(too bad I didn't know that Hz output was common for mach spindle interfaces when I wrote stepgen ..)
cradek: are you going to post the wood ball vid, or should I?
hmm, this is my chance to not get emails about it
* jmkasunich plays with kino (vid editor)
EMC: 03cradek 07TRUNK * 10emc2/src/emc/rs274ngc/gcodemodule.cc: show block delete in the preview
EMC: 03cradek 07TRUNK * 10emc2/src/emc/usr_intf/axis/scripts/axis.py: show block delete in the preview
EMC: 03cradek 07TRUNK * 10emc2/share/axis/images/axis.ngc: show block delete in the preview
EMC: 03cradek 07TRUNK * 10emc2/lib/python/rs274/glcanon.py: fix highlighting for arcs/traverses
rats - youtube barfed on the video
I reformatted it from asf to mpeg4 and rotated it to be upright in the process - mplayer plays it, but youtube can't figure it out
fenn_ is now known as fenn
you're up early :)
yeah. kinda sucks
Heh - are you home?
That is good. ;)
indeed it is
though it would be better if the water heater hadn't died in the storms Sunday night
or mostly died anyway
gas or electric?
must be gas. No real electronics in electric
rented though, so they have to fix it
other than the lack of hot shower thing, yeah, it is nice ;)
I guess I should look at it some more, then call them
sorry for early disturbing, but I digged into emc code and saw that HAL pins are all useless copied form motion/traj/emcio to HAL and reverse every servo cycle..
could it be that all those memory structures exist in HAL shared memory ?
(it saves a lot of CPU cycles)
I don't quite understand your description of the "copies"
all HAL pins are in shared memory, as are signals
there is no copying when pins or signals are accessed
ah - you're talking about how pin values are read and stuffed into other structures (or the reverse)
emcmot_hal_data <-> emcmotStatus
these are copied in both ways every cycle
could be one structure in HAL ?
I was looking at the puma kins source file. I was deciding that I don't understand it. You can set 4 length - but it really doesn't explain what the distances are. (I am sure if I understood the math...) I have been doing searches - might just have to change the lengths and see what happens
Status is something that gets sent via NML to wherever it needs to go. it can't be the same as HAL because NML messages may need to travel over a network
I assume it is the arm and offset for the first 2 joints.. but not sure.
On the way home there was a rv that was going the opposite direction shreaded the drivers side tire - crossed the median and came to rest in the passing lane
sizeof(emcmot_status_t) < 2kb. copying 2kb less than once per ms is a tiny tiny amount of work
my 64-bit machine is down; I wonder if it was the storm last night, or if the kernel is still unstable
thanks for that advice
no problem. just let me know if you need more
yes, there's another issue on which I could use some advice.
jepler: ok copying is tiny, is there any known bottleneck of emc?
emc performs quite well enough for me, so I haven't searched for them.
what about axis ?
if I was on a low-resource machine I'd run a different front-end; again, it works fine for me.
seb_kuzminsky: let me read your message again, and continue the conversation over here
seb_kuzminsky: hal_5i2x is jmkasunich's work, and I don't know whether he plans to resume it. I don't think it's in a state where it makes sense to package it, but I wouldn't remove it from the source tree either.
i'll be happy to ignore it :-)
seb_kuzminsky: so basically you see mesa-hostmot2 as the future for all mesa cards, but preserve m5i20 for now due to the large installed base?
that's basically it
it prolly makes sense to leave the current m5i20 around and keep supporting it for a longish time
for the few 7i43_hm2 users in 2.2, how extensive arre the hal changes to switch to mesa-hostmot2?
is it mostly a search-and-replace, or is some intelligence required?
the string before the first "." in all the pin, param, & funct names changes
the set of functions changed
that's all i think
oh, also firmware needs to be loaded with bfload before loading the driver now
I think the plan sounds fine
how big are all the firmwares put together? I wonder if it's enough to justify having a separate package for them.
uh oh, actual work calls
currently 1.4 MB, for 5i20 and 7i43, final will be maybe 5 MB or so
ok enjoy rl, we'll talk later
cradek: thanks for those bugfixes!
argh the return of the "touch off dialog doesn't work"
what is it that our unique snowflakes do to cause it?
I think the highlight one is not quite right. it sometimes shows an extra line in cone.ngc when selecting some of the swirly rapids
I haven't yet reproduced the incorrect exceeds-limits warning
I thought it would be easy. but I guess I need to copy the mazak limits and var and use that program
in the preview it's easy to get an incorrect red box around an extents number, but I haven't been able to get the warning.
cradek: I have a ngc file that gives me the exceeds limits warning
hm, according to callgrind, most of the cycles inside task are inside emcIoUpdate (about 69%), compared to 5% for emcMotionUpdate and 8% for NML::write
BigJohnT: I duplicated the mazak setup and I have a test case now, I'm trying to whittle it down.
