Back
[07:43:47] <skunkworks> hmm - how do I get lstopo?
[07:44:48] <skunkworks> it isn't in /bin
[07:46:58] <skunkworks> ah - hwloc
[07:51:35] <skunkworks> seb_kuzminsky,
http://pastebin.com/XbN0zZni
[07:52:07] <skunkworks> makes sense as L2 is shared between 2 cores.
[07:52:54] <skunkworks> I would probably get the same results with isolating 0,1 vs 2,3
[07:53:04] <skunkworks> overnight 12us
[08:32:59] <skunkworks> I wonder how the rt_preempt would preform using those settings.
[08:33:40] <skunkworks> or perform even...
[09:04:53] <seb_kuzminsky> skunkworks: cool, thanks
[09:05:09] <skunkworks> Am I understanding it right?
[09:09:21] <seb_kuzminsky> your statement matches my understanding as well
[09:09:31] <jepler> no, I don't think you would get the same result isolating 0,1
[09:09:43] <jepler> because linuxcnc will run realtime threads on the top CPU number all the time
[09:09:57] <jepler> linuxcnc (rtapi) doesn't look at what CPUs are subject to isolcpus
[09:10:03] <seb_kuzminsky> ah, right
[09:10:10] <seb_kuzminsky> well, i agree with this statement: < skunkworks> makes sense as L2 is shared between 2 cores.
[09:10:31] <seb_kuzminsky> vcpus 0 & 1 share one L2, and vcpus 2 & 3 share another L2
[09:12:42] <jepler> there's a libhwloc. somebody should build it into a program that tells you what isolcpus= setting to use
[09:15:41] <skunkworks> ah. would these setting have any effect on rt_preempt?
[09:16:16] <jepler> the only other thing I have to mention is that I swear I saw one system where the pairs weren't 2*k and 2*k+1 but k and k+n according to /proc/cpuinfo
[09:16:30] <jepler> skunkworks: yes, isolcpus= seems to also affect rt_preempt performance
[09:16:37] <jepler> I use isolcpus=3 on my odroid
[09:17:37] <skunkworks> I will try it
[09:22:30] <jepler> looks lke I tried to write this script once before
http://emergent.unpythonic.net/files/sandbox/isol.sh
[09:24:49] <jepler> http://paste.debian.net/130552/
[09:25:17] <skunkworks> wow
[09:26:05] <jepler> and on my odroid:
http://paste.debian.net/130554/
[09:27:07] <skunkworks> that could be part of the latency test.. could even add it to the kernel line.. :)
[09:27:56] * skunkworks waits for 'patches welcome'... :)
[09:29:12] <jepler> skunkworks: do you have a github account?
[09:31:45] <jepler> *crickets*
[09:32:48] <archivist> that is the sound of him typing on his keyboard
[09:33:39] <jepler> I was thinking about the thread "Probably a documentation error". Anybody with a github account could prepare a patch to the documentation, just within the github website
[09:33:57] <jepler> .. skunkworks would be a good guinea pig for this process
[09:34:19] <jepler> you don't even need to 'git clone' or have git on your computer to do doc changes
[09:34:39] <jepler> if we could convince users to fix the docs for us, then linuxcnc documentation would be perfect
[09:34:42] <jepler> right?
[09:35:16] <archivist> I usually mention to JT when I spot something (cheating)
[09:35:32] <jepler> yeah, though he can only do so much
[09:41:17] <skunkworks> I could play with that in the future.
[09:41:28] <skunkworks> I have to make a note to future self
[09:43:51] <jepler> I'm not sure how well this idea actually works in practice, so it would be interesting to find out
[09:44:08] <jepler> if it worked well I'd totally push it in every thread which is ultimately about a documentation issue
[10:00:49] <skunkworks> ok - readline-common is installed..
http://pastebin.com/iGphVtjw
[10:01:58] <bjmorel_work> I think you need readline-dev
[10:03:45] <skunkworks> it doesn't exist in my world...
