[Xenomai-help] Xenomai + PREEMPT_RT

[Xenomai-help] Xenomai + PREEMPT_RT

[Daniel Walker]

Hi all!

With PREEMPT_RT getting more and more mature, it seems natural that
PREEMPT_RT could be used with Xenomai. Has anyone considered this
already? Or even better, is anyone already doing work with these two
technologies combined?

Daniel

Re: [Xenomai-help] Xenomai + PREEMPT_RT

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 996 bytes --]

Daniel Walker wrote:
> Hi all!
> 
> With PREEMPT_RT getting more and more mature, it seems natural that
> PREEMPT_RT could be used with Xenomai. Has anyone considered this
> already?

A clear yes: this is part of the long-term roadmap we are currently
preparing, just stay tuned. In fact, Philippe expressed this goal
already in earlier announcements.

Having native (patch-less) real-time will indeed be an interesting
option for the future, besides co-scheduling.

> Or even better, is anyone already doing work with these two
> technologies combined?

There once existed a combo patch (that I don't find ATM) of an ancient
-rt version + old Adeos, a combination that will more or less
automatically come again when PREEMPT_RT is mainline. Porting the
nucleus over PREEMPT_RT so that the skins can run "natively" is another
task yet to begin. Anyway, we are already including this perspective in
current design decisions when subsystems are revised or enhanced.

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]

Re: [Xenomai-help] Xenomai + PREEMPT_RT

[Daniel Walker]

On Wed, 2006-09-06 at 12:50 +0200, Jan Kiszka wrote:
> > Or even better, is anyone already doing work with these two
> > technologies combined?
> 
> There once existed a combo patch (that I don't find ATM) of an ancient
> -rt version + old Adeos, a combination that will more or less
> automatically come again when PREEMPT_RT is mainline. Porting the
> nucleus over PREEMPT_RT so that the skins can run "natively" is another
> task yet to begin. Anyway, we are already including this perspective in
> current design decisions when subsystems are revised or enhanced.

So I take it that designing the combined version is still taking place?

It sounds like the current direction is more of a straight merge. 
Since PREEMPT_RT and Xenomai are somewhat parallel in purpose would
there be any room to eliminate duplication between the two?

Daniel

Re: [Xenomai-help] Xenomai + PREEMPT_RT

[Gilles Chanteperdrix]

Jan Kiszka wrote:
 > There once existed a combo patch (that I don't find ATM) of an ancient
 > -rt version + old Adeos,

Here:

http://download.gna.org/adeos/patches/attic/v2.6/i386/combo/candidates/adeos-linux-2.6.12-rc2-RT-V0.7.44-03-i386-r10c3.patch

-- 


					    Gilles Chanteperdrix.

Re: [Xenomai-help] Xenomai + PREEMPT_RT

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 1338 bytes --]

Daniel Walker wrote:
> On Wed, 2006-09-06 at 12:50 +0200, Jan Kiszka wrote:
>>> Or even better, is anyone already doing work with these two
>>> technologies combined?
>> There once existed a combo patch (that I don't find ATM) of an ancient
>> -rt version + old Adeos, a combination that will more or less
>> automatically come again when PREEMPT_RT is mainline. Porting the
>> nucleus over PREEMPT_RT so that the skins can run "natively" is another
>> task yet to begin. Anyway, we are already including this perspective in
>> current design decisions when subsystems are revised or enhanced.
> 
> So I take it that designing the combined version is still taking place?

You can do so. It's a mid- to long-term goal while we also have to take
care of immediate user requirements.

> 
> It sounds like the current direction is more of a straight merge. 
> Since PREEMPT_RT and Xenomai are somewhat parallel in purpose would
> there be any room to eliminate duplication between the two?

There are certainly duplications, both wrt/ technology and application
scenarios, though none of both projects is a subset of the other. For
sure it would be nice if we can find common grounds, make components
interchangeable or merge features into mainline. If you already have any
rough ideas, please let us know.

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]

Re: [Xenomai-help] Xenomai + PREEMPT_RT

[Philippe Gerum]

On Wed, 2006-09-06 at 07:51 -0700, Daniel Walker wrote: 
> On Wed, 2006-09-06 at 12:50 +0200, Jan Kiszka wrote:
> > > Or even better, is anyone already doing work with these two
> > > technologies combined?
> > 
> > There once existed a combo patch (that I don't find ATM) of an ancient
> > -rt version + old Adeos, a combination that will more or less
> > automatically come again when PREEMPT_RT is mainline. Porting the
> > nucleus over PREEMPT_RT so that the skins can run "natively" is another
> > task yet to begin. Anyway, we are already including this perspective in
> > current design decisions when subsystems are revised or enhanced.
> 
> So I take it that designing the combined version is still taking place?
> 
> It sounds like the current direction is more of a straight merge. 
> Since PREEMPT_RT and Xenomai are somewhat parallel in purpose would
> there be any room to eliminate duplication between the two?

The current direction is about allowing the Xenomai core and interfaces
to run indifferently over the current co-kernel technology, or a native
preemption one, to get the real-time predictability.

Xenomai is all about the real-time interfaces it provides to
applications: emulation of legacy RTOS core APIs (e.g. VRTX, pSOS+,
VxWorks, ...) and standard specifications (e.g. POSIX 1003.1b, uITRON).
Additionally, the RTDM interface abstracts the underlying RTOS services
needed for implementing our real-time drivers for industrial devices
(e.g. CAN).

In other words, POSIX aside, all the rest would remain relevant,
including over a PREEMPT_RT kernel, particularly as a migration tool
from non-Linux environments.

-- 
Philippe.