On Fri, 30 Mar 2012, Charles Steinkuehler wrote:
Date: Fri, 30 Mar 2012 14:42:05 -0500
From: Charles Steinkuehler <[email protected]>
Reply-To: EMC developers <[email protected]>
To: EMC developers <[email protected]>
Subject: Re: [Emc-developers] The preempt_RT vs RTAI question
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On 3/30/2012 2:28 PM, Viesturs L??cis wrote:
2012/3/30 Charles Steinkuehler <[email protected]>:
I do real-time on Windows, with no kernel patches! :)
Please, do not get me wrong, I am not an IT guy by any means, but
I seriously doubt that non-patched windoze can be real-time
(although it depends on the definition of "realtime").
By real-time, I mean the system does everything required without
missing any hard real-time deadlines. This is possible for us because
of our fairly long timing windows (10 mS to 15 mS) based around video
timing, with anything needing fine timing control handled by hardware.
This would not really work for LinuxCNC, as the feedback cycle has to
happen much faster, and if your process timing jitter becomes a
significant portion of the overall feedback cycle period, bad things
start to happen! :)
LinuxCNC is about being smart controller on dumb hardware, it is
designed to do all the thinking part of the machine controls so
that hardware does the technical tasks - generate signals, count
incoming pulses etc, but all the intelligence, like closing
feedback loop for servos, all the kinematic calculations etc - that
is done inside LinuxCNC.
I understand that, but with FPGA boards available, having to rely on
extremely low latency figures for decent software control seems
unnecessary. Peter indicated earlier versions of his hardware had
support for assisting the software with timing issues, but that
LinuxCNC didn't make use of the feature. Maybe I found something I
can try to tackle!
I should make it clear that this is not necessarily a desired feature.
LinuxCNC real time architecture is a based on a single control locus/time
domain system with no buffering of real time tasks. Anything that compromises
this simple model is likely have lots of undesired side effects, and for most
real controls, the current latencies (10->20 Usec is easy with currently
avalable hardware) better timing is not needed. This is why my suggestions
are things that will work with the current timing model.
Peter Wallace
Mesa Electronics
(\__/)
(='.'=) This is Bunny. Copy and paste bunny into your
(")_(") signature to help him gain world domination.
------------------------------------------------------------------------------
This SF email is sponsosred by:
Try Windows Azure free for 90 days Click Here
http://p.sf.net/sfu/sfd2d-msazure
_______________________________________________
Emc-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-developers
