EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/ladder/classic_ladder.lyx: minor fix
EMC: 03bigjohnt 07v2_2_branch * 10emc2/docs/src/hal/comp.lyx: minor fix
EMC: 03bigjohnt 07v2_2_branch * 10emc2/docs/src/Master_Integrator.lyx: that looks better
EMC: 03jmkasunich 07TRUNK * 10emc2/TODO: note to self: clean out abandoned development
skunkworks is now known as skunkworks__
skunkworks__ is now known as skunkworks___
I see a Mshaver config in IDParams.vhd :)
swp: I was hoping I could do what you just said probably isn't possible
which, use external #defines or replace the two lines?
use external stuff
there isn't a #define in vhdl (I don't think)
well, it is possible if the c preprocessor can be used
the lines in hostmot2 that are relevant are:
(which says, I need the package IDROMParms"
ThePinDesc: PinDescType := PinDesc_JDosa66;
TheModuleID: ModuleIDType := ModuleID_JDosa66;
which take two constants from that package
the "use" line does NOT say "I need file IDparms.vhd"
the project file says which files are to be used
I think by definition, all files in a directory are considered "part of the project"
if I have 20 different files that define PinDesc, and put one of those names in the project file, it should? work
in addition to any specifically listed libraries (like IEEE.std_logic...)
no, I think you need to change the name
then why do they bother with a project file?
then change the two lines in the generic map
you can combine VHDL and Verilog (and schematics ...) in a single project
the rules are a little different there
also, it includes specific pin constraint files, etc.
so what happens if you have two files in a directory that both define the same symbol
say, stepgen.vhd, and stepgen-backup.vhd
I'm not sure
hmmm. that could be part of the problem I had way back with the analog board stuff
I had problems where the SPI module wouldn't change, and I may have made a backup copy of the source file before trying the changes
definitely check the docs though, I'm not 100% sure I'm right on this
what docs? the toolchain docs are clear as mud (and hiding somewhere)
I think I learned that tidbit about the "working directory" either online or in a course at ESC
ah, you have a default logical library called "work", which I believe is your working directory
i'm not here yet but i will be in an hour or two
ok, mr not-here
[01:54:47] <SWPadnos> http://toolbox.xilinx.com/docsan/xilinx7/help/iseguide/mergedProjects/xsim/html/xs_r_vhpcomp.htm
I wonder if vhpcomp is the program you want?
[01:55:12] <SWPadnos> http://toolbox.xilinx.com/docsan/xilinx7/help/iseguide/mergedProjects/xsim/html/xs_cl_vhpcomp.htm
no, it is a simulator
at least it has command line options :)
I've used ghdl to do the same thing, but Free
all the tools have command line options
in fact, I just ran a make, and it appears to have worked
using the existing arrangement for specifing the config
ok, so manually edit then remake
manyally edit I20Hostmot.vhd if you want to change the pindesc/moduledesc
you mean just to see what it does?
then remake (and get another hostmot2 with different stuff in it)
(my goal here is to NOT do that)
"using the existing arrangement for specifing the config"
wonder how many people will answer Tom's question
at least 1
your turn jmkasunich
I should have waited a few minutes.
I wonder if we'll all have the same answer
it's so easy to answer because I bet we have all done that.
who, me. never!
someday, all trucks will have fuel injection
that would be tricky in systems with more than one thread
yeah I don't know how you'd know what thread to use
ok, "someday, auto-addf-assist"
halcmd: loadrt abs
Would you like to add the function abs.0 to the servo thread?
that is useless
it needs to look at what signals are inputs to abs, and what signals it drives, and figure out the proper order
you'd have to pick one function to put in a thread (probably motmod's functions)
sure, the slowest thread that has a user of the output
then it would follow dependencies as far as it could, both upstream and down
is the thread it goes into, some time before that function runs (the consumer function)
how would you determine ordering?
