Back
[07:21:15] <jthornton> anyone know what ini variables emccalib.tcl uses?
[07:21:27] <jthornton> I've looked at the code but it is not clear
[08:32:42] <cradek> do you mean which ones does it let you tune if they're found?
[08:36:34] <cradek> wow I can't figure it out either
[08:37:05] <cradek> anything that has a setp in a hal file, that's substituted from the ini file?
[08:38:16] <cradek> the scan starts at line 172
[08:49:56] <lair82> Good morning guys,
[08:53:22] <lair82> Going back to my predicament, would it be worth posting my config files to pastebin?
[08:53:49] <jthornton> ah I bet the setp is the key
[10:06:33] <cradek> lair82: couldn't hurt
[10:07:04] <cradek> lair82: also trying AXIS would be smart for narrowing it down
[10:07:22] <lair82> ok, I will post my files
[10:07:50] <lair82> is it easier for troubleshooting?
[10:08:03] <cradek> it's known to work right
[10:08:17] <cradek> it's been stable for years
[10:34:59] <cradek> seb_kuzminsky: are comps that fail the new test totally broken?
[10:35:45] <KGB-linuxcnc> 03Chris Radek 05cradek/multi-tlo 79e2e5b 06linuxcnc 10src/emc/rs274ngc/interp_array.cc 10src/emc/rs274ngc/interp_check.cc 10src/emc/rs274ngc/interp_convert.cc 10src/emc/rs274ngc/interp_internal.hh Allow use of several simultaneous tool offsets with G43.2 H * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=79e2e5b
[10:51:38] <lair82> http://pastebin.com/MkmUqAje http://pastebin.com/6pYs3dZY The first one is my INI file, the second is my HAL file
[10:57:18] <lair82> http://pastebin.com/33dzSUv8 This is a printout of the "halcmd show thread"
[11:38:11] <cradek> wonder what's taking all that time!
[11:38:37] <cradek> oh holy cow, look at your classicladder, 150 rungs 120 inputs 600 symbols?
[11:38:43] <cradek> is your ladder massively complex?
[11:39:50] <cradek> yep from the looks of it :-)
[11:40:39] <lair82> Yes it is, my machiine is almost completely controlled thru ladder
[11:41:17] <cradek> I suspect that's what's taking all your cpu time, but I'm not sure how to confirm that
[11:41:39] <cradek> are all those classicladder_rt numThis numThat all set as low as possible?
[11:42:43] <lair82> I will go thru and chisel it down as far as possible, I know that the symbols can go down and the rungs can go down.
[11:42:47] <cradek> ferror limit 5 inches!? that's really dangerous
[11:43:04] <cradek> is your tuning really bad and you're ignoring it?
[11:43:26] <cradek> the ferror is the main thing that will disable the machine when you crash it
[11:44:03] <lair82> Thats wrong I forgot to turn them back down, Whoops
[11:44:37] <cradek> you could increase your servo cycle a bit more maybe, but not too far with accel=15
[11:44:45] <cradek> check to see what impact it has on following
[11:45:03] <cradek> sure looks like you're out of cpu
[11:45:12] <lair82> M turning centers will run at about .0005 in" error for every 1 IPM of feed
[11:48:20] <lair82> Just ran out and cranked them back down. Thanks for noticing that.
[11:48:51] <lair82> what would be a recommendable option?
[11:49:51] <cradek> like I said - try turning down the ladder stuff, then if that isn't enough try increasing the servo cycle and see if it hurts following, and if that isn't enough maybe it's time to consider more cpu
[11:50:12] <cradek> it'd be nice if you'd get your servo cycle run time down to half the period
[11:50:44] <cradek> depending on what all your ladder does, perhaps you could run it in its own (slower) thread
[11:51:04] <cradek> IF my theory is correct: that it's ladder taking up all the time
[11:51:36] <cradek> bbl
[12:10:21] <seb_kuzminsky> cradek: yes, .comps that are rejected by the new comp.py would fail to load with 'halcmd loadrt' when build by the old comp
[12:20:52] <skunkworks_> I have seen on some of the older atoms - ladder takes a lot of overhead..
[12:37:35] <seb_kuzminsky> i wonder what's happened in the upstream classicladder project since we forked in ... 2005
[12:44:58] <lair82> How would i go about putting it in its own thread? Do you mean I would I need a faster CPU?
[12:50:02] <lair82> After looking at the actual numbers, I have 110 rungs, 386 symbols, 10 sections, 109 inputs, and 89 outputs in that machine.
