#linuxcnc-devel Logs

Jul 05 2017

08:32 AM jepler: jthornton: https://emergent.unpythonic.net/files/sandbox/0001-for-jt-an-iso-with-mate-and-master.patch
08:49 AM JT-Shop: jepler: thanks I'll give that a whirl in the morning during free time
08:52 AM jepler: you might also just want to review hooks/customise.sh because it is basically just a shell script that runs on a freshly installed debian OS (without even a DE)
08:52 AM jepler: perhaps you'll figure out why your steps/script never found the buildbot master branch debs
09:04 AM jthornton: ok thanks I'll look at that too
10:15 AM skunkworks: 2nd gen i5 - quad core. <5us
10:15 AM jepler: skunkworks: I think that number is too good to be true
10:16 AM jepler: you should put a parport on it, set it to toggling at 50kHz, and use a scope to find out the actual jitter
10:16 AM skunkworks: jepler, I was going to - today :)
10:16 AM skunkworks: just to see.
10:16 AM jepler: yay
10:17 AM skunkworks: well - I don't have a way to measure jitter... But I wanted to see if linuxcnc would actually run a say 20us base period consistantly
10:17 AM skunkworks: or 25us
10:17 AM skunkworks: that would be a normal rtai style setup
10:18 AM skunkworks: I am running a base thread and servo thread on this system... (latency-histogram)
10:18 AM skunkworks: both are less than 5us
10:19 AM jepler: skunkworks: btw "hwloc-nox" will print information that lets you understand which isolcpus= numbers refer to HT pairs
10:20 AM jepler: "hwloc-nox" is the package, "lstopo" is the program
10:20 AM skunkworks: thanks - I was using hwloc which gives you a graphical.
10:20 AM jepler: ah that's fine too
10:20 AM jepler: it's the same program
10:20 AM skunkworks: this system shares l4 cache..
10:21 AM skunkworks: over all 4 cores but doesn't seem to be that great of a problem
10:21 AM jepler: huh my system doesn't have L4 cache. But its L3 cache is shared among all CPUs. I think that's typical.
10:21 AM jepler: or all CPUs on one die anyway
10:22 AM jepler: $ egrep '^processor|^core id|^$' /proc/cpuinfo
10:23 AM skunkworks: the j1900 that I talked about yesterday has no shared cache between the 0,1 and 2,3 cpus
10:23 AM skunkworks: I think it only has up to l3
10:27 AM jepler: https://en.wikichip.org/wiki/intel/microarchitectures/silvermont
10:28 AM jepler: OK, you're right. on silvermont the last level cache is shared between 2 CPU cores
10:28 AM jepler: I'm still not sure if you're using the same cache level numbers as everyone else though
10:33 AM skunkworks: it is running a 30us base thread
10:33 AM skunkworks: 20,25 caused realtime delays
10:35 AM mozmck: So isolcpus helps on preempt-rt! For some reason I was under the impression that it was not needed or would not help.
10:45 AM skunkworks: jepler, yes - you are correct. even though the latency test seems to run for hours at <5us - you cannot run a 35us base period in linuxcnc
10:45 AM skunkworks: without realtime errors
10:55 AM pcw_home: I don't see any latency improvement on this system with isolcpus ( with either Preempt-RT 4.1.38 or 4.9.27)
10:55 AM pcw_home: (core duo dual core 3.16 GHz isolcpus=1) always around 30 usec eventually if i run videos etc
11:22 AM jepler: OK, so I have a hostmot2 stepgen commanded in velocity mode, velocity-cmd=1, position-scale=1000000, steplen=stepspace=250ns
11:22 AM jepler: no dpll
11:22 AM jepler: I read back position-fb, ddt, subtract 1
11:23 AM jepler: this last value I call "err", because it should be 0 if everything is perfect
11:23 AM jepler: .. but it actually tells me something about when the PCI reads occur
11:24 AM jepler: usually it's around -.002 which I assume is a difference between nominal and real system clocks
11:25 AM jepler: anyway, if I see e.g., a spike of err to +.06 in one cycle, can I tell something about the latency of the PCI read in that cycle?
11:25 AM jepler: .. is it 6% of 1ms ~= 60us?