and it would yell "help" when it couldn't resolve something
cradek: this assumes a significant HAL change, such that each pin of a comp has some metadata, saying "function foo reads this pin" or "function bar writes this pin"
the trouble is that you would have to fire off that scheduling function manually, unless you change the ordering with eack link
any function that has no read pins can be executed early in a thread
any function that has no write pins can be executed at the end of a thread
I think I see what you mean - you should be able to figure it out, or pick one reasonable alternative
SWPadnos: I don't think of this as a "live" thing
more like, loadrt, link stuff, run a program, it spits out a pile of addf commands
make haldepends > addf_file :)
oh hey - an interesting PBS show. bbl
SWPadnos: I think this is the doc I used to figure out the workflow and command lines: http://toolbox.xilinx.com/docsan/data/alliance/dev/dev.htm
jmkasunich, interesting. there is a way of specifying a chip type on the command line (as one would hope)
oh, maybe not
chip type is specified in the .scr file, which is input to the synthesis step
ok, the MAP program
at least that chooses an architecture type to map elements to
I have makefile code (from last year) that generates a .scr file and passes the chip info into the toolchain
good thing. it's not something you can scan the docs for ;)
I'm leaning toward an approach that uses a .in file
I20HostMot2.vhd.in + myconfig.in thru a python prog to make myconfig.vhd
myconfig.in is nothing more than one of the many stanzas in IDParms.vhd
I wonder if it's possible to split that giant file into separate files, each with one pindesc/moduledesc
and the common stuff in a separate file that's used by all those
that is the specific goal I've been talking about all evening
I still wonder if it's possible :)
occam's razor - that file started out with only the common stuff and one stanza of config in it
if that worked, then I should be able to return to that
I didn't look at the first 300 lines to see if there are things that might conflict or get duplicated with multiple files using (a) one copy included somehow or (b) multiple copeis of that common stuff in the work dir
the common stuff will remain in that file
right - it's a question of how to USE or #include that in multiple files that I don't know
I guess I should download the xilinx tools tonight (it'll probably take all night now)
myconfig.stuff + I20hostmot2.vhd = myconfig_i20hostmot2.vhd
myconfig_i20hostmot.vhd will be labeled as an autogenerated file (and make will regenerate it if either of the source files changes)
damn. I have an email with an authorization code for ISE 10.1, but I can't find the (2GB) file, which I'm sure I must have downloaded
find / -size ?
(dunno if there is a -size)
find on which computer? :)
how many do you have to try?
I checked my desktop and my laptop. I think it's on the new workstation
searching for size > 1.5G should narrow it down quite a bit
if you only count computers then 3 or 4, but it's more if you count hard drives
I think I might remember doing it on the new machine, but I don't know if that's before I installed 8.10 or after
or if I upgraded that machine rather than reinstalling
EMC: 03seb 07TRUNK * 10emc2/src/emc/rs274ngc/ (interp_internal.hh rs274ngc.hh):
EMC: made some error-checking/stack-managing/logging macros sane, to make gcc fortify happy
EMC: i dont *think* i changed what the code does...
i guess paul took his buildslave and went home
it is very late there
where paul lives?
his buildslave went down 7 hours ago and he hasnt been responding to email
i guess if he's in england that explains it
gcc's fortify is a big step forward for code quality, i think if paul doesnt get his act together i might try to set up a sid or intrepid buildslave
that would be nice to have
hm, intrepid and jaunty both have rtai packages in the ubuntu universe package repository...
is gcc 4.3 available for 8.04?
8.04 is like a million years old
come on backports guys, get your head in the game
um, you mean Hardy Heron, the latest distribution we have? :)
the very same, sir
LTS is nice and all, but it definately comes at a price
has anyone tried ubuntu's rtai packages for hardy?
hardy or intrepid?
I think it's just a source package - it's pretty small IIRC
[03:50:46] <seb_kuzminsky> http://packages.ubuntu.com/search?suite=default§ion=all&arch=any&searchon=names&keywords=rtai
* seb_kuzminsky apt-get installs rtai
I may soon. the box I have ISE on is an intrepid machine
unfortunately, the bastiges don't give away the 64-bit tools for free
oh it's only available for i386, and i'm on amd64
yes, for intrepid
ok, then I'll have the same problem
SWPadnos: the good stuff seems to be in rtai-source, 2.7 MB
crap. even the EDK I have a license for isn't available in 64-bit
"This package contains the patches for modifying the kernel, and sources for building the kernel modules, module dependent runtime tools and the testsuite. "
SWPadnos: ise the xilinx environment? the free one for linux runs just fine on my machine in x86_64 mode. iirc I copied from an install on a 32-bit machine and then changed 'lin64' to 'lin' at one place in settings.sh.