[12:50:56] <cradek> seb_kuzminsky: then sure, I think it belongs in 2.5
[12:51:03] <cradek> seb_kuzminsky: thanks for fixing it
[12:53:37] <seb_kuzminsky> ok, i'll push it there
[12:53:58] <seb_kuzminsky> doesn't "halcmd show funct" tell you how much time each function is using?
[12:54:10] <cradek> it didn't on my sim build
[12:54:14] <cradek> I thought it did too
[12:56:01] <cradek> 4 s32 RO 46 and2.0.time
[12:56:01] <cradek> 4 s32 RW 1001 and2.0.tmax
[12:56:04] <cradek> they are params
[12:56:28] <cradek> 7 s32 RO 288 classicladder.0.refresh.time
[12:56:28] <cradek> 7 s32 RW 816 classicladder.0.refresh.tmax
[12:56:30] <cradek> aha!
[12:56:33] <cradek> lair82: ^^
[12:58:27] <seb_kuzminsky> oh that's ok then
[13:00:12] <cradek> that's mine, not his
[13:00:49] <seb_kuzminsky> i meant it's good we have the times *somewhere*, even if it's not where any of us first guessed
[13:00:58] <cradek> ah, right
[13:01:46] <lair82> I will go look at that now,
[13:02:05] <cradek> grab a "show all"
[13:09:11] <lair82> http://pastebin.com/w4kHXAhe results from "halcmd show"
[13:09:25] <lair82> Ladder is jacked
[13:13:31] <lair82> 16 s32 RO 1031382 classicladder.0.refresh.time
[13:13:49] <lair82> 16 s32 RW 1098495 classicladder.0.refresh.tmax
[13:14:39] <cradek> yep, smoking gun
[13:14:52] <cradek> did you trim down the numbers of things yet?
[13:15:33] <lair82> not yet, the guy just started running it, will try to within the next hour
[13:15:52] <cradek> it looks super consistent, so that's good at least
[13:16:46] <cradek> are your servo amps velocity mode?
[13:18:07] <lair82> As far as I know yes, they are older than I am 34Yrs old
[13:18:36] <cradek> then I bet you could have a slower servo cycle and nothing would care
[13:19:05] <lair82> 2000000???
[13:19:21] <cradek> yeah try it. then classicladder would only run half the time.
[13:19:29] <cradek> that leaves you with half a computer for your gui
[13:22:08] <cradek> seb_kuzminsky: what became of wj200_vfd?
[13:38:47] <lair82> I cranked it up to 1500000, and trimmed the values in my ladder program. See how things go. I would have investigated further, but the shop supervisor was breathing down my back, literally, waiting to make some chips.
[13:43:12] <cradek> well now you know what the problem is, and how to measure it
[13:44:14] <pcw_home> its a shame these are not supported yet:
[13:44:15] <pcw_home> http://www.newegg.com/Product/Product.aspx?Item=N82E16813157498&cm_re=j1800-_-13-157-498-_-Product
[13:44:17] <pcw_home> these run circles around the D525s
[13:46:04] <lair82> what do you mean by that, not supported?
[13:46:24] <pcw_home> Too new
[13:46:50] <lair82> Linux will nut run it I take it?
[13:47:03] <lair82> not run on it
[13:47:09] <pcw_home> needs a newer Linux kernel than available with RTAI
[13:48:01] <lair82> Would this work with machinekit?
[13:48:36] <pcw_home> well I should qualify that the 2.6.whatever RTAI patched kernel on the live CD works but no graphics acceleration
[13:50:06] <pcw_home> It probably will work with machine kit with a Xenomai or Preemt-RT patched kernel (you need linux 3.12 or > for the accelerated graphics support)
[13:51:13] <cradek> do you mean hardware-accelerated opengl? or is all regular GUI stuff bad?
[13:53:35] <pcw_home> Not sure just noticed that screen updates were quite slow with 2.6
[13:53:36] <pcw_home> (and first 3.x kernel tha worked was 3.12)
[13:54:50] <pcw_home> (earlier 3.x hernels had garbled video)
[13:55:20] <lair82> I might give it a try.
