#linuxcnc-devel | Logs for 2015-01-12

[04:35:14] <alex_joni> pcw_home: if it's documented it's a feature
[08:37:03] <skunkworks> Good morning!
[08:40:08] <micges-dev> ahoy
[08:47:50] <skunkworks> micges-dev, how goes it?
[08:48:31] <micges-dev> good thanks
[08:48:54] <micges-dev> real life dragged me in ;)
[08:49:06] <skunkworks> heh
[08:49:30] <micges-dev> and you?
[08:51:48] <skunkworks> Good - busy
[08:51:49] <seb_kuzminsky> morning fellas
[08:51:56] <skunkworks> Hi seb!
[09:09:17] <JT-Shop> hola Seb
[09:43:30] <pcw_home> micges-dev: problem with 7I83 is not extra packet but something that takes ~80 usec more in the recv section
[09:43:32] <pcw_home> it may be just a sserial bug (maybe there's a bug with no returned data)
[09:44:39] <pcw_home> thats about 240,000 instructions, a long time for the CPU to be twiddling its thumbs...
[09:59:07] <pcw_home> ( dual core 3.16 GHz CPU )
[10:13:49] <skunkworks> what would be different?
[10:36:23] <pcw_home> not sure maybe the sserial.c recv part has an issue when theres no recv data? I'll have to poke at this some more
[10:37:16] <pcw_home> (7I83 is a output only device though I guess it can read field voltage)
[10:40:01] <micges-dev> pcw_home: hmm, weird
[10:41:06] <micges-dev> pcw_home: you have 2+1 packets in master and 2.7?
[10:42:11] <pcw_home> well 2+1 in 2.7 at work and home
[10:43:09] <pcw_home> I thought it was 4 packets but i think its was just the power of suggestion when the read time went from 80 usec to 160 usec when I added the 7I83
[10:45:55] <pcw_home> I check at work with 2.7pre2 and its just 3 packets but something is making the read part slow
[10:45:57] <pcw_home> (packet sizes are mostly unchanged its like theres a bug in the read part of the code thats wasting a relativelyl huge amount of time)
[10:46:02] <pcw_home> I checked
[10:47:20] <pcw_home> I will try and duplicate with PCI because it may not be hm2_eth related at all
[10:50:32] <micges-dev> could be
[10:55:23] <pcw_home> may need some timer function calls in the code to see where its spending its time (or maybe standard profiling tools can be used since its Preemt-RT)
[10:58:40] <micges-dev> pcw_home: check if 7i84 is ok or not ok, I've got one so we can compare tests results
[11:00:09] <micges-dev> if it's ok*
[11:02:24] <pcw_home> OK I will in a few minutes
[11:19:11] <skunkworks> pcw_home, could mesa ship a partial? (dad is only waiting for the panel interface)
[11:22:04] <pcw_home> sure but it may take a bit (inventory this week)
[11:24:38] <Tom_itx> is that a sserial thing just on the ethernet card?
[11:36:31] <alex_joni> hi guys
[11:36:36] <alex_joni> HNY2015
[11:37:12] <seb_kuzminsky> hi alex_joni
[11:37:19] <alex_joni> hey seb
[11:37:24] <alex_joni> been a while :)
[11:38:13] <seb_kuzminsky> yeah man!
[11:38:16] <seb_kuzminsky> what's new?
[11:38:30] <alex_joni> hmm.. life's busy :/
[11:38:37] <alex_joni> not much time for lcnc
[11:46:08] <skunkworks> alex_joni, is that good or bad? ;)
[12:37:20] <PCW> OK on my test system adding _any_ second sserial device to a 7I76E more than doubles the indicated read time
[12:37:21] <PCW> from ~240000 = ~75 usec to ~540000 = ~165 usec
[12:41:34] <micges-dev> PCW: do you have 7i76e IO powered?
[12:59:50] <PCW> yes
[13:01:20] <PCW> Ill try adding more remotes (they really should only add a few usec each)
[13:02:04] <PCW> whats funny is this did not happen on my test machine at home so its pretty odd
[13:05:03] <PCW> test machine at home has same CPU but different MB/Ethernet MAC
[13:06:48] <PCW> will try PCI next
[13:52:21] <micges-dev> PCW: are you checking clean 2.7 or with my latest sserial patch?
[13:59:32] <PCW> with patch
[14:00:45] <PCW> Its does seem related to hm2_eth
[14:02:20] <PCW> with hm2_pci adding another sserial card adds about 3 usec
[14:05:32] <PCW> (with patch)
[14:11:59] <micges-dev> 2.7 w/o patch = read 156us, write 32us
[14:12:10] <micges-dev> 7i76e IO
[14:17:19] <micges-dev> 2.7 w/o patch (7i76e IO + 7i73) = read 258us, write 33us
[14:24:54] <micges-dev> PCW: I see only slightly longer packets, both cases have 2+1 all the time
[14:26:11] <PCW> yeah unpatched 2.7 ~75 usec W/O 7i73 170 usec with
[14:26:33] <PCW> so patch makes no difference
[14:27:19] <PCW> is it possible that this is because of the recv polling delay (what is it now?)
