Back
[06:05:57] <jepler> seb_kuzminsky: I think dewey rebased again
[06:53:42] <jepler> Aborting
[06:53:42] <jepler> could not detach HEAD
[06:53:44] <jepler> silly git
[06:55:08] * Tom_itx offers an axe
[07:05:36] <skunkworks> Nearly headless git
[07:30:14] <KGB-linuxcnc> 03Jeff Epler 05jepler/master/move-nongenerated-manpages cb1c7e3 06linuxcnc 10(122 files in 6 dirs) manpages: Remove useless dates from .TH lines * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=cb1c7e3
[07:30:14] <KGB-linuxcnc> 03Jeff Epler 05jepler/master/move-nongenerated-manpages a074e71 06linuxcnc 10src/hal/utils/halcompile.g halcompile: don't include useless date in generated manpages * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=a074e71
[07:30:14] <KGB-linuxcnc> 03Jeff Epler 05jepler/master/move-nongenerated-manpages 1087678 06linuxcnc 10(377 files in 8 dirs) docs: move all non-generated manpages to docs/src * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=1087678
[07:30:21] <jepler> sigh, about to have another episode of support fatigue reading the forum
[13:35:21] <PCW> Just tested Preemt-RT 4.4.6-rt13 on the G3258, same issues as 4.4.x on the core duo
[13:35:23] <PCW> bad latency (4.1.x is much better on both) or hard crashes (depending on exact version)
[13:37:41] <mozmck> PCW: thanks for testing and reporting - it is helpful
[13:38:28] <mozmck> PCW: Maybe you could mention it over on #linux-rt?
[17:09:14] <skunkworks> zlog
[17:09:44] <skunkworks> spindle rpm
http://electronicsam.com/images/matsuura/20160405_163500.jpg
[17:40:36] <jepler> sigh, why won't the people who want to "send a stream of positions" write their own open source CNC control, instead of wishing aloud that linuxcnc did not have the fundamental design decisions that it has
[17:41:00] <jepler> skunkworks: context?
[17:41:05] <jepler> (for your screenshot)
[17:43:30] <skunkworks> jepler: the matsuura gear tooth encoder hooked back into the 7i48 -> 7i80
[17:43:45] <skunkworks> http://electronicsam.com/images/matsuura/sensors.jpg
[17:43:46] <jepler> skunkworks: ah yes, thanks
[17:44:12] <skunkworks> 69 tooth - so 276 count
[17:47:42] <skunkworks> it doesn't pop around too much - I thought I would have to low-pass it
[19:14:39] <CaptHindsight> mozmck: there's 1 person in that channel #linux-rt
[19:17:03] <jepler> irc.oftc.net's where #linux-rt apparently lives, if this cursory googling is right
[19:17:33] <jepler> aka irc.debian.net
[20:58:15] <mozmck> ah, yes - it's on a different IRC server.
[21:01:22] <mozmck> irc.oftc.net is how I'm connected to it.
[21:13:51] <linuxcnc-build> build #4029 of 0000.checkin is complete: Failure [4failed] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/4029 blamelist: Jeff Epler <jepler@unpythonic.net>, Norbert Schechner <nieson@web.de>
[21:17:17] <linuxcnc-build> build #644 of 1520.rip-jessie-amd64 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1520.rip-jessie-amd64/builds/644 blamelist: Jeff Epler <jepler@unpythonic.net>, Chris Morley <chrisinnanaimo@hotmail.com>, John Thornton <bjt128@gmail.com>, Norbert Schechner <nieson@web.de>, Jan Mr?zek
[21:17:17] <linuxcnc-build> <email@honzamrazek.cz>
[21:17:36] <linuxcnc-build> build #644 of 1510.rip-jessie-rtpreempt-i386 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1510.rip-jessie-rtpreempt-i386/builds/644 blamelist: Jeff Epler <jepler@unpythonic.net>, Chris Morley <chrisinnanaimo@hotmail.com>, John Thornton <bjt128@gmail.com>, Norbert Schechner <nieson@web.de>,
[21:17:36] <linuxcnc-build> Jan Mr?zek <email@honzamrazek.cz>
[21:17:58] <linuxcnc-build> build #644 of 1530.rip-jessie-rtpreempt-amd64 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1530.rip-jessie-rtpreempt-amd64/builds/644 blamelist: Jeff Epler <jepler@unpythonic.net>, Chris Morley <chrisinnanaimo@hotmail.com>, John Thornton <bjt128@gmail.com>, Norbert Schechner
[21:17:58] <linuxcnc-build> <nieson@web.de>, Jan Mr?zek <email@honzamrazek.cz>