SWPadnos: I'd already prepared the machine to run 32-bit stuff (skype and flash plugin)
I'm sure it runs OK, but FPGA work is one of the few places where 64 bit support might actually be useful
yer just bein' a snob
SWPadnos: is that something i'd need more than a gig of ram to seriously believe?
heh - I'm not sure
for the 5i20 it shouldn't matter, for something that actually uses a significant portion of the 5i22 it might
oh, isn't that funny. EDK is supported in 64-bit on Linux, but not on Windows
that is how it should be
it's just strange that it is :)
EMC: 03seb 07TRUNK * 10emc2/src/hal/halmodule.cc: the typedef is not needed in C++, this commit shuts up a g++ warning
yay! I finally found out how to install ISE on a 64-bit system
all you need is the 32-bit version of libuuid, which isn't installed by default. download, extract, copy to /usr/lib342, and voila
of course, now it wants to download 1.5GB of updates
to go from 10.2 (minor point rev unknows) to 10.1.03
is there a list somewhere of "things that will break when upgrading to 2.3"?
besides your monitor?
if you use s32 innies and outies in classicladder 2.2 it will break when you go to 2.3 cause the name changes
is there someplace in trunk to add this info?
EMC: 03bigjohnt 07v2_2_branch * 10emc2/docs/src/ladder/classic_ladder.lyx: add info on s32's and fix default table
nm I found it
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/hal/comp.lyx: fix section title
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/UPDATING: update the update file
BigJohnT: should be here: http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?UpdatingConfigurationsForDevelopmentVersions
not sure how up to date that is
for one seb didn't mention hm2 pin name changes
I doubt anybody's been updating UPDATING
interesting item of the day: http://beagleboard.org/
runs linux, has fpu, about 20 GPIOs, and maybe sub-30uS latency with "rt-preempt" patches
UPDATING in trunk looks like a carry over from 2.2.x
I see that Chris has updated the wiki page on classicladder
jepler: cool :)
where does AXIS get the info to display Vel: on the backplot?
it comes from the TP
but it refers to current tooltip velocity
can that be used by HAL?
if not directly, then we can export a pin from the motion controller that has the same value
it's actual velocity, which is available in HAL anyway
it's not the commanded velocity
but like SWPadnos says you can get that
BigJohnT: current (actual) velocity referrs to the velocity it's travelling any given monet
ok that is what I want
you can get that even per joint basis
look at the sim config
there are 3 ddt's in place
ddt + hypot = vel :)
yeah, but a ddt / axis = axis_vel
ddt+hypot = 3d vel
is commanded velocity in hal anywhere?
BigJohnT: actual/current velocity is way different from commanded vel. though
not right now
commanded velocity gets scaled 3/4 times before it reaches HAL levels :)
yes, what I need is a corner lock for plasma
explain a bit more
you want a comparator commanded vel/actual vel to trigger corner lock?
read back a little :)
if commanded velocity is less than tolerance from actual turn off a bit that allows torch height control to work
I read that, but it made my head spin a bit..
BigJohnT: you mean the other way around
actual is less than commanded ..
ok :) then it's all good :)
here's a thought.. did you look how dallur did it?
a while back I did
I'm sure you'll understand lots more now
there's an odt somewhere in CVS in one of the configs
yes, I was hoping to come up with a "built in" solution so to speak so it will be easy for anyone to use a THC
I can understand that
[15:37:04] <alex_joni> http://cvs.linuxcnc.org/cvs/emc2/configs/plasma-thc/THCDesign.odg
darn, I have to go and show a customer engineer how to change a timer value in a plc :/
I'll have to look at when I get home... no linux boxes here
talk to later
there is OO for win too :)
hmmm. so here's a weird one
testing the G540, I can run G-code loops all day without missing any steps (assuming I don't lie to stepconf about latency)
* alex_joni senses a but
yet when I jog 0.0001", with SCALE set to 10000, half the time the motor doesn't move
(changed but to yet based on comments from the peanut gallery)
and even stranger is that sometimes, it seems like it waits until there's a full step before moving (not 10 microsteps, but "a few" step pulses), then it moves a fair amount
not necessarily the distance it should move, just a big step
I'm about to get the scope out to make sure the pulses are actually getting out the port
you do that.. but get halscope out :)
I'm sure they're getting out
I'm sure HAL thinks they're being output, I'm more suspicious of an electrical problem
do you actually lose steps though? Like if you run 1000 .0001 is it in the right place?
