On Monday, January 30, 2012 10:57:23 PM Kent A. Reed did opine: > On 1/30/2012 8:36 PM, gene heskett wrote: > > Greetings all; > > > > Just in the last day it has come to my attention that both > > emc-2.6.0-pre (latest emc from buildbot, master-rt branch) and the > > slightly newer renamed linuxcnc-2.6.0-pre (latest from buildbot) are > > both running exclusively on cpu0, and about tapping it out, even at > > 65 u-sec base loop, keyboard response from either is in the 1+ second > > area, and a G39.2 z2.0 virtually crushed the pcb contact the full > > 1/16" of clearance under it, putting a good mark and an extremely > > visible sag in the middle before it got stopped. > > > > So I rebooted the atom box& immediately ran the newest emc from the > > buildbot, no change, its all on cpu0, and this was in the dmesg output > > when I looked: > > > > [ 22.713987] CPU0 attaching NULL sched-domain. > > [ 22.749773] CPU0 attaching sched-domain: > > [ 22.749782] domain 0: span 0-1 level CPU > > [ 22.749809] groups: 0 1 > > [ 22.749822] CPU1 attaching sched-domain: > > [ 22.749827] domain 0: span 0-1 level CPU > > [ 22.749832] groups: 1 0 > > [ 22.750253] CPU0 attaching NULL sched-domain. > > [ 22.750262] CPU1 attaching NULL sched-domain. > > [ 22.780105] CPU0 attaching sched-domain: > > [ 22.780113] domain 0: span 0-1 level CPU > > [ 22.780120] groups: 0 1 > > [ 22.780131] CPU1 attaching sched-domain: > > [ 22.780136] domain 0: span 0-1 level CPU > > [ 22.780141] groups: 1 0 > > [ 25.155333] CPU0 attaching NULL sched-domain. > > [ 25.155345] CPU1 attaching NULL sched-domain. > > [ 25.184112] CPU0 attaching sched-domain: > > [ 25.184120] domain 0: span 0-1 level CPU > > [ 25.184127] groups: 0 1 > > [ 25.184138] CPU1 attaching sched-domain: > > [ 25.184143] domain 0: span 0-1 level CPU > > [ 25.184148] groups: 1 0 > > [ 25.184510] CPU0 attaching NULL sched-domain. > > [ 25.184518] CPU1 attaching NULL sched-domain. > > [ 25.216149] CPU0 attaching sched-domain: > > [ 25.216157] domain 0: span 0-1 level CPU > > [ 25.216164] groups: 0 1 > > [ 25.216175] CPU1 attaching sched-domain: > > [ 25.216179] domain 0: span 0-1 level CPU > > [ 25.216184] groups: 1 0 > > [ 105.425985] I-pipe: Domain RTAI registered. > > [ 105.425999] RTAI[hal]:<3.8.1> mounted over IPIPE-NOTHREADS 2.6-03. > > [ 105.426005] RTAI[hal]: compiled with gcc version 4.4.3 (Ubuntu > > 4.4.3-4ubuntu5) . > > [ 105.426087] RTAI[hal]: mounted (IPIPE-NOTHREADS, IMMEDIATE > > (INTERNAL IRQs DISPATCHED), ISOL_CPUS_MASK: 0). > > > > ^^^^^^^^^^^^^^^^^^^^ > > > > <...> > > What does the ^^^^^^^^^^^^^^ above tell the guru's? > > > > Cheers, Gene > > Gene: > > I don't know squat about the internals of RTAI. When I start (LinuxCNC) > 2.4.6 on Ubuntu 10.04LTS (installed from the LiveCD and then updated), I > get the same "RTAI ...ISOL_CPUS_MASK: 0". On the other hand, early on , > dmesg reports the kernel booted with "isolcpus=1". > > My ASUS AT5NM10-I MB is not connected to a CNC controller so I'm running > the sim/axis_mm demo configuration as I report this. I don't see a slow > down in LinuxCNC response. > > As an aside, I'm not loaded up with a lot of ordinary (e.g., > non-realtime) programs. I have seen a slow down in their response in the > past like when I was trying to stress the machine while running > latency_test. In a nutshell, I supposed that a single Atom cpu isn't all > that powerful but perhaps there's something else going on as well. > > What's your evidence that LinuxCNC is running exclusively on cpu0? Don't > forget that all the realtime stuff is running on the isolated cpu. It's > the userspace stuff like axis that is running on the nonisolated cpu and > that shows up in process listings (ps, top, etc). top only shows one cpu, fairly busy,
htop shows 2 cpu's with the 2nd one sitting at 0.0% use. gkrellm shows cpu0 as being quite busy, but cpu1 is sitting steadfastly at 0% usage. A week ago, gkrellm showed cpu1 usage only when emc was running, so something has definitely changed in the last few days. Both machines have now been rebooted, it seems you can't reboot just one end of an NFS link as the un-rebooted end of it will lock solid, needs a reset button push to reboot within about 10 minutes. But that is secondary. Faint memories remind me that it happened several times before & may be one of the reasons I didn't really want to use nfs. But it otherwise is working well for the usage I need. On the old single core athlon, running at the same speed, I had no problems running several konsoles, 1 instance of Konversation (irc client) and a copy of firefox so I could go get stuff by google searching, and a copy of synaptic in case I needed something from the repo. All this while emc was carving something with a base thread of 35 u-secs. Now suddenly, I am running at half the speed for a base thread, around 65 u-secs, and the system is lagging exactly like the old system if I tried to pull the base thread below 29 u-secs. I've been reading up on the cpuset things (this is tied to isolcpus according to my grepping in the kernel src tree) and I did a mkdir like the manpage shows and mounted it so I can browse what is there once its mounted, but I've no clue what exactly it is that I am looking at so I haven't tried to change anything, just proving its there. And obviously this is a show stopper, when a 38.2 goes so far past contact it bent the contact pcb down about a 16th in the middle before the machine knew it had contact. Another 16th and it would have damaged more than the scrap of pcb as the pcb would have been in contact with a 1x2x3 alu post under the contact pcb, probably breaking a small drill. Last night, running those same commands in MDI mode, I was getting G38.2 trips with a max deviation of about .00075", motor shut offs were instant, lower by a few microns if it was moving at 10 ipm as opposed to 1 ipm. So at the moment, I'm back to being bumfuzzled again. No clue what I changed that would have screwed up the older emc-2.6.0-pre after its been reinstalled. Take both our minds off it for a few. Hows the missus doing? > Regards, > Kent Cheers & thanks Kent, Gene. -- "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author) My web page: <http://coyoteden.dyndns-free.com:85/gene> If a can of Alpo costs 38 cents, would it cost $2.50 in Dog Dollars? ------------------------------------------------------------------------------ Keep Your Developer Skills Current with LearnDevNow! The most comprehensive online learning library for Microsoft developers is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3, Metro Style Apps, more. Free future releases when you subscribe now! http://p.sf.net/sfu/learndevnow-d2d _______________________________________________ Emc-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-users
