Task execution time

Task execution time

[Lionel.DRAGHI]

There is a general agreement that the only feature missing in Ada to fully
code flexible schedulers is a way to get task execution time.

Is there any ongoing proposal for the next language revision?


-- 
Lionel Draghi

Re: Task execution time (writing a scheduler)

[Larry Kilgallen]

In article <mailman.4.1056375821.8204.comp.lang.ada@ada.eu.org>, Lionel.DRAGHI@fr.thalesgroup.com writes:
> There is a general agreement that the only feature missing in Ada to fully
> code flexible schedulers is a way to get task execution time.

In what forum was that general agreement reached ?

After 24 years on the market, VMS has rather extensive hooks for coding
such schedulers, but it took many years to get to that point.  There are
many IO-related OS primitive measurements needed to achieve fairness.

I presume you intended that the schedulers be adequate to meet needs
rather than merely being flexible.  I can make a scheduler much more
flexible by adding an larger color palatte from which to enhance the
logo it displays on a screen.

Re: Task execution time

[Martin Dowie]

<Lionel.DRAGHI@fr.thalesgroup.com> wrote in message
news:mailman.4.1056375821.8204.comp.lang.ada@ada.eu.org...
> There is a general agreement that the only feature missing in Ada to fully
> code flexible schedulers is a way to get task execution time.
>
> Is there any ongoing proposal for the next language revision?

http://www.ada-auth.org/AI-SUMMARY.HTML

Is AI-307 what you're after?

RE: Task execution time (writing a scheduler)

[Lionel.DRAGHI]

| -----Message d'origine-----
| De: Kilgallen@SpamCop.net [mailto:Kilgallen@SpamCop.net]
...
| In article 
| <mailman.4.1056375821.8204.comp.lang.ada@ada.eu.org>, 
| Lionel.DRAGHI@fr.thalesgroup.com writes:
| > There is a general agreement that the only feature missing 
| in Ada to fully
| > code flexible schedulers is a way to get task execution time.
| 
| In what forum was that general agreement reached ?
| 
| After 24 years on the market, VMS has rather extensive hooks 
| for coding
| such schedulers, but it took many years to get to that point. 
|  There are
| many IO-related OS primitive measurements needed to achieve fairness.
| 
| I presume you intended that the schedulers be adequate to meet needs
| rather than merely being flexible.  I can make a scheduler much more
| flexible by adding an larger color palatte from which to enhance the
| logo it displays on a screen.

I am not sure i understand your point, but yes, "flexible" is vague. 
I meant mainly EDF schedulers.

-- 
Lionel Draghi

Re: Task execution time

[Martin Dowie]

"Martin Dowie" <martin.dowie@no.spam.btopenworld.com> wrote in message
news:bd746a$54p$1@titan.btinternet.com...
> <Lionel.DRAGHI@fr.thalesgroup.com> wrote in message
> news:mailman.4.1056375821.8204.comp.lang.ada@ada.eu.org...
> > There is a general agreement that the only feature missing in Ada to
fully
> > code flexible schedulers is a way to get task execution time.

Oops - I see you don't literally mean "get task execution time" :-)

RE: Task execution time (writing a scheduler)

[Larry Kilgallen]

In article <mailman.6.1056380390.8204.comp.lang.ada@ada.eu.org>, Lionel.DRAGHI@fr.thalesgroup.com writes:
> 
> 
> | -----Message d'origine-----
> | De: Kilgallen@SpamCop.net [mailto:Kilgallen@SpamCop.net]
> ...
> | In article 
> | <mailman.4.1056375821.8204.comp.lang.ada@ada.eu.org>, 
> | Lionel.DRAGHI@fr.thalesgroup.com writes:
> | > There is a general agreement that the only feature missing 
> | in Ada to fully
> | > code flexible schedulers is a way to get task execution time.
> | 
> | In what forum was that general agreement reached ?
> | 
> | After 24 years on the market, VMS has rather extensive hooks 
> | for coding
> | such schedulers, but it took many years to get to that point. 
> |  There are
> | many IO-related OS primitive measurements needed to achieve fairness.
> | 
> | I presume you intended that the schedulers be adequate to meet needs
> | rather than merely being flexible.  I can make a scheduler much more
> | flexible by adding an larger color palatte from which to enhance the
> | logo it displays on a screen.
> 
> I am not sure i understand your point, but yes, "flexible" is vague. 
> I meant mainly EDF schedulers.

To communicate better, you should explain what you mean by EDF.

Re: Task execution time (writing a scheduler)

[Ole Kristensen]

Kilgallen@SpamCop.net (Larry Kilgallen) writes:

<snip>

> To communicate better, you should explain what you mean by EDF.

Earliest deadline first.