[21:18:16] <linuxcnc-build> build #644 of 1500.rip-jessie-i386 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1500.rip-jessie-i386/builds/644 blamelist: Jeff Epler <jepler@unpythonic.net>, Chris Morley <chrisinnanaimo@hotmail.com>, John Thornton <bjt128@gmail.com>, Norbert Schechner <nieson@web.de>, Jan Mr?zek
[21:18:16] <linuxcnc-build> <email@honzamrazek.cz>
[21:23:39] <linuxcnc-build> build #4018 of 1306.rip-precise-amd64 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1306.rip-precise-amd64/builds/4018 blamelist: Jeff Epler <jepler@unpythonic.net>, Chris Morley <chrisinnanaimo@hotmail.com>, John Thornton <bjt128@gmail.com>, Norbert Schechner <nieson@web.de>, Jan Mr?zek
[21:23:39] <linuxcnc-build> <email@honzamrazek.cz>
[21:23:58] <linuxcnc-build> build #2177 of 1403.rip-wheezy-amd64 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1403.rip-wheezy-amd64/builds/2177 blamelist: Jeff Epler <jepler@unpythonic.net>, Chris Morley <chrisinnanaimo@hotmail.com>, John Thornton <bjt128@gmail.com>, Norbert Schechner <nieson@web.de>, Jan Mr?zek
[21:23:58] <linuxcnc-build> <email@honzamrazek.cz>
[21:24:27] <linuxcnc-build> build #2176 of 1400.rip-wheezy-i386 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1400.rip-wheezy-i386/builds/2176 blamelist: Jeff Epler <jepler@unpythonic.net>, Chris Morley <chrisinnanaimo@hotmail.com>, John Thornton <bjt128@gmail.com>, Norbert Schechner <nieson@web.de>, Jan Mr?zek
[21:24:27] <linuxcnc-build> <email@honzamrazek.cz>
[21:24:51] <linuxcnc-build> build #1842 of 1401.rip-wheezy-rtai-i386 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1401.rip-wheezy-rtai-i386/builds/1842 blamelist: Jeff Epler <jepler@unpythonic.net>, Chris Morley <chrisinnanaimo@hotmail.com>, John Thornton <bjt128@gmail.com>, Norbert Schechner <nieson@web.de>, Jan
[21:24:51] <linuxcnc-build> Mr?zek <email@honzamrazek.cz>
[21:26:07] <linuxcnc-build> build #4016 of 1300.rip-precise-i386 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1300.rip-precise-i386/builds/4016 blamelist: Jeff Epler <jepler@unpythonic.net>, Chris Morley <chrisinnanaimo@hotmail.com>, John Thornton <bjt128@gmail.com>, Norbert Schechner <nieson@web.de>, Jan Mr?zek
[21:26:07] <linuxcnc-build> <email@honzamrazek.cz>
[21:26:47] <linuxcnc-build> build #2372 of 1404.rip-wheezy-rtpreempt-amd64 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1404.rip-wheezy-rtpreempt-amd64/builds/2372 blamelist: Jeff Epler <jepler@unpythonic.net>, Chris Morley <chrisinnanaimo@hotmail.com>, John Thornton <bjt128@gmail.com>, Norbert Schechner
[21:26:47] <linuxcnc-build> <nieson@web.de>, Jan Mr?zek <email@honzamrazek.cz>
[21:27:05] <linuxcnc-build> build #1687 of 1402.rip-wheezy-rtpreempt-i386 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1402.rip-wheezy-rtpreempt-i386/builds/1687 blamelist: Jeff Epler <jepler@unpythonic.net>, Chris Morley <chrisinnanaimo@hotmail.com>, John Thornton <bjt128@gmail.com>, Norbert Schechner
[21:27:05] <linuxcnc-build> <nieson@web.de>, Jan Mr?zek <email@honzamrazek.cz>
[21:27:32] <linuxcnc-build> build #3227 of 1301.rip-precise-rtai-i386 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1301.rip-precise-rtai-i386/builds/3227 blamelist: Jeff Epler <jepler@unpythonic.net>, Chris Morley <chrisinnanaimo@hotmail.com>, John Thornton <bjt128@gmail.com>, Norbert Schechner <nieson@web.de>, Jan
[21:27:32] <linuxcnc-build> Mr?zek <email@honzamrazek.cz>
[21:36:55] <linuxcnc-build> build #2210 of 1405.rip-wheezy-armhf is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1405.rip-wheezy-armhf/builds/2210 blamelist: Jeff Epler <jepler@unpythonic.net>, Chris Morley <chrisinnanaimo@hotmail.com>, John Thornton <bjt128@gmail.com>, Norbert Schechner <nieson@web.de>, Jan Mr?zek
[21:36:55] <linuxcnc-build> <email@honzamrazek.cz>
