#linuxcnc-devel | Logs for 2014-03-21

[12:25:39] <pcw_home> has anyone tried the 3.4.55 RTAI kernel on Ubuntu 14.04?
[12:37:38] <skunkworks> I thought 14.04 was beta
[12:38:40] <pcw_home> it it but release is April 17
[12:39:42] <skunkworks> oh - cool
[12:40:01] * skunkworks needs to get a ssd for his laptop - then install 14.04
[12:40:28] <pcw_home> testing on my new Baytrail MB Its really neat
[12:40:41] <skunkworks> cool
[12:40:49] <skunkworks> *neat
[12:41:28] <pcw_home> ~8 usec latency with FF/flash videos running, 4 KHZ servo loops and all
[12:41:49] <skunkworks> this is rtai?
[12:41:52] <pcw_home> maybe 2/3 times faster than the Atom MBs
[12:42:05] <pcw_home> yes RTAI
[12:42:13] <skunkworks> on 12.04 ?
[12:42:14] <pcw_home> no tweaks yet
[12:42:30] <pcw_home> 12.04 with 2.6.122
[12:42:33] <skunkworks> nice
[12:42:50] <pcw_home> cooler than the Atoms
[12:43:36] <pcw_home> (you can rest your hand on the passively cooled heatsink comfortably)
[12:43:59] <skunkworks> wow - link?>
[12:44:08] <pcw_home> and the parallel port works with our EPP cards
[12:44:54] <pcw_home> http://www.newegg.com/Product/Product.aspx?Item=N82E16813128688
[12:45:43] <pcw_home> I had to update the BIOs (well EFIOs) to get Ubuntu to boot
[12:46:06] <pcw_home> they have some with PCI instead of PCIE but not out yet
[12:47:22] <pcw_home> http://www.gigabyte.com/products/product-page.aspx?pid=4895#ov
[12:50:07] <pcw_home> This looks nice also:
[12:50:08] <pcw_home> https://www.asus.com/Motherboards/J1800IA/
[12:52:57] <pcw_home> They seem about as fast as Core2 Duos of the same clock speed (they are 2 core 2.4 GHz) so much snappier than the Atoms
[12:53:19] <skunkworks> nice
[12:53:26] <skunkworks> that should help a lot...
[12:53:42] <skunkworks> have you tried them with rt-preempt?
[12:54:02] <pcw_home> Not yet
[12:55:19] <skunkworks> well - throw one of you hd in there!
[12:55:39] <pcw_home> I had such a Frankenstein system on the hard drive I just started again from scratch so will first try the 3.4.55 RTAI kernel
[12:56:13] <pcw_home> 10.04 updated to 12.04 updated to 14.04
[12:57:28] <skunkworks> yikes
[12:58:40] <pcw_home> I have a clean 14.04 on it now, i will see if that can be made to work, other wise I'll start with a clean12.04
[13:00:43] <skunkworks> you should have just kept the hd I sent you :)
[13:01:38] <pcw_home> Probably :-)
[13:07:06] <pcw_home> The nice thing about the J1800 MBs is there are about 10 different ones now but since
[13:07:08] <pcw_home> its effectively a SOC, all the MBs are pretty much the same except for minor details
[13:12:36] <skunkworks> that should help to keep latency consistant
[13:14:58] <pcw_home> I got the ~8 usec with power management enabled (but it may have never gotten hot enough to do anything)
[13:28:38] <skunkworks> heat gun... ;)
[13:35:39] <pcw_home> I dont want to kill it just yet...
[13:36:06] <pcw_home> (I did before I upgraded the BIOS)
[13:59:20] <skunkworks> what happed with the old bios?
[14:00:05] <jthornton> skunkworks, who is working on the new trajectory planner? I forget...
[14:00:43] <skunkworks> Robert Ellenberg
[14:01:13] <jthornton> ok, is he working on any rotary axes?
[14:01:35] <jthornton> he just uses the mail list IIRC?
[14:02:53] <pcw_home> skunkworks: no boot with Ubuntu no matter what I tried
[14:10:21] <skunkworks> jthornton: I have asked him about it - but if you wanted to start an email thread in devel - it might get him re-thinking about it :)
[14:10:53] <jthornton> ok, I have a friend that could really use it
[14:12:42] <skunkworks> this is what he had replied when I asked about abc look ahead
[14:12:45] <skunkworks> Doing anything other than line-line blends in 9D is a nightmare and probably not worth it, but line-line is basically the same math, once you have the unit vectors and angles and such.
[14:12:47] <skunkworks> If axes have dissimilar acceleration limits, the acceleration would be limited by the most sluggish axis. I can overcome this limit in 3D somewhat with a mathematical trick, but I'm not sure if the same is possible in 9D. (I could just
[14:12:48] <skunkworks> Another possibility is to normalize the coordinates (i.e. scale everything so that the accelerations are equal, then "unscale" it when getting a position. It's not perfect since you could have an axis that had high acceleration but low maximum velocity.
[14:12:50] <skunkworks> We'd have to express arc-length and velocity as the 9D equivale
[14:13:34] <skunkworks> which is close to bla, bla, bla to me
[14:13:49] <jthornton> ok, I just started an email
[14:14:20] <skunkworks> Yay - (I get tired of asking all the questions) :)
[14:15:58] * jthornton goes back to the shop
[14:16:29] <skunkworks> (I sound like an eager school kid)
[14:16:38] <skunkworks> 'brown noser'
[20:17:51] <dgarr> i am ready to merge the demo_appsv6 branch to master if there are no (big) objections
[20:19:07] <andypugh> Not from me, I don't even know what you are describing.
