#linuxcnc-devel | Logs for 2015-06-18

[06:10:52] <skunkworks> https://github.com/mhaberler/asciidoc-sandbox/wiki/Generating-clock-interrupts-from-Mesanet-cards
[07:28:46] <jepler> nobody has ever articulated to me what is gained by using a different clock source for the thread timing
[07:30:30] <jepler> is there a bug filed in the mesaflash tracker about the supposed 64-bit problems? I would have told you I was confident I used it on a 64-bit system myself...
[07:34:51] <jepler> why yes there is https://github.com/micges/mesaflash/issues/4
[07:45:44] <jepler> hm I don't have a PCI card with EEPROM
[07:45:50] <jepler> 7i80 works fine on 64 bits
[07:53:41] <jepler> cradek: do you have a spare pci mesa card with eeprom (e.g., 5i25)?
[07:55:33] <skunkworks> jepler, welcome back!
[07:56:42] <skunkworks> Heh - I swapped out the processor on this computer running mach4. 2.4 celeron -> 2.4 pentium4.
[07:57:32] <skunkworks> from this. http://electronicsam.com/images/KandT/testing/grbl/mach4halscopemorer.png
[07:57:34] <skunkworks> to
[07:57:59] <skunkworks> http://electronicsam.com/images/KandT/testing/grbl/mach4P4more.png
[07:58:08] <skunkworks> Still a bunch of overages..
[08:03:30] <jepler> skunkworks: thank you
[08:03:34] <jepler> skunkworks: it was a good trip, but I'm happy to be home
[08:04:06] <skunkworks> I know the feeling
[08:06:17] <skunkworks> Mach4 path (which doesn't have any path following control yet...)
[08:06:19] <skunkworks> http://electronicsam.com/images/KandT/testing/grbl/mach4.png
[08:06:25] <skunkworks> and mach3
[08:06:28] <skunkworks> http://electronicsam.com/images/KandT/testing/grbl/mach.png
[08:06:49] <skunkworks> I don't see much difference for it being touted as much improved..
[08:07:39] <skunkworks> I ment this one.. (for mach4) http://electronicsam.com/images/KandT/testing/grbl/mach4P4plot.png
[08:13:00] <jepler> hm over at the right side there, is it actually going that far off the path or is it an artifact of axis's backplot sampling? I'm surprised to see a nice arc cut off like that but it is very similar in both images
[08:14:05] <jepler> I see the accelerations are much higher in the one plot as compared to the other
[08:17:07] <skunkworks> jepler, mach violates slightly less..
[08:17:13] <skunkworks> *mach3
[08:17:46] <skunkworks> I was wondering that also - but at that point the machine is slowing down to make the corner - I would think that the plot would follow.
[08:19:10] <skunkworks> and anything that I throw at it using linuxcnc follows as expected.
[08:21:19] <skunkworks> http://electronicsam.com/images/KandT/testing/grbl/linuxcnc.png
[08:21:38] <skunkworks> It rounds thos few areas as expected..
[08:21:47] <skunkworks> (strait G64)
[08:22:52] <skunkworks> I could try linuxcnc with 150+ accelleration and see if I get similar artifacts..
[08:30:25] <skunkworks> 150in/s^2 - path looks identical on linuxcnc
[08:30:34] <skunkworks> I think the plot is accurate.
[08:32:13] <jepler> yes, linuxcnc seems to take considerably fewer liberties with the path
[08:32:21] <jepler> though there are still a few I spot at that scale
[08:32:59] <jepler> if you don't care what path the machine actually follows, it's possible to get done with a part program much faster
[08:34:38] <skunkworks> sure. But with linuxcnc I can say - I want to follow the path within x.xxx... Mach4 has none of that yet. (and mach3 is really crappy IMHO on how it sets path following)
[08:39:27] <skunkworks> mach3 has 2 main settings. At smaller than what angle - go to exact stop. and how far from the end of the segment to start the blend.
[08:59:38] <skunkworks_> zlog
[09:02:26] <jepler> I was just thinking of standing still for a few seconds before announcing victory
[09:14:17] <pcw_home> Linux pcw-G41M-Combo 4.0.5-rt4 #1 SMP PREEMPT RT Wed Jun 17 08:21:17 PDT 2015 i686 i686 i686 GNU/Linux
[09:14:18] <pcw_home> /me is a preempt-rt version hipster
[09:15:01] <pcw_home> \me is a preemt-rt version hipster
[09:15:29] <pcw_home> and obviously a irc newbee
[09:17:10] <pcw_home> the first 4.0 version had pretty large latencies but 4.0.5-rt4 seems fine
[09:26:21] <Roguish> pcw_home: hey, is there a .iso for the preempt-rt linuxcnc?