(jepler, seb_kuzminsky) is the hostmot2 email on the list, or was it private?
I don't see anything that I would immediately link to your earlier conversation
SWPadnos: it was private, I'll bounce it to developers
I was thinking about the number of bitfiles that were recently checked in - I'm not sure I like the trend :)
yeah that was one of the things on my mind
ooooh.. triangle for vismach.py
err.. it was jmkasunich :)
yeah thank the right guy
SWPadnos, jepler: the bitfiles have been in TRUNK for a while now, I just installed them in the debian package
Hey alex! :)
so who would I have to thank for loading of .stl's or the like ?
i could put them in another package (emc2-mesa-firmware or something) if you think that's better
alex_joni: that remains to be seen
seb_kuzminsky: not at the moment
* alex_joni prepares mentally for a long thank-you note to jepler
alex_joni: is stl the format that you already made me write a convert for once?
no, not stl
the format was something properietary, but I extended the converter to write .obj's
I think obj's are fairly easy to read..
I see cradek once wrote an ascii stl "parser/converter"
yeah, also for me ;)
there is a very good triangle-app out there
there == sourceforge
it's called meshlab, and reads/writes loads of mesh-formats
it also allows some basic interaction (closing holes, colouring,filters, etc)
so I wouldn't be too worried about which format we'd support..
well.. see you guys later :)
(long days at the beach are exhausting :)
I'll probably pop in the next evenings
seb_kuzminsky, I just happened to notice the firmware files when you added them to the .deb :)
that was one thing that jmkasunich and I were trying to avoid with mesa_5i2x
too bulky? clutter up your downlink?
in a sense, yes - I'm hosting the repository ;)
(or atleast providing the hosting on DH_
The repo is about 31 MB, of which 1.4 MB is hostmot2 firmware files
it was up around 1.5TB last month (repo + ISO), which is starting to be a noticeable fraction of my bandwidth allotment
when all the anyio boards are supported, there might be 5-10 MB of bitfiles
sure - I don't think it's a problem at the moment, it just occurred to me that it might be at some point
that's a lot for a single driver
i agree, it's a lot...
that 99% of users don't use ;)
i could maintain a separate repo of bitfiles somewhere
maybe mesa would host it
seb_kuzminsky: it could be in the same repo, but as a different package
mesa_5i2x was supposed to get around that by having tools to stick configuration information into the bitfile, and also program that onformation into a RAM block in the FPGA
SWPadnos: all the possible configs in one bitfile, and block ram to enable/disable what you dont want?
so HM2 would be one bitfile per supported device (7i43, 5i20, 5i22 ...), and the RAM block would tell the driver what to enable/disable
more or less
so the driver knows what pins to export at load time
i dont know for sure, but i get the impression from pcw that he's running out of gates on some boards
ie, the 200K 7i43 might not be able to hold 8 stepgens, 8 encoders, and 8 pwmgens all at the same time
could be - especially on the 200k boards
so different bitfiles for different configs would be needed...
alex_joni: you mean a different CVS module in the same repo? or a different debian package built from the same cvs module?
8 stepgen+8encoder+8PWM should fit into 200k just fine, though I haven't tested that theory :)
same repo, different .deb
if its the same repo it wont help your bandwidth issue
it would if only the people who manually install it get updates to it
consider that the Ubuntu repos have thousands of programs that most of us don't install, and we don't download them every time we update
i see what you mean. yes.
most of your bandwidth is in .debs, not in cvs updates?
cvs isn't on dreamhost, Jeff hosts that :)
gotta go, rl calls
EMC: 03jepler 07TRUNK * 10emc2/lib/python/vismach.py: rudimentary support for stl format
the ISO is a lot of it though - something like 1300 downloads last month
once again that whiner gets what he asks for
alex_joni, that is
uh - you can draw objects in cad - export as stl and inport them into vismach?
err - nice
hm the lighting doesn't seem quite right
[20:06:53] <jepler> http://emergent.unpy.net/files/sandbox/5axis-magnolia.png
uses a model from http://people.scs.fsu.edu/~burkardt/data/stla/stla.html
the surface normals are probably wrong
yeah, it looks like they could have the wrong sign or something
though that usually ends up making invisibile surfaces, doesn't it?
