#linuxcnc-devel | Logs for 2014-09-12

[01:09:46] <NTU> jepler: kbuild rtai undefined variables are back, not sure why
[09:40:47] <skunkworks> So - I grabbed one of the other motherboards that dad had (all the same asus model) and this one has been running all morning under 16us.
[09:40:59] <skunkworks> it does have a different processor (single core it seems)
[09:41:31] <skunkworks> running glxgears and streaming youtube video
[09:47:02] <skunkworks> over a wifi usb dongle
[10:50:03] <skunkworks> well it just jumped to 200us
[10:50:05] <skunkworks> darn
[10:51:34] <micges-dev> skunkworks: jitter?
[10:51:41] <skunkworks> yes
[10:53:00] <micges-dev> any specific setup or jus testing new mb?
[10:53:57] <skunkworks> well - this is hardware that is in our big mill and seems to run 10.04/rtai just fine.
[10:54:06] <skunkworks> testing with the debian hybrid iso.
[10:54:13] <micges-dev> ah ok
[10:56:01] <skunkworks> just trying to see if I can figure out why.
[10:57:03] <pcw_home> preemt-rt might do better :-)
[10:57:19] <skunkworks> have not gotten that far yet :)
[10:58:50] <pcw_home> preemt-rt may have lower peak latency then 200k but latency has higher spectral density
[10:59:21] <skunkworks> sure - but normally maxes at 70us or so from my testing
[10:59:25] <pcw_home> must be something busted to get 200 K though
[11:00:13] <skunkworks> well - a lot of the onboard stuff is nvidia which probably doesn't help
[11:00:26] <skunkworks> but I am running a ati video card
[11:00:58] <micges-dev> pcw_home: I got Asrock D1800 for tests
[11:01:49] <pcw_home> OK good that works reliably for me (though was a PITA to get it to boot linux the first time)
[11:01:55] <skunkworks> http://www.asus.com/Motherboards/M2N68AM_PLUS/specifications/
[11:06:03] <pcw_home> I learned some pid-stepper tricks:
[11:06:04] <pcw_home> 1: P gain ~= servo thread frequency is a good rule of thumb
[11:06:06] <pcw_home> 2. I can get acceptable performance down to 250 Hz (meet ferror limit of .0002" at 1200 IPM/200IPS^2)
[11:11:48] <micges-dev> I didn't tried it yet, machine was busy
[11:34:18] <jepler> this weekend I intend to see how far from working my little mill is (servos) and hopefully run the motors around with the u3
[11:35:08] <jepler> (I don't think I've run it for 2+ years at this point)
[11:39:12] <pcw_home> That sounds like fun
[11:39:45] <cradek> having working machinery causes projects that need working machinery to happen
[11:39:54] <skunkworks> onboard video - 200us with rt-preemt
[11:39:59] <skunkworks> +
[12:29:32] <memleak> RTAI + linuxcnc build scripts (source, no packages) part 1: http://dpaste.com/3MZ80VQ.txt part 2: http://dpaste.com/2YQPMKZ.txt part 3: http://dpaste.com/07M3HVX.txt May use wget && chmod a+x on bash scripts
[12:32:48] <memleak> kernel is part 1, rtai userspace (/usr/realtime) is part 2, linuxcnc is part 3, skunkworks the default kernel config should boot on your system, its not quite as big as the debian / ubuntu config but should at least mount rootfs on most machines and distros
[12:33:21] <zq> hm
[12:33:29] <memleak> all networking drivers built as modules, all common SATA and PATA drivers built in, and full USB support
[12:33:59] <memleak> i had some trouble using the stock ubuntu and debian kernel configs due to no initrd
[12:34:37] <zq> how useful do you guys think an http version of the rs274 interpreter would be?
[12:34:40] <memleak> once seb_kuzminsky posts his rtai package building instructions (if he already has i cant find them) ill build .debs using my shiny new RTAI tree
[12:35:19] <memleak> (with full blown kernel + initial ramfs)
[12:36:23] <memleak> things that are missing is full I2C I20 and some other busses
[12:36:38] <memleak> drm modesetting works + gpu accel, etc.
[12:42:04] <memleak> you may need to uninstall the linuxcnc package first in order to use the rip-environment binaries, you can fall back to /usr/bin by closing and opening a new terminal window and installing linuxcnc package again via synaptic
[12:43:35] <memleak> one thing i can do too is create a tarball to decompress to / and overwrite system-wide linuxcnc + RTAI but thats not safe.
[13:32:54] <jepler> memeak: no, don't do that. geez wtf
[13:33:15] <jepler> memleak:
[13:50:21] <skunkworks> rt-preempt has been running for a few hours at 60us...
[13:51:11] <skunkworks> tab complete getting lazy?
[13:53:32] <NTU> heh jepler..
[13:55:52] <NTU> ./configure --prefix=/usr && make PREFIX=output install && tar output to file && tar zxf rtai_and_linuxcnc.tar.gz -C /
[13:56:02] <NTU> its not _THAT_ deadly
[13:56:19] <NTU> simple apt-get reinstall <package> reverts it (mostly.)
[13:57:06] <cradek> that's fine to do to your own system, but unfortunately people will often do the thing you tell them, so be sure to tell people only good ideas
[13:57:28] <NTU> thats how gentoo works..
[13:57:46] <NTU> and archlinux
[13:58:21] <NTU> arch is configure to /usr install to folder and compress to package
[13:58:24] <cradek> whether gentoo does a thing and whether that thing is a good idea to advise are different
[13:59:16] <NTU> but yet again archlinux is unstable as hell.. so you're probably right come to think of it
[14:17:22] <PCW> micges-dev: hm2-ser card name fixed
[14:17:52] <micges-dev> thanks
[14:18:09] <PCW> (well all of space 7 fixed it had a off-by-one error)
[14:18:14] <skunkworks> rt_preempt seems to work as expected
[14:19:11] <KGB-linuxcnc> 03Norbert Schechner 052.6 ed090f8 06linuxcnc 10lib/python/gladevcp/tooledit_widget.py 10src/emc/usr_intf/gmoccapy/gmoccapy.glade 10src/emc/usr_intf/gmoccapy/gmoccapy.py 10src/emc/usr_intf/gmoccapy/release_notes.txt gmoccapy_1_2_2 - support now also matchbox-keyboard * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=ed090f8
[14:22:02] <PCW> fixed bit file
[14:22:04] <PCW> freeby.mesanet.com/7i90_ser_justio.bit
[14:22:35] <micges-dev> got it
[14:27:06] <micges-dev> it's working
[14:27:20] <micges-dev> even space 6 data are correct
[14:27:43] <micges-dev> packets received: all 52, good 51, bad 0
[14:27:43] <micges-dev> packets sended: all 51, good 51, bad 0
[14:41:58] <PCW> I'll eventually fix the hm2-ser code so it gets the name from the IDROM
[14:44:08] <micges-dev> good idea
[17:52:18] <jepler> micges-dev: verifying ROM still worked, I assume it's fine
[17:53:07] <micges-dev> that's good, thanks
[18:57:50] -hobana.freenode.net:#linuxcnc-devel- *** Notice -- TS for #linuxcnc-devel changed from 1410562219 to 1326726487
[18:57:50] hobana.freenode.net changed topic of #linuxcnc-devel to: http://linuxcnc.org | Latest release: 2.6.3
[21:05:47] <jepler> bounce bounce