certi-devel
[Top][All Lists]
Advanced

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

Re: CERTI Java-binding ?; Was: [certi-dev] certi performance comparison


From: Eric Noulard
Subject: Re: CERTI Java-binding ?; Was: [certi-dev] certi performance comparison
Date: Wed, 2 Jul 2008 16:27:07 +0200

2008/7/2 Gotthard,Petr <address@hidden>:
>> > I think this is the best way to do HLA  bindings to
>> different languages.
>>
>> Pierre will be glad to ear you :=)
>> The only trouble here is that the Java Federate has to launch
>> a native executable (RTIA). So Java Federate depends on a
>> RTIA running on the targeted platform.
>
> To me this doesn't matter. There is always some platform and I have to
> compile the RTIG anyway. (Unless you want to have Java RTIG as well. But
> this leads us to Portico.)

Nope, currently you may run RTIG anywhere.
You may have RTIG on Linux 32
                     a Federate on Solaris 64
                     another Fedarate on exotic platform using Java binding

>> We discuss the fact that we may implement RTIA in java too,
>> this way the Federate is 100% java and RTIA in java may be
>> java thread with potentially better performance than
>> (TCP) socket communication with the C++ RTIA.
>
> But even the C++ federates communicate with RTIA via socket.

Yes almost, unix RTIA use UNIX socket as a default
whereas Windows and preliminary Java use TCP Socket.

If we focus on High-Performance RTI (my long term goal)
we may switch to shared memory instead of socket
(or even multi-thread but I don't want to shout that too loud because
 it may generate a flame :-)
this has been done in the past for a project named HP-CERTI
which was designed for SMP cluster.
http://www.sisostds.org/index.php?tg=fileman&idx=get&id=2&gr=Y&path=Simulation+Interoperability+Workshops%2F2004+Fall+SIW%2F2004+Fall+SIW+Papers+and+Presentations&file=04F-SIW-014.pdf

but the resulting work has never been merged to CERTI trunk :-(

>> The conclusion was that implementing RTIA in java is not an
>> easy task, so we started with a "simple" libRTI in java.
>
> Implementing a new RTIA would duplicate the effort: all RTIA
> extensions/bugfixes would have to be implemented twice.

That the tough point, you are right.
That's why it's wise not to begin with this solution and examine
pro and cons before hitting the keyboard.
I totally agree with you on this point.

-- 
Erk




reply via email to

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