where did you get the SMP RTAI kernel that you used for the power supply project?
linuxcnc.org I believe
ISTR playing with that myself shortly after I got my core2
I'd like to try it here - seems a shame to have a core sleeping
lemme see if I can find it again
hmmm. maybe cradek was hosting it ??
I couldn't find anything named "*smp*deb"
I did some googling earlier
I found a few IRC logs where we were discussing SMP and do-nothing tasks in passing, but not the actual discussion the day I was testing it
there'd probably be a link to the kernel near there
you'd think so
eventually, grep will finish looking at my local logs :)
ah - now I know why you have terabyte drives
aha - I think I found it
and not in my local logs :)
[01:56:59] <SWPLinux> http://www.linuxcnc.org/experimental/
use the 2.6.20 versions, I think 2.6.17 was the one with the kkeeyybbooaarrdd rreeppeeaattiinngggggg problem
glad I found it - I need to install it on another machine :)
to install those, I point sources.list at that repo?
I downloaded them and then installed with dpkg
or do I download the files
but gdebi will work also
if you download in firefox (on that machine), it will ask you if you want to open them with gdebi
it just did
you'll have to recompile emc2, of course (with the right realtime dir selected)
right - compiling is a given, since I'm doing hal-y stuff and will be compiling my own comp(s) as well
so far I can choose (at grub menu) between RTAI and normal ubuntu kernels, and at the login menu between xfce and gnome
another permutation should be fun
it's no biggie
too bad there isn't an easy way to have a switch (physical, connected e.g. to the parallel port) select which kernel to boot
which reminds me, I need to figure out my grub problem
apparently you missed the word "easy" :)
when I boot, it reads the MBR from the usb disk (the one I booted for the initial install), then fetches grub, from I don't know which disk, then fetches menu.lst from the hard disk, gives me a menu, and finally fetches a kernel from the hard disk
if the USB disk isn't in, it won't boot - either the HD MBR is broke, or there isn't a valid grub on the HD
actually, with multiple drives you can do it, you just need to use a DPDT and flip the cable select wires around (to switch which is HD0 vs HD1), but that doesn't work with SATA
did that last night
didn't seem to work
it's not 100% clear how the devices are enumerated
yeah, I was just thinking that
but it's USB, not a CF slot?
how much of a BIOS is there?
ie, can you turn off USB legacy support?
haven't read all the fine pring
I can turn usb boot on and off, or move it in the boot order
this screen is cramped
three of four smp kernel packages done
(getting the source too)
I was going to say ...
so, was this hard disk used before or blank?
it was infested with windows 98
that has been fixed
ok, so it should have a valid MBR
you could try booting it in a desktop machine
too much like work
(the disconnecting and reconnecting and such)
presumably you looked at the boot flag ...
ISTR I looked yesterday
I haven't messed with the "needs usb to boot" issue yet today
too many open tasks
I guess the worst solution is to leave the USB drive plugged in all the time
because it "works", I moved on to RT kernels and such
but I need (want) to fix it
the packages I just (think I) installed don't appear in synaptic
maybe because that repo isn't in its list?
dpkg -l shows linux-image, -headers, and rtai-modules for the standard RT kernel (2.6.24-16)
but only shows rtai-modules for the smp one (2.6.20-14)
that's cause the smp one is called -magma, not -rtai, and grep can't read my mind
bah, that kernel didn't add itself to menu.lst, and I forgot to edit it
no joy - added it to the list, it appears on the menu, but "file not found" when I select it
the image and initrd are in the /boot directory
wonder if I typo'ed something dumb
all the other kernels are 2.6.24-16-blahblah, the SMP one is 126.96.36.199-blahblah
. instead of -
gotta be something
so here's a funny one: http://www.newegg.com/Product/Product.aspx?Item=N82E16883220004
note the quantity discounts
jmkasunich: iirc, grub-install '(hd0)'
yes with the parens for some reason
$ cat /boot/grub/device.map
^ check this first
I have some uncertainty about tha...
the groot in menu.lst refers to hd1,0, which is uncommon - hence the uncertainty
didn't think to check the map
mine says 'root (hd0,0)'
experience has taught me to solve this booting stuff before spending much time on configuring the system...
what's that? ;-)
"mine" = your map? or your menu.lst?
