On Tuesday 28 August 2012 16:00:29 Lars Segerlund did opine: > Actually, for the debian rt-preempt there is an apt repo to add .... > all of rt-preempt is not mainlined yet, but it's the goal. > But I have been running an rt-preempt kernel on my laptop for a > while, and quite arguably it doesn't affect performance much, until I > do the tuning of the system, in which case it doesn't sleep and so on. > > Also there seems to be some confusion on rt-preempt and RT-Linux .... > they are different beasts. > > I don't think it's the high end board thing, it's more how the system > handles context switches, and the device drivers, the main problem is > locking in the rest of the kernel. > > RT-Preempt runs the interrupts as threads, in order to do priority > inversion and such, so things got a lot better after the BKL went > away. > > There are problems with proprietary grafix drivers also, ( NVidia ) > .... but on the whole it's a lot more of a regular linux kernel, most > kernels are running preempt at the moment, it's just the rt part that > is specific. > > Like any other system it's sensitive to tuning, but I think there is > a project for profiling problems, check out the rt-wiki > https://rt.wiki.kernel.org/index.php/Main_Page , there are a lot of > tools and hints, so please have a look before bashing. > There are also a bit of examples, it's not 100% up to date but nice > reading. > > The mailing list is really good, if you have a solid problem, ask and > you shall recieve .... I suggest we check with the mutex bug on that > list. > > / regards, Lars Segerlund. >
Lars S. & the rest of the developers: Perhaps we should, after getting linuxcnc to actually limp w/o crashing on a semi-stock -rt version of a recent kernel, develop a test scenario that could be run on any multicore cpu. 1: do the usual stuff at boot time like turn off the hyperthreading, and set the last cpu core as reserved, like we do for the dual core atoms with the isolcpus=1, or whatever bit mask in needed to select the 3rd core on my phenom. 2. If latency-test does not now, but I believe it does already, train it to run itself on the isolcpus=core. 3. From that, we can develop a baseline that tells us how far we have to go, either in terms of buying a hardware stepgen board, or how much farther we need to go to be able to run software step generation. 4. If its obvious that we will need hardware stepgens, no doubt there will be some who will fall by the wayside because of that expense, but ATM I can probably afford to follow this path. How many of the older & retired folks among us that can afford this remains to be seen. However, speaking for myself, I can probably swing it unless Dee's final expenses deplete our $. That time is approaching as I'm having trouble getting enough food into her to hold her already lifetime slim 100-110 lb weight. My $0.02 (adjust for inflation since 1934) :) Cheers, 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> is up! There are two kinds of egotists: 1) Those who admit it 2) The rest of us ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Emc-developers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-developers