[09:27:01] <Roguish> not that i'm lazy or anything, just a whole lot quicker that way.....
[09:27:23] <seb_kuzminsky> Roguish: yes, sort of
[09:27:43] <Roguish> where might one find it?
[09:27:54] <Roguish> i'm game to give it a try.
[09:28:07] <seb_kuzminsky> http://linuxcnc.org/docs/2.7/html/getting-started/index.html#_installing_on_debian_wheezy_with_preempt_rt_kernel
[09:28:39] <Roguish> THANKS.
[09:39:59] <cradek> jepler: 6i24?
[09:42:58] <jepler> cradek: probably would do, yes. the machine I'd test in as a pci-e slot.
[09:43:13] <jepler> has a pci-e slot
[09:47:09] <jepler> cradek: want to meet for a coffee sometime today at meadowlark?
[09:53:39] <jepler> apparently there's an rt kernel back in sid .. https://packages.debian.org/sid/linux-image-4.0.0-2-rt-amd64
[09:53:47] <jepler> based on 4.0.5
[09:55:58] <skunkworks> is sid the next wheezy?
[09:56:59] <mozmck> I think it's debian's unstable or testing branch
[09:57:05] <skunkworks> oh
[09:57:25] <skunkworks> jessie
[09:57:37] <mozmck> Jessie is the current stable release.
[09:57:41] <skunkworks> does jessie have a rt-preempt kernel?
[09:58:09] <skunkworks> pcw_home, how does 4. perform?
[09:58:22] <mozmck> No, but I bet you can use the one from sid
[10:01:07] <mozmck> skunkworks: https://www.debian.org/releases/
[10:03:32] <pcw_home> skunkworks: seems good. Ive run it a day on the DC7800 at 3 KHz with normal desktop activities and no issues
[10:04:07] <pcw_home> (previous version didn't work as well)
[10:04:15] <skunkworks> 7i73?
[10:04:43] <pcw_home> still at assy but shoud be here in the next few days
[10:04:57] <mozmck> pcw_home: is there any improvement over 3.18?
[10:04:59] <skunkworks> great!
[10:05:15] <mozmck> Any 7i92IDC yet? ;)
[10:05:24] <pcw_home> mozmck: to early to tell
[10:05:28] <cradek> jepler: that would be nice. early afternoon?
[10:05:36] <skunkworks> mozmck, did you get your usb issue figure out?
[10:05:41] <pcw_home> I ve started kitting it for the next build
[10:05:59] <mozmck> skunkworks: yeah, quit using the DC7800 :)
[10:06:01] <cradek> jeez another monsoon is coming
[10:06:12] <skunkworks> oh - motherboard change?
[10:06:35] <mozmck> I need to go back and try a different ethernet card.
[10:06:45] <mozmck> No, changed computers.
[10:06:50] <skunkworks> https://groups.yahoo.com/neo/groups/mach1mach2cnc/conversations/topics/148006
[10:07:07] <skunkworks> isn't that a motherboard change?
[10:07:07] <mozmck> I'm running it on a lenovo now and it runs great
[10:07:15] <skunkworks> :)
[10:07:20] <mozmck> Well, yes, and a bit more too :)
[10:09:42] <skunkworks> is the lenovo a laptop?
[10:09:52] <jepler> cradek: sure
[10:10:11] <mozmck> No, it's a desktop
[10:50:10] <cradek> dear everyone on the forum: if your linuxcnc install works for years and then one day it stops working right and gets all flaky IT IS YOUR HARDWARE.
[10:50:17] <cradek> thank you that is all
[10:51:31] <skunkworks> heh
[13:01:23] <skunkworks> linuxcnc log of a certain section of gcode
[13:01:25] <skunkworks> http://electronicsam.com/images/KandT/testing/grbl/linuxposlog.png
[13:01:38] <skunkworks> mach3 (notice the acc scale change)
[13:01:49] <skunkworks> http://electronicsam.com/images/KandT/testing/grbl/mach3poslog.png
[13:02:22] <skunkworks> mach4
[13:02:24] <skunkworks> http://electronicsam.com/images/KandT/testing/grbl/mach4poslog.png
[13:02:56] <skunkworks> it definatly goes faster...