sorry, my menu.lst
map is exactly what I pasted (one line)
# groot=(hd1,0) <--- one of those automagic comments
root (hd1,0) <--- every kernel in the menu
what's hd1 in device.map?
jmkasunich@robot:/boot/grub$ cat device.map
that would imply booting from sdb
no wonder it can't boot
so I wonder why is the hard disk sdb
Filesystem 1K-blocks Used Available Use% Mounted on
/dev/sda1 9275996 4312016 4496492 49% /
that is the hdd
ok, seems menu.lst and device.map are both wrong
err, device.map is fine
here is another bit of oddness
EMC: 03seb 07TRUNK * 10emc2/configs/hm2-servo/ (16 files): replaced m7i43_th with configs for all AnyIO boards
fix menu.lst to say (hd0,0) and run grub-install '(hd0)'
wait a minute
it looks like fstab is borked
/dev/sda1 / ext3 rw,relatime,errors=remount-ro 0 0
UUID=d251d562-180e-4a9e-9d16-47d54516c402 / ext3 relatime,errors=remount-ro 0 1
I just realised that the sdb part is a comment
the UUID refers to the HD
kernel/boot/vmlinuz-188.8.131.52-magma root=UUID=d251d562-180e-4a9e-9d16-47d54516c402 ro nosplash
has UUID _and_ a root
root is the 'bios drive', UUID is the magic search for the root partition
so, fix the comment in fstab to avoid hurting my lil' brain
fix menu.lst root
remove the line in device.map that refers to the usb stick
and then do grub install
IMO, yes, all disclaimers apply
broke it good
still no grub at all with the usb stick out
I get grub with the stick in, and it loads menu.lst from the HD, but when I attempt to boot a kernel from the HD, I get file not found
it must be enumerating the devices wrong somehow
you can edit the grub settings before it boots - I forget how - read the screen
you mean to get back to where I was, booting with stick, using kernel from disk?
I mean getting it booted so you can try again
yeah, changed hd0,0 to hd1,0, and it is booting
did you check the bios settings for bogons about drive enumeration?
it's only sort of booting - X failed, I'm in a text mode root prompt
sorry I "helped"
bios has "hard drive order" under the boot menu
hard disk is first, usb second
(it actually identifies them by make, model, etc
restoring the 2nd line of device.map got me back to X
apparently the x login screen uses that mapping to find its parts
I don't understand that
EMC: 03seb 07TRUNK * 10emc2/configs/hm2-servo/ (8 files): get rid of CVS version clutter
I think I have two independent problems
1) not booting at all from hard disk
2) grub trying to fetch kernels from the wrong place
should probably fix #1 first
not sure you've managed to install the mbr there yet
grub-install is supposed to do that
but if the enumeration is wrong, maybe I just installed it on the usb
fdisk says no partitions are bootable
gotta fix that, right?
changed and written, lets see what happens
that should not have worked, but it did
(it should have gotten into grub, then failed because of the hd1 thing)
oh, I never "unfixed" menu.lst
it still has hd0,0 as root
so it should have worked, and it did!
actually, I should start a latency test
glxgears cruising smoothly (300+ fps), latency 14760
opened foxfire, tried youtube but I don't have the player
opened gimp, no change
what system is this?
the little atom
jmk-robot, is this with the SMP kernel?
smp kernel won't drive the network interface
I may boot into it tomorrow and try the test, I don't need a net to run the maze
ok, I wasn't sure that was the end result
seems to be known bugs in the 2.6.20 version of that driver
yeah, or video (X) really
If I want an smp kernel, I'll have to build a newer one
restarted the test - if I don't start glxgears, I'm at 8510
glxgears instantly pops it up to 14xxx (14819 this time, 14760 last time)
have a url for the board?
