#linuxcnc-devel | Logs for 2013-07-04

[09:42:41] <jepler> I installed debian 7 with linux-image-3.2.0-4-rt-686-pae on my test machine which is good for rtai under hardy and lucid. Unfortunately, the cyclictest results are terrible (max 2827 after just a few seconds)
[09:43:10] <jepler> so I guess I don't have a machine where I can usefully participate in debugging any linuxcnc-specific preempt-rt problems :-/
[09:45:39] <jepler> hmmm, results are better with -y rr
[09:46:33] <jepler> 221 after 30s
[09:48:56] <jepler> 394
[09:49:25] <jepler> 583
[09:49:30] <jepler> :-/ so still probably not usable
[09:55:24] <CaptHindsight> I guess they are writing some kernel debugger to trace all the problems with RT-Preempt and RTAI
[10:03:44] <CaptHindsight> http://www.3ders.org/articles/20130703-just-released-fedora-19-comes-with-3d-printing-software.html
[10:04:17] <CaptHindsight> heh, is somebody working on Linuxcnc for fedora?
[11:00:42] <jepler> 'latency-test 1ms 1ms' results actually seem to be better than cyclictest results
[11:00:53] <jepler> 27us after 30s
[11:14:50] <IchGuckLive> hi is the side of the namechange issue online somewere
[12:05:13] <jepler> linuxcnc latency-test on preempt-rt is max jitter 56us after 1 hour
[12:05:21] <jepler> latency-test 1ms 1ms
[12:05:27] <jepler> that's actually quite usable
[12:06:03] <jepler> this is rtos-integration-preview3-merged-into-master from git.l.o + RTLD_NOW fix
[12:06:34] <jepler> mhaberler: hostmot2 doesn't actually load in this version, though it compiles. is there a different branch I should look at where hostmot2 works in userspace, or should I contribute a patch to fix it?
[12:23:03] <jepler> hostmot2: dlopen: /usr/src/linuxcnc/rtlib/hostmot2.so: undefined symbol: kzalloc
[13:02:22] <jepler> hm, got a 1ms jitter :-/
[13:46:14] <mhaberler> jepler: hostmot2 load error - some missing symbol?
[13:47:11] <zultron> CaptHindsight, I saw the Fedora 3D printing announcement through another channel. Looks like they have a spin with tools to download/design/slice/preprocess/whatever, then they expect you to upload the result to the 3D printer controller, presumably an Arduino.
[13:47:23] <zultron> So no RT capabilities needed. :P
[13:47:28] <mhaberler> I've forward merged all rtos-integration* branches and related work into rtos-master-v0; as soon as John gives me the goahead this will be folded into hist ub work branch, and then that result will be the merge candidate
[13:49:54] <mhaberler> so if any patches rtos-master-v0 would be great but it really dont matter, since the differences to rtos-integration-preview3-merged-into-master are non RTAPI-related (mostly); could be the branch on git.l.o could be old, but then this has worked since a long time
[13:50:26] <zultron> mhaberler, jepler, I guess another week before the ub branch is done. That's next on the list, now that the r8168 driver is stuffed into the autobuild scripts (and unrelatedly, bcfg2 Fedora pkgs updated)
[13:51:34] <mhaberler> jepler: I will pick applicable fixes from you (configure, boost, rtld_now etc) into git.mah.priv.at rtos-master-v0
[14:59:47] <jepler> whee, the kzalloc issue was the only one
[15:03:44] <jepler> so anyway .. rt_preempt has bad latencies under both cyclictest and linuxcnc on this machine that is good for rtai under hardy and lucid
[15:09:02] <jepler> mhaberler: I have not checked whether they are applicable at the tip of rtos-master-v0 but this is what was needed in preview3-merged: http://emergent.unpythonic.net/files/sandbox/rt_preempt-mesa.patches.mbox
[15:10:18] <jepler> or maybe you want it in userpci/slab.h as kzalloc. I prefer rtapi_ prefixes on APIs that rtapi has promised to provide...
[15:10:20] <mhaberler> ah, thanks, will add that; funny, I thought we already did that one
[15:10:44] <jepler> preview3 is pretty old, but you said something about it being more suited to testing than rtos-master-v0 so that's where I'm testing.
[15:10:45] <mhaberler> need to read back in history, maybe lost in a different branch
[15:11:26] <mhaberler> in my repo: see http://git.mah.priv.at/gitweb?p=emc2-dev.git;a=shortlog;h=refs/heads/rtos-integration-preview3-merged-into-master
[15:11:31] <mhaberler> second from top
[15:11:50] <jepler> oh yeah I probably made the mistake of using the git.l.o branch
[15:12:05] <jepler> fresh machine, just the one remote. sorry to waste your time.
