#linuxcnc-devel Logs

May 13 2020

10:17 AM rene_dev_: jepler the new rendering pipeline is getting along good, but depends on pyglm, which isnt available :(
10:22 AM rene_dev_: currently its only focused on gtkglarea, do you want to look into the gl stuff for tk, or shall I do it?
11:12 AM andypugh: jepler: What did you do to provoke the RTAI crash using only RTAI code and no LinuxCNC?
11:13 AM andypugh: And is there any reason to think it is the same crash as we see with LinuxCNC?
11:15 AM skunkworks: andypugh:
11:15 AM skunkworks: root@debian10rtai:/usr/realtime-4.14.174-rtai-amd64/testsuite# while sync; do /usr/realtime-4.14.174-rtai-amd64/bin/rtai-load /usr/realtime-4.14.174-rtai-amd64/testsuite/.runinfo; sleep .1; /usr/realtime-4.14.174-rtai-amd64/bin/rtai-load cleanup; echo Iteration $((++i)) done; done
11:15 AM skunkworks: I think
11:21 AM skunkworks: andypugh: http://tom-itx.no-ip.biz:81/~tom-itx/irc/logs/%23linuxcnc-devel/2020-05-07.html
11:23 AM andypugh: I just trried to set up userspace RTAI but I am not sure it works
11:47 AM andypugh: It seems that user-space RTAI does work, and fairly well.
12:04 PM jepler: andypugh it's the same kernel stack in both cases for me. It matched one of the 4 that Seb had captured
12:05 PM andypugh: I am not seeing anything like that, peresumably because I am doing everything via ssh
12:06 PM andypugh: (The test PC is upstairs in the spare room / hobby workshop)
12:06 PM andypugh: I am trying the repeated abs.0 test with rtai uspace, just to see.
12:07 PM andypugh: It seems to work, though takes an odd amount of time to start up.
12:08 PM andypugh: (ie latency runs at about 4µS after an initial 1 second latemcy)
12:09 PM andypugh: Is there a use-case for RTAI-uspace? (hm2-eth maybe?)
12:16 PM jepler: Hm2-eth won't work, it uses regular Linux network drivers
12:19 PM jepler: In theory, the uspace+rtai lxrt approach lets us - serve rtai and preempt-rt users with one single build (packaging work needed)
12:19 PM jepler: - no need for kernel modules, no direct dependency on kernel version as long as it is either preempt-rt or rtai
12:19 PM jepler: - rtai never needs to enable kernel-mode math for us again; no need to worry about the cflags rtai wants for kernel interfering with our needs
12:20 PM jepler: - can transition to using C++ features in HAL modules, where they are not incompatible with realtime guarantees
12:20 PM jepler: but that assumes that uspace+rtai lxrt has the same reliability and latency has kernel-model rtai and we just don't know yet
12:20 PM jepler: (a 1s latency at startup? ow)
12:22 PM andypugh: Is it a latency or a setup time? I am not clear.
12:23 PM andypugh: For whatever reason the repeated abs.0 test is running a lot slower. It has been running about half an hour and has done 670 iterations.
12:24 PM jepler: not sure
02:33 PM rene_dev_: any glade expert here? I cant get glade3 to use python3 when loading the catalog
03:02 PM pcw_home: does anyone know the magic that makes the spindle tuning tab show up in the "calibrate" utility?
03:08 PM Tom_L: http://linuxcnc.org/docs/html/config/stepconf.html
03:08 PM Tom_L: pcw_home,
03:08 PM Tom_L: This page only appears when Spindle PWM is chosen in the Parallel Port Pinout page for one of the outputs.
03:09 PM Tom_L: or were you looking for something else?
03:10 PM pcw_home: Something else: what determines the tabs that the "calibrate" utility displays
03:10 PM Tom_L: mmm, i've never used it
03:12 PM pcw_home: its quite handy for servo configurations because it allows you to tune "live" and then save the changed settings back to the ini file
03:12 PM Tom_L: yeah i can see it as a time saver
03:15 PM pcw_home: A forum user asked how to add the spindle to the calibrate tabs And I assumed it was just the ini file section header because it works for me with "SPINDLE_9" but it appears to be something else
03:15 PM pcw_home: maybe a LinuxCNC version issue, not sure
03:19 PM jepler: pcw_home: there is a process by which the calibration script tries to detect items in the inifile which are used by halcmd setp. I doubt anybody is really familiar with the details at this time.
