Re: Compare Swarm with Repast

[Marcus G. Daniels]

Gulyas Laszlo wrote:

> On Mon, 12