Re: Ada and real-time

Re: Ada and real-time

[Rick Hudson]

>>>>> On 13 Oct 92 15:32:17 GMT, russ@alliant.com (Russell McFatter) said:

Russ> (3)  No self-imposed nondeterminism:

Russ>      Can a language be nondeterministic?  Sure it can.  Take LISP, for
Russ>      example...

Russ> 	"Whoa!  That missile flew right over our heads and this thing
Russ> 	 didn't even try to intercept it!"

Russ> 	"Bummer of a time to do a garbage collection.  Sorry 'bout your
Russ> 	 city, man."


Actually, there have been techniques for doing real time GC around since the
seventies. Most Lisp systems do not utilize these techniques due to the total
overhead required (a bounded number of extra instruction on each memory read).
The fact that all Ada systems fail to offer GC is an implementation decision.
There is nothing in the Ada language definition that would prevent implementati
on
of a full blown generational copy collector and I suspect this was not by
accident. There is nothing in the Lisp language definition that prevents
real time garbage collection either.

Other than this we have considerable common ground when it comes to describing
what real time means.
--

                Richard L. Hudson, Research Associate
                University Computing Services
                Lederle Graduate Research Center
                University of Massachusetts
                Amherst, MA  01003
                (413) 545-1220; Hudson@cs.umass.edu

Re: Ada and real-time

[dog.ee.lbl.gov!hellgate.utah.edu!caen!zaphod.mps.ohio-state.edu!sol.ctr.c]

determinism is very important.  we've ditched Ada's nondeterministic
task and have instead implemented our own, deterministic tasking.

Ada's tasking is not suitable for embedded real-time systems.

Re: Ada and real-time

[mcsun!uknet!yorkohm!minster!mjl-b]

In article <1992Oct14.143733.842@saifr00.cfsat.honeywell.com> walls@saifr00.cfs
at.honeywell.com (Gerald Walls) writes:
>determinism is very important.  we've ditched Ada's nondeterministic
>task and have instead implemented our own, deterministic tasking.

This is misleading. If all your tasks have distinct priorities, then the
scheduling of those tasks is entirely deterministic.

>Ada's tasking is not suitable for embedded real-time systems.

I disagree.

Mat

| Mathew Lodge                      | "A conversation with you, Baldrick,     |
| mjl-b@minster.york.ac.uk          |  and somehow death loses its sting..."  |
| Langwith College, Uni of York, UK |  -- Blackadder II                       |

Re: Ada and real-time

[cis.ohio-state.edu!news.sei.cmu.edu!jbg]

In article <1992Oct14.143733.842@saifr00.cfsat.honeywell.com>,
walls@saifr00.cfsat.honeywell.com (Gerald Walls) writes: 

|> determinism is very important.  we've ditched Ada's nondeterministic
|> task and have instead implemented our own, deterministic tasking.
|> 
|> Ada's tasking is not suitable for embedded real-time systems.

By "non-deterministic tasking" do you mean the fact that it can be difficult
or impossible to write down a timeline for a set of Ada tasks, given
variations in execution times and the effect of preemption?  If so, this is
not, in itself, sufficient reason to conclude that Ada tasking is unsuitable
for embedded real-time systems.  

These issues are discussed in the paper "Real-time Scheduling and Ada" in the
April 1990 issue of COMPUTER magazine.  In essence, the paper shows how rate
monotonic theory allows you to ignore the "non-determinism" of Ada tasking by
showing whether each task will get enough compute time to meet its deadline.
The theory has been used on delivered embedded systems, so it is actually
applicable to real systems, and it is being used on real-time systems that are
using Ada tasking in its "non-deterministic" sense.

John B. Goodenough					Goodenough@sei.cmu.edu
Software Engineering Institute				412-268-6391

Re: Ada and real-time

[iggy.GW.Vitalink.COM!cs.widener.edu!eff!sol.ctr.columbia.edu!The-Star.hon]

In article <1992Oct15.144135.24045@sei.cmu.edu> jbg@sei.cmu.edu (John Goodenoug
h) writes:
>
>By "non-deterministic tasking" do you mean the fact that it can be difficult
>or impossible to write down a timeline for a set of Ada tasks, given
>variations in execution times and the effect of preemption?  If so, this is
>not, in itself, sufficient reason to conclude that Ada tasking is unsuitable
>for embedded real-time systems.  
>

the FAA requires this in avionics software.  Ada tasking is not
suitable for real-time avionics software.