[14:28:12] <micges-dev> delay is 10us after each check if data arrived
[14:29:05] <micges-dev> try with debug=1, should write waiting times
[14:29:36] <PCW> ahh that should be helpful
[14:30:35] <PCW> I know the 7I76e is not adding any measurable delay
[14:31:18] <PCW> ( looking at TP 1 and TP2 with a scope )
[14:32:36] <PCW> I means theres not a significant difference in the read-request --> send read data times with or without the 7I73
[14:33:25] <micges-dev> ok
[14:35:28] <micges-dev> I don't have ebough power to enable debug with 7i73
[14:36:48] <PCW> OK trys is ~5 without 7I73 and ~12 with (rxpacket size went from 80 to 96 bytes)
[14:36:48] <micges-dev> PCW: can you try it?
[14:37:17] <PCW> what the heck???
[14:37:54] <micges-dev> I've got 72 bytes
[14:38:02] <micges-dev> send me your config
[14:39:29] <micges-dev> and maybe bitfile you're using to be sure
[14:39:33] <PCW> probably because i have mode 2 selected on the 7I76
[14:41:39] <micges-dev> ah
[14:45:22] <PCW> OK I dont think its a hm2_eth or sserial problem
[14:46:05] <PCW> it appears to be a strange feature of either the network stack or the driver
[14:46:29] <PCW> sorry for the wild goose chase
[14:46:54] <PCW> the same timing shows up with ping!
[14:47:35] <PCW> 80 bytes ~65 usec 90 bytes 160 usec!
[14:47:50] <PCW> arghh
[14:47:56] <micges-dev> what...
[14:48:45] <micges-dev> I'm on atom d525
[14:49:03] <PCW> but same MAC Ill bet
[14:49:35] <micges-dev> what mb you have?
[14:49:58] <PCW> This is a intel D41G (core duo e8500)
[14:50:17] <PCW> but it has a RTK mac
[14:51:25] <micges-dev> yeah here too
[14:51:39] <micges-dev> RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller
[14:54:04] <PCW> yeah, wonder if there are any driver knobs to tweak
[14:54:39] <PCW> let me try an ancient 3com 100bt card...
[15:00:53] <PCW> yep 8111 driver...
[15:01:29] <PCW> 3com card is slower but ~100 usec with no 7I73, 108 with
[15:02:37] <micges-dev> yeah, sane numbers
[15:05:05] <PCW> weird, i wonder where the next discontinuity is...
[15:05:46] <skunkworks> pcw wheezy or ubuntu?
[15:05:57] <PCW> wheezy
[15:06:14] <KGB-linuxcnc> 03Norbert Schechner 052.6 c175812 06linuxcnc 10src/emc/usr_intf/gmoccapy/gmoccapy.py 10src/emc/usr_intf/gmoccapy/release_notes.txt gmoccapy_1_5_1 - solved a bug in counts handling * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=c175812
[15:06:14] <KGB-linuxcnc> 03Norbert Schechner 052.6 aca8334 06linuxcnc Merge branch '2.6' of ssh://norbert@git.linuxcnc.org/git/linuxcnc.git into 2.6 * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=aca8334
[15:06:23] <micges-dev> I'm on ubuntu 12.04
[15:06:53] <KGB-linuxcnc> 03Norbert Schechner 052.7 f6699a3 06linuxcnc Merge branch '2.6' into 2.7 * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=f6699a3
[15:07:35] <PCW> probably not terribly important unless you want >1KHz thread but weird
[15:07:37] <KGB-linuxcnc> 03Norbert Schechner 05master ef81ece 06linuxcnc Merge branch '2.7' * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=ef81ece
[15:16:01] <PCW> "Your shipment [rand20 has been overnighted"
[15:16:03] <PCW> sad when spammers cant even get their scripts right
[15:18:25] <micges-dev> I don't see any options in driver besides debug=0..16
[15:20:11] <PCW> how about unbug = 1
[15:23:51] <PCW> probably some optimization somewhere...
[18:49:40] <KGB-linuxcnc> 03Michael Geszkiewicz 052.7 155959d 06linuxcnc 10src/hal/drivers/mesa-hostmot2/sserial.c hostmot2: if sserial realtime link with remote fail, set hal pin reflecting sserial running state to zero * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=155959d
[20:06:03] <KGB-linuxcnc> 03Michael Geszkiewicz 052.7 1d3c769 06linuxcnc 10src/hal/drivers/mesa-hostmot2/sserial.c 10src/hal/drivers/mesa-hostmot2/sserial.h hostmot2: warning when remote sserial device has firmware version lower than r14. * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=1d3c769
[20:12:08] <KGB-linuxcnc> 03Michael Geszkiewicz 052.7 fb177c3 06linuxcnc 10src/hal/drivers/mesa-hostmot2/sserial.c 10src/hal/drivers/mesa-hostmot2/sserial.h hostmot2: warning when remote sserial device has firmware version lower than r14. * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=fb177c3
[22:12:03] <skunkworks> logger[psha]:
[22:12:03] <logger[psha]> skunkworks: Log stored at http://psha.org.ru/irc/%23linuxcnc-devel/2015-01-13.html