#linuxcnc-devel | Logs for 2015-07-20

[09:25:34] <cradek> my enco spam says they have 10% + free shipping
[09:28:22] <skunkworks_> hmm - is it really spam then?
[09:28:51] <cradek> it's not really spam; I might have actually agreed to it, to get the deals.
[09:29:53] <archivist> its spam when you cannot afford to get something during a promotion
[09:32:09] <skunkworks_> heh
[10:00:57] <seb_kuzminsky> jepler: did i hear you say the performance impact of adding memory barriers is 50 cycles/barrier on x86?
[10:03:20] <seb_kuzminsky> wow, a lot more stuff showed up in rob-ellenberg/feature/tangent-improvement-2.7
[10:03:45] <seb_kuzminsky> skunkworks_: does that branch have your blessing now?
[10:04:38] <jepler> seb_kuzminsky: yes that's what I measured on a core i5
[10:07:08] <skunkworks_> I have been running a bunch of gcode with no issues so far. It has an extra ini setting - ARC_BLEND_KINK_RATIO
[10:07:31] <skunkworks_> I didn't see what he defaulted it to - he says .1 should be close to master behavior.
[10:07:41] <skunkworks_> (I don't quite understand it yet)
[10:08:02] <seb_kuzminsky> ok thanks for being on top of it
[10:08:31] <seb_kuzminsky> do you know if he's ready to merge or if he wants to work more on it?
[10:08:54] <skunkworks_> I can ask. I think he has run it through all of his testing too.
[10:09:12] <seb_kuzminsky> yeah i saw that he added some more #424-specific stuff to his test suite
[10:11:04] <skunkworks_> yes
[10:12:00] <skunkworks_> it feels right ;)
[10:15:21] <seb_kuzminsky> :-)
[11:58:48] <seb_kuzminsky> jepler: i think that's an acceptable perf hit in order to get correctness on arm, so i think we should add that to 2.7
[11:59:03] <seb_kuzminsky> but the hal-streamer stuff should probably go in master instead of 2.7
[12:04:04] <jepler> seb_kuzminsky: OK, I'll produce a 2.7-appropriate patch for your review some day soon.
[12:28:15] <seb_kuzminsky> thx
[12:38:02] <seb_kuzminsky> hi jthornton
[12:38:41] <seb_kuzminsky> what state are the docs in? are they ready for the 2.7.0 release or is there still work to do?
[12:53:59] <pcw_home> I have run2.7 with hm2_eth multicard support for a day or so now without issue (just one card though)
[12:57:53] <seb_kuzminsky> using the split reads?
[12:58:08] <pcw_home> I notice the hm2_eth manual still mentions the board_mac=xxxx parameter which is gone AFAIK
[12:59:35] <pcw_home> dont know about the split reads
[13:00:15] <seb_kuzminsky> yep, you're right, the hm2_eth manpage lists board_mac but the driver doesn't have it
[13:00:56] <seb_kuzminsky> i mean using the .read-request hal function in addition to .read
[13:05:50] <skunkworks_> god - 2.7 is going to be awesome
[13:06:43] <seb_kuzminsky> yeah, lots of good new stuff :-)
[13:23:20] <Tom_itx> is andy's toolchange comp going into 2.7?
[13:49:53] <jepler> pcw_home: whoops, will put that on a mental list of things to fix. yes, the mac is no longer specified...
[15:19:57] <PCW> just tested hm2_eth with multiple cards, seems to have some sort of data problem (nearly immediate following errors even if set really wide= 1 inch)
[16:23:23] <jepler> hm, darn, ok
[16:24:07] <jepler> stepgens?
[16:27:16] <jepler> position coming from card 0 or card 1?
[16:30:18] <jepler> .. in my test setup I looked at stepgen feedback position on card 0, and then just now looked at gpio feedback coming in on card 2
[16:30:24] <jepler> because that's easy
[16:35:42] <PCW> I just hacked my existing test setup with a 7I76E and added a 7I80 (with nothing but GPIO enabled and no hal access to the 7I80 except the read and write addf's)
[16:36:51] <PCW> so basically the same as your setup (first card a working stepgen setup, second card pretty much unused)
[16:39:17] <PCW> (two sserial remotes on the 7I76E: 7I76 and 7I90)
[17:28:44] <jthornton> hi seb_kuzminsky I've just about caught up from my trip and I see 3 more things left on my list
[17:37:11] <skunkworks> zlog
[18:00:17] <seb_kuzminsky> jthornton: ok cool
[18:01:09] <seb_kuzminsky> rob ellenberg is finishing up the tangent-arc fix, i think that and the docs is the final thing before 2.7.0
[18:21:38] <jepler> seb_kuzminsky: pcw's testing shows multiple ethernet cards is broken, sadly. and I'd like to get the barrier calls in, but that's so minor I don't care if it misses 2.7.0 and waits for .1.
[18:21:48] <jepler> I'll be looking at the ethernet stuff this evening
[18:22:54] <seb_kuzminsky> i think .0 is at least a week away
[18:23:11] <seb_kuzminsky> i only tested hm2-eth-multiple with 1 pci card, and i only used .read
[18:23:15] <seb_kuzminsky> so, not a very good test
[19:05:11] <PCW> Is it possible that I have a config error and read-request is needed when multiple Ethernet cards are used?
[19:13:14] <PCW> also on the man page supported boards should be 7I80HD, 7I80DB, 7I76E, 7I92
[19:23:46] <jepler> PCW: unfortunately I haven't reproduced a problem here
[19:24:16] <jepler> I adapted hm2_stepper sample config to 7i80 and then also configured the 7i92 on ethernet
[19:24:31] <jepler> I tried with and without the read-request function
[19:24:41] <jepler> I tried with the 7i80 configured first, and with the 7i92 configured first
[19:26:57] <jepler> without read-request http://paste.debian.net/284319/ / with read-request http://paste.debian.net/284320/
[19:28:07] <jepler> testing on 2.7 branch at ref v2.7.0-pre6-205-g4249427
[19:33:52] <jepler> it seems like low probability that it's specific to 7i76e, none of the code changes were specific to a card
[19:44:17] <micges> PCW: can you paste your setup?
[19:48:40] <jepler> this patch gets you the config I'm testing with: http://emergent.unpythonic.net/files/sandbox/0001-WIP-2-board-ethernet-sample-config.patch
[19:49:26] <micges> jepler: you've got also 7i76e and 7i80?
[19:50:14] <micges> (sorry asked before link opened, nm)
[19:51:07] <jepler> micges: no, unfortunately I don't have a 7i76e for testing
[19:52:05] <micges> I've got 7i76e and 7i80 so I can try same stuff as Peter
[19:52:28] <micges> got also 7i92 so trying yours
[21:37:19] <pcw_home> I can paste my config up but tomorrow, its at work and there's no remote access to that machine