Re: Ada and real-time

[kronos.arc.nasa.gov!iscnvx!news]

------------------------- Original Article -------------------------
   I have heard of rate-monotonic scheduling.  Are there any other scheduling
theories which might be applicable to Ada tasking and which also provide a
means of conclusively demonstrating that some or all time constraints will
be satisfied?  If so, could you please point me towards the appropriate
references.

                            Trey Haddad (ghaddad@lmsc.lockheed.com)
I found these opinions on my doorstep, could you please give them a good home?

Re: Ada and real-time

[cis.ohio-state.edu!zaphod.mps.ohio-state.edu!uakari.primate.wisc.edu!rela]

In article <1992Oct15.163026.7899@saifr00.cfsat.honeywell.com>, walls@saifr00.c
fsat.honeywell.com (Gerald Walls) writes:
|> In article <1992Oct15.144135.24045@sei.cmu.edu> jbg@sei.cmu.edu (John Gooden
ough) writes:
|> >
|> >By "non-deterministic tasking" do you mean the fact that it can be difficul
t
|> >or impossible to write down a timeline for a set of Ada tasks, given
|> >variations in execution times and the effect of preemption?  If so, this is
|> >not, in itself, sufficient reason to conclude that Ada tasking is unsuitabl
e
|> >for embedded real-time systems.  
|> >
|> 
|> the FAA requires this in avionics software.  Ada tasking is not
|> suitable for real-time avionics software.

But, the FAA is requiring the use of Ada.  Ada tasking is deterministic
given a set of circumstances.  The external inputs to the Ada program are
not deterministic.  Ada running under Unix will not be deterministic.  This
is because of the external effects of the operating system, but, then 
again, Unix is not a suitable real-time operating system.

Brad

Re: Ada and real-time

[cis.ohio-state.edu!news.sei.cmu.edu!ajpo.sei.cmu.edu!griest]

In article <1992Oct15.163026.7899@saifr00.cfsat.honeywell.com> walls@saifr00.cf
sat.honeywell.com (Gerald Walls) writes:
>In article <1992Oct15.144135.24045@sei.cmu.edu> jbg@sei.cmu.edu (John Goodenou
gh) writes:
>>
>>By "non-deterministic tasking" do you mean the fact that it can be difficult
>>or impossible to write down a timeline for a set of Ada tasks, given
>>variations in execution times and the effect of preemption?  If so, this is
>>not, in itself, sufficient reason to conclude that Ada tasking is unsuitable
>>for embedded real-time systems.  
>>
>
>the FAA requires this in avionics software.  Ada tasking is not
>suitable for real-time avionics software.


Show me the regulation that prohibits "this".  I know of commercial
avionics software that does not have purely deterministic scheduling.
(Any software that has asynchonous interrupts probably fits into
this category.)

   Tom Griest

Re: Ada and real-time

[cis.ohio-state.edu!zaphod.mps.ohio-state.edu!darwin.sura.net!cs.ucf.edu!c]

The summary says it all.  No one has mapped Ada tasking constructs
to _general_purpose_ hardware.  As I recall (don't have my LRM handy),
the language is required to execute tasks AS IF THEY WERE EXECUTING ON
SEPARATE CPU's.  This does not map directly on to any general purpose
hardware that I know of.  I've often wondered why.  Any compiler gurus
care to enlighten the universe (or are we back to the days of DP-priests
in white coats? :-).

The Ada compilers with which I've worked closely (Alsys 80286, Tartan
C30) have a large run-time penalty imposed by using the language's
tasking.  In the 2 Alsys-80286 jobs, the overhead was within
acceptable limits (the requirement was for soft real time (I bet
that raises some heckles :-)); the Tartan C30 (TI 320C30) compiler
was unable to mix interrupt handling with tasking under our timing
constraints, so we wrote procedural code called from the interrupt
handler (the requirement was for hard real time).

Jim Crigler
------------------------------------------------------------------------
If the 24th century is so great, how come they don't have a cure for
baldness?