#linuxcnc-devel | Logs for 2016-04-19

[07:56:16] <jthornton> I installed the rtai kernel in linuxmint like I've done a bunch of times but when I boot up to it I get a garbage screen with a bunch of vertical lines. I also get during boot up PXE-E01: PCI: Vendor and Device ID's do not match... not having much fun with the BP
[08:29:15] <JT-Shop> on the Dell if I plug the DB25 cable into the 5i25 the PC won't boot up
[08:33:11] <JT-Shop> next item up for bid is the rt-preempt hard drive and the ethernet thingy
[08:42:52] <JT-Shop> hmm restoring the bios to defaults got rid of that error message so I must have turned something off that I should not have
[08:48:06] <skunkworks> zlog,
[09:02:32] <JT-Shop> do you even worry about latency with rt-preempt?
[09:09:56] <pcw_home> if plugging the 5I25 cable into the PC makes the PC not boot, I would bet you have 5V power problem
[09:10:28] <pcw_home> most likely feeding external 5V power back onto the PC
[09:11:47] <JT-Shop> like a jumper wrong?
[09:11:59] <pcw_home> yeah
[09:13:08] <pcw_home> PC power supplies wont start if you leak external power in to their power rails
[09:14:12] <skunkworks> I try for latency <100us
[09:17:19] <JT-Shop> yep the jumpers were set wrong
[09:17:24] <pcw_home> On RT Preempt (and our Ethernet or other cards with DPLL firmware installed)
[09:17:25] <pcw_home> you mainly need to make sure the servo period tic can be met without more than
[09:17:27] <pcw_home> say a 1/2 period delay (say 500 us at 1 MS thread period) But if you have 500 usec latencies
[09:17:28] <pcw_home> you may well have much worse ones occasionally so skunkworks advice is good
[09:20:56] <skunkworks> over the weekend the hp i5 had a realtime delay
[09:21:20] <skunkworks> So not a perfect machine
[09:21:29] <pcw_home> what thread rate?
[09:21:35] <skunkworks> 1khz
[09:22:22] <pcw_home> how bad a delay?
[09:22:55] <skunkworks> I forgot to look
[09:22:58] <skunkworks> sorry
[09:23:33] <pcw_home> The h97 pro4 has been flawless at 4 KHz for about a year uptime
[09:23:41] <skunkworks> cool
[09:24:45] <pcw_home> my $40 core Duo HP is also flawless but 2 KHz max, it will run a week or so at 4 KHz and then get an error (300 usec or so)
[09:26:19] <pcw_home> I think its related to the likely-hood of a Ethernet read delay happening at the same time as a motion-command-handler delay
[09:27:04] <JT-Shop> sweet, I'm back in business and the touch screen works
[09:27:48] <skunkworks> I will run it again - oh - it was 2khz
[09:28:14] <JT-Shop> and the touch screen looks good, it was fuzzy with the d525
[09:29:20] <pcw_home> I dont think these are much different than the problems that I see in Atom D525s
[09:29:22] <pcw_home> (you cannot run a servo thread much faster than 1 KHz on these despite the fact that they show 10 usec range latencies)
[09:29:26] <skunkworks> The j1900 that we are using in the matsuura has been left on for weeks without a realtime delay
[09:30:51] <pcw_home> to see if you are close to the edge, watch servo-thread.tmax
[09:31:00] * JT-Shop thinks he will go dig in the dirt to celebrate
[09:31:18] <JT-Shop> before the rains get here
[09:32:07] <pcw_home> the off lease business class Dells and HPs are nice in that they seem much better made than generic clone hardware
[09:33:39] <pcw_home> Already did my digging for the week (rototilling the garden area)
[09:35:36] <skunkworks> set it to 1khz - letting it run
[09:36:39] <skunkworks> most of the time (until the overrun the thread max is around 200us max)
[09:36:48] <pcw_home> watch servo-thread.tmax
[09:37:02] <skunkworks> that is what I do
[09:37:20] <skunkworks> I will let it run and see what it maxes at
[09:37:53] <skunkworks> we have been buying these 2nd gen i5's here for workstations. they seem really nice
[09:37:58] <skunkworks> about $200
[09:38:19] <skunkworks> 4gb ram, win 7pro, 1tb hd
[09:38:41] <skunkworks> small form factor and will do 2 screens stock
[09:39:17] <pcw_home> Yes, I bought one (HP SFF) to replace Kathys windows machine
[09:39:18] <pcw_home> its quite nice
[09:42:40] <skunkworks> they have a printer port and 2 serials :)
[09:42:56] <pcw_home> Don't think she would appreciate me commandeering it for a latency test :-)
[09:43:22] <skunkworks> heh
[14:47:50] <andypugh> Interesting. If you search Youtube for LinuxCNC videos uploaded in the last week the answer is 94. That’s a lot, and suggests that we may have more than 100 users :-)
[14:58:57] <cradek> I do wonder what order of magnitude our number of users is
[15:00:55] <cradek> 7800 forum users
[15:01:13] <andypugh> 10^4 then
[15:01:18] <cradek> so probably either 10,000 or 100,000
[15:02:45] <andypugh> > 31,622 would be 10^5 :-)
[15:04:16] <cradek> I subscribe to the local city paper, and when I look at the "e edition" on my phone, I get ads covering the articles even after I log in.
