#linuxcnc-devel | Logs for 2014-06-04

[10:14:05] <mozmck> Interesting article on memory requirements of various desktop environments. http://flexion.org/posts/2014-03-memory-consumption-of-linux-desktop-environments.html
[10:53:25] <mozmck> hahaha! I don't quite know what he's saying but googles translate makes it sound funny: "...but I think that'll be it bang on the head only while it is "crack"." http://www.linuxcnc.org/index.php/english/forum/49-basic-configuration/27900-thc-config-that-works#47677
[10:54:06] <mozmck> google's translation that is... maybe I need to use it :)
[10:56:13] <pcw_home> I was trying to figure that one out myself
[10:58:44] <mozmck> Sounds like he's saying Mach4 is a commercial gimmick that is not that good. Maybe the rest is that it will be knocked out (banged on the head) while it is new? soon?
[11:05:52] <pcw_home> maybe...
[11:26:12] <cradek> settings->cycle_i = -block->p_number;
[11:26:12] <cradek> settings->cycle_j = -block->q_number;
[11:26:15] <cradek> this is cool
[11:26:43] <cradek> with G5 if you don't specify the new IJ you get smoothness with the previous cubic spline
[11:27:28] <cradek> this is very nice functionality -- unfortunately all the documentation is written in C
[11:43:01] <pcw_home> The X and Y are control points and I/J are offsets?
[11:43:11] <pcw_home> (g5.1)
[11:43:55] <cradek> with G5.1, XY are the endpoint. IJ are incremental from start point to the (single) control point
[11:44:24] <cradek> G5.1 is quadratic (one control point)
[11:44:30] <pcw_home> Ahh
[11:45:50] <cradek> truetype-tracer generates G5.1 codes because truetype uses quadratic curves
[11:46:09] <pcw_home> so just like a normal motion gcode
[11:46:32] <pcw_home> never really looked at them
[11:46:48] <cradek> yes our G5.1 is no more weird than G2/3
[11:47:19] <cradek> they'll see more use if they're in the docs and gcode quickref (I'm working on it)
[11:47:39] <cradek> kind of need pictures though. I'm not sure how to draw them.
[11:47:48] <pcw_home> yeah the quick referece is a bit too quick
[11:48:02] <cradek> well each thing on it is a link to the full docs
[11:48:12] <cradek> the full docs should have the pictures
[11:48:26] <cradek> the quickref is my favorite piece of our documentation
[11:49:47] <pcw_home> I see a link on the NURBS gcode but not g5.1 maybe I'm looking at old docs
[11:54:23] <cradek> http://timeguy.com/cradek-files/emc/g5.1.png
[11:54:35] <cradek> possibly clear as mud
[11:57:35] <pcw_home> The example is a great addition
[11:58:13] <cradek> it's easy to think of most obvious quadratic. not sure what example to use for a cubic.
[11:58:39] <cradek> we don't have ANY docs for G5 (cubic) yet
[12:02:43] <pcw_home> It seems those would be most useful if CAM programs would generate them
[12:02:54] <cradek> yeah clearly
[12:03:40] <pcw_home> https://www.mail-archive.com/emc-developers@lists.sourceforge.net/msg10415.html
[12:03:41] <cradek> truetype-tracer is a good example of how they can be useful and they make the gcode simpler
[12:03:59] <cradek> postscript/type1 fonts are made of cubics
[12:06:52] <cradek> pcw_home: that is not the implementation we currently have
[12:06:53] <pcw_home> I didnt realize that truetype-tracer makes use of G5.1, thats neat
[12:07:12] <cradek> unfortunately nobody picked that up
[12:08:10] <cradek> pcw_home: G5.1 for truetype and G5.2 for type1 (truetype-tracer also can do type1 fonts)
[12:08:29] <cradek> so ... the name isn't perfect anymore
[12:54:48] <skunkworks> logger[psha],
[12:54:48] <logger[psha]> skunkworks: Log stored at http://psha.org.ru/irc/%23linuxcnc-devel/2014-06-04.html
[14:22:26] <KGB-linuxcnc> 03Chris Radek 05v2.5_branch e137bd0 06linuxcnc 10docs/html/gcode.html 10docs/src/gcode/gcode.txt Document G5, G5.1, G5.2, G5.3 NURBS/spline commands * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=e137bd0
[14:23:52] <skunkworks> nice!