[10:04:38] <skunkworks> ah - maybe libreadline
[10:04:49] <seb_kuzminsky> it's spelled libreadline-dev
[10:10:34] <skunkworks> got it.
[10:24:29] <KGB-linuxcnc> 03Sebastian Kuzminsky 05liblinuxcnc-ui 260e5e6 06linuxcnc 10src/emc/usr_intf/emcrsh.cc linuxcncrsh: use lui for setting task mode (manual/auto/mdi) * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=260e5e6
[10:24:29] <KGB-linuxcnc> 03Sebastian Kuzminsky 05liblinuxcnc-ui 5cb466f 06linuxcnc 10src/liblinuxcnc-ui/Submakefile 03src/liblinuxcnc-ui/coolant.cc 10src/liblinuxcnc-ui/linuxcnc-ui.h lui: add mist/flood on/off * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=5cb466f
[10:24:29] <KGB-linuxcnc> 03Sebastian Kuzminsky 05liblinuxcnc-ui 4ad18a6 06linuxcnc 10src/emc/usr_intf/emcrsh.cc linuxcncrsh: use lui for mist/flood on/off * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=4ad18a6
[10:24:31] <KGB-linuxcnc> 03Sebastian Kuzminsky 05liblinuxcnc-ui fcef840 06linuxcnc 10src/emc/usr_intf/xemc.cc xemc: use lui for flood/mist on/off * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=fcef840
[10:33:01] <skunkworks> Umm - wow...
http://electronicsam.com/images/KandT/testing/Screenshot%20-%2011062014%20-%2006:11:42%20PM.png
[10:33:24] <skunkworks> idle=poll and isolcpus=2,3
[10:35:18] <seb_kuzminsky> skunkworks: wow!
[10:35:51] <seb_kuzminsky> heh, "introduction to linuxcnc", "electronics overview for mill conversion with linuxcnc", "making a tardis"
[10:36:49] <skunkworks> odd
[10:37:20] <jepler> obviously a steep learning curve there
[10:37:35] <skunkworks> oh - the video.. (it is just long so I use it for testing..)
[10:37:53] <archivist> some very poor "related" on many of the sites these days
[10:39:44] <skunkworks> about 50 frames a second on the full screen glxgears
[10:41:15] <skunkworks> up to 10us
[10:44:53] <seb_kuzminsky> skunkworks: is this your j1900?
[10:45:05] <skunkworks> yes
[10:45:11] <skunkworks> I also have a J1800
[10:45:16] <kb8wmc> I was directed over here while looking to download camview-emc from psha's site, it was suggested that someone had mirrored psha's site...can anyone here confirm and direct me to such a mirror
[10:45:50] <skunkworks> I think CaptHindsight, was talking about it...
[10:46:07] <kb8wmc> ok
[10:46:23] <skunkworks> (I noticed psha's site was down today..)
[10:47:39] <kb8wmc> yes and it must have been down for the last couple days as I have attempted to reach it for d/l for last two days without success
[10:48:21] <archivist> has archive.org got a copy too
[10:48:43] <kb8wmc> don't know, did not look there
[10:51:42] <alex_joni> they don't mirror attachments though
[10:52:49] <kb8wmc> I did not find it there at archive.org
[11:10:53] <skunkworks> still 10us
[11:31:28] <skunkworks> pcw, you should try that ^
[11:59:21] <skunkworks> I will let it run over night
[12:00:12] <skunkworks> so - for better realtime performance - you sould look for cache layouts that are not shared... - interesting
[12:00:17] <skunkworks> bbl
[12:14:11] <KGB-linuxcnc> 03John Thornton 052.6 29bbf2d 06linuxcnc 10docs/src/gcode/gcode.txt Docs: add text describing example code * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=29bbf2d
[12:45:10] <skunkworks> the_wench: where is your log?
