On 2/13/2013 10:28 PM, Tom Easterday wrote: > I spent the day playing with this on the Intel Atom D2700MUD. I rebuilt the > system using the rtos-integration-preview3-merged-into-master > branch. I rebuilt it for another reason (stupid graphics fubar) and figured > I would try this branch for the heck of it. > > Neither no-halt nor idle=poll had any effect (good or bad) on the system as > far as I could tell in the short testing intervals I did today (5 minutes or > so generally). I played with Hyperthreading, which definitely needs to stay > off on my system, and isolcpus which helps latency some (I think) but really > makes the machine slow for doing anything else other than Linuxcnc - but > that's the point isn't it? In general I am seeing between 15,000-22,000 but > I will let the machine run over night or a weekend soon for more concrete > numbers. > > If I run update-grub, whether I have made a change to grub or not, the number > shoots up to over 170,000. Obviously this isn't something one would do while > using Linuxcnc, like Kent's messing in the /proc area, and shows you should > keep your mitts out of where they don't belong. :-) > > Tom > > On Feb 13, 2013, at 2:29 PM, sam sokolik <[email protected]> wrote: > >> For this system (amd) the only thing that seems to make a difference is >> the idle=poll (no-halt seems to have no effect) >> >> I go from 200+us latency to 7us top. >> >> sam >>
Regarding the AMD Athlon II X4-640 system I've mentioned several times, I got a few moments to try a restricted set of combinations. Just first impressions. --- Some trivia: Still Michael's i686 3.2.21-xenomai+ kernel from late last fall on a Ubuntu 10.04LTS system. Always isolcpus=3 (e.g., the last one reported by the bios) Five-minute latency tests with both 25us base-period, 1ms servo-period threads running on isolated cpu 3. No special attempt was to "torture" the system. Just random browsing. "cpu hog" means the one-line script from the Wiki. "us" means microsecond --- Results: case 1. no other boot parameters, no cpu hog servo jitter 6us / base jitter 9.5us --- case 2. no other boot parameters, 1 cpu hog running servo jitter 5.9us / base jitter 6.2us --- case 3. no-hlt and idle=poll, no cpu hog servo jitter 4.0us / base jitter 5.0us [avg of 3 runs] servo jitter 2.8us / base jitter 3.3us [a later single run; gives a sense of the variability] --- case 4. no-hlt and idle=poll, 1 cpu hog running servo jitter 3.2us / base jitter 5.1us [I believe this is an anomalous result caused by me playing with the sensors/ksensors packages. Forgot to go back and check. Sorry.] --- case 5. nohlt and idle=poll, no cpu hog servo jitter 2.5us / 3.1us --- case 6. nohlt and idle=poll, 1 cpu hog running <no test> --- case 7. only idle=poll, no cpu hog servo jitter 2.7us / base jitter 3.4us --- case 8. only idle=poll, 1 cpu hog running essentially same as case 7 within some tenths of a microsecond. ------------- Comments: 1. My jitter is so low with this CPU/motherboard combination that none of this matters practically:-) 2. I recorded jitters to tenths of a microsecond but all indications are that I can get 1-2us variability without doing much of anything. 3. Like Tom and Sam, I don't see a strong dependence on either no-hlt or nohlt 4. Even at these low latencies, idle=poll has a beneficial effect and appears to make the cpu-hog approach irrelevant (not exactly new news...) 5. Reported temperature rises nearly 20degC with the extra boot parameters and a cheap house-brand heatsink/fan. Caveat: could be worse than this. The lm-sensors package and the ksensors package both report some temperature I don't believe is the CPU itself; at least the number they show is nearly 10degC lower than the one I see if I quickly reboot and look at the readings in the BIOS. Sigh. ------------- Regards, Kent ------------------------------------------------------------------------------ Free Next-Gen Firewall Hardware Offer Buy your Sophos next-gen firewall before the end March 2013 and get the hardware for free! Learn more. http://p.sf.net/sfu/sophos-d2d-feb _______________________________________________ Emc-developers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-developers