11:32 AM jepler: the good news is, on my system the error from stepgen feedback (err) is a lot like the error from timedelta's timestamps (err2) https://emergent.unpythonic.net/files/sandbox/latency-vs-mesa.png
11:35 AM jepler: hal file https://emergent.unpythonic.net/files/sandbox/lat.hal -- you'll need to change the hm2_pci load line probably
11:35 AM jepler: .. I have a 5i24 that enumerates first and a 5i20 that enumerates second, hence the ",firmware="
11:57 AM skunkworks: hmm - I only have mesa ethernet boards here at the moment.
12:07 PM jepler: the result would still be interesting, assuming you turn off the dpll
01:10 PM skunkworks: jepler, you ran it as a strait hal file?
01:22 PM jepler: skunkworks: yeah with halrun -I
01:22 PM jepler: so then I interactively loadusr halscope or whatever
01:37 PM skunkworks: thanks
01:49 PM jepler: cradek: http://www.machinedesign.com/cables-connectors-enclosures/when-flexible-cable-doesn-t-flex-long
01:51 PM jepler: skunkworks: did you use any special wire when rewiring your heated bed? what wire gage did you use for the bed heater itself?
01:52 PM jepler: skunkworks: I've printed and assembled the drag cable but I haven't actually started the rewiring. I think first I'll run for a few dozen hours with the drag cable assembly installed to make sure it doesn't bind or self destruct.
01:56 PM skunkworks: jepler, I used 18g wire I scarfed from a power supply for the heater. The sensor I just used the smaller wire from the same loom..
01:57 PM skunkworks: *computer power supply
01:57 PM skunkworks: the heater takes less than 2 amps
01:57 PM jepler: oh that's less than I assumed
01:57 PM jepler: I figured the bed would account for half of the 10A rating of the PSU
01:58 PM jepler: so it's like 5.5 to 6ohm resistance measured at the bed?
01:58 PM skunkworks: I don't remember the resistance.. I just measured current inline.
01:58 PM jepler: sure.
02:00 PM skunkworks: sorry - 2.9a
02:00 PM skunkworks: Less than 3 amps :)
02:00 PM jepler: less than 2, less than 3, less than 6 ...
02:00 PM skunkworks: 35w
02:04 PM jepler: the table with the most google juice says 16AWG is OK for power transmission up to 3.7A so that seems fine
02:05 PM jepler: my colleague on the left is just worried about the wire lifetime as it's flexed 1000s of times
02:10 PM skunkworks: eh - at the radius the cable chain uses - I think you will be fine
02:12 PM skunkworks: just don't use solid wire ;)
02:14 PM jepler: did you save the original board-side connectors and splice onto them, or what?
02:18 PM jepler: [TFW the "detailed specifications" link takes you back to the store's front page]
02:19 PM skunkworks: yes - I just cut them long enough that I could splice them - used heat shrink tubing
02:22 PM skunkworks: running a 25us base thread on this installed i5
02:30 PM skunkworks: going to let it run over night
05:52 PM Tom_itx: pcw_home, does the 7I90 store the user bitfile onboard? if so is there a way to save the file off it?
05:56 PM pcw_mesa: Yes the 7I90 stores the bitfile onboard
05:56 PM pcw_mesa: I dont think currently that Mesaflash can read bitfiles from a device
05:56 PM pcw_mesa: but it would be a pretty simple feature to add since Mesaflash already has a verify option
05:57 PM pcw_mesa: (which reads the flash and compares it to s specified file)
05:58 PM Tom_itx: yeah
05:58 PM Tom_itx: just trying to recover from a crashed hdd the least painful way
05:58 PM Tom_itx: still looking to find the source and bitfiles elsewhere
05:59 PM pcw_mesa: mesaflash -- readhmid should get you a good idea of whats there
05:59 PM Tom_itx: yeah
05:59 PM Tom_itx: i'm a ways from that right now
06:00 PM seb_kuzminsky: i'm back in the world, i poked the crashed buildslave and things are moving forward again
06:00 PM JT-Shop: yea!
06:01 PM seb_kuzminsky: now it's time to take all these sunburnt kids to get ice cream! see y'all later!
06:17 PM jepler: seb_kuzminsky: thank you!
06:18 PM jepler: I'm not *sure* whether ice cream is an effective sunburn remedy but it's worth a try
06:18 PM jepler: the cold and the fat together shuold soothe the burns at least temporarily
06:18 PM jepler: maybe try mint?
06:27 PM Tom_itx: pcw_home, the 7I90 would be a 72 pin bitfile right? looking thru these backup files i have here...