if I run a program, no lost steps
if I jog, lost steps all the time
I noticed it because it's possible that the last loop I ran lost 1 step
so I went to jog the laser dot back to the right place, and the thing didn't move
if you put jog step 10 times higher.. how does it behave?
* skunkworks_ thinks SWPadnos has a bunch of black dots painted on the walls around his room
burned not painted
I'm using axis in all emc machines
lemme check that alex
and I have build machine for 3d engraving
gcode files for this machine have about 300.000 lines
and even on Pentium 4 and 1GB ram preview of axis has about 0.8 fps
it definitely loses some of the steps, even on longer jogs
micges: depends on the video driver
nvidia geforce 2
and jogging with keyboard have high latencies
I have noticed that AXIS seems to slow down a lot more that it used to (like the lastest user list email says)
I don't know what made that change, but even a few hundred segments of backplot make it noticeably slower (on non-speed demon machines)
I saw that too
BigJohnT: here's something for you to document: the special comments (AXIS,hide) (AXIS,show) (AXIS,stop) change how AXIS previews the gcode
BigJohnT: between (AXIS,hide) and (AXIS,show), the lines are skipped in the preview
that doesn't affect backplot though, does it?
BigJohnT: when (AXIS,stop) is encountered, the preview ends even before M2 or % is encountered
SWPadnos: your bug report would be useful if it told me when the slowness was added
I'll try to figure it out
I don't remember when I put the slowness in, so someone needs to narrow it down
what, you don't keep notes?
apparently I wrote something else in the check-in message
oh, code words
"add reload tool table menu item" really means "reduce redraw speed due to excessive CPU power
i think i hurt paul's feelings... :-/
I'm sure you can't do that
I doubt he has any
seb_kuzminsky: you mean because his buildslaves are offline?
his buildslaves have been offline for about a day
when he took it down i asked if it was to reboot to a kernel that can compile our drivers (needs CONFIG_PCI_LEGACY)
he refused for some BS reason and i called him on it
I'm curious what the reason was ..
he refuses to enable CONFIG_PCI_LEGACY
we need that
so we can't use his help
(if it's at least a plausible reason for disabling CONFIG_PCI_LEGACY)
"No.. Why would I want to do that ?" && "pci_find_device and pci_find_slot are deprecated"
that's his stated reasons
[18:26:44] <alex_joni> http://lkml.org/lkml/2008/11/21/19
the proper thing to do (as i think all of us, including paul, know) would be to change the code, then change the build requirements
not a lot of users of that option
wait - I thought paul was the name of the build machine..
skunkworks_: run by a guy with the same name :)
(at least used to run..)
i'm not opposed to updating our code, but let's do it in a way that doesnt hork everything
seb_kuzminsky: making 2 drivers not build at all would probably be fine while we fix them
the m5i20 should be removed anyways.. so it's probably not worth fixing :)
didn't jmk just make a commit to the TODO file to that effect?
i'm all for it
* alex_joni looks at hordes of users running towards seb_kuzminsky for support
we can't remove those drivers from 2.2, right? they'll need to be updated in that branch rather than removed
or having the 2.2.x branch fail on pauls buildslave
that's good enough for me :)
we have four drivers that use pci_find_device:
I don't know that there was ever a working driver in the m5i2x tree
oh yes, and hal_5i2x
so after we fix those in 2.2 and fix or remove them in trunk, *then* we can change our advertised & tested build requirements
2.2 has all those except opto_ac5 i think
does hostmot2 use the new PCI calls?
ok, so we can also copy your code into the other drivers :)
nah.. way easier
the new code isn't a simple 1-to-1 replacement
get seb to do it :)
oh right - seb, get on it.
it's all tied into the hotplug system
get paul to do it -- he's the one who is pretending to care if emc builds under these conditions
you're the "new PCI interface" expert!