[15:35:34] <PCW> Hmm looks like the OSADL people know something I don't about Preempt-RT kernel configs (not surprising)
[15:35:36] <PCW> Stole their kernel config file for a N3150 and it seems to fix the large Ethernet latencies I
[15:35:37] <PCW> was seeing on the Zotac N3150 based mini PC I'll give it a few days with you tube videos and see if its OK
[15:53:08] <mozmck> Hmm - you have a link for that config?
[15:57:44] <PCW> https://www.osadl.org/Profile-of-system-in-rack-4-slot-7.qa-profile-r4s7.0.html
[15:58:12] <mozmck> thanks!
[15:58:20] <PCW> pretty sure they have the config file for all their test rack systems
[15:58:49] <mozmck> Our little J1900 board seems to work well, but when we ran latency-histogram on it we got max latencies up to 151us
[16:00:05] <mozmck> The i3 box though is really bad, so I might need to look at some config settings.
[16:01:07] <PCW> latency is less important the the servo-thread.tmax
[16:01:24] <PCW> than the
[16:05:00] <PCW> I really have no idea why the H97 PRO4 is so good, without any fancy tweaks
[16:05:48] <mozmck> how do you monitor servo-thread.tmax? halshow?
[16:06:17] <PCW> yeah, hal watch
[16:06:50] <mozmck> Ok, I'll have to start looking at that. What is an acceptable max?
[16:08:20] <PCW> As a very rough guess, I would say 1/2 of the servo thread time
[16:08:53] <PCW> (after beating on it for a couple days)
[16:10:07] <PCW> (you can look at motion.servo.last-period for the full thread time in CPU clocks)
[16:10:30] <PCW> (those numbers are all in CPU clocks)
[16:10:47] <mozmck> Ok, so by 1/2 do you mean 1.5 times the servo thread time?
