On 2/19/26 12:59, Bari wrote:

On 2/18/26 2:25 PM, Bertho Stultiens wrote:
On 2/18/26 8:01 PM, Bari wrote:
Since RTAI has reached the end of its life the way forward might be Xenomai 4 https://evlproject.org/overview/
How many users might find this useful?

The few that require systems with lower latency than preempt_rt can provide. There are a few of us.
And there are tricks to make preempt-rt better. I've invented a third IRQ thread back when I was trying to use an RPI3b to drive my lathe. I still use that for the much faster rpi4b.  Actually real simply done. I defined a new. slower because humane hands aren't that fast, 200 hz thread and moved all my hand inputs to it, like the encoder dials that effectively replace the cranks that left with the removal of the factory apron. I replaced the apron by tapping for 8 1/4" bolts into the front face of the carriage and replaced the apron gearbox with a panel of 1/2"  7076 ALU . With a central window for the slider the X motor lives in, equpt with a belt tightening adjuster.  And a push switch (used to trigger the motion per click adjustment in hal) and 100 PPR $22 encoders that Jameco.com had for a long time. They are fed into the axis decoders in a 7i90HD and wired up in .hal as replacements for the cranks that used to be.  All that stuff runs at a lower priority that the 1kilohertz main loop. There is no fast loop, the 1khz is the fastest.   And it still can be moved faster than I can twist a dial.  Dead smooth, no clicks from data starvation like the rpi3b did originally. And it all runs on 21 watts when the lathe power is off.

I've considered replacing those hungry wintel computers,  350 watt burning off lease Dell's on the other 3 machines, but that kilowatt I'd save at a cost of around $350 per machine for the whole conversion would shift that 350 watts per machine, to the pair of electric heaters that heat my well insulated garage, 6" walls full of cocoon and a foot of it on the ceiling. So the net power saving would be zip.  But except for that consideration I now wish I had used something like a bananapi-m5 to do all my now wintel powered machines.

That pi is the most stable thing running medium powered machines out there.  Back when it had a ups, (its battery died a year ago, the replacement UPS I bought uses more power heating the battery than the pi uses) uptimes could be years. 4 3d printers are also using pi clones & my backup server is one of the bananapi-m5's.  So yes, I would be interested in an evl/Xenomai version for ARM64's.  If I don't miss roll call first.

Is there a pressing reason or major advantage to use Xenomai instead of using the non-kernel module approach with stock RT-PREEMPT kernels?
I tried it once, a decade ago when I had a stack of atoms running things, made 5 or so attempts but it wouldn't build on the atoms then. I don't recall why now.

I don't think many have been using (older) Xenomai. There are things referring to it in the LCNC build and source code, but I have a feeling some bitrot may have happened. Or are there people being busy keeping LCNC on Xenomai up-to-date?

I think that most were just using RTAI since it has been working for decades with little effort using a low cost LPT card and BOB. If Xenomai had been easily available, then it would of had a following as well.

it will be lots of work and we aren't sure yet how to support Mesa
cards with it yet.
At least some Mesa cards, all Ethernet cards and currently anything used on ARM, is userspace only. They can't be built as a kernel module. I'm not sure it would be worth the effort trying because it requires a *lot* of work and may be impossible for some without a rewrite. There is a port of Xenomai to ARM, but I'm still not sure whether it is worth the effort.

Keeping two RT systems alive in the LCNC source also requires much better CI because we can't currently test anything kernel based. So breakage will probably be more rule than exception.

Faster systems might require RDMA to keep up. But these are edge cases and not hobbyist projects.
You are ignoring the speed of the pi when its running the interface on rpspi interface. 42 megabit writes, 25 megabit reads in 32 bit packets to/from a 7i90HD and a triplet of 7i42TA;s for 72 i/o lines. I'm doing a lot on that lathe but have only used about half of those 72 lines for the lathe.
Xenomai already works with EtherCat.

_______________________________________________
Emc-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-developers

Cheers, Gene Heskett, CET.
--
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author, 1940)
If we desire respect for the law, we must first make the law respectable.
 - Louis D. Brandeis
Don't poison our oceans, interdict drugs at the src.



_______________________________________________
Emc-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-developers

Reply via email to