[13:56:14] <pcw_home> here's a similar (though more expensive) one with a PCI slot
[13:56:16] <pcw_home> http://www.superbiiz.com/detail.php?name=MB-J1800IA
[13:57:09] <CaptHindsight> but then you're supporting Intel :)
[13:57:56] <pcw_home> unfortunately AMD has nothing that really compares
[13:58:29] <CaptHindsight> AMD is way faster on the GPU side
[13:58:40] <pcw_home> doesnt help linuxcnc
[13:58:54] <CaptHindsight> linuxcnc doesn't need it
[13:59:13] <CaptHindsight> if I need more CPU I use the FX APU's
[14:00:12] <CaptHindsight> my image processing takes up all the cpu vs linuxcnc
[14:01:15] <pcw_home> Just sayin that the baytrail MBs are a big win in terms of CPU speed vs
[14:01:16] <pcw_home> the Atoms at the same (or lower) power levels and also have excellent RT latency
[14:01:38] <CaptHindsight> but then you're stuck with Intel :)
[14:02:57] <pcw_home> well good to be stuck with cool and fast processors :-)
[14:03:16] <lair82> Good lookin out pcw_home, I almost bought the ASROCK board, not realizing my 5i23 wasn't plug into it.
[14:04:01] <CaptHindsight> if I need more cpu then I'd use something like
http://www.newegg.com/Product/Product.aspx?Item=N82E16813157465
[14:04:46] <pcw_home> Yeah but needs a fan (or a huge passice cooler)
[14:04:53] <pcw_home> passive
[14:05:12] <CaptHindsight> and Intel doesn't get rewarded
[14:07:42] <CaptHindsight> I've never seen Linuxcnc need much CPU
[14:07:49] <pcw_home> http://www.newegg.com/Product/Product.aspx?Item=N82E16813128698&cm_re=j1900-_-13-128-698-_-Product
[14:07:51] <pcw_home> is another baytrail MB with PCI slot, but the Gigabyte models will need BIOS upgrade (and much hair pulling) to boot linux
[14:07:52] <cradek> in both my industrial machines I've used a full size motherboard with plenty of slots
[14:09:02] <pcw_home> CaptHindsight: lair82 is out of CPU
[14:11:49] <cradek> it would be interesting to profile classicladder
[14:12:16] <cradek> I'd guess probably lots of string manipulation (labels, formulae)
[14:12:48] <pcw_home> Yeah is suspect its not very efficient
[14:13:02] <CaptHindsight> lair82: how did you do that? running the Mesa drivers?
[14:13:40] <CaptHindsight> you're supposed to run the CPU stuff on the CPU and the GPU stuff on the GPU
[14:18:48] <cradek> CaptHindsight: it's classicladder
[14:19:38] * cradek bombards the devel list
[14:24:02] <CaptHindsight> lair82: what board+cpu are you currently using?
[14:34:57] <lair82> Gigabyte GA-D525TUD, MESA 5I23, 7I44, 7I49,
[14:36:40] <lair82> We also have one of the three that is using an INTEL D525MW with an Intel ATOM processor, with the identical mesa components behind it
[14:37:35] <lair82> All of which exhibit the same identical problems
[14:37:41] <CaptHindsight> I'd like to test your ladder
[14:38:45] <lair82> I can send you an email with the files if you like, I don't know if pastebin will work for that.
[14:39:38] <cradek> I loaded classicladder_rt with your command line and it runs fast
[14:39:46] <CaptHindsight> sounds like a good abuse test
[14:39:47] <cradek> so it's clearly something about your ladder
[14:40:23] <cradek> I bet there's one particular thing that's horrible
[14:43:04] <pcw_home> does classic ladder have any profiling tools, like settable timestamp tokens or the like?
[14:43:32] <cradek> I don't know of anything like that
[14:43:50] <cradek> lair82: hey please send me your ladder too, what the heck
[14:44:52] <lair82> Can I post it on pastebin? and will it work on your end? I think it would be better to email a zipped folder to you guys of my entire config
[14:45:25] <cradek> you could try pastebin, but I kind of doubt it will work
[14:49:49] <seb_kuzminsky> cradek: i made the wj2000_vfd branch work, haven't touched it since then
[14:50:05] <cradek> eh, why not merge it and move on then
[14:50:15] <lair82> http://pastebin.com/D79L97cg Post gui file,
http://pastebin.com/AMJLViBz Ladder file
[14:51:07] <lair82> Gotta Roll guys, Talk to you tomorrow, And Thank ALL of guys for the help,
[14:53:17] <lair82> If that didn't work email me, and I will send you my files. rick@superiorroll.com
[14:53:31] <cradek> I think it loads here
[14:54:16] <cradek> err no, it looks kind of bogus
[14:54:20] <cradek> lots of "???" labels
[14:54:24] <seb_kuzminsky> heh, there's a combination of CamelCase and snake_case in the wj200 hal presence
[14:54:31] <seb_kuzminsky> pin out bit is_atSpeed
[14:54:46] <seb_kuzminsky> and no docs, but what the hey
[14:54:55] <seb_kuzminsky> cradek: do you want it in 2.5? otherwise i'll merge it into 2.6
[14:55:36] <cradek> hal pin naming spec says only words-and-dashes are allowed
[14:56:28] <cradek> 2.6 please, since it's a nontrivial build with those library dependencies
[14:57:56] <seb_kuzminsky> ok
[14:59:55] <cradek> docs/src/hal/general_ref.txt
[15:00:22] <cradek> HAL components should not use underscores or MixedCase. Use only lowercase letters and numbers in names.
