On Monday 23 June 2014 14:23:56 Bari did opine And Gene did reply: > On 06/23/2014 01:09 PM, Gene Heskett wrote: > > On Monday 23 June 2014 13:49:51 Bari did opine > > > > And Gene did reply: > > >From the resident old fart :) > > > > >> Just posting this info since this gets asked about often in IRC > >> > >> For RTAI to work with more recent kernels than 3.4 with Linuxcnc: > >> > >> 1) fix the scheduler in RTAI > > > > URL for that patch? > > There is no patch yet. It has to get fixed with a patch.
So this is a todo list. :) > >> 2) math support in kernel kconfig has changed and no options seem to > >> work with Linuxcnc > > > > Are you saying that forcing the math to be done by LCNC by leaving > > out any fast math in the kernel is the only way that works? Or that > > nothing newer than 3.4 works? > > Sorry to clarify: the Kconfig for RTAI (not the kernel) in regards to > math support is broken, no option seems to > work with Linuxcnc. > > Ever since newlib and glibc math options were added to Kconfig it > stopped working with Linuxcnc. > > >> 3) Linuxcnc needs to fix /proc in order for any 3.10 or newer RTAI > >> kernels to work > > > > URL for that patch? > > Needs to be written. The todo list again. :) > >> One workaround for the broken scheduler is with isocpu= you need to > >> leave the last cpu core out of the list. For example isolcpus=1,2,3 > >> for a 4 core and it should work or a single core cpu should work > >> fine. > > > > This seems backwards from what I've been told ever since the isolcpus > > thing became available. Its my understanding that if I have > > isolcpus=1 in the kernel command line of the grub file on my 2 core > > atom boards, with hyperthreading turned off as its a huge time > > killer, then LCNC sees it and runs the base_thread stuff on cpu1, > > and the rest of it on cpu0. And everything else I run (firefox, > > konversation, gedit etc) can only see cpu0. And the system runs > > well even when there is network traffic. > > Yes, it's backwards. That is the problem with RTAI. A comit was pushed > to the RTAI tree a few months ago that broke the way isolcpus works. > This is yet to be fixed. Just loverly I say, loverly. And Paolo M. seems less than "interested" in fixing any of that, I am also on that mailing list, FWTW. I've seen patches go by, but seldom accepted. Doesn't seem to me to be a very good way to run a train... Is there a problem with someone else trying to get their head around RTAI, and just plain forking what Paolo doesn't appear to care about anymore? Or do his copyrights preclude that? FWIW, I played with xenomai a year or so back, seemed to work although I wasn't impressed with the latency's on a 1400 mhz turion in my lappy, but its also memory bound at 1Gb and single core, 10 year old HP. Others have built it to use xenomai too, but no one has said why there is a lack of interest in pursuing it further, or I've missed the messages. Sim runs here, on 3.13.6, on a quad core phenom without isolcpus but thats a far cry from the real thing. Thanks Bari. Cheers, Gene Heskett -- "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author) Genes Web page <http://geneslinuxbox.net:6309/gene> US V Castleman, SCOTUS, Mar 2014 is grounds for Impeaching SCOTUS ------------------------------------------------------------------------------ HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing & Easy Data Exploration http://p.sf.net/sfu/hpccsystems _______________________________________________ Emc-developers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-developers
