#linuxcnc-devel | Logs for 2014-02-03

Back
[09:48:57] <micges> skunkworks: hi
[09:51:03] <micges> I'm running 7i76e with 7i73, 20us write, 95us read
[12:36:30] <skunkworks> cradek, I love that gif
[13:56:23] <skunkworks> wow linuxcnc compiles a lot faster on my i7 vs celeron... wonder why?
[13:57:03] <memleak> never heard of someone outside the gaming world with an i7!
[14:03:30] <skunkworks> I have a first gen i7 in a dell laptop. the thing gets so hot!
[14:03:47] <skunkworks> served me well
[14:05:36] <skunkworks> bought it referbished quite a few years ago
[14:06:15] <memleak> heh i hear that a lot from i7 users
[14:08:38] <pcw_home> memleak: ISTR did some latency testing with preemt_rt kernels, are any available for testing?
[14:09:02] <memleak> ISTR?
[14:09:28] <pcw_home> I Seem To Recall
[14:10:00] <memleak> i didn't package anything..
[14:10:05] <memleak> all built by hand
[14:19:02] <memleak> i mean i could post my kernel configs if anyone is interested
[14:20:16] <micges> memleak: yes please
[14:20:22] <CaptHindsight> preemt_rt kernels on the asrock E350M1 ended up with ~40uS max jitter after several days of use
[14:20:33] <memleak> it was basically just a preempt_rt kernel trimmed down to it's minimum and linuxcnc ubc-3 branchh
[14:21:09] <memleak> s/remove trailing h// (my sed skills are gone atm)
[14:22:56] <memleak> after an entire week latency is at 42059 on servo thread
[14:23:14] <CaptHindsight> so far starting Firefox has the greatest impact on jitter
[14:23:19] <memleak> one moment while i post kernel config
[14:23:57] <CaptHindsight> you can run Stress with memory nearly maxed out for several hours and it doesn't blink
[14:24:41] <memleak> google-chrome and glxgears do not affect latency either
[14:24:48] <CaptHindsight> but after several minutes at 20-30uS starting Firefox usually kicks it up to its max of 42uS
[14:26:08] <memleak> i've tried doing quite a lot of stuff trying to get it to change and it has never gone beyond 42, its great!
[14:27:01] <pcw_home> video (or maybe big memory block allocation) seems to be what interferes with Ethernet timing also
[14:28:10] <pcw_home> (if you set axis update rate to 0.100 seconds you can see regular .1 second Ethernet latency bumps)
[14:28:19] <memleak> latency just jumped up to 48721, heh finally. pastebin.com/zgRt1p2E
[14:31:38] <pcw_home> you must be running a 64 bit kernel (or you would have noticed the max jitter going down)
[14:32:05] <pcw_home> I guess this is fixed now
[14:32:16] <memleak> this is 64-bit
[14:32:21] <memleak> as noted by kernel config
[14:32:57] <memleak> entire system built with -march=btver1
[14:52:34] <micges> latency-test is fixed in master, not yet in ubc3 and ubc3-7i80
[14:53:45] <memleak> does base thread in master? its gone in ubc3
[14:54:08] <memleak> s/does/is/
[14:56:17] <micges> yes