SWPadnos_ is now known as SWPadnos
awallin: i've tried dot for hal schematics and it gives horrible results...
mhaberler: i've played a bit with graphviz...
and i really don't find way to draw pretty looking graphcs
had a chance to look at the signals thing, and maybe completing hal widgets -> gpin?
I think it's really minimal what's missing, and I could update docs and make programs more elegant, and it will wash the dishes too!
i've not neigher emc2-sim here nor gcc :)
uh, you're really challenged then... ok. I keep drilling this toolchang
so playing with some minor things :)
it's ok to install graphviz over gprs but not emc2-sim :)
understanding this NML message thing is, uh, challenging
Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emcdevel/2010-12-11.txt
wow, 240kb/s over mobile phone...
they put hdsp capable access point somewhere nearby...
huh. I wonder what happens if there's an abort during M1xx execution?
or a following error
depends on M1xx script i think
i guess there is no kill involved in M1xx
I mean what happens in EMC. I'm wondering if there's a way to terminate a running M1xx script in the event of an error
heh, kill -TERM; sleep 10; kill -KILL
kill --with-children :)
huh, kill -CONT too to be picky :)
hm, TV is created to create zombies from humans
bet it is like a failed tool change your stuck in limbo
or worse, the M1xx program may still be twiddling HAL bits ...
maybe some like -INT will be suitable?
if user whant to handle inerrupts it's still possible but most of M1xx commands will be killed
as long as you can kill a whole process tree, maybe
remember that M1xx can spawn other programs
you have to create process group after fork prior to exec
and then hope that M1xx won't create new one :)
it's not difficult really
but as i mentioned already if m1xx creates another pgroup - no chances :)
until cgroups will land in mainline kernel :)
hmm. wasn't that the recent change that generated so much press?
(for better desktop performance and all that stuff)
based on cgroups mech
so maybe around 12.04 we'll have an RTAI patch that works well with cgroups ;)
more interseting cgroups provide way to kill process tree
systemd init system is using that feature
but i've not inspected that in details :)
I just got my mothers computer running more or less how I want. man, it's fast
I'm amazed at how quickly it boots
i've used pgroups at my work for reliably killing groups of child processes and it works
SWPadnos: what operating system?
the Ubuntu graphic startup screen (the one with the dots that light up) sometimes doesn't even appear on the screen
Ubuntu, of course :)
it's like 3 seconds or less to the desktop, from the time GRUB decides to actually boot
yes - I have said the k&t boots really fast. I think the amount of time that the thing is in the bios splash is about how log the os takes to boot
it's the Crucial SATA 6GB SSD
nice - this is just TB hd. iirc
it was fast with a 2TB spinning drive, but the SSD just blows it away
I also managed to get a working CUDA installation, which is a first for me :)
now I have to duplicate my efforts on my own machine :)
WTF is CUDA? Sounds like a snake or something
or a fish
you put a car in your moms computer?
[15:07:08] <SWPadnos> http://tinyurl.com/csj4ym
I dont do blind urls
[15:07:27] <SWPadnos> http://lmgtfy.com/?q=cuda
Bitch, wikipedia doens't help
[15:08:04] <SWPadnos> http://en.wikipedia.org/wiki/CUDA
you have some intensive computations?
I looked the first time you whined about it
my mother plans to do massive computations, yes
ok, so again WTF is cuda?
in a nutshell. all I saw was something about nvidia
and stopped reading
wasn't it the third link down?
Jymmm, that's what it is
SWPadnos: video driver?