I don't see why you're letting him waste our time
so the structure of the driver changes somewhat
jepler's probably right
jepler: you're right
* alex_joni goes home
it's hard to tell if that option (and the old interface) will be removed though
i asked paul in email several times if he was going to help fix the things he was expecting his build environment to break
i told him the supported platforms all had champions among the developers, and new build platforms needed to come with new champions
he ignored me the first couple of times i asked him (just didnt reply to that part of the email)
then the third time he said: "ff.. YES."
i'm not sure what he meant by ff... but the YES was enough for me to give him a buildslave account on our buildmaster
i'm done dealing with him for now, if he comes back with something useful i'll listen
may have been the beginning of "ffs"
oh hey - here's a Xilinx ISE question: do you know how to use the hostmot2_import.tcl script to create a project?
I'm not sure that's still possible in ISE 10.1
I think there's "import tcl" in one of the menus
but I don't have ise handy
yeah, I looked for it, but I think it may have been removed (there's a note that could say that in a "changes" section of some document I saw)
you can still export, but I don't see an import option anywhere near it
oh good. Gigabyte thinks that my board should be RMA'ed (after I sent them a photo of the crispy FET)
they'll just refill the magic smoke
you could do it yourself if you had the right smoke fittings.
I'm not sure what the problem is though - it's not necessarily the FET itself
(though that's obviously bad now)
the right smoke is more problematic
I have smkow pumps, but they seem to be stuck pumping in the wrong direction :)
get a vacuum pump, and plug it reversed into a socket
I just tried running it upside down
SWPadnos: have you scoped the gecko yet?
I'd rather test a gecko like this: http://www.wiesmann-auto-sport.de/ewirstart.html
Urban 17,6 l / 100 km <- /me lost interest
wow. that would even be bad for an american car
hm apparently my car is 8.4l/100km for city driving. crazy europeans and they way they measure fuel efficiency!
yeah - I estimated it at ~15 and also lost interest
well, that and the EUR48k price tag for used ones
mine does about 6l/100km
diesel or gasoline?
that's what I used to get with my Plymouth Horizon
now that's an ugly car :D
about the same as a VW rabbit ;)
[19:41:18] <alex_joni> http://www.turbinecar.com/Scamp1.jpg
<- like that?
ewww. no way
[19:41:49] <SWPadnos> http://www.edmunds.com/pictures/VEHICLE/1990/Plymouth/4875/1990.plymouth.horizon.8918-396x249.jpg
that's what google said for plymouth horizon
same car, different color (same year even)
well it was wrong
slighly better :D
lots of room (4 doors plus hatchback)
got 40MPG normally, even got 30MPG towing a trailer
[19:42:55] <alex_joni> http://image.automobilemag.com/f/future_cars/2008/6699476/0609_x+2009_volkswagen_rabbit+front.jpg
<- rabbit looks a bit nicer
we'd probably still have it if it hadn't been in an accident
look at a 1990 rabbit
[19:43:58] <SWPadnos> http://ebeaa.org/graphics/578.jpg
btw.. the name rabbit is strange :)
we use vw golf over here ..
[19:45:53] <alex_joni> http://www.auto-tuning.cz/cg_digi/337a.jpg
<- 1974 golf
golf and rabbit aren't the same thing
they were both sold here, I think at the same time
I know rabbit was some model available only in the US I think
"The Volkswagen Golf is a compact car/small family car manufactured by Volkswagen since 1974 and marketed worldwide across six generations, in various body configurations and under various nameplates -- prominently as the Volkswagen Rabbit in the United States and Canada (Mk1 and Mk5), and as the Volkswagen Caribe in Mexico (Mk1)."
