Real-time scheduling

Real-time scheduling Is it possible to make real-time scheduling with LW ?

To test it, I tried to make a simple chronometer in milliseconds. It runs almost correctly... There is a little granularity - generally acceptable – but witch can sometime result on loss of 10, 20 or even 30 ms steps (witch is really excessive...)

Probably the GC is the cause of that, but I didn’t find a good solution to avoid this effect with the different controllers of GC activity...

Do you know some source examples witch can help me ?

Thanks for infos

Denis

----------------------------------------------------
Denis Pousseur
6 clos du Drossart
1180 Bruxelles, Belgique

Mail : denis.pousseur@compositeurs.be
Website : http://compositeurs.be
----------------------------------------------------

Re: Real-time scheduling

Which OS are you running? I could be the OS, maybe its doing some system 
things.
My Windows PC clock regularly loses large amounts of time (a few seconds 
a day).

My intial reaction to your real-time needs (< 10ms) is, yikes!

I do think however that LW does have a "real-time" GC floating around 
somewhere, though
I am not sure where I read that.

Wade

Denis Pousseur wrote:

> Is it possible to make real-time scheduling with LW ?
>
> To test it, I tried to make a simple chronometer in milliseconds. It 
> runs almost correctly... There is a little granularity - generally 
> acceptable – but witch can sometime result on loss of 10, 20 or even 
> 30 ms steps (witch is really excessive...)
>
> Probably the GC is the cause of that, but I didn’t find a good 
> solution to avoid this effect with the different controllers of GC 
> activity...
>
> Do you know some source examples witch can help me ?
>
> Thanks for infos
>
> Denis
>
>
> ----------------------------------------------------
> Denis Pousseur
> 6 clos du Drossart
> 1180 Bruxelles, Belgique
>
> Mail : denis.pousseur@compositeurs.be
> Website : http://compositeurs.be
> ----------------------------------------------------
>

Re: Real-time scheduling

Thanks for your answer.

I run on Mac OS 10.4.
Excuse me, as you can see my English is very poor : what's mean  "yikes"
exactly? Is it the expression of a scepticism (witch I suspect...) ?

I was myself very sceptic about the possibility of doing real-time
scheduling in Lisp, but someone told me that, with native multi threads,
it's possible to make something very efficient. He told me about some
examples on OpenMcl. This is why I made these tests.

Best

Denis


Le 10/01/06 21:54, « Wade Humeniuk » <whumeniu@telus.net> a écrit :

> 
> Which OS are you running? I could be the OS, maybe its doing some system
> things.
> My Windows PC clock regularly loses large amounts of time (a few seconds
> a day).
> 
> My intial reaction to your real-time needs (< 10ms) is, yikes!
> 
> I do think however that LW does have a "real-time" GC floating around
> somewhere, though
> I am not sure where I read that.
> 
> Wade
> 
> Denis Pousseur wrote:
> 
>> Is it possible to make real-time scheduling with LW ?
>> 
>> To test it, I tried to make a simple chronometer in milliseconds. It
>> runs almost correctly... There is a little granularity - generally
>> acceptable  but witch can sometime result on loss of 10, 20 or even
>> 30 ms steps (witch is really excessive...)
>> 
>> Probably the GC is the cause of that, but I didn¹t find a good
>> solution to avoid this effect with the different controllers of GC
>> activity...
>> 
>> Do you know some source examples witch can help me ?
>> 
>> Thanks for infos
>> 
>> Denis
>> 
>> 
>> ----------------------------------------------------
>> Denis Pousseur
>> 6 clos du Drossart
>> 1180 Bruxelles, Belgique
>> 
>> Mail : denis.pousseur@compositeurs.be
>> Website : http://compositeurs.be
>> ----------------------------------------------------
>> 
> 
> 



----------------------------------------------------
Denis Pousseur
6 clos du Drossart
1180 Bruxelles, Belgique

Mail :  denis.pousseur@compositeurs.be
Website : http://compositeurs.be
----------------------------------------------------

Re: Real-time scheduling

On Jan 10, 2006, at 2:55 PM, Denis Pousseur wrote:

>
> Thanks for your answer.
>
> I run on Mac OS 10.4.
> Excuse me, as you can see my English is very poor : what's mean   
> "yikes"
> exactly? Is it the expression of a scepticism (witch I suspect...) ?

