certi-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: RE : Re: RE : RE: [certi-dev] Real-Time HLA simulation?


From: Eric Noulard
Subject: Re: RE : Re: RE : RE: [certi-dev] Real-Time HLA simulation?
Date: Wed, 30 Apr 2008 10:42:53 +0200

2008/4/30 HADJ AMOR HASSEN <address@hidden>:
> Hello again,
>
> Thanks Petr and Erik for answering. I learned carefully your answers but
> some things are fuzzy.
>
> Erik can you explain me more this sentence?
>
> (freezing may simply be calling TAG and waiting for the RTI to say YES.)

This is a typo.
You send TAR to the next simulation step you want.
The RTI will block you in your "tick loop" since your
are waiting for TAG (the TAG is the TAR callback)

> I searched in the code implementation of TimeAdvanceGrant() for blocking
> functions like sleep() but nothing. I want to know exactly (the code) how
> the federate is freezed when he requests TAG.

My mistake, tick may is the blocking function which makes you wait
between your TAR and the TAG RTI answer.

tick is blocking in select system call.

moreover in CERTI 3.2.7 and CERTI CVS there is a Petr contribution
which adds timeout version of the tick which may be handy for RT simulation.
see https://savannah.nongnu.org/bugs/index.php?22798

> What i understood :
>
> 1- do local computation.
> 2- send update of attributes.
> 3- Calculate the difference between RealTime and simulated time.
4- TAR to the difference calculated before
5- tick with timeout
6- When you receive TAG you may have to wait (yourself) until RealTime
== Current FedTime

>
> I am not sure that is right. Can you confirm?

This is a possible scheme.

> Finally, another idea is to create a Federate Manager which is both
> regulating and constrained and which request time advance. If the federate
> Manager didn't request time advance, this will freeze all the federation.

I think this almost the same scheme but you have a single Federate which
is regulating and others are "only" contrained.


-- 
Erk




reply via email to

[Prev in Thread] Current Thread [Next in Thread]