03:19 PM jepler: tcl/bin/emccalib.tcl:proc find_ini_refs
03:20 PM pcw_home: Ahh thanks I'll take a look.
03:20 PM pcw_home: wonder if that changed recently
03:20 PM jepler: the code is hairy, sorry
03:37 PM andypugh: I seem to recall a recent commit on this subject
03:39 PM andypugh: Ah, yes: https://github.com/LinuxCNC/linuxcnc/commit/740c4252aa09be3e16eed5629c2892e3d2643a2d
03:39 PM andypugh: [TUNE]
03:39 PM andypugh: Not an answer to the queston, though.
03:41 PM andypugh: jepler: I am running the RTAI load/unload script from the other day. How many cycles do you typically get before failure?
03:41 PM pcw_home: The forum user had [SPINDLE_9] And I though that worked for quite a while
03:52 PM andypugh: I wonder why Chris chose _9 for the spindle?
03:53 PM linuxcnc-build: build #45 of 1660.rip-buster-python3 is complete: Failure [4failed compile runtests] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1660.rip-buster-python3/builds/45 blamelist: Chris Morley <chrisinnanaimo@hotmail.com>
04:20 PM linuxcnc-build: build #6841 of 0000.checkin is complete: Failure [4failed] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/6841 blamelist: Chris Morley <chrisinnanaimo@hotmail.com>
04:46 PM jepler: tests/interp/m98m99/12-M99-endless-main-program failed to print a final "disable". Flaky test, or actual problem?
04:47 PM jepler: USRMOT: ERROR: command timeout
04:47 PM jepler: shmctl(1835532294, IPC_STAT, ...): Invalid argument
04:47 PM jepler: shmctl(1835499525, IPC_STAT, ...): Invalid argument
04:47 PM jepler: libnml/os_intf/_shm.c 412: rcs_shm_nattch: shmctl failed: Invalid argument
04:47 PM jepler: seems a bit unexpected at any rate
04:55 PM skunkworks: pcw_home: it stays around 400us unless you open a browser
05:00 PM pcw_mesa: Thats better than mine, maybe kernel/setup issue
05:00 PM andypugh: jepler: That does seem odd. There was a recent commit that peripherally touched M99
05:02 PM andypugh: https://github.com/LinuxCNC/linuxcnc/pull/852 fixed a problem that M99 created.
05:02 PM skunkworks: pcw_home: i am isolcpus 1,2,3
05:03 PM skunkworks: and 18000mhz
05:08 PM pcw_mesa: I think I may be 1500 MHz (and isolcpus 1,2,3)
05:08 PM skunkworks: ok
05:09 PM pcw_mesa: videos cause trouble
05:10 PM skunkworks: yes.
05:11 PM skunkworks: instant over 1ms running yt. although at 500hz it might be fine. maybe that is why i was running that period.
05:12 PM pcw_mesa: did you notice any performance difference at 500 Hz? ( crunchier accel etc )
05:13 PM skunkworks: no
05:19 PM pcw_mesa: adding acceleration to the stepgen should allow lower update rates
05:23 PM pcw_mesa: though I think the dpll quits around 200 Hz
05:27 PM skunkworks: heh. like i say. did all the poly testing thinking i was running 500hz
05:35 PM memfrob: andypugh, saw your email. rtai-config and the make bzImage lines have nothing to do with each other.
05:35 PM memfrob: OpenSUSE and Debian did something incredibly stupid to GNU make and essentially broke building a kernel and the modules, without running the following commands:
05:36 PM memfrob: make all && make bzImage && make && make all && make modules && make && make modules && make modules_install && make && make modules_install
05:37 PM memfrob: That will pretty much guarantee everything gets built. I have no idea what the hell they did but building kernels and modules sometimes requires several calls to `make` over and over again.