I got mine from newegg, someone told me they're out now
there are several boards that are similar to it, but no idea on latencies
somebody robbed my xorg.conf
would be neat to get one for my car
[05:10:05] <SWPadnos> http://www.newegg.com/Product/Product.aspx?Item=N82E16813121359
there is nothing of substance in there
yeah, 8.04 (and later) xorg.conf files have no examples or markers for where you need to change things
not just no markers
there is nothing in there to tell you what it is using
no sections whatsoever
Identifier "Configured Monitor"
Identifier "Default Screen"
Monitor "Configured Monitor"
Device "Configured Video Device"
yeah, it auto-detects every time unless you tell it not to
/var/log/Xorg.0.log will tell you
fscking gawd damned microsoft followers!
keep your bloody hands off my computer!
huh. if you search NewEgg for "945GC", there are 31 items
and some are quite inexpensive - like $40
lower with rebates
not small tho?
mini-ITX, and some are shuttle mini systems
oops, the inexpensive ones are micro ATX, which I think is just slightly smaller than ATX
(and larger than mini ITX)
it's not as small as I thought
the non-2 unit from Intel (same PN, no 2 at the end) is there, $69.99 with CPU
I think that is single core
her's my new favorite (but not for EMC): http://www.newegg.com/Product/Product.aspx?Item=N82E16883220004
it was reviewed in the same article in EDN
looks like it is using the "i915" video driver
might be interesting to see what happens with vesa
holding at 15773
btw, make -j<some-reasonalbe-number> doesn't crash the machine
and with the non-RT SMP kernel, compiles go about twice as fast with j3 as without j
and on that note, goodnight
that makes sense
EMC: 03seb 07TRUNK * 10emc2/configs/hm2-stepper/ (16 files):
EMC: replace hm2-stepper config with one that handles all AnyIO boards
EMC: (and is tuned for the steppers on my workbench)
seb_kuzminsky: whee.. that looks nice
it's probably lots easier on users aswell
EMC: 03alex_joni 07TRUNK * 10emc2/tcl/tkemc.tcl: apply fix from Dewey Garett - fix angular jogspeed value bug (if no DEFAULT_ANGULAR_VELOCITY was given any angular jogs were dropped)
When I've set motion.probe-input while homing there are two messages about error, "Probe tripped during homing motion.
Probe tripped during a jog."
better than no message :D
those are both features. they would have saved your probe by stopping motion
cradek_ is now known as cradek
I suspect the incorrect message comes from the free mode controller
but .. better 2 messages than none at all
oh you get them both at the same time?
that is a bug (minor) if so
cradek: yes minor, but bug
alex_joni: yeah thanks for putting that suggestion in terms i could understand ;-)
micges_emc: what version?
hm, sometimes I get only one error, sometimes 2 or 4
up to two for each moving joint I think
now there are 18 messages...
EMC: 03cradek 07TRUNK * 10emc2/src/emc/motion/control.c: make sure the user gets the correct error, and only one error, when aborting due to an unexpected probe trip.
seems like the last few months have really added tons of features/fixes. Very cool. Nice work.
there has certainly been a burst of activitiy
I hope that's good...
(adding big features late is sometimes a bad idea)
na - just means they are in there if they need to be fixed. :)
seb_kuzminsky: hi again
was wondering if you have some thoughts about the 7i64
I understand there's some support for it in the hostmot2 firmware, so the driver would need work, etc
cradek: that's why I built in a month for no new features before the betas and before creating the new branch
true (and smart)
yup, if people keep testing/fixing stuff :)
alex_joni: that's right, the hm2 firmware supports spi uarts, there'd need to be driver support
peter and i talked about this a bit
was wondering if there's a timeframe..
there's no real good way to probe for the slave at the end of the spi bus, but the driver needs to know what it is
it's not on my short list ... :-(
I'm not trying to push you..
the driver would have to have a 7i64 sub-driver (like it currently has encoder, stepgen, etc)
a 'long period' is just as good for me atm
the user would have to tell the 7i64 driver to attach itself to a particular spi port
is the SPI p2p ?
point 2 point?
nothing too complicated software wise, though it makes the load-time configuration string even more awkward
i think so but i'm not sure
ok.. then it's a bit easier than point 2 multipoint :)
having more than one 7i64 on one spi port..
Alex: for 7I64 it will probably always have only one SPI per per card
PCW: ok, cool.. thanks
update speed gets slow if you chain a lot of them
Yes that why they are not chained
do you recall offhand how many 7i64 are supported in the stock hostmot2 configs?
SWPadnos: 0 in the BIT files we currently have
As many as you want basically. We have a little breakout board that supports 6 from one 50 pin FPGA conn
I thought you had meant before that they're in the hardware but not supported in the driver yet
ok. and what's the update rate/cycle time?