[21:36:56] <linuxcnc-build> build #4030 of 0000.checkin is complete: Failure [4failed] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/4030 blamelist: Jeff Epler <jepler@unpythonic.net>, Chris Morley <chrisinnanaimo@hotmail.com>, John Thornton <bjt128@gmail.com>, Norbert Schechner <nieson@web.de>, Jan Mr?zek <email@honzamrazek.cz>
[21:48:30] <linuxcnc-build> build #4019 of 1306.rip-precise-amd64 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1306.rip-precise-amd64/builds/4019 blamelist: Norbert Schechner <nieson@web.de>
[21:49:50] <linuxcnc-build> build #4017 of 1300.rip-precise-i386 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1300.rip-precise-i386/builds/4017 blamelist: Norbert Schechner <nieson@web.de>
[21:52:20] <linuxcnc-build> build #3228 of 1301.rip-precise-rtai-i386 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1301.rip-precise-rtai-i386/builds/3228 blamelist: Norbert Schechner <nieson@web.de>
[21:52:27] <linuxcnc-build> build #2178 of 1403.rip-wheezy-amd64 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1403.rip-wheezy-amd64/builds/2178 blamelist: Norbert Schechner <nieson@web.de>
[21:52:28] <linuxcnc-build> build #2177 of 1400.rip-wheezy-i386 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1400.rip-wheezy-i386/builds/2177 blamelist: Norbert Schechner <nieson@web.de>
[21:53:37] <linuxcnc-build> build #645 of 1530.rip-jessie-rtpreempt-amd64 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1530.rip-jessie-rtpreempt-amd64/builds/645 blamelist: Norbert Schechner <nieson@web.de>
[21:53:40] <linuxcnc-build> build #645 of 1510.rip-jessie-rtpreempt-i386 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1510.rip-jessie-rtpreempt-i386/builds/645 blamelist: Norbert Schechner <nieson@web.de>
[21:54:42] <linuxcnc-build> build #645 of 1520.rip-jessie-amd64 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1520.rip-jessie-amd64/builds/645 blamelist: Norbert Schechner <nieson@web.de>
[21:54:42] <linuxcnc-build> build #2373 of 1404.rip-wheezy-rtpreempt-amd64 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1404.rip-wheezy-rtpreempt-amd64/builds/2373 blamelist: Norbert Schechner <nieson@web.de>
[21:54:48] <linuxcnc-build> build #1688 of 1402.rip-wheezy-rtpreempt-i386 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1402.rip-wheezy-rtpreempt-i386/builds/1688 blamelist: Norbert Schechner <nieson@web.de>
[21:55:06] <linuxcnc-build> build #645 of 1500.rip-jessie-i386 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1500.rip-jessie-i386/builds/645 blamelist: Norbert Schechner <nieson@web.de>
[21:55:10] <linuxcnc-build> build #1843 of 1401.rip-wheezy-rtai-i386 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1401.rip-wheezy-rtai-i386/builds/1843 blamelist: Norbert Schechner <nieson@web.de>
[21:59:43] <linuxcnc-build> build #2211 of 1405.rip-wheezy-armhf is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1405.rip-wheezy-armhf/builds/2211 blamelist: Norbert Schechner <nieson@web.de>
[22:01:32] <linuxcnc-build> build #4031 of 0000.checkin is complete: Failure [4failed] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/4031 blamelist: Norbert Schechner <nieson@web.de>
[22:07:58] <KGB-linuxcnc> 03Jeff Epler 052.7 9d6a5bc 06linuxcnc 10src/emc/task/taskintf.cc interp: fix build error on Ubuntu 16.04 * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=9d6a5bc
[22:07:59] <KGB-linuxcnc> 03Jeff Epler 052.7 b8ca05c 06linuxcnc 10docs/src/gui/gladevcp.txt docs: fix build error * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=b8ca05c
[22:07:59] <KGB-linuxcnc> 03Jeff Epler 052.7 5954dcf 06linuxcnc 10docs/src/Submakefile build: make failure copying images an error * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=5954dcf
[22:15:22] <jepler> https://sourceware.org/bugzilla/show_bug.cgi?id=5781#c9
[22:15:30] <jepler> > I have also come across a very similar issue on i7 Intel platforms, please see bug 16531. Calls to cos can take around 0.15 ms, 1000 times their normal time, which is a serious problem for the real-time system we are developing.