05:38 PM memfrob: On my Gentoo systems, it's simply just make && make modules_install and everything just works.
05:39 PM memfrob: Then I use `cp` and install the kernel, config and System.map to /boot, done
05:39 PM andypugh: I didn’t think that they were necessarily the same problem, but rtai-config seems to say that the modules are in a different place to where make install puts them. As far as I can say the makefile has nothing in it for make modules or modules install. I can try it later, but currently the system is 20,000 cycles in to the “Jepler test”
05:40 PM andypugh: Ah, you know, I am an idiot?
05:40 PM Tom_L: skunkworks, you overclocking your rpi4?
05:40 PM memfrob: Your rtai-config might be broken due to a remnant rtai-config in a different dir
05:40 PM memfrob: $ type rtai-config
05:40 PM andypugh: That’s the explanation for all of this
05:41 PM memfrob: Yeah, andypugh you need to be more careful with this stuff and double-check.
05:41 PM andypugh: I am looking at the kernel build instructions, not the RTAI build instructions.
05:42 PM memfrob: I recently made RTAI's configure.ac more robust and fool-proof in terms of installation prefix and kernel directory
05:42 PM skunkworks: Tom_L: just a little
05:42 PM memfrob: March 16: https://github.com/NTULINUX/RTAI/commit/31a656ee6f57cfd18cc20bda6ab9f5d4e8c21491#diff-67e997bcfdac55191033d57a16d1408a
05:42 PM Tom_L: skunkworks, does it run hot?
05:43 PM Tom_L: 1800Mhz? really?
05:43 PM andypugh: memfrob: I think you might be right
05:43 PM andypugh: andypugh@rm-one:~$ which rtai-config
05:43 PM andypugh: /usr/local/bin/rtai-config
05:44 PM memfrob: yeah it's going to be harder to install there now.
05:44 PM memfrob: you'd need to do --prefix=/usr/local now
05:44 PM Tom_L: :/
05:45 PM Tom_L: just when i thought i had it figured out
05:45 PM memfrob: The rpi4 needs a very good heatsink, we had overheating issues too.
05:45 PM Tom_L: i got one andy recomended
05:46 PM memfrob: Are you trying to get 4K working?
05:46 PM Tom_L: 4k?
05:46 PM memfrob: 4K screen resolution
05:46 PM Tom_L: i don't have it yet
05:47 PM Tom_L: arriving tomorrow but no i dont need 4k
05:47 PM memfrob: Good lord we spent weeks trying to get 4K working, and mind you I'm a kernel and U-Boot guy, and eventually we gave up.
05:47 PM Tom_L: and here i thought you were gonna tell me how easy it was
05:47 PM memfrob: One of the problems was that the RPI4 has a non-standard HDMI controller, and it got our 4K Samsung monitor in a weird state
05:48 PM andypugh: I want a 6k monitor.
05:48 PM Tom_L: aren't there 8k out now?
05:48 PM memfrob: Yes.
05:48 PM memfrob: And 8K VR.
05:51 PM andypugh: (I already have 5120x2880 in my iMac)
05:52 PM andypugh: And it is rather lovely, in fact it is the main reason that I sit dowstairs at the Mac and ssh all my dev work. That and the whole “trying to crash the system” thing.
05:53 PM memfrob: I'm just waiting for a miniLED 4K 120 Hz TV :)
05:54 PM memfrob: Rather have higher FPS than higher resolution for gaming
05:55 PM memfrob: TCL is coming out with one that has 25,000 backlights
05:56 PM skunkworks: Tom_L: it is in one of those heatsink cases.. it is warm
05:57 PM Tom_L: i thought they ran stock at 700Mhz?
05:58 PM Tom_L: 1800? really?
05:58 PM pcw_mesa: I just have a 1"x1" stick-on heatsink and its fine
05:58 PM Tom_L: https://www.amazon.com/Metal-Raspberry-Heatsink-Thermal-Adhesive/dp/B07VMD25SK
05:59 PM Tom_L: getting one of those for it
06:02 PM memfrob: There ya go!
06:03 PM skunkworks: Tom_L: that is what i am using
06:04 PM Tom_L: what about the GPU?