06:37 PM pcw_mesa: yes the pinout source files will be pin_xxxx_72.vhd
06:38 PM linuxcnc-build: build #2682 of 1402.rip-wheezy-rtpreempt-i386 is complete: Failure [4failed install-missing-build-dependencies compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1402.rip-wheezy-rtpreempt-i386/builds/2682 blamelist: Jeff Epler <jepler@unpythonic.net>
06:38 PM linuxcnc-build: build #5023 of 0000.checkin is complete: Failure [4failed] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/5023 blamelist: Jeff Epler <jepler@unpythonic.net>
06:50 PM linuxcnc-build: build #2160 of 1903.clang-wheezy-amd64 is complete: Failure [4failed install-missing-build-dependencies compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1903.clang-wheezy-amd64/builds/2160 blamelist: Jeff Epler <jepler@unpythonic.net>
06:51 PM Tom_itx: thanks
07:02 PM linuxcnc-build: build #5006 of 1300.rip-precise-i386 is complete: Failure [4failed install-missing-build-dependencies compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1300.rip-precise-i386/builds/5006 blamelist: Jeff Epler <jepler@unpythonic.net>
07:02 PM linuxcnc-build: build #5009 of 1306.rip-precise-amd64 is complete: Failure [4failed install-missing-build-dependencies compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1306.rip-precise-amd64/builds/5009 blamelist: Jeff Epler <jepler@unpythonic.net>
07:06 PM linuxcnc-build: build #2159 of 1902.clang-wheezy-rtai-i386 is complete: Failure [4failed install-missing-build-dependencies compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1902.clang-wheezy-rtai-i386/builds/2159 blamelist: Jeff Epler <jepler@unpythonic.net>
07:06 PM linuxcnc-build: build #4218 of 1301.rip-precise-rtai-i386 is complete: Failure [4failed install-missing-build-dependencies compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1301.rip-precise-rtai-i386/builds/4218 blamelist: Jeff Epler <jepler@unpythonic.net>
07:07 PM linuxcnc-build: build #3167 of 1400.rip-wheezy-i386 is complete: Failure [4failed install-missing-build-dependencies compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1400.rip-wheezy-i386/builds/3167 blamelist: Jeff Epler <jepler@unpythonic.net>
07:08 PM linuxcnc-build: build #2683 of 1402.rip-wheezy-rtpreempt-i386 is complete: Failure [4failed install-missing-build-dependencies compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1402.rip-wheezy-rtpreempt-i386/builds/2683 blamelist: Jeff Epler <jepler@unpythonic.net>
07:08 PM linuxcnc-build: build #3168 of 1403.rip-wheezy-amd64 is complete: Failure [4failed install-missing-build-dependencies compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1403.rip-wheezy-amd64/builds/3168 blamelist: Jeff Epler <jepler@unpythonic.net>
07:12 PM linuxcnc-build: build #3369 of 1404.rip-wheezy-rtpreempt-amd64 is complete: Failure [4failed install-missing-build-dependencies compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1404.rip-wheezy-rtpreempt-amd64/builds/3369 blamelist: Jeff Epler <jepler@unpythonic.net>
07:13 PM linuxcnc-build: build #2834 of 1401.rip-wheezy-rtai-i386 is complete: Failure [4failed install-missing-build-dependencies compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1401.rip-wheezy-rtai-i386/builds/2834 blamelist: Jeff Epler <jepler@unpythonic.net>
07:21 PM jepler: these failures are still from the "buildprofiles" branch, which I learned after I pushed that it uses a feature not present in most versions of apt & devscripts and requires a pbuilder configuration change even where it is supported.
07:22 PM jepler: https://wiki.debian.org/HowToCrossBuildAnOfficialDebianKernelPackage is potentially interesting for anybody who wants to get into cross-building kernels for terrible embedded systems. :shivers:
07:31 PM linuxcnc-build: build #3177 of 1405.rip-wheezy-armhf is complete: Failure [4failed install-missing-build-dependencies compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1405.rip-wheezy-armhf/builds/3177 blamelist: Jeff Epler <jepler@unpythonic.net>
07:32 PM linuxcnc-build: build #5024 of 0000.checkin is complete: Failure [4failed] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/5024 blamelist: Jeff Epler <jepler@unpythonic.net>
09:06 PM Tom_itx: pcw_home, i use the seveni90.xise project file for the 7i90 parallel port mode right?
09:36 PM jepler: seems plausible
09:36 PM Tom_itx: building it now