[16:11:27] <PCW> I mean 1/2 the total time (so 500 usec on a 1 ms thread)
[16:21:10] <PCW> if the Ethernet error hardening gets done, it should make occasional overruns / dropped packets relatively harmless
[16:23:56] <PCW> they seem quite rare even on "bad" systems
[16:23:58] <PCW> as long as you have no multi ms delays even a skipped cycle
[16:23:59] <PCW> _should_ not be terribly harmful (they are harmful now because of poor timeouts/ error handling)
[16:36:24] <KGB-linuxcnc> 03Jeff Epler 05master 4e4e964 06hostmot2-firmware 10build.py 03iselib.py iselib: factor out some stuff from build.py * 14http://git.linuxcnc.org/?p=hostmot2-firmware.git;a=commitdiff;h=4e4e964
[16:36:24] <KGB-linuxcnc> 03Jeff Epler 05master 930a61b 06hostmot2-firmware 10(6 files in 2 dirs) pin/xml generation: don't require ghdl * 14http://git.linuxcnc.org/?p=hostmot2-firmware.git;a=commitdiff;h=930a61b
[16:36:24] <KGB-linuxcnc> 03Jeff Epler 05master b25d9e0 06hostmot2-firmware 10iselib.py Drop support for pre-sourcing settings.sh * 14http://git.linuxcnc.org/?p=hostmot2-firmware.git;a=commitdiff;h=b25d9e0
[16:36:25] <KGB-linuxcnc> 03Jeff Epler 05master 200190b 06hostmot2-firmware 10README Update README * 14http://git.linuxcnc.org/?p=hostmot2-firmware.git;a=commitdiff;h=200190b
[16:36:29] <KGB-linuxcnc> 03Jeff Epler 05master d803bfc 06hostmot2-firmware 04README 03README.md 10debian/rules README: turn into markdown * 14http://git.linuxcnc.org/?p=hostmot2-firmware.git;a=commitdiff;h=d803bfc
[16:36:33] <KGB-linuxcnc> 03Jeff Epler 05master 0867f67 06hostmot2-firmware 10(805 files in 3 dirs) clean up top directory * 14http://git.linuxcnc.org/?p=hostmot2-firmware.git;a=commitdiff;h=0867f67
[16:36:36] <KGB-linuxcnc> 03Jeff Epler 05master c2e3e30 06hostmot2-firmware 10(18 files in 3 dirs) continue topdir cleanup: move python stuff to scripts/ * 14http://git.linuxcnc.org/?p=hostmot2-firmware.git;a=commitdiff;h=c2e3e30
[16:36:40] <KGB-linuxcnc> 03Jeff Epler 05master 7c6b6e0 06hostmot2-firmware 04DESCRIBE 10Makefile 03scripts/DESCRIBE cleanup: move DESCRIBE to scripts/ * 14http://git.linuxcnc.org/?p=hostmot2-firmware.git;a=commitdiff;h=7c6b6e0
[16:36:44] <KGB-linuxcnc> 03Jeff Epler 05master 9b2bd7d 06hostmot2-firmware 10scripts/DESCRIBE DESCRIBE: use git describe --dirty * 14http://git.linuxcnc.org/?p=hostmot2-firmware.git;a=commitdiff;h=9b2bd7d
[16:36:48] <KGB-linuxcnc> 03Jeff Epler 05master 9e2a383 06hostmot2-firmware 10Makefile error early if settings.sh was loaded * 14http://git.linuxcnc.org/?p=hostmot2-firmware.git;a=commitdiff;h=9e2a383
[16:51:27] <KGB-linuxcnc> 03Jeff Epler 05master 1daa3c6 06hostmot2-firmware 10Makefile clean topdir: create the firmware .mk file in fw/ * 14http://git.linuxcnc.org/?p=hostmot2-firmware.git;a=commitdiff;h=1daa3c6
[16:51:28] <KGB-linuxcnc> 03Jeff Epler 05master f514647 06hostmot2-firmware 10src/idrom_tools.vhd xml/PIN files: describe pktuart properly * 14http://git.linuxcnc.org/?p=hostmot2-firmware.git;a=commitdiff;h=f514647
[16:51:28] <KGB-linuxcnc> 03Jeff Epler 05master 3f144f2 06hostmot2-firmware 10src/idrom_tools.vhd remove outdated comment * 14http://git.linuxcnc.org/?p=hostmot2-firmware.git;a=commitdiff;h=3f144f2
[16:51:29] <KGB-linuxcnc> 03Boris Skegin 05master 5d091a0 06hostmot2-firmware 03src/PIN_PktUART1inst_34.vhd Example of one PktUART instance * 14http://git.linuxcnc.org/?p=hostmot2-firmware.git;a=commitdiff;h=5d091a0
[17:02:46] <hm2-buildmaster> build #49 of build is complete: Failure [4failed] Build details are at http://buildbot.linuxcnc.org/hm2-buildbot/builders/build/builds/49 blamelist: Jeff Epler <jepler@unpythonic.net>