I've only seen the same car merketed under different names when it was also different brand names, such as the Dodge Omni which was the same as a Plymouth Horizon
but I'm quite sure there was a Golf here, so maybe it was a different car, or they just figured people would think they had a choice :)
or it was the imported model :)
I understand in Mexico there is a factory still making the rabbit with the original body style (80s) - they got the tooling when production of that model stopped in the US
the newer golf's weren't available until 2006 in the US
started in 2003 in europe
so maybe they did lots of imports
(the reverse happened with the new beatle)
[19:51:36] <alex_joni> http://en.wikipedia.org/wiki/Vw_rabbit
cradek: wikipedia says africa
oh sweet, there's a 4WD version of my car
I want I want I want
cars are getting so ugly
actually I think the 92-99 version looks best of all fo them
the Mkx x>4 are harder to get used to :)
MK4 is nice imo
does that mean ugly?
mine is Mk4
at first, after a while you get indiferent, after another while you start to like them
I really liked our jetta
Mk6 looks like some kind of suppository
mk6 seems angry
hmm.. the wikipedia page is odd.. you can buy MK6 over here
[20:03:09] <cradek> http://www.canadiandriver.com/articles/rp/03gtivr6.htm
^ mine exactly
if they really mean 39 mpg, they are very clearly wrong
I've never had over 32
they probably mean 29
isn't yours red?
GTI's have a tendency to lift the inside rear wheel off the tarmac when cornering
haha I don't think I've had that problem
maybe I drive differently somehow
7.3 l/100 km => 32 MPG
their conversion is wrong
"When driven in the safe-and-sane manner prescribed by the rules of the road however, the GTI offers exceptional everyday road manners."
there is a tdi forum that used to have as its main picuture a vw cornering on the track with the inside wheel off the ground.
ok it was both inside tires
[20:29:32] <skunkworks_> http://www.tdiclub.com/
good night all
see ya Alex
jepler: I assume this is for trunk?
off to the house with me
well, the steps are going out to the G540, but it doesn't always move
you have the charge pump enabled or disabled?
I just changed my test program so it only moves 0.0001 forward then back to 0, and the motor moves in spurts
charge pump has to be enabled or the G540 doesn't run
no it doesn't
(unless that's what that hidden switch is for :) )
yep that is the one
nonetheless, it's enabled and the motor moves
just not all the time when I tell it to
try it with the charge pump disabled
I'm pretty sure I moved the switch, and it
it's doing the same thing
try a different axis
any particular reason, or just suggesting troubleshooting ideas?
any idea on how this special comment works for AXIS (AXIS,hide) (AXIS,show) (AXIS, stop)
there are 4 drives on the 540
it'll be the same motor and cable, but I suppose it would be good to eliminate the drive itself
as for the (AXIS,*) comments, if you put them in your G-code, they will tell axis to not put segments in the preview, to continue putting segments into the preview, or to stop previewing altogether
if you have a G-code file that's supposed to loop forever, or is dependent on external input to stop looping, it's useful to be able to tell AXIS to not try to preview that :)
I can't seem to make it work here and I upped this morning
ok it is for the preview only
got it now boss
I'll get my dirt out of your hole
Y axis is about the same
or I should say "second driver is about the same"
hmm, you have 3.3 or 5v parallel port?
looks like about 3 at the end of the cable
wonder if that has anything to do with it? just fishing in the dark...
could be. I'd be surprised if this is a 3.3V parallel port though - it's on an old Celeron 500 computer
I think it might have an ISA slot ;)
do you have another computer to try it on?
yes. I will shortly
I may even connect it up to the 5i20 in this machine to see what happens ;)
oh hey - did you notice the big comment in the gs2_vfd driver?
it's about usage (hint hint) :)
no, I missed it
on here or #emc
in the source file :)
I should make that into a --help or -h usage text to
ok how would you complete this sentence... loadrt loads a HAL component into...
loadrt loads a realtime HAL component.
and how does that differ from loadusr?
loadusr loads a userspace program
realtime components are kernel modules on RT systems
(on sim systems, they're both userspace programs, but they're still conceptually different)
is realtime the only one that you can have different threads on?
realtime components need to have functions added to threads to do anything, whereas userspace programs are their own separate processes, which optionally talk to other HAL components via pins and parameters
ok that is what I was looking for...
where is the gs2_vfd driver at ?