[15:10:09] <skunkworks> http://www.nvidia.com/object/what_is_cuda_new.html
it's Nvidia's system for using the GPU for general computing
and this is better than multiple cores?
psha: wasn't russia hackers using it to break wireless security?
it's multiple cores ^2
SWPadnos: what, you installed multiple video cards ?
her card has 448 cores (it's a GTX 480), running at 700MHz+
no, just one good one :)
skunkworks: no, russian hackers are breaking wireless security with thermorectal cryptoanalisis
but it could use 4 if I put them in
448 cores?! wth
it's a video card
and a GPU has more balls than a cpu?
they all have loads of parallel processors in them, but the newer ones can actually do full double-precision floating point math
Jymmm: GPGPU has extremly high 'peak' performance but memory access is not obvious
psha: yeck (I had to google it)
FP usage I get. Damn dev's
oops, 480 core
SWPadnos: And she knew she needed this how?
does about 1.06 trillion floating ops/sec, or ~530billion double precision
she does math for fun
and programs for fun
SWPadnos: And she *knew* she needed this how?
so this is kind of a combo for her
you ask silly questions
also it's SIMD system - you load instruction set to all cores and provide partitioned data to them
SWPadnos: You opened your big mouth and volunteered to do this didn't you?
she wanted a new computer (to replace the Pentium 133?), and left the details to me ;)
hey, at least I get to play with it
and it'll probably reduce the number of calls I get, since she'll be busy
it's a win-win
axis.[0-3].* pins belongs to Axis or somewhere else?
oops, large one
[15:45:40] <psha> http://psha.org.ru/tmp/hal.svg
this one is smaller
i've killed his connection!! :(
psha: axis.N are emc's motion controller
psha: AXIS ui has some pins too, named axisui.*
cradek: so - this http://imagebin.ca/img/W9djKuh.png
that is 60hz hum
100ms - 6 cycles - 1/(.1/6)=60hz
if I did the math right
so - I may have some grounding issues.
(although it is in the <.0001 error
it looks like 5.5 or so cycles per division to me
yeah, a little under 22 valleys in the rightmost 4 divisions
(you don't count the end one, that's zero)
hmm - so not exatly 6?
doesn't look it to me
if you count over multible 100ms it is multible of 6
I started with the valley on the right edge as 0, then counted through 4 divisions and got almost 22 additional valleys (since I can see the one on the edge, it's slightly to the left of the end of the division, so I decided to count the one just to the left of the 4th division line)
doesn't look it to me :)
I think it is way too early
for you, sure
ok - you are right
so my grouding is good - tuning is bad. (although it is a lot better - I don't think it is totally gone)
the peak-peak oscillation looks like about 40-50u
other than the big blip, of course
that is acc/decel
you sure? :)
you have it at 200u/division
oh, for the accel/decel part
yep, +/- 0.00017-ish
.00034 P-P, which makes the oscillation look even less bad
right - I think it could be better - but that is pretty good
for a newbie tuner
what 'net signal pin' means?
pin is connected as input or output to signal?
or if i want to connect input pin i have to explicitly tell 'net signal <= pin'?
or it depends on pin type?
the symbols <= and => are only for helping humans
so if i write 'net signal pin1\nnet signal pin2' i'll get similar result when i swap pin1 and pin2?
net is shorthand for creating a signal of a certain type and connecting one or more pins to it
if you write net mysig pin1 \n net mysig pin2 yes, its same as net mysig pin2 \n net mysig pin1 , but better yet use net mysig pin1 pin2 , those 2 pins are connected by the conduit 'mysig'
nwe name pin1 pin2 pin3 will create a net called "name" woth the type of pin1, and connect all 3 pins to it
it checks first that all the pins exist, have the same type, and there is no more than one writer
if a signal called name already exists, then those pins will be connected to it (assuming that no rules are broken, like multiple writers or different types)
i mean that consideration about input/output connection is based on pin type and not on net syntax?
to clarify a bit my interest - i'm writing .hal to .dot convertor
so you may examine your hal setup in image :)
net name out1 in1 in2 in3 is equivalent to net name in1 out1 in2 in3
woof! how do you get dot object with multiple connections
(mult connections that dont get tangled )
now, why would a spam title be "Big penis like a girl"?
[18:49:03] <psha> http://psha.org.ru/tmp/nist-lathe.svg
graphviz is not suitable for cyclic graphs...