[14:24:54] <cradek> we'll see if I get all the linking stuff right
[14:25:06] <cradek> quickref, anchors, etc
[14:25:17] <cradek> I think the buildbot docs end up on www.linuxcnc.org?
[14:25:49] <skunkworks> you talking about http://www.linuxcnc.org/docs/
[14:25:53] <cradek> yes
[14:25:53] <skunkworks> ?
[14:26:03] <skunkworks> then yes :)
[14:45:45] <cradek> that's awesome
[15:25:54] <CaptHindsight> has anyone tried toggling the GPIO on the headers of AMD boards to see if they may be used to replace LPT?
[15:50:24] <mozmck> CaptHindsight: what are some boards with this feature?
[15:54:44] <CaptHindsight> http://www.jetwaycomputer.com/NF82.html only 8 GPIO but the actual drivers on the pins are fast enough for PCIe and SATA control signals
[15:55:29] <CaptHindsight> I have the NDA docs and they can be configured as IO in user space
[15:55:41] <mozmck> http://lkml.iu.edu/hypermail/linux/kernel/0902.2/02547.html
[15:55:48] <mozmck> interesting.
[15:56:36] <CaptHindsight> there are actually many GPIO pins that are muxed but the 8 have become semi-standard on several APU boards
[15:57:23] <CaptHindsight> from the embedded vendors like Jetway, axiomtek etc
[15:57:25] <mozmck> Looks like there is a driver in the linux kernel for them as well, but I don't know if that could be used for realtime output.
[15:57:51] <mozmck> Are those APUs pretty good?
[15:58:05] <CaptHindsight> I've been using them for the past few years
[15:58:16] <mozmck> With linuxcnc?
[15:58:20] <CaptHindsight> 10uS jitter
[15:58:27] <CaptHindsight> yes
[15:58:27] <mozmck> interesting
[16:00:12] <CaptHindsight> as far as real time goes the registers are there in user space, the hardware has fast drivers
[16:01:41] <CaptHindsight> they are muxed with other high speed signals, so if BIOS configures them and the board has the headers they are there
[16:04:29] <CaptHindsight> I could make my own board and bring out all the unused GPIO to headers
[16:04:45] <mozmck> How many can you get?
[16:04:50] <mozmck> gpio that is.
[16:05:16] <cradek> I got it right!
[16:16:26] <CaptHindsight> mozmck: there are ~230 muxed, so if you leave out things like VGA you could end up with quite a few, maybe 32 or more
[16:16:41] <CaptHindsight> I'll have to look at the docs again
[16:17:06] <cradek> pcw_home: I'd be interested to know whether you can make sense of the new spline/NURBS docs
[16:17:14] <cradek> http://linuxcnc.org/docs/2.5/html/gcode.html
[16:19:57] <CaptHindsight> mozmck: looks like 24 are GPIO handled by GEvent pins
[16:29:59] <CaptHindsight> mozmck: ftp://kolibrios.org/users/art_zh/doc/public/A50_RRG.pdf an open version of the docs. Starts at page 251
[16:34:21] <mozmck> thanks!
[16:40:18] <CaptHindsight> that's for the older hubs and you didn't get it from me :)
[16:46:39] <mozmck> ;)
[17:31:13] <micges> skunkworks_: looks familiar?
[17:31:17] <micges> skunkworks_: http://ibin.co/1Oo6SAH8wQqP
[17:37:55] <CaptHindsight> if we do just write/write/write, we can get up to 1.2 MHz on the GPIOs
[17:38:14] <CaptHindsight> we did a little testing - it looks like with a read/modify/write of the gpios on the SB800 platform, we can get about 400KHz
[17:39:46] <CaptHindsight> so possibly 1.2MHz stepping on the GPIO's
[17:40:31] <CaptHindsight> well 1/2 that since thats write/write/write high/low/high
[17:47:29] <CaptHindsight> mozmck: so this might work!
[17:47:55] <mozmck> It could! Sounds interesting anyhow.
[17:52:31] <CaptHindsight> oh, some coreboot devs are getting upset "write "i will not use a general purpouse computer as a microcontroller" on the chalkboard 100 times"
[17:52:55] <CaptHindsight> and we're weenies for using PC's :)
[17:53:26] <mozmck> heh!
[17:54:13] <CaptHindsight> I might have finally found the source for the recrap philosophy
[20:46:39] <pcw_home> cradek: looks much better, I'll have to play with some examples
[21:08:42] <skunkworks_> micges: G64 tolerance set too high?
[21:50:14] <skunkworks_> good to see andy...