[13:05:27] <PCW> skunkworks: i tried isolcpus=1 and idle-poll, I dont see much difference (latency test results look good but actual jitter is still around 30 usec)
[13:06:41] <seb_kuzminsky> do you mean "idle=poll"?
[13:06:48] <PCW> yes
[13:07:50] <PCW> I really dont think the latency test means much except for noticing gross problems
[13:09:59] <PCW> now 30 usec is a lot better than the D525s ~75 usec latency
[13:12:05] <PCW> (which explains why you cant use more than about a 1.5 KHz servo thread on a a D525 without getting RT errors)
[13:17:01] <skunkworks> PCW, with the j1900 and idle=poll isolcups=2,3 I was getting a max of 10us (but will run it overnight) the best I was getting before was around 70us
[13:18:08] <PCW> Maybe quad core difference, I always seem to get around 10-11 usec worst case on the j1800
[13:18:22] <PCW> (under RTAI)
[13:21:06] <PCW> let me go back to no idle=poll and no isolcpus=1 and run the test for an hour or so
[13:23:37] <skunkworks> that was rt_preempt...
[13:37:12] <PCW> ok that _does_ improve the preemt-rt latency
[13:38:29] <PCW> at least the test (I was getting about 50 usec is a few minutes with youtube videos running, now its about 15 usec)
[13:41:52] <skunkworks> on what system?
[13:45:59] <PCW> Gigabyte J1800
[13:46:39] <skunkworks> the j1900 shares cache between pairs of cores. 0,1 and 2,3
[13:47:58] <PCW> but... while the idle-poll and isolcpus improve the latency test, they do not seem to affect the actual max jitter (as measured with hardware)
[13:47:59] <PCW> it mainly makes the system slower
[13:48:30] <skunkworks> heh - well that sucks
[13:50:29] <PCW> well it no different than before, code that runs in the servo or base thread may be delayed by random times that do not show up in the latency test
[13:51:14] <bjmorel_work> PCW: did you try the isolcpus on a 4-core? I could see it not having an effect on the 2-Core J1800 because of the shared cache, but it might work ok on the 4-core.
[13:51:44] <PCW> It has an effect (but only on the latency test)
[13:52:39] <PCW> I dont have a 4 core CPU that I can use to test ATM
[13:56:29] <skunkworks> I can play with it tomorrow..
[13:57:44] <skunkworks> for some reason - I thought you had the 1900 also
[13:59:18] <PCW> No, but I may buy this one for testing:
[13:59:19] <PCW> http://www.newegg.com/Product/Product.aspx?Item=N82E16813157565
[14:00:42] <skunkworks> the gigabit j1900's have 2 nics.
[14:06:04] <PCW> Yeah but i wanted a different brand to try
[14:06:43] <PCW> (I do have Asrock J1900 at home but for HTPC)
[14:11:35] <skunkworks> ah
[14:16:38] <CaptHindsight> isolcpus= behaves it self in the new RTAI tree, in fact you don't even need to use it anymore since the scheduler is fixed
[14:54:26] <kwallace1> In my opinion, that's a pretty darn nice board above. 2 x PCI, PCIe, a parallel port, serial port, HDMI, good latency and low cost (not too much more than a Beagle). I wonder why PCI (or parallel port) hasn't been killed off yet. I'm not a fan of Intel graphics, I wonder how well OpenGL gets along with this board.
[14:55:29] <kwallace1> Fan-less too?
[14:55:48] <PCW> fanless and graphics seem fine
[14:56:07] <kwallace1> Cool
[14:56:23] <PCW> most of the new J1800,J1900 MBs have a parallel port
[14:56:37] <PCW> (often on a header)
[14:59:27] <kwallace1> Offhand, is there a write-up on a hardware latency test?
[15:02:59] <PCW> Im cheating and using the hm2dpll
[15:38:36] <CaptHindsight> kwallace1: like send some data to PCIe device, crunch some numbers, respond to an interrupt, receive some data over PCIe, rinse repeat?
