#linuxcnc-devel Logs

Nov 22 2019

11:08 AM CaptHindsight: why is OpenCN recommending nvidia or intel gpu hardware yet having issues with GPU DMA?
11:11 AM CaptHindsight: the trajectory planner might be an improvement
11:11 AM CaptHindsight: they used the ipipe version of Xenomai
11:11 AM CaptHindsight: they seem to be unaware of the recent RTAI kernels
11:13 AM CaptHindsight: ouch ugly buildroot used there as well
11:15 AM CaptHindsight: RTAI is up to 4.14.148
11:16 AM CaptHindsight: looks like Ethercat was the main goal
11:18 AM CaptHindsight: the bullet points in their video are strange
11:19 AM CaptHindsight: new kernels are not an issue
11:19 AM CaptHindsight: LCNC is not tied to Debian
11:20 AM CaptHindsight: TurBoss: did not notice you in here :)
11:20 AM TurBoss: hello
11:22 AM CaptHindsight: we have been able to dedicate individual cores to RT and non-RT
11:22 AM CaptHindsight: not sure why that is some new bullet point
11:23 AM TurBoss: can the traj planner be backported if any better?
11:23 AM CaptHindsight: if they used AMD GPU's they could help from the Radeon devs directly
11:23 AM CaptHindsight: not going to happen with nvidia or intel
11:25 AM CaptHindsight: TurBoss: have to take a look at how they integrated it
11:26 AM CaptHindsight: https://gitlab.com/mecatronyx/opencnc/opencn
11:29 AM CaptHindsight: based on the video I'm not expecting much
11:30 AM TurBoss: there is no python interface?
11:31 AM TurBoss: interesting
11:36 AM TurBoss: https://gitlab.com/mecatronyx/opencnc/opencn/blob/master/buildroot/CHANGES
11:37 AM CaptHindsight: back to 2009
11:39 AM CaptHindsight: don't have time to try it for a while
12:10 PM mozmck: no python interface is a plus :-)
12:11 PM mozmck: I suspect the TP could be backported, but it sounds like it would be non-trivial.
03:16 PM CaptHindsight: "Qt : yes the new GUI is written in C++ (although I like very much Python"
03:18 PM CaptHindsight: 100uS thread for EtherCAT in OpenCN
03:19 PM CaptHindsight: ok so nvidia and Intel GPU's gave them issues with 10KHz!?
03:20 PM CaptHindsight: oh well, I'd be interested to see any latency testing they did
05:29 PM skunkworks: I might be too intoxicated to figure out how to even build it..
05:33 PM hazzy-m: skunkworks: well, according to the Ballmer Peak you may be in just the right state to try building it
05:34 PM hazzy-m: https://xkcd.com/323/
05:35 PM skunkworks: Heh - Hilarious.
05:36 PM skunkworks: hazzy-m: you jumping ship then? ;)
05:38 PM hazzy-m: nah, I'm too lazy
05:40 PM skunkworks: Jerk limiting always gives me a little arousal... but so far - no one has created a solution that obeys constraints...
05:48 PM hazzy-m: and for machine tools its not that big of an advantage. For a router or printer it would be nice though!
05:50 PM jthornton: laser and plasma may benefit from jerk control, but how to you keep a jerk from running a machine?
05:52 PM skunkworks: Umm - passwords?
05:52 PM jthornton: enter password = NoJerksAllowed23
05:54 PM pcw_mesa_: I would think that if you are cutting at a constant speed and material removal rate, the jerk would be controlled by the gcode more than the machine
05:55 PM jthornton: what's a good current motherboard?
05:56 PM pcw_mesa_: Dont know, haven't bought one in a while
05:56 PM jthornton: me neither
05:57 PM pcw_mesa_: I think the last one I bought was a J3355 or J3455 a couple years ago
05:57 PM Tom_L: dunno how current but i've been using gigabyte H110M-S2H-GSM & B250M-DS3H
05:57 PM pcw_mesa_: well other than a RPI4
05:58 PM jthornton: yea that's the last one I bought too a Rpi4
05:59 PM pcw_mesa_: skunkworks do you have a image of a RPI4 SD card (with RT kernel)?
05:59 PM hazzy-m: only one I've EVER bought is a PPI4 lol
05:59 PM * hazzy-m is a scavenger
06:00 PM jthornton: how can you make an image of your Rpi4?
06:00 PM pcw_mesa_: I used the "etcher" utility
06:00 PM skunkworks: It is official.. Horzog is a control snob...
06:00 PM jthornton: ah I have that
06:01 PM Tom_L: https://blog.cloudkernels.net/posts/rpi4-64bit-image/
06:01 PM skunkworks: jthornton: I don't have a image...
06:01 PM * jthornton ponders how to use etcher without a cd drive...
06:01 PM pcw_mesa_: I was trying to avoid kernel building on the RPI
06:02 PM jthornton: pcw_mesa_, I made one from scratch that is very basic
06:03 PM pcw_mesa_: I guess I could try that, the SPI conneced boards work reliably at 1 KHz on the RPI3 so maybe even better on the RPI4
06:04 PM jthornton: https://gnipsel.com/linuxcnc/uspace/
06:04 PM jthornton: step by step
06:32 PM skunkworks: I have used those directions more than once...
06:33 PM skunkworks: I did have a following error a couple nights ago... Still not 100% trusting the platform yet
06:41 PM jthornton: stepper?
06:46 PM jthornton: https://www.youtube.com/watch?v=v5OCL6rL5OY&feature=youtu.be
06:50 PM skunkworks: yes
06:50 PM skunkworks: what is that?
06:51 PM skunkworks: https://www.youtube.com/watch?v=ei5XQRW8aaE
07:12 PM pcw_mesa_: skunkworks: do you have the packet error hal code?
07:14 PM pcw_mesa_: (so if a packet times out it doesn't to the FE test nor apply PID corrections)
08:00 PM skunkworks: pcw_mesa_: yes. but the packet error doesnt increase
08:00 PM skunkworks: i still think it is some sort of frequency scaling.... i have not tested that though..
08:05 PM Tom_L: pcw_mesa_, i'm wondering if anything has significantly changed on making bit files that i would need to update my page for that