RE: [certi-dev] certi with HLA_USES_UDP / tick

[Vaibhav Bansal]

Hello,

1. Yes, MAK/Portico run smoothly.
2. All attributes of a single object are being updated at one time. All
objects belong to same class.
3. Yes, I am using 3.3.2, and I compiled it myself(Visual Studio 2008), and
I would be glad to try it after the patch too.
4. Sorry but I would not be able to share the source code :(. 
5. Same results with antivirus/firewall disabled.

Thanks and Regards
Vaibhav Bansal

-----Original Message-----
From: address@hidden
[mailto:address@hidden On Behalf Of
Eric Noulard
Sent: Monday, August 31, 2009 3:56 PM
To: CERTI development discussions
Subject: Re: [certi-dev] certi with HLA_USES_UDP / tick

2009/8/31 Vaibhav Bansal <address@hidden>:
> Hello again,
>
> 1. Yes, both the federates are running on same windows host,
>        Though the behavior does not differ much if two machines are used.

That's good news.
Do your Portico/MAK test run smoothly in the two machine test case?

> 2. Yes, the federates also display objects, and the objects are nothing
but
> simple spheres

OK.

> 3. Only the receiving federate lags. (why would the sending federate lag?)

The sending one should effectively not lag at all as long as you have enough
CPU power to run all the processes.

> 4. About 300 messages per second, (60Hz, 5 attributes updates per frame)

That should be sustainable, however one question why don't you
update all attribute at once? Do they belong to different object class?

> 5. Yes, the ratio between TICK_REQUEST and REFLECT_ATTRIBUTE_VALUES should
> be about 1:5,
>        I think that only one callback for reflectAttributes is being
> processed, even after using tick(0,1).

That's definitely a problem.
Could you confirm you are using CERTI 3.3.2?
Did you compile CERTI yourself or did you take the provided Windows
installer?
Would you be able to recompile CERTI CVS if we provide some patch?

> 6. Sorry, but it would not be possible to run my case on linux box.
>        (btw does CERTI for windows differs too much from linux version?)

Not that much however the scheduling behavior of the underlying  OS does :-)
With CERTI you get 2 processes for each federate  + the RTIG process,
the sheduling behavior of those process "may" differ on Linux and Windows
cases.
In fact we did do kind of realtime simulation experiment on Linux but
not windows.
see http://mirrors.linhub.com/savannah/certi/papers/08E-SIW-061.pdf

Would you be able to share you test federate source code with us in order
to makes it possible (no promise though)  for us to try to reproduce
the problem here
(either on Windows and/or Linux)

I have one more question, do you have any firewall/anti-virus thing
running on the machine?
If yes would you be able to stop it an rerun?

-- 
Erk
Membre de l'April - « promouvoir et défendre le logiciel libre » -
http://www.april.org


-- 
CERTI-Devel mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/certi-devel