:( url not found 404, yes i tried graphviz a lot
uploaded to incorrect dir :)
heh, if you are confident with graphviz maybe you may tune params?
cant open with inkscape synfig or imagemagick
hm, ok with browser, eog
no, i had no success making graphic diagrams to hal ( no final, got some done in GEDA and micges did it with ... a closed source eda tool )
and eog failed here too, but i understand the idea of an inspectable, zoomable, gui representaion of a hal fle , and one that can be translated into a .hal file
it a great idea, esp for people like me that are myopic, cant see thru a mass of data ( like the emc2 code base :)
hm, check again
maybe you've catched it in the middle
yes, haha i just looked at the xml i had gotten, its just the error msg in xml format, now is fine ( and BIG )
i'm trying to tune params but without any success :)
pretty, i can see it now ( had to zoom way back) what do you mean by tune params?
garphviz parms? what would you change on the svg you posted?
yes, graphviz params
how to ask it to fit into some reasonable resolution :)
<svg width="4645pt" height="2719pt"
it's a bit large one :)
i will now try the ZGRViewer for zoomable (its SVG so the browsers are >supposed< to handle ^+ ^- )
it = your file (not ZGRViewer)
i view it just with browser )
does teh browser handle the wwc specs for svg? can you zoom?
try ^+ and ^- in firefox
ffox namaroka 3.16.3pre
? not enough zoom level?
yes :) it's chromium (core of google-chrome)
often opera was better at svg
limited to 50%
i just noticed ^mousewheel zooms too, yes the smallest render is often still too big ( i think thats what you mean ) and vs:vs the largest may not allow the detail you like.
the 'myopic' problem strikes again, you cant see the chip when viewing the machine, you cant see the spindle when viewing the button on the control panel.
some is the viewing tool and some is just difference of magnitudes.
i wanted to highlight nets, that allows the view size to be less important, you could scroll around looking for the ends of the net with the visual clue of a highlight
all edges may be drawn in own color :)
or the widget in imagemagick that shows a thumbnail with a rectangle over it. field of view vs whole image ( where you are in the overall view )
'all edges may be drawn in own color' i think you have to find the edges & nodes connected to the net and change some visual attribute ( color width line type)
micges did some nice work with same idea in eagle
i suspect it's not simple to check which 'edge' was selected...
it's simple to draw svg with cairo surface but i don't know how to get feedback
yes, eda tools looks suitable for hal schematics
it's simple schematics :)
is this of any use? http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?HalSchematicsUsingGschem
yes, i've seen that page
iirc i never got good dragging, and was my code was very kludgy with the threads etc
im trying tintfu right now, but its a java app ( java has always been iffy for me ) i need some dot files to test with
graphviz dot files?
there is nist-lathe.dot near .svg file
just replace extension
well tintfu blew up on 1st simple .dot file from wikipedia
tom3p: if you wish to take a look onto hal -> dot conversion script it's on http://psha.org.ru/cgit/psha/haldot.git/
ok avoid tinfu, the fiels open with dotty just fine, but tinfu tries to make each word in the comments a node :( http://imagebin.org/127457 http://imagebin.org/127458
thanks i will
it's not cgit but git
but cgit works too )
git got it ok
usage is simple: ./parsehal.py < halfile > dotfile
not what i expected, not a real .dot file, heres the terminal session and a cat of the output http://pastebin.com/HcvcMeiN
it's valid dot file i think, what's wrong?
i meant doty cant render it for me (blank screen )
oops! scale hits again! its ok
feed it to dot -Tsvg to get svg image
look in Makefile for example
i dragged those sliders a long ways and saw nothing, but it was really there
that's what i was talking about
if you force 'size="x,y"' in graph header you'll get bad results too
i'll have to go now
if you'll get some results - catch me tomorrow
but i think it's better to stick with geda...
btw 'dot -Tsvg largeHal.dot >largeHal.svg' is what worked for me and it renders ok in eog
but i dont think eog is showing svg ( blurry pixelated at hi zoom )
and in ffox it stays crisp
SWPadnos_ is now known as SWPadnos
mhaberler_ is now known as mhaberler
ries_ is now known as ries
SWPadnos_ is now known as SWPadnos
SWPadnos_ is now known as SWPadnos