[22:15:34] <jepler> ouch
[22:17:20] <jepler> I documented something similar at $DAY_JOB, and I think I talked about it here, where I found that under glibc's exp() the worst case could be 2000x the average case
[22:18:26] <jepler> .. certainly while doing circular interpolation I know motion will be calling sin and cos. I wonder if that is the cause of pcw's observation that there are very rare spikes in the runtime of motion
[22:19:01] <cradek> is he on x86_64?
[22:19:26] <pcw_home> Yeah, I wonder if that explains motion-command-handlers 60 usec spikes
[22:19:38] <jepler> that poster doesn't say. I think back when I was fiddling around I was using 64-bit code but I'm not sure
[22:20:50] <cradek> this bug says it's 64 bit only
[22:21:01] <jepler> yes, elsewhere in the thread it says that 32-bit x86 is using x87 for math
[22:22:00] <jepler> pcw_home: 64-bit userspace on your system?
[22:22:10] <pcw_home> no, 32
[22:22:32] <cradek> how rare is very rare?
[22:24:19] <jepler> http://paste.debian.net/424581/
[22:24:30] <pcw_home> motion-command handler median time on this machine is about 100 ns spikes a couple times a day are 60 usec
[22:24:42] <jepler> my findings ^^
[22:25:07] <jepler> I seem to have misplaced the program I ran to look for problem cases, which was just a random search and a print every time it found a new highest-time input..
[22:25:30] <jepler> plus something fancy to call the same exp() a bunch of times in a row to make sure I wasn't counting a task switch
[22:26:06] <pcw_home> let me plot it
[22:26:29] <jepler> aha
https://emergent.unpythonic.net/files/sandbox/mathbench.cc (compile with g++ -std=c++11 -O or so)
[22:27:55] <jepler> hm I see it doesn't even print the culprit values :-/ derp
[22:28:51] <jepler> anyway with -m64 the first printout (1 million trials) it already clocks some runs in the 2^13 range, while with -m32 it doesn't clock any above 2^8 within 1 minute (same core i7 system)
[22:31:23] <jepler> .. updated with version that prints last value seen in each bin ..
[22:42:27] <jepler> for 32-bit targets, I also tested sin, atan2, hypot and didn't find any outlier behavior
[22:42:54] <jepler> so no, if it's a 32-bit system this is probably not the explanation for occasional thread time spikes
[22:48:33] <jepler> I didn't grep to see what other libm functions are used, that's all guessing on my part
[22:57:18] <pcw_home> https://imagebin.ca/v/2cmST4sB12z8
[22:57:20] <pcw_home> the biggest spikes there are about 20 usec
[22:57:21] <pcw_home> but occasionally get to 60 usec
[23:03:19] <pcw_home> ( and the spikes seem to be dependent on the gcode being run )
[23:07:49] <pcw_home> https://imagebin.ca/v/2cmVnlYB6EAQ
[23:07:50] <pcw_home> big flurry of activity at start of motion
[23:20:15] <skunksleep> Calculating the path? (Look ahead)
[23:22:00] <pcw_home> maybe
[23:22:01] <pcw_home> I just dont see how you get to 60 usec (~180000 inst)
[23:22:41] <linuxcnc-build> build #1173 of 1903.clang-wheezy-amd64 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1903.clang-wheezy-amd64/builds/1173 blamelist: Jeff Epler <jepler@unpythonic.net>
[23:24:55] <skunksleep> You could set the look ahead lower for grins and see if it improves
[23:30:07] <pcw_home> Ill try disabling the new TP o see if that makes any difference
[23:30:21] <linuxcnc-build> build #1173 of 1902.clang-wheezy-rtai-i386 is complete: Failure [4failed compile] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1902.clang-wheezy-rtai-i386/builds/1173 blamelist: Jeff Epler <jepler@unpythonic.net>
[23:31:40] <skunksleep> I dont know how that works.. The trajectory planner in real-time plans the path every servo cycle. How does it know how much time it has? Or does it add only 1 segment every servo cycle (because it is unlikely you will have segments that take 1 servo cycle so you gain on the que)
[23:43:02] <pcw_home> the spikes do appear to be TP related (old TP seems to have much lower max times)