[13:02:58] <skunkworks> :)
[14:11:28] <cradek> skunkworks: what was maxaccel supposed to be?
[14:12:06] <jepler> root@rat:/home/jepler/src/mesaflash# ./mesaflash --device 5i24 --verify SVST8_4.BIT
[14:12:09] <jepler> Checking file... OK
[14:12:11] <jepler> File type: BIT file
[14:12:14] <jepler> Boot sector OK
[14:12:17] <jepler> hmm mesaflash problem on 64-bit platform not reproduced here
[14:19:34] <jepler> pcw_home: how different are 5i25 and 6i24 from the pov of mesaflash? https://github.com/micges/mesaflash/issues/4 is reported as failing with a 5i25, but wfm with a 6i24.
[14:25:25] <skunkworks> jepler: 30in/s^2
[14:25:54] <skunkworks> you see linuxcnc obeys that pretty close.. So I trust the numbers...
[14:27:01] <cradek> so you're getting about 110 out of 30 in mach4, yuck
[14:27:09] <cradek> they must not have a good way of testing?
[14:27:29] <cradek> I remember before emc2 it was quite hard to guess what was working
[14:27:47] <cradek> no halscope, not even a backplot to compare with a preview!
[14:28:40] <skunkworks> linuxcnc is awesome
[14:28:50] <cradek> jepler: oops I guess I also have a 5i24
[14:28:53] <skunkworks> it looks to have gained atleast 1 second on linuxcnc
[14:29:02] <skunkworks> in that short run
[14:30:41] <cradek> wonder if someone not rob needs to look into our tp bug
[14:30:52] <skunkworks> I just emailed him
[14:31:13] <skunkworks> 'I'm glad you could isolate it to that particular code section. I'll take a look early next week (I'm moving my fiancee's stuff this weekend, so I won't have time before then). If past experience is any indicator, it will be a quick fix.'
[14:31:48] <cradek> sweet
[14:32:01] <skunkworks> yes - then 2.7!! :)
[14:32:06] <cradek> if he can't look into it, I'll try
[14:33:10] <skunkworks> if we don't see anything by the end of next week maybe
[14:33:45] <skunkworks> mach 3 and 4 act pretty identical.
[14:34:25] <cradek> must be the same code
[14:34:44] <cradek> you don't write the same bugs twice - in a rewrite you write all new and refreshing bugs
[14:34:47] <skunkworks> pretty close it seems.
[14:35:54] <skunkworks> the gross errors are in the de-accelleration.. maybe that is a feature by design...
[14:36:23] <skunkworks> I bet it is more - oh crap - I have to slow down and I don't have enough time.
[14:37:56] <PCW> jepler: all the PCI/PCIe cards with flash should be identical as far as flash programming goes
[14:39:21] <jepler> PCW: ok, hm. I wonder what's biting mah and not me then.
[14:40:33] <PCW> Wonder if running Xenomai might make something different
[14:41:36] <jepler> could be I suppose
[14:41:41] <cradek> in linuxcnc definitely (different math library) but surely not in mesaflash
[14:41:52] <cradek> it doesn't even use any rt stuff does it?
[14:42:02] <PCW> its weird, some card access must work since it finds the hm2 cookie
[14:42:15] <PCW> no RT at all
[14:42:28] <PCW> runs fine on stock linux
[14:43:19] <skunkworks> cradek: have you looked at the new tp? does it make any sense to you?
[14:43:30] <jepler> I have this vague memory of some kind of problem verifying the bootloader that I encountered
[14:43:35] <jepler> whether on arm or amd64 I don't know
[14:43:35] <PCW> it looks like it fails to read or compare the boot sector correctly
[14:43:42] <jepler> but I can't find any patches from me that went into mesaflash to cure it
[14:44:03] <cradek> skunkworks: I've only fixed it that one time, and the bug was very overt
[14:44:57] <cradek> skunkworks: I'm not sure I have much of a big-picture understanding of it, so I can probably only do the kind of fixes that are possible without really understanding the code (which are more than you might think)
[14:46:41] <cradek> so, I guess I don't know, because I haven't tried, because I haven't had to try
[14:52:54] <skunkworks> he has a lot more debugging in there - but I don't know what it means.
[22:56:45] <cradek> so I guess cnc workshop is going on
[23:07:27] <cradek> I guess I didn't go