[15:12:11] <jepler> (and mine)
[15:38:38] <skunkworks> one of the lathes dad got still has the cosmolee(sp) on it...
[16:26:03] <skunkworks> jepler: defiantly try idle=poll in the kernel line
[16:28:04] <jepler> I will do it defiantly!
[16:30:23] <skunkworks> damn spell check
[16:30:47] <skunkworks> It should know what I mean
[16:35:54] <cradek> haha, I thought that's what you meant (even though it'll probably not help, boldly do it anyway)
[16:45:17] * skunkworks should make up a hal file for the lathes.
[16:46:07] <skunkworks> maybe I will have some time sunday to try it. I am trying to get dad to trace the printer port pins to the board. I don't know which pin the 'clock' is.
[21:16:43] <skunkworks> This is what we have gleamed so far
[21:16:51] <skunkworks> x step pin 2
[21:16:53] <skunkworks> x dir pin 3
[21:16:54] <skunkworks> z step pin 4
[21:16:56] <skunkworks> z dir pin 5
[21:16:57] <skunkworks> index pin 12
[21:16:59] <skunkworks> estop pin 11
[21:17:00] <skunkworks> 74ls374 enable pin 14
[21:17:02] <skunkworks> 74ls374 clock pin 1
[21:17:03] <skunkworks> (printer port pins)
[21:18:18] <skunkworks> I think the only thing we don't know is a) can we can get it to work and b) what the input for the 100 line single channel encoder is.
[21:19:39] <skunkworks> but that should be pretty easy once things are hooked up. at the maximum rpm - the 100 line will only be 4666pps still workable with a 50us base period.
[21:22:05] <skunkworks> I think this is doable.
[21:22:38] <skunkworks> maybe ;)
[21:25:01] <skunkworks> the only iffy thing is if the step and dir change can clock at the same time. I hope so.
[21:25:37] <skunkworks> you wouldn't think they would be that fancy to clock the dir first - then clock the step.
[21:29:30] <skunkworks> I guess that would just work out.. if I need the dirhold/dirsetup that step cycle would happen in multible base cycles - sending multible set/reset clocks
[21:30:56] <skunkworks> no - really - I do always talk to myself. ;)
[21:31:10] <skunkworks> logger[mah]:
[21:31:10] <logger[mah]> skunkworks: Log stored at http://linuxcnc.mah.priv.at/irc/%23linuxcnc-devel/2013-07-05.html
[21:33:51] <skunkworks> oh - it is the 4th. heh - I should look outside
[22:07:55] <jepler_> apparently it's common for 3d printers to have two controlled temperatures (nozzle and bed). While thinking about a single temperature, it seemed natural to use the S-number and spindle-at-speed .. but this doesn't scale to two numbers
[22:09:48] <Tom_itx> yes
[22:10:01] <Tom_itx> and some run multiple nozzles
[22:10:13] <Tom_itx> up to 3 i've heard of
[22:10:43] <jepler_> in that case are all the nozzles kept hot & ready?
[22:11:10] <jepler_> I guess you are typically using more than one filament in the same layer and you wouldn't want to wat..
[22:11:13] <jepler_> wait
[22:11:22] <Tom_itx> have you looked at the marlin software?
[22:11:37] <jepler_> no
[22:11:39] <Tom_itx> i think that's the one alot are using on them
[22:11:50] <Tom_itx> for a mega2560 avr
[22:11:53] <Tom_itx> or such
[22:12:15] <jepler_> you could use motion.feed-hold as long as you don't care about not doing any initial G0 moves while waiting for temps to be reached
[22:13:23] <Tom_itx> https://github.com/ErikZalm/Marlin
[22:13:32] <Tom_itx> there's also a branch for the delta for it
[22:13:35] <Tom_itx> somewhere
[22:13:46] <jepler_> hah thanks
[22:13:55] <Tom_itx> most of the settings are in configuration.h
[22:15:32] <Tom_itx> some use rubber heat mats for the bed (mains voltage)
[22:15:57] <Tom_itx> most use the circuit board heater thingy
[22:17:57] <jepler_> so it appears that marlin has explicit "wait for bed temp" and "wait for nozzle temp" m-codes
[22:19:09] <jepler_> it just feels like linuxcnc could make it impliclt like spindle-at-speed
[22:39:21] <Tom_itx> yeah, it will pause quite some time before it begins a print as i understand it
[22:51:51] <Tom_itx> did you visit with the guy at the fest running linuxcnc on his mendlemax bbb?