I'd say "yikes" is an expression of amazement, shock or fear. My  
French is poorer than your English, but I'd venture that "yikes" fits  
in somewhere between "sacre bleu" and "merde".
>
> I was myself very sceptic about the possibility of doing real-time
> scheduling in Lisp, but someone told me that, with native multi  
> threads,
> it's possible to make something very efficient. He told me about some
> examples on OpenMcl. This is why I made these tests.
>
> Best
>
> Denis
>
>
> Le 10/01/06 21:54, « Wade Humeniuk » <whumeniu@telus.net> a écrit :
>
>>
>> Which OS are you running? I could be the OS, maybe its doing some  
>> system
>> things.
>> My Windows PC clock regularly loses large amounts of time (a few  
>> seconds
>> a day).
>>
>> My intial reaction to your real-time needs (< 10ms) is, yikes!
>>
>> I do think however that LW does have a "real-time" GC floating around
>> somewhere, though
>> I am not sure where I read that.
>>
>> Wade
>>
>> Denis Pousseur wrote:
>>
>>> Is it possible to make real-time scheduling with LW ?
>>>
>>> To test it, I tried to make a simple chronometer in milliseconds. It
>>> runs almost correctly... There is a little granularity - generally
>>> acceptable – but witch can sometime result on loss of 10, 20 or even
>>> 30 ms steps (witch is really excessive...)
>>>
>>> Probably the GC is the cause of that, but I didn’t find a good
>>> solution to avoid this effect with the different controllers of GC
>>> activity...
>>>
>>> Do you know some source examples witch can help me ?
>>>
>>> Thanks for infos
>>>
>>> Denis
>>>
>>>
>>> ----------------------------------------------------
>>> Denis Pousseur
>>> 6 clos du Drossart
>>> 1180 Bruxelles, Belgique
>>>
>>> Mail : denis.pousseur@compositeurs.be
>>> Website : http://compositeurs.be
>>> ----------------------------------------------------
>>>
>>
>>
>
>
>
> ----------------------------------------------------
> Denis Pousseur
> 6 clos du Drossart
> 1180 Bruxelles, Belgique
>
> Mail :  denis.pousseur@compositeurs.be
> Website : http://compositeurs.be
> ----------------------------------------------------
>
>
>
>

Re: Real-time scheduling

Denis Pousseur <denis.pousseur@compositeurs.be> writes:

> I was myself very sceptic about the possibility of doing real-time
> scheduling in Lisp, but someone told me that, with native multi threads,
> it's possible to make something very efficient. He told me about some
> examples on OpenMcl. This is why I made these tests.

Japanese lisp legend Takeuchi claims that he has achieved mean wake up
times of lisp threads of 14 microseconds, with 119 microseconds as the
worst delay - on his vintage (1993) lisp machine!

(See his JLUGM 2000 paper on
 http://jp.franz.com/jlug/en/jlugm2000/index.html)
-- 
  (espen)

Re: Real-time scheduling

At 10/01/2006 22:55, Denis Pousseur wrote:

>Thanks for your answer.
>
>I run on Mac OS 10.4.
>Excuse me, as you can see my English is very poor : what's mean  "yikes"
>exactly? Is it the expression of a scepticism (witch I suspect...) ?
>
>I was myself very sceptic about the possibility of doing real-time
>scheduling in Lisp,

According to the time constraint, it'is EVEN 
possible with standard CL implementations. I have 
done this about 10 years ago, using ACL on a Sun 
sparc, time limit less than 1 sec - (1/2 sec if I remember). Look at OPA here:
http://algo.be/ref-projects.htm

In the real-time sub-system, the global GC was 
disabled, after the loading of the knowledge 
bases. The application proved to be viable 
regarding the real time constraint (the main problem was KB validation).

Bien à vous,

Francis

>but someone told me that, with native multi threads,
>it's possible to make something very efficient. He told me about some
>examples on OpenMcl. This is why I made these tests.
>
>Best
>
>Denis
>
>
>Le 10/01/06 21:54, « Wade Humeniuk » <whumeniu@telus.net> a écrit :
>
> >
> > Which OS are you running? I could be the OS, maybe its doing some system
> > things.
> > My Windows PC clock regularly loses large amounts of time (a few seconds
> > a day).
> >
> > My intial reaction to your real-time needs (< 10ms) is, yikes!
> >
> > I do think however that LW does have a "real-time" GC floating around
> > somewhere, though
> > I am not sure where I read that.
> >
> > Wade
> >
> > Denis Pousseur wrote:
> >
> >> Is it possible to make real-time scheduling with LW ?
> >>
> >> To test it, I tried to make a simple chronometer in milliseconds. It
> >> runs almost correctly... There is a little granularity - generally
> >> acceptable  but witch can sometime result on loss of 10, 20 or even
> >> 30 ms steps (witch is really excessive...)
> >>
> >> Probably the GC is the cause of that, but I didn¹t find a good
> >> solution to avoid this effect with the different controllers of GC
> >> activity...
> >>
> >> Do you know some source examples witch can help me ?
> >>
> >> Thanks for infos
> >>
> >> Denis
> >>
> >>
> >> ----------------------------------------------------
> >> Denis Pousseur
> >> 6 clos du Drossart
> >> 1180 Bruxelles, Belgique
> >>
> >> Mail : denis.pousseur@compositeurs.be
> >> Website : http://compositeurs.be
> >> ----------------------------------------------------
> >>
> >
> >
>
>
>
>----------------------------------------------------
>Denis Pousseur
>6 clos du Drossart
>1180 Bruxelles, Belgique
>
>Mail :  denis.pousseur@compositeurs.be
>Website : http://compositeurs.be
>----------------------------------------------------