers mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/emc-developers ------------------------------------------------------------------------------ Everyone hates slow websites. So do we. Make your web apps faster with AppDynamics Download AppDynamics Lite for free today: http://p.sf.net/sfu/appdyn_d2d_nov _______________________________________________ Emc-developers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-developers