[20:32:20] <skunkworks_> http://www.electronicsam.com/images/puma/pumaarm.jpg
the normals come from the stl file; I wonder if they use the opposite convention of opengl or something.
oh well, somebody else can clean up after me :-P
or maybe they need normalizing
skunkworks_: 1.5kW on a 110 circuit! yeah, best make that dedicated
at first I thought the file was bad, but it turned out to be stupid CR/LF line endings are neede by STLWorks
tomp: no control :)
stupid question - where is the file that vismach uses for the puma?
hm the time my hardy/amd64/rtai system crashed could have been correlated with the thunderstorm
I'll say it was that :-P
[22:50:07] <seb_kuzminsky> http://scan.coverity.com/
-- free coverity testing for open source projects
yep, talked to them about that
a couple of years ago
we just started using coverity at my work, it's sweet
did you run it through?
i see emc2 isnt on the scan list right now
no. I don't recall what the issue was, but there was one
rtai build environment perhaps?
maybe sim would be easier
it was a technical issue rather than a political problem, i hope?
no, I don't think it's that. I think it had to do with who should be notified of a problem or something of that sort
kind of a crossover ;)
but I don't remember for sure
I don't know how well it can check networked apps either, which EMC is (by virtue of NML)
also, it's got in-kernel parts, so checks for linking behavior are difficult
(though their code grew out of the Stanford Checker, which was used for a long time to check the kernel itself)
anyone mind if i try to get that set up?
the nays have it! :-)
regarding the m5i2x driver subdir.....
I think much but not all of that has been superseded by hostmot2
the part that has not is the stuff that was intended to support people doing their own VHDL
or putting together predefined pieces of vhdl
the hm2 stuff also doesnt do the cool blockram config stuff you did
for example. I have a makefile in there that invokes the xliinx tools
I never got the user interface bits going
the main advantage over the m5i2x architecture was the ability to use one bitfile with many combinations of enabled/disabled hardware
to be honest, that was _not_ my primary goal
yes... assuming it can all fit in the fpga, some of the hm2 configs are stretching the 200K parts in particular
I was more interested in being able to mix and match chunks of vhdl - since I planned (and still plan) to do some functions that peter hasn't done, and probably isn't interested in doing
and the hm2 vhdl isnt set up well for that kind of hacking? (i know nothing about vhdl & havent looked at it)
the vhdl is semi-modular
but the driver arch isn't so much
I wanted both vhdl and driver to be modular
I also had these grandiose ideas about letting only moderatedly advanced users reconfig their board
that was really what brought me to a halt
right, by putting firmware config info in the blockram, right?
(well, that and getting divorced)
swp and I had come up with the idea of 4 classes of users
i remember swp telling me about that
1) vhdl hacker - writes his own vhdl, and matching driver
2) vhdl newbid - mixes and matches existing vhdl
3) advanced user without xilinx tools - turns on and off modules in an existing bitfile
4) basic user - loads the hal driver and uses whatever pins it gives him
i want to talk about this with you, but i have to go pick up the kids from preschool now, will you be around this evening, say in 3-4 hours?
3 anyway, that will be 10pm here
ok see you then maybe
sounds like a good time for me too (hopefully)
SWPadnos: do you know anything about video encoding?
precious little, especially if you're looking for mencoder command line options :)
mencoder -endpos 0:59 -oac copy -ovc lavc -lavcopts vbitrate=1800 -vf rotate=1 CLIP6405.ASF -o rotated.avi && mplayer -ao null rotated.avi
this rotates a video but the quality suffers no matter what vbitrate I chose
(get rid of -endpos)
if that's the problem you're stuck on
I got the rotate part down
not sure what else you'd like to do
mencoder CLIP6405.ASF -oac mp3lame -ovc lavc -vf rotate=1,harddup -lavcopts vcodec=mpeg4 -ofps 30 -of mpeg -o test3.mpeg
that works, but youtube doesn't like it
last year when I did the festcam vid (assembling a movie from frames) I used vcodec=msmpeg4
which must have worked (unless skunkworks re-encoded it - I didn't do the upload to youtube)
but when I try msmpeg4 now, I get a 2K file that barfs when I try to play it
did you use -of mpeg then? I think it's unusual to put mpeg4 in a .mpeg container, no matter how logical that sounds. ('d consider getting rid of -of and producing a .avi file instead
looks like I did avi back then
hmmm. I wonder if I inadvertently gave jmkasunich a couple of power strips
I haven't unpacked that box yet
I started with 12, no idea how many are there now
I just noticed that I'm missing the one I usually plug the laptop into
so I must have brought it there, but I don't know if I brought it back
(and there's another one I'm pretty sure I brought there, but I don't remember it coming back either)
jepler: I found endpos in the man page, but not "startpos" - is there such a thing?
(I'd like to drop the first couple of seconds)
jmkasunich: -ss I think
ah ha - I found one, and may not have brought the other
-ss, of course, how obvious
how could I miss it on such a short and simple man page
-ss == --short-and-simple
uploading, we'll see how it goes
wow kde4 sure managed to piss me off right quick
[23:49:56] <jmkasunich> http://www.youtube.com/watch?v=1EYaM4FkASA
hell, even gnome lets you register keyboard shortcuts to launch applications. I guess the dirty kde hippies were too lazy to include it in kde4. http://bugs.kde.org/show_bug.cgi?id=161010
weird. I'm pretty sure it was there in 3
yes it was
too bad we didn't record that at night with no talking
the talking is no big deal - its not like the machine whining is a great piece of music that is being interrupted
well that's true enough
maybe emc should put in the input device permissions "fix" by default