[15:43:36] <CaptHindsight> kwallace1:
http://www.ise.pw.edu.pl/~rrom/SPIE/SPIE8454-WILGA2012/SPIE8454-WI12-cd/DATA/8454_60.PDF
[15:43:42] <kwallace1> Oops, I recall that there was a setup with two PCs. The PC being tested sent out a signal, the other measured the latency. Thinking more, I'm not sure if the current latency test is effectively the same or not.
[15:44:24] <CaptHindsight> the linuxcnc latency doesn't test with actual IO
[15:44:49] <CaptHindsight> it's just a tool to see what ballpark the system is in
[15:48:34] <kwallace1> I was curious how involved measuring actual I/O latency was. If it were not too hard, I'd like to set a PC aside to do testing.
[15:50:03] <kwallace1> Thank you for the link above.
[15:57:28] <CaptHindsight> did anyone ever just use a PCI/e fpga card and use that to test latency? use HM2 as is, but add a block to generate interrupts with and independent clock
[15:57:39] <CaptHindsight> and/an
[17:16:04] <KGB-linuxcnc> 03Sebastian Kuzminsky 05liblinuxcnc-ui 3602ab1 06linuxcnc 10src/emc/usr_intf/emcsh.cc emcsh: use lui instead of shcom for coolant * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=3602ab1
[17:16:04] <KGB-linuxcnc> 03Sebastian Kuzminsky 05liblinuxcnc-ui 375d408 06linuxcnc 10src/emc/usr_intf/halui.cc halui: use lui for coolant * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=375d408
[17:34:17] <linuxcnc-build_> build #780 of 1400.rip-wheezy-i386 is complete: Failure [4failed compile runtests] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1400.rip-wheezy-i386/builds/780 blamelist: Sebastian Kuzminsky <seb@highlab.com>
[17:51:17] <linuxcnc-build_> build #2634 of 0000.checkin is complete: Failure [4failed] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/2634 blamelist: Sebastian Kuzminsky <seb@highlab.com>
[18:13:49] <seb_kuzminsky> hmm
[18:19:50] <KGB-linuxcnc> 03Sebastian Kuzminsky 05liblinuxcnc-ui 8c9c2e7 06linuxcnc 10tests/halui/jogging/halui.ini 10tests/halui/mdi/halui.ini 10tests/liblinuxcnc-ui/lui-test.c halui tests: turn on debug output * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=8c9c2e7
[18:35:19] <linuxcnc-build_> build #781 of 1403.rip-wheezy-amd64 is complete: Failure [4failed compile runtests] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1403.rip-wheezy-amd64/builds/781 blamelist: Sebastian Kuzminsky <seb@highlab.com>
[18:35:50] <linuxcnc-build_> build #972 of 1404.rip-wheezy-rtpreempt-amd64 is complete: Failure [4failed compile runtests] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1404.rip-wheezy-rtpreempt-amd64/builds/972 blamelist: Sebastian Kuzminsky <seb@highlab.com>
[18:36:21] <linuxcnc-build_> build #781 of 1400.rip-wheezy-i386 is complete: Failure [4failed compile runtests] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1400.rip-wheezy-i386/builds/781 blamelist: Sebastian Kuzminsky <seb@highlab.com>
[18:37:38] <linuxcnc-build_> build #292 of 1402.rip-wheezy-rtpreempt-i386 is complete: Failure [4failed compile runtests] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1402.rip-wheezy-rtpreempt-i386/builds/292 blamelist: Sebastian Kuzminsky <seb@highlab.com>
[18:38:14] <linuxcnc-build_> build #439 of 1401.rip-wheezy-rtai-i386 is complete: Failure [4failed compile runtests] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1401.rip-wheezy-rtai-i386/builds/439 blamelist: Sebastian Kuzminsky <seb@highlab.com>
[18:44:35] <linuxcnc-build_> build #811 of 1405.rip-wheezy-armhf is complete: Failure [4failed compile runtests] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1405.rip-wheezy-armhf/builds/811 blamelist: Sebastian Kuzminsky <seb@highlab.com>