[17:02:49] <hm2-buildmaster> build #50 of build is complete: Failure [4failed compile] Build details are at http://buildbot.linuxcnc.org/hm2-buildbot/builders/build/builds/50 blamelist: Jeff Epler <jepler@unpythonic.net>, Boris Skegin <boris.skegin.de@gmail.com>
[17:22:47] <jepler> hmm first build error looks like it's buildbot's fault, second is mine
[17:23:57] <KGB-linuxcnc> 03Jeff Epler 05master 786f003 06hostmot2-firmware 10Makefile build: must create fw/ before placing files there * 14http://git.linuxcnc.org/?p=hostmot2-firmware.git;a=commitdiff;h=786f003
[17:44:55] <hm2-buildmaster> build #51 of build is complete: Failure [4failed build-debian-package] Build details are at http://buildbot.linuxcnc.org/hm2-buildbot/builders/build/builds/51 blamelist: Jeff Epler <jepler@unpythonic.net>
[17:57:54] <KGB-linuxcnc> 03Jeff Epler 05master 9106c72 06hostmot2-firmware 10Makefile build: fix package build error * 14http://git.linuxcnc.org/?p=hostmot2-firmware.git;a=commitdiff;h=9106c72
[18:13:43] <JT-Shop> j1900 has +-37us latency running all afternoon
[18:19:30] <hm2-buildmaster> build #52 of build is complete: Failure [4failed] Build details are at http://buildbot.linuxcnc.org/hm2-buildbot/builders/build/builds/52 blamelist: Jeff Epler <jepler@unpythonic.net>
[18:22:27] <jepler> OK now back to a bit seb has to look into
[18:22:36] <jepler> seb_kuzm1nsky: ^^^ if you have a chance, can you look into this?
[18:28:03] <seb_kuzminsky> whoops, i know what that is
[18:28:08] * seb_kuzminsky goofed
[18:36:10] <jepler> seb_kuzminsky: thank you for looking into it
[18:36:23] <jepler> you can see from the above that I did my share of goofing too
[18:40:05] <seb_kuzminsky> it happens
[18:41:13] <seb_kuzminsky> ok, i poked the hm2 buildmaster config and triggered a rebuild of the most recent build, i'll check in on it from the hackspace in a few hours
[18:41:16] <seb_kuzminsky> bbl
[19:22:07] <dgarr> jepler: in case you want to look at it: http://www.panix.com/~dgarrett/stuff/0001-glcanon.py-is_lathe-is-a-function.patch
[20:13:42] <jepler> dgarr: right, thanks
[20:15:07] <jepler> dgarr: OK, you're right. thanks for waiting.
[20:15:29] <jepler> it's is_foam that is a function one place and an attribute another :)
[20:15:30] <jepler> :(
[20:51:38] <seb_kuzminsky> hm2-buildmaster: force build --branch=master build
[20:51:39] <hm2-buildmaster> build #54 forced
[20:51:39] <hm2-buildmaster> I'll give a shout when the build finishes
[21:13:12] <hm2-buildmaster> Hey! build build #54 is complete: Success [3build successful]
[21:13:12] <hm2-buildmaster> Build details are at http://buildbot.linuxcnc.org/hm2-buildbot/builders/build/builds/54
[21:16:02] <seb_kuzminsky> woo
[21:22:11] <KGB-linuxcnc> 03Dewey Garrett 05master b1fb19c 06linuxcnc 10src/hal/components/lowpass.comp lowpass.comp: man page help for setting gain pin * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=b1fb19c
[21:29:09] <jepler> seb_kuzminsky: thanks
[21:29:12] <jepler> *raises glass*
[21:31:44] <skunkworks> if the drives enables are enable-not - what are the chances of me being able to just hook the 7i48's enable to it?
[21:33:56] <skunkworks> (+ to the enable and - to common)
[21:48:14] <seb_kuzminsky> cheers
[22:46:12] <pcw_home> if they are active low enables, that should work
[22:48:05] <pcw_home> for the crazy AMC ground-to-disable drives, you need a relay (or a normally on OPTO=formB, turns out these do exist)
[23:04:06] <dgarr> pcw_home: see AMC ap note an-009.pdf regarding jumper J1 which is (sometimes?) available to alter the inhibit (it was on my 30A8's)
[23:04:09] <dgarr> http://search.freefind.com/find.html?oq=glossary&id=52826417&pageid=r&_charset_=UTF-8&bcd=%C3%B7&scs=1&query=j1+jumper&Find=Search&mode=ALL&search=all
[23:05:04] <pcw_home> Yeah I guess thats available on some, I had forgotten