Re: Task execution time (writing a scheduler)

[Larry Kilgallen]

In article <m3smq076jr.fsf@linux.local>, Ole Kristensen <ole@linux.local> writes:
> Kilgallen@SpamCop.net (Larry Kilgallen) writes:
> 
> <snip>
> 
>> To communicate better, you should explain what you mean by EDF.
> 
> Earliest deadline first.

Well, that rule would certainly make for a rather simplistic scheduler,
allowing a task with an early deadline to fully lock out other tasks
with intermittant CPU needs to miss their deadline because the first
of their IO operations (which automatically yield the CPU) were not
allowed to start until the early deadline task finished.

RE: Task execution time (writing a scheduler)

[Lionel.DRAGHI]

| -----Message d'origine-----
| De: Kilgallen@SpamCop.net [mailto:Kilgallen@SpamCop.net]
...
| 
| To communicate better, you should explain what you mean by EDF.
OK, but within this context, EDF is widely used.

I have checked AI-307, and this is exactly what i was looking for, thank you
Martin.

It's tagged:
!priority High
!difficulty Hard
I hoppe compiler vendors will endorse the proposal in spite of the
difficulty.


-- 
Lionel Draghi

Re: Task execution time

[Marin David Condic]

I took a peek at this briefly and I think it would be a nice feature to 
have. But it could be a tough thing to do in a variety of platforms 
where it would count most. (Incredibly useful in realtime embedded 
systems with no OS and where you have to keep track of CPU utilization 
to stay on budget.) Important to this would be an accurate method of 
determining total CPU utilization versus wall time. I don't know if 
AI-307 covers that since I just looked at the summary & didn't want to 
pull the full text from the CVS tree.

MDC

Martin Dowie wrote:
> 
> 
> http://www.ada-auth.org/AI-SUMMARY.HTML
> 
> Is AI-307 what you're after?
> 
> 


-- 
======================================================================
Marin David Condic
I work for: http://www.belcan.com/
My project is: http://www.jast.mil/

Send Replies To: m c o n d i c @ a c m . o r g

     "In general the art of government consists in taking as
     much money as possible from one class of citizens to give
     to the other."

         --  Voltaire
======================================================================

Re: Task execution time

[Martin Dowie]

"Marin David Condic" <nobody@noplace.com> wrote in message
news:3EF841BD.7000906@noplace.com...
> I took a peek at this briefly and I think it would be a nice feature to
> have. But it could be a tough thing to do in a variety of platforms
> where it would count most. (Incredibly useful in realtime embedded
> systems with no OS and where you have to keep track of CPU utilization
> to stay on budget.) Important to this would be an accurate method of
> determining total CPU utilization versus wall time. I don't know if
> AI-307 covers that since I just looked at the summary & didn't want to
> pull the full text from the CVS tree.

Yes, that's my take on it too. I think something useful could be added
to, say VxWorks (taskHookLib) by a user implementation. But to
acheive really good accuracy, I suspect will take the compiler vendor
and OS provider to provide.

For Win32, access the Task Manager Performance could be a useful
implementation and again, for real accuracy, I suspect it's  up to the
compiler vendor and OS provider...

Re: Task execution time (writing a scheduler)

[@ 2003-06-24 13:58 UTC (permalink / raw)]

Lionel.DRAGHI@fr.thalesgroup.com wrote:
> 
> | -----Message d'origine-----
> | De: Kilgallen@SpamCop.net [mailto:Kilgallen@SpamCop.net]
> ...
> | 
> | To communicate better, you should explain what you mean by EDF.
> OK, but within this context, EDF is widely used.
> 
> I have checked AI-307, and this is exactly what i was looking for, thank you
> Martin.
> 
> It's tagged:
> !priority High
> !difficulty Hard
> I hoppe compiler vendors will endorse the proposal in spite of the
> difficulty.
> 
> 

Last minute information about this and other proposals (from Alan Burns):

"There was a WG9 and an ARG meeting last week as part
of Ada Europe. I thought it might be interesting
to you all to get an update on the progress
of real-time AIs

Ravenscar AIs (249 and 305) - have already passed through WG9

Partition Elaboration Policy (265) now agreed by WG9

Execution Timers (budget timing) (307) agreed by ARG,
under editorial review - will pass to WG9

Non-Preemptive Dispatching (298) agreed by ARG,
under editorial review - will pass to WG9

Rewording of D.2 (321) agreed by ARG,
under editorial review - will pass to WG9

Timing Event (297) agreed by ARG,
under editorial review - will pass to WG9

Task Termination Procedure (266) still alive but best
way to meet varying demand still not certain.

Dynamic Ceiling (327), well received, work to be done
but looks likely to succeed.

All in all very good progress is being made - let get
more together at the forthcoming workshop in September.

Alan
-- Professor Alan Burns, Head of Department Department of Computer 
Science University of York Heslington York, UK YO10 5DD"