Woody, Tom, JohnUh, yea. I put copies of some posts re this at;
The file specification you provided above did not work for me.
I haven't tried a black box linux yet but I plan to show one off at NAMES
in three months. NIST sees NML as a fully capable real time communication
system. They say;"To achieve efficient communication between distributed real-time
processes, it is desirable to both choose the best protocol for each
ommunication path and limit variation to improve software portability.
These divergent goals can be satisfied through the use of a uniform
pplication programming interface (API) that hides the details of specific
protocols from programmers. The Neutral Message Language (NML) is a
uniform API to communication functions that includes many popular
protocols: interprocess shared memory; interprocessor backplane global
memory; and Internet networking. NML implements a mailbox model for
ommunication, with both queued- and non-queued access, blocking- and
non-blocking reads and writes, and multiple readers and writers. NML
provides language bindings for both C++ and Java. The protocol
parameters are contained in configuration files that are read at run time, so
hat a system's allocation of processes to processors can be deferred as late
s desired and modified dynamically. NML handles mutual exclusion for
data integrity, and converts between native machine format and neutral
data encoding when necessary."William P. Shackleford, Frederick M. Proctor, and John L. Michaloski
National Institute of Standards and Technology
ABSTRACT
--Now since most all of the EMC communicates using NML you should be able to
split up the task most any place. So long as the NML channels exist
between processors the EMC should not care where the tasks start, how they
pipe, or where they end.Perhaps we would like to share this task by selecting a common SBC and
some other hardware. I've done some looking at pc104 but a pentium based
SBC seems like it would do the job at lower cost but without the hassle of
changing mother boards more often than some change socks.
I only mentioned the PC104 because of the current Embeded Linux
Journal design contest using this type of SBC. I agree with
the Pentium based SBC, the smaller the better. My idea is to put
the motion critical stuff on the SBC and the interpretation, calculation
and user interface stuff on the control PC. EMC is a place to
start, but what I think is really needed for this idea to take off is a
well documented command set for motion control on the SBC that can be interfaced
to by just about anyone who wants to write a gcode or hpgl or what ever
interpreter and user interface on just about any type of control PC.
I think the ultimate would be pulse generator software and the attendant
operating system in ROM or Flash ROM running in a diskless environment.
You turn it on and it is ready to accept simple ascii commands from the
control PC to control some number of axis, ideally more than 4. It
would also be nice for the SBC to monitor axis position through encoders
and look after the home and limit sensors. My design goal for the
pulse generator would be something smart enough to handle the above functions,
but did NOT requre mass storage, display or keyboard.
I think you could do this now, without EMC, by using IndexerLPT, some form of DOS based single board computer, and a rather simple program to read commands from the serial port and pass them over to IndexerLPT. I know, it would not be low cost, public domain or anything like that, but the concept could be proven quite easily.
Most of us on this list are obviously knowledgeable enough to deal with Linux, RT-Linux and EMC, but if low cost CNC for the average hobby machinist or small commercial shop is to really grow in use, setup needs to be much easier than it is today.
thanks
John Guenther
Thanks for bringing this up.
--Ray
On Fri, 08 Dec 2000, Woody wrote:
> (Dang, I've unlurked and now I can't stop...)
>
> If you look at the NIST EMC/RCS/NML stuff, they claim this should be
> a 'feature' using something like TCP/IP/UDP and modifying the NML
> configuration. I'd be _real_ interested if someone has (or would) try this.
>
> I'm _VERY_ new to EMC and still struggling with RCS/NML and haven't
> taken more than the cursory look at the division of duties between io,
> mot and task.
>
> If anybody could throw me some detailed info on that sort of thing and
> perhaps how the UI interfaces with EMC (i'm definitely not a Tkl guru),
> maybe a tutorial for us all that would be most excellent.
>
> -Woody
>
> ----- Original Message -----
> From: John Guenther <jguenthe-at-nafis.fp.trw.com>
> To: Multiple recipients of list <emc-at-nist.gov>
> Sent: Friday, December 08, 2000 11:39 AM
> Subject: Can EMC be run accross two PC's
>
>
> >
> > Hi All,
> >
> > Here is my question, can EMC be split up and run on two PC's? The idea
> > is to put the emcio and emcmot functions on a dedicated processor, maybe
> > even a single board computer running LInux. I would then be running the
> > emctask functions, the gui and what ever else on another computer.
> > Connection between the two computers would / could be by high speed
> > serial interface.
> >
> > My theory is that this might allow the emctask functions to be performed
> > by a Windoze computer with a familiar user interface. The time critical
> > and complicated functions would be performed on a dedicated computer
> > running rt-linux.