#linuxcnc-devel Logs

Jul 12 2017

06:45 AM -!- #linuxcnc-devel mode set to +v by ChanServ
09:24 AM skunkworks_: wow - forgot this workstation was running the latency histogram.. Been running for a week or so - 8.8us latency
10:17 AM skunkworks: jepler, I might not know what I am looking at...
10:17 AM skunkworks: http://electronicsam.com/images/KandT/testing/Screenshot_2017-07-12_09-44-12.png
10:17 AM skunkworks: err2 is ac-coupled
10:18 AM skunkworks: the spirts of noise on it are me moving the mouse
10:18 AM skunkworks: 7i92
10:19 AM skunkworks: compared to yours.. https://emergent.unpythonic.net/files/sandbox/latency-vs-mesa.png
10:19 AM jepler: "err2" is the actual time between thread invocations reported by timedelta, the thing used by latency-test
10:21 AM jepler: "err" is supposed to give an idea of "real" latency actually communicating with the mesa board
10:21 AM skunkworks: so 3us +/-?
10:21 AM skunkworks: give or take?
10:22 AM jepler: I'm not sure of the units of "err"
10:22 AM jepler: what I wanted to see was whether it is correlated with err2
10:22 AM jepler: and it's not particularly correlated with it in your case
10:22 AM skunkworks: right - I have a whole lot of ethernet latency
10:23 AM skunkworks: is err2 in clocks or time
10:24 AM jepler: I think err is in "fractions of one period", so an error of 100m = .1 is 100us ?
10:24 AM jepler: err2 is in time
10:24 AM skunkworks: ok
10:25 AM skunkworks: 100us+ would make sense for ethernet
10:25 AM skunkworks: off and on
10:25 AM jepler: yeah
10:27 AM skunkworks: uspace seems pretty amazing on new hardware atleast
10:27 AM skunkworks: newish
10:27 AM skunkworks: (this is a 3rdgen i5
10:33 AM skunkworks: so - with pci yours is showing about the same
10:34 AM skunkworks: 5m - 5us/dev - so 10ish us
10:34 AM skunkworks: +/-
10:34 AM skunkworks: maybe 15
10:41 AM seb_kuzminsky: when thermal engineers cry: http://www.phoronix.com/scan.php?page=news_item&px=AMD-ARM-In-My-Hands
10:48 AM KGB-linuxcnc: 03Sebastian Kuzminsky 052.7 7b66c8e 06linuxcnc Merge remote-tracking branch 'origin/2.7-asciidoc-manpages' into 2.7 * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=7b66c8e
10:48 AM KGB-linuxcnc: 052.7-asciidoc-manpages 03fb4ad 06linuxcnc 04. branch deleted * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=03fb4ad
10:52 AM KGB-linuxcnc: 03Sebastian Kuzminsky 05master bb5e8a4 06linuxcnc 10src/hal/drivers/mesa-hostmot2/pins.c 10src/hal/drivers/mesa-hostmot2/sserial.c 10src/hal/drivers/mesa-hostmot2/sserial.h Merge remote-tracking branch 'origin/2.7' * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=bb5e8a4
11:06 AM skunkworks: seb_kuzminsky, seems they should have removed the sticker
11:12 AM jepler: the fan seems right
11:12 AM jepler: it forces air down, and it leaves at the left and right of the heatsink
11:13 AM archivist: with the fan sucking in the sides and out the top, you can brush when powered :)
12:41 PM skunkworks: http://electronicsam.com/images/matsuura/20170712_120923.jpg
12:41 PM skunkworks: pretty good shot of jitter
12:41 PM skunkworks: that is a 30us base period using parport reset
12:42 PM skunkworks: 5us time base
01:33 PM skunkworks: http://electronicsam.com/images/matsuura/20170712_130107.jpg
01:33 PM skunkworks: with a 5us reset (5000)
01:33 PM skunkworks: scope isn't calibrated very good. the timebase is off - that is with an adjustment to make the 1khz read close
01:34 PM skunkworks: now it looks like 30usISH base period
01:44 PM jepler: skunkworks: nice way to visualize it, though you wouldn't see a truly long delay that way
01:44 PM skunkworks: right - more of an average
01:45 PM skunkworks: 30khz stepgen from the printer port would be just fine for a lot of people.
01:45 PM skunkworks: (although they should all buy mesa...) :)
02:03 PM mozmck: heh! check out the proposed fix for this systemd problem (attached to comment 10) https://bugzilla.redhat.com/show_bug.cgi?id=1469670#c10
02:15 PM jepler: if(!strcmp(userid, "0day")) { ... }
02:15 PM jepler: oh wait different problem
04:56 PM * jepler forkbombs his latency-histogram just for fun
04:57 PM jepler: having read this article about a Windows performance bug when many programs exit at the same time
04:57 PM jepler: it does seem to have pushed my 10minute latency all the way up to 12us..
04:57 PM jepler: load average: 1992
04:59 PM jepler: though it has provoked a problem where latency-histogram has stopped updating(?) I can see from 'halcmd show' that latency-max is 12078 and servo max_time is 33015
05:00 PM jepler: ah there it finally updated. maybe a problem with my remoting software (xpra) was involved in the visual freezing too
05:14 PM -!- #linuxcnc-devel mode set to +v by ChanServ
05:26 PM skunkworks: forkbombs
05:46 PM seb_kuzminsky: borkfombs
08:18 PM MrHindsight is now known as CaptHindsight