[20:28:15] <memleak> http://git.linuxcnc.org/gitweb?p=linuxcnc.git;a=shortlog;h=refs/heads/dgarr/demo_appsv6
[20:35:03] <memleak> Is Tcl/Tk very difficult to modify or am I just doing it wrong (text editor) are there graphical toolkits for Tcl/Tk or something?
[20:35:44] <memleak> helper utilities etc to manage the source for you to an extent?
[20:59:54] <ju-emb> Anyone any idea what's wrong with my fresh LCNC sim on 12.04.
[20:59:56] <ju-emb> I used
[20:59:58] <ju-emb> http://wiki.linuxcnc.org/cgi-bin/wiki.pl?Bhttp://wiki.linuxcnc.org/cgi-bin/wiki.pl?Build_A_Simulator_Manuallyuild_A_Simulator_Manually
[21:00:02] <ju-emb> It doesn't run. Outputs on:
[21:00:04] <ju-emb> http://pastebin.com/p2j99nJz
[21:05:47] <skunkworks> does glxgears run? it is pointing to broken opengl
[21:15:11] <skunkworks> I had a similar problem with ati using the fglrx driver. (works fine either using the open source mesa driver or the catlyist(for sim)
[21:15:44] <memleak> catalyst is fglrx
[21:15:55] <memleak> well... the config tool for it
[21:19:32] <skunkworks> all I know is the one ubuntu wants to install always breaks my opengl (fglrx) and the catylist driver from ati's site works. (or the open source driver - but that makes this laptop run too hot)
[21:20:06] <skunkworks> but I would not use the catylist driver for realtime machines
[21:20:33] <memleak> ah yes, i love how the drivers perform so much differently when you install amd drivers from official website and when you install via synaptic, even though it's the same driver versions... forgot about that with debian based distros!
[21:20:58] <memleak> closed source from AMD yet perhaps somehow modified? :/
[21:22:02] <skunkworks> no clue
[21:23:25] <memleak> for the record i think it's safe to say not only to avoid amd's closed drivers, but closed drivers in general for real-time. i've seen latency spikes with closed source blobs for broadcom's NICs and we all know how linuxcnc performs with nvidia blobs :)
[21:24:31] <skunkworks> agreed
[21:33:57] <ju-emb> so what are the recommended driver for my box. It's an AMD64 with ATI Chipset
[21:39:24] <ju-emb> starting glxgears brings same errors as the start of LinuxCNC
[21:39:26] <ju-emb> any ideas how to fix it?
[21:42:37] <skunkworks> what video driver are you using - my luck has been (if it is just for simulation) go with the catylist driver from ati's site
[21:46:49] <ju-emb> I remember that the catalyst driver runs up to a specific kernel version.
[21:48:25] <andypugh> What motherboard?
[21:48:52] <memleak> ju-emb, that is correct.
[21:49:15] <memleak> i'd check the release notes of the driver before installing as you can end up with borked X that will not start.
[21:49:16] <andypugh> There was a message on the mailing list or forum in the last coiuple of weeks where someone ended up working through old versions of drivers to find the best one
[21:49:59] <memleak> http://support.amd.com/en-us/kb-articles/Pages/latest-linux-beta-driver.aspx that page says kernels up to 3.13
[21:50:47] <memleak> amd driver 13.12 (latest linux non-beta) says up to kernel 3.11
[21:54:43] <ju-emb> andypugh: I'm trying to find out what motherboard that is, but no luck until now, the box took the way in the garbage long time ago
[21:55:04] <andypugh> boot screen?
[21:55:32] <ju-emb> ok, so I have to restart the box
[21:56:04] <ju-emb> rebooting ....
[22:00:27] <ju-emb> whao andypugh knows where to loockup informations.
[22:00:29] <ju-emb> I didn't know that the MB-Version comes up on bootscreen
[22:01:06] <ju-emb> it's a A760G M2+ (A76GM113 BF)
[22:03:42] <ju-emb> stopping the chat and reboot I've lost the chat history.
[22:03:44] <ju-emb> someone can post the link to the catalyst 14xx driver again, thanks
[22:04:58] <memleak> http://support.amd.com/en-us/kb-articles/Pages/latest-linux-beta-driver.aspx
[22:04:59] <andypugh> It looks like no-one has ever mentioned that board on LinuxCNC
[22:05:35] <ju-emb> I'm not using it for realtime, it's my developer machine
[22:07:26] <ju-emb> I've to write some g-code to make gears on the 4th axis, and it's better do that on the desk in a nice chair than in the shop standing in front of the machine
[22:07:59] <ju-emb> so I tried the sim on my devel machine
[22:10:24] <ju-emb> LSPCI say's [AMD/ATI] RS780L [Radeon 3000]
[22:10:26] <ju-emb> so the catalyst Beta driver will not help
[22:17:14] <memleak> for older GPUs (pre HD series) you'll need these: http://support.amd.com/en-us/download/desktop/legacy?product=Legacy2&os=Linux+x86
[22:19:05] <memleak> *pre 5000 HD series
[22:21:23] <ju-emb> That driver is sitting on my hdd in the download folder already,
[22:21:25] <ju-emb> means some time ago I tried it but as I remember it slows the system down so that it was impossible to use
[22:23:06] <ju-emb> OK, it seems there is no easy way to make that work on my machine.
[22:23:08] <ju-emb> So I'll do a night-shift ssh-ing into the shop machine and doing the gear programming