[18:46:44] <linuxcnc-build_> build #2623 of 1300.rip-precise-i386 is complete: Failure [4failed compile runtests] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1300.rip-precise-i386/builds/2623 blamelist: Sebastian Kuzminsky <seb@highlab.com>
[18:47:44] <linuxcnc-build_> build #2625 of 1306.rip-precise-amd64 is complete: Failure [4failed compile runtests] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1306.rip-precise-amd64/builds/2625 blamelist: Sebastian Kuzminsky <seb@highlab.com>
[18:48:54] <linuxcnc-build_> build #2624 of 1200.rip-lucid-i386 is complete: Failure [4failed compile runtests] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1200.rip-lucid-i386/builds/2624 blamelist: Sebastian Kuzminsky <seb@highlab.com>
[18:49:05] <linuxcnc-build_> build #1825 of 1301.rip-precise-rtai-i386 is complete: Failure [4failed compile runtests] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1301.rip-precise-rtai-i386/builds/1825 blamelist: Sebastian Kuzminsky <seb@highlab.com>
[18:50:02] <linuxcnc-build_> build #2624 of 1202.rip-lucid-amd64 is complete: Failure [4failed compile runtests] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1202.rip-lucid-amd64/builds/2624 blamelist: Sebastian Kuzminsky <seb@highlab.com>
[18:53:18] <linuxcnc-build_> build #2624 of 1201.rip-lucid-rtai-i386 is complete: Failure [4failed compile runtests] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1201.rip-lucid-rtai-i386/builds/2624 blamelist: Sebastian Kuzminsky <seb@highlab.com>
[18:53:19] <linuxcnc-build_> build #2635 of 0000.checkin is complete: Failure [4failed] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/2635 blamelist: Sebastian Kuzminsky <seb@highlab.com>
[21:50:46] <seb_kuzminsky> oh man
[21:53:15] <seb_kuzminsky> oh, my bad
[21:53:26] <seb_kuzminsky> i shouldn't check in my deliberately broken test of the test :-/
[21:58:20] <KGB-linuxcnc> 03Sebastian Kuzminsky 05liblinuxcnc-ui f24c5a7 06linuxcnc 10tests/liblinuxcnc-ui/lui-test.c lui-test: revert the bogus previous commit * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=f24c5a7
[22:02:21] <skunkworks> I was | | close to testing a basic image (circle) registration setup but the supereye camera didn't like the stepper noise and would disconnect after a random amount of time...
[22:02:58] <KGB-linuxcnc> 03Sebastian Kuzminsky 05liblinuxcnc-ui 8fd1f63 06linuxcnc 10tests/liblinuxcnc-ui/lui-test.c lui-test: test flood & mist coolant * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=8fd1f63
[22:07:22] <kb8wmc> skunkworks: is there any chance that you would know where I can obtain camview-emc and related files as psha's link seems to still be down?
[22:15:01] <seb_kuzminsky> jepler: luigetter.py claims lube and lube_level are unused, but it looks like ioControl.cc sets them and several guis access them (halui, shcom, emcrsh, etc), what's up with that?
[22:15:54] <skunkworks> kb8wmc: sorry. I am playing with opencv.. (very loosely as I am pretty remedial..)
[22:16:42] <kb8wmc> ok...tnx
[22:39:30] <CaptHindsight> kb8wmc: which files do you need?
[22:48:44] <kb8wmc> hey there....I need camview-emc and the related utility files
[22:49:40] <kb8wmc> CaptHindsight: sorry I was away for a while
[22:54:45] <kb8wmc> be back in a few....
[23:41:46] <CaptHindsight> kb8wmc: I think I have the deb for testing (squeeze)
[23:42:21] <CaptHindsight> and halio.c
[23:43:11] <CaptHindsight> and his emcfiles.zip