[15:00:36] <cradek> dots for heirarchy, dashes to separate words
[15:01:53] <cradek> that's trivial to fix
[15:01:57] <cradek> I'd just fix it before merging
[15:02:30] <cradek> or if you want to push it to the wj200_vfd, I'll do it
[15:02:34] <cradek> ... branch
[15:02:49] <seb_kuzminsky> i'll do it tonight
[15:03:33] <seb_kuzminsky> the funny library build requirements are properly handled by our Build-Depends and src/configure, but yes it's still a little unusual
[15:03:56] <cradek> 2.6 is virtually ready anyway, right?
[15:04:00] <cradek> right right?
[15:04:19] <seb_kuzminsky> we could release it now on lucid & precise, yes
[15:04:49] <seb_kuzminsky> if we want wheezy we have to copy the kernel deb from precise to wheezy,
[15:04:58] <seb_kuzminsky> and if we want live cds we need new rtai kernel debs
[15:05:22] <seb_kuzminsky> and there's a pile of open bugs on the tracker, but probably nothing that needs to block the release
[15:06:17] <cradek> how soon do you think you'll be able to redo the kernel package?
[15:06:58] <seb_kuzminsky> i'll be able to work on it the nights of next week
[15:07:07] <cradek> that's great
[15:07:31] <cradek> then maybe building the cd will be simple for me, from there
[15:07:37] <seb_kuzminsky> cool
[15:07:43] <cradek> so maybe we can release in 10 days if all goes well
[15:08:06] <seb_kuzminsky> i think a live cd is good to have, but not required for the release
[15:08:08] <cradek> uh-oh, I should never say a target date in a public channel
[15:08:24] <seb_kuzminsky> it's pretty easy to go from a vanilla install to a linuxcnc-on-rtai install
[15:09:42] <cradek> unfortunately history tells us people don't like to do that
[15:09:46] <cradek> they love those cd images
[15:10:07] <seb_kuzminsky> yeah...
[15:10:27] <cradek> in 05 we (I) actually ordered a bunch of pretty CDs from ubuntu and handed them out, along with very simple instructions for installing linuxcnc afterward
[15:10:28] <seb_kuzminsky> it'd be easy to update the lucid livd-cd with the new linuxcnc debs
[15:10:41] <cradek> everyone whined about it
[15:10:51] <seb_kuzminsky> but we probably want to encourage people to get off lucid
[15:10:54] <cradek> well that's true - I did just that for 2.5
[15:10:55] <cradek> yep
[15:11:04] <seb_kuzminsky> no good deed goes unwhined-about
[15:11:11] <cradek> for me wheezy is the clear choice
[15:11:11] <seb_kuzminsky> bbl
[15:11:22] <cradek> for the project, I don't know what the clear choice is
[15:23:42] <KGB-linuxcnc> 03Chris Radek 052.6 b5426d5 06linuxcnc 10src/emc/usr_intf/touchy/mdi.py We have full tool offsets now. * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=b5426d5
[15:23:42] <KGB-linuxcnc> 03Chris Radek 05cradek/multi-tlo c6b589d 06linuxcnc 10src/emc/rs274ngc/interp_array.cc 10src/emc/rs274ngc/interp_check.cc 10src/emc/rs274ngc/interp_convert.cc 10src/emc/rs274ngc/interp_internal.hh Allow use of several simultaneous tool offsets with G43.2 H * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=c6b589d
[15:23:42] <KGB-linuxcnc> 03Chris Radek 05cradek/multi-tlo 4528848 06linuxcnc 10src/emc/usr_intf/touchy/mdi.py Touchy support for mdi G43.2 * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=4528848
[16:31:45] <KGB-linuxcnc> 03Sebastian Kuzminsky 05v2.5_branch 90acb15 06linuxcnc 10src/hal/utils/comp.g comp: reject .comp files whose names dont match the component name * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=90acb15
[16:31:45] <KGB-linuxcnc> 03Sebastian Kuzminsky 05v2.5_branch 12f3f9e 06linuxcnc 10(5 files) tests: verify that comp rejects .comp files whose names dont match the component name * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=12f3f9e
[17:21:19] <skunkworks_> logger[psha]:
[17:21:19] <logger[psha]> skunkworks_: Log stored at
http://psha.org.ru/irc/%23linuxcnc-devel/2014-05-29.html
[17:21:28] <skunkworks_> cradek:
http://youtu.be/HPzow8L1dxw
[17:26:20] <skunkworks_> http://youtu.be/djlUG3ZF_tE
[17:42:30] <seb_kuzminsky> the wj200 branch doesn't produce a manpage
[17:42:49] <seb_kuzminsky> and when i coax it into doing it, comp puts the manpage in section 9, not section 1 where it belongs
[17:47:33] <seb_kuzminsky> this affects other .comps, like gladevcp
[17:50:32] <seb_kuzminsky> that might be the only one
[17:52:04] <seb_kuzminsky> wait, gladevcp doesn't use comp
[17:52:13] <seb_kuzminsky> it's just a git-tracked manpage in the wrong place
[17:52:16] <seb_kuzminsky> never mind me...