06:04 PM Tom_L: do you overclock it too?
06:04 PM skunkworks: yes
06:04 PM skunkworks: a litlle. just makes the experience a bit better
06:04 PM skunkworks: ;)
06:05 PM Tom_L: what? cooking hot dogs?
06:06 PM skunkworks: gui responsiveness
06:06 PM Tom_L: how far is a little?
06:06 PM skunkworks: give me a few to look
06:20 PM skunkworks: Tom_L:
06:20 PM skunkworks: gpu_freq=600
06:20 PM skunkworks: over_voltage=6
06:20 PM skunkworks: arm_freq=1800
06:20 PM skunkworks: that - I think - voids you warrantee..
06:20 PM skunkworks: You can go to 1500 without voiding warranty
06:21 PM Tom_L: heh has anyone made good on a warranty anyway?
06:21 PM skunkworks: you also want force_turbo=1
06:21 PM Tom_L: why overvoltage?
06:21 PM skunkworks: to get to 1800 I needed to do that.
06:22 PM Tom_L: that's not volts, rather a setting right?
06:22 PM Tom_L: i'll probably play with it a while before i try to cook it
06:23 PM skunkworks: right
06:23 PM skunkworks: a scaling
06:27 PM skunkworks: https://www.tomshardware.com/reviews/raspberry-pi-4-b-overclocking,6188.html
06:30 PM Tom_L: so i have the go ahead to get a flir so i can monitor it right?
06:55 PM memfrob: a flir?
06:57 PM skunkworks: you need a what is it micro hdmi to hdmi cable
06:57 PM Tom_L: the kit comes with one
06:58 PM Tom_L: i was wisely advised
06:58 PM Tom_L: memfrob, thermal imaging
07:01 PM skunkworks: 51deg c
07:01 PM skunkworks: currently
07:01 PM skunkworks: the heatskink is warm to touch
07:01 PM Tom_L: can you inerrogate it for temp?
07:01 PM Tom_L: interrogate
07:01 PM skunkworks: yes
07:02 PM skunkworks: vcgencmd measure_temp
07:02 PM Tom_L: i wonder if there's a 'cpu-z' for pi
07:02 PM Tom_L: i ran it on my android and it showed all the active sensors it has
07:09 PM Tom_L: https://www.geeks3d.com/20191004/gl-z-0-5-0-released-raspberry-pi-cpu-and-gpu-monitoring-improved/
07:09 PM Tom_L: similar
07:12 PM skunkworks: neat!
07:12 PM Tom_L: skunkworks, did those frequencies require new firmware?
07:27 PM skunkworks: ummm.. I dont know..
07:29 PM Tom_L: it's likely newer than that article anyway
09:08 PM memfrob: CPU-Z to measure CPU frequency? I tried seeing what the actual MHz was on the CPU in /proc/cpuinfo and it wasn't there
09:08 PM memfrob: Only bogomips in dmesg
09:09 PM Tom_L: measures alot of the vitals
09:10 PM memfrob: vitals as in temps or freq?
09:10 PM Tom_L: both
09:10 PM Tom_L: did you open that link?
09:10 PM memfrob: lm_sensors might cover the first half
09:10 PM Tom_L: possibly
09:11 PM memfrob: ARM: Some things might actually work compared to x86
09:11 PM Tom_L: i actually used that once on a parallel port to measure some lm74? chips around the house
09:14 PM memfrob: What I find silly is that lmsensors needs to be calibrated in order to properly work in Linux. You're off by 20c out of the box and need to figure out how much enormous offset you need to be accurate.
09:15 PM memfrob: Step 1.) Use a laser thermometer
09:17 PM memfrob: GL-Z looks closed source.
09:19 PM memfrob: Application may not start on other distros depending on how many shared libraries it uses.
09:28 PM skunkworks: Tom_L: I remember updating the firmware at some point - but I don't know when
09:37 PM skunkworks: you guys rock. I cannot believe how much work has been done over the past few weeks.
09:40 PM Tom_L: i 2nd that
09:47 PM memfrob: I didn't do anything!
09:47 PM memfrob: lol