On Thursday 06 December 2012 10:44:53 Kent A. Reed did opine: > On 12/6/2012 9:15 AM, EBo wrote: > > <...> > > > > From that I can see that > > > > there is a small group of outliers at ±20us. From this information we > > might be able to look at the traces to see what is happening around > > those events and cleaning up some of the latency. I hear one of the > > previous posters about "my latency is better than yours" thing, but > > these plots (and possible statistical analysis) tells me how good my > > setup is and points me where to look for possible other latency gains. > > Those outliers were always there of course and could be seen with a > little wasted time staring at latency-test and latencyplot, but the > histogram approach demonstrates the value of pictures. > > I was the "mine is better than yours" poster so I have to quibble with > the "but these plots...." That was exactly my point to Gene. Focus on > understanding and on improving the performance, not on the absolute > numbers. > > > Thanks guys for providing this functionality. > > Yup. They get my tip of the hat. > > Regards, > Kent > > > PS - thanks everyone for not mentioning the mistake in my offhanded > remark to Gene about reducing the sizes of the captured-sample files by > reducing the precision of the sample. I was like Wile E Coyote (a Warner > Brothers cartoon character for those who don't know). I got up a head of > steam and ran off a cliff but didn't know it until I looked down. Sorry. > Fortunately, Dewey, et al. have made it a moot point. > Chuckle.. NP Kent, but I would point out that it only takes one out-lier of an extra 50 u-s, in a 25 u-s loop, to cause a stepper stall and a wrecked part or tool. To me I could care less if its 2 u-s or 7 u-s because they aren't show stoppers. But that 2x a day 50 u-s lag IS a show stopper because that is the puppy that will cheerfully eat your lunch.
Since latencyplot records the max values, after 60,000 samples, the peak was 19 u-s on this box. But I don't have an isolcpus statement in my grub.conf here either, and this is a one size fits all machine where those machines are optimized to run linuxcnc on real hardware. I ran it on the lathes box with an ssh -Y login, and its display is funkity, the timing scale reads from -5 to +5 on the left edge, but its like a split screen with servo thread baseline sitting at -5, and base thread baseline at 0. But a 5 is the worst its recording (so far). Servo's baseline is -5, peak is now at 0. Base thread is on 0, peak at +5. Obviously there is a difference in those 2 scripts. I haven't run the updater in 3 or 4 days on those boxes so this one is marginally fresher. :) Cheers, Gene -- "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author) My web page: <http://coyoteden.dyndns-free.com:85/gene> is up! Neutrinos have bad breadth. I was taught to respect my elders, but its getting harder and harder to find any... ------------------------------------------------------------------------------ LogMeIn Rescue: Anywhere, Anytime Remote support for IT. Free Trial Remotely access PCs and mobile devices and provide instant support Improve your efficiency, and focus on delivering more value-add services Discover what IT Professionals Know. Rescue delivers http://p.sf.net/sfu/logmein_12329d2d _______________________________________________ Emc-developers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-developers