[17:53:30] <seb_kuzminsky> i gues wj200_vfd.comp is the only one affected
[17:53:57] <seb_kuzminsky> it declares 'option userspace;', i guess that should clue comp into putting the manpage in section 1
[18:53:12] <seb_kuzminsky> cmorley: you added a gladevcp manpage in section 9, and then later michael haberler added another one in section 1
[18:53:52] <seb_kuzminsky> section 1 is the right place for the gladevcp manpage, but i'm not sure how to merge the contents of gladevcp.9 into gladevcp.1, do you have any input on this?
[19:06:13] <KGB-linuxcnc> 03Sebastian Kuzminsky 052.5-comp-manpages fe635ea 06linuxcnc 10src/hal/utils/comp.g comp: put manpages for userspace components in section 1 * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=fe635ea
[19:06:13] <KGB-linuxcnc> 03Sebastian Kuzminsky 052.5-comp-manpages 1738ead 06linuxcnc 10(10 files in 2 dirs) tests: rename comp name tests * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=1738ead
[19:06:13] <KGB-linuxcnc> 03Sebastian Kuzminsky 052.5-comp-manpages 1861889 06linuxcnc 10(5 files) tests: add a test of comp(1) manpage generation * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=1861889
[19:06:33] <seb_kuzminsky> i think 2.5-comp-manpages belongs in 2.5
[19:06:53] <seb_kuzminsky> before that branch, comp puts all manpages in section 9, which is for kernel modules
[19:07:25] <seb_kuzminsky> after that branch, comp checks for 'option userspace' in the .comp and puts *those* manpages in section 1 (non-option-userspace manpages still go in 9)
[19:07:48] <seb_kuzminsky> we have no .comps with 'option userspace'in 2.5 so it doesnt matter much
[19:08:10] <seb_kuzminsky> the wj200 .comp is our first userspace .comp, it needs this change
[20:53:30] <cmorley> seb: what the difference between man 1,3 and 9? I can look at the gladevcp man pages this weekend
[21:51:47] <skunkworks_> she passed out around 9:00
[21:51:56] <skunkworks_> keeps going and going and going
[21:52:26] <skunkworks_> heh
[21:52:31] <skunkworks_> wrong window
[22:01:42] <seb_kuzminsky> heh
[22:04:02] <seb_kuzminsky> cmorley: the 'man' manpage (
[22:04:17] <seb_kuzminsky> ("man man") has a good description of what should go in the different manpage sections
[22:20:19] <KGB-linuxcnc> 03Bence Kovacs 052.6 94cc151 06linuxcnc 10src/hal/drivers/hal_gm.c Fix zero pointer in case of PCI card with DAC version 1. * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=94cc151
[22:39:06] <seb_kuzminsky> cradek: do you mind if i push 2.5-comp-manpages to 2.5? rationale above ^^^
[23:43:19] <seb_kuzminsky> wow, elon musk claims that the super draco engine on the dragon v2 spacecraft is "fully printed" (meaning 3d printed, i'm pretty sure, from context)
[23:43:29] <seb_kuzminsky> http://new.livestream.com/accounts/142499/events/3040228
[23:55:04] <seb_kuzminsky> https://twitter.com/elonmusk/status/375737311641628672?screen_name=elonmusk