They are on the hardware, but not ant of the bitfiles EMC uses ATM
SWPadnos: afaik, the hm2 firmware source supports spi uarts, but no bitfiles exist that include them, and the driver doesnt support them
ok, my misunderstanding - no problem
About 3/4 uSec for 24in/out
(3 or 4 )
how long a cable do you suppose can be used to connect them to the anyIO board?
About 3 feet or so, its terminated TTL levels
of course it's an update speed tradeoff
lower clock rate may allow longer cable
does the SPI block have adjustable read skew?
Not yet Every time I look at it I find something else to do...
only a few crazies like me need it anyway :)
Actually I do want read skew adjust for the 7I65 anyway just have to grit teeth a bit-wrangle
Still need to finish autosend for buffered SPI interface also
hmmm. what A/D is used on the 7i65?
I wonder if there's a 16-bit pin-compatible replacement
I kind of doubt it
not at 1MSPS of course
Probably something similar though
hmmm. 8 channels / bipolar / 16 bits is a tough one
there's a 14-bitter
The 7I65 is not designed as a HS DAQ system it uses shared SPI data for cheapness, the 1 MSPS is not really usable
I just liked the AD7329s input ranges and shift clock
OK. Im also making a little test A-D for THC
BigJohn will test it
well, that's good news and bad news. the bad news is that I can't just buy some 7i65 from you to replace the analog board I designed. the good news is that I actually make money on those boards :)
The 7I65 is really designed for servo use
The A-D is just a added gimcrack
? a showy object of little use or value
Well I doesn't add much to the 7I65 cost and several people have asked for A-D capability
with A/D it's possible to use analog tachs and close the velocity loop separately from the position loop
not if you have only one
the 7i65 has 8
8 analog outputs, 8 analog inputs, 8 quadrature encoder inputs
missed the 8 x AD input
The 7I65s muxed encoders will need driver support as well (encoder counter register map is same but I/O allocation is different)
how does the muxing work anyway?
I saw you mention that earlier, but didn't understand where the muxing takes place
(can't see how it could be the input pins, possibly overlaid registers ...)
7I65 has 2-1 muxing so 8 encoder inputs use 13 wires instead of 24
12 even/odd muxed encoder signal to FPGA and one mux drive out from FPGA
Demuxing in FPGA before counters hides any difference to software
(other than I/O pin setup)
if you have a 7i65, can those extra encoder input pins be used as GPIO?
(with the muxing still enabled)
that would be a PITA for the software :)
Not easliy but I have considered a virtual Input and virtual output pin PinDesc would make the accessible
that would potentially be useful for the 7i64 as well
since you get more I/Os than there are pins on the AnyIO board
or pins used anyway
Well back to layout
but you have to know the mapping of physical board pins to external virtual ones
oh darn, I missed him
I was just reading this http://www.linuxcnc.org/docview/html//motion_kinematics.html.
'axes' is correct according to Websters and Dictionary.com
axes is defined as the plural of axis
maybe it s/b axi or axii but it seems axes is suitable :)
stuste1: can you be more specific what you mean?
"When he is coming over, I'm always careful to hide the axes"
I've always been troubled that axes and axes are spelled the same
at the bottom of the .html page is states axes is incorrect usage
stuste1: yeah in favour of joints
not some other spelling
stuste1: I think the objection is to using axis to refer to one motor/slide (joint)
it's not about axes being an incorrect plural of axis
I am just reading to edukate myself on the kinematic data flos
The word .axes. is also commonly (and wrongly) used when talking about CNC machines, and referring to the moving directions of the machine.
moving directions of a CNC are joints
even if carthesian axes are aligned to the same directions
yes I understand the difference
I have had a crash course in kinematics the last few weeks
I can imagine
when it seems I am almost there I find the end of the tunnel moved farther away
heck did you run into another problem?
gdb is sweet - I get the values I want for most of it
not in the C or kinematics - in the flow of data through the control
hmm.. how come?
when I get a better handle on it I will be back with coherent questions
I have to run
ok.. see you later
I will be in Topeka this afternoon
bbl - thanks
EMC: 03jepler 07TRUNK * 10emc2/src/emc/usr_intf/axis/scripts/axis.py: allow the filename ~/.axisrc to be overridden by inifile [DISPLAY]USER_COMMAND_FILE (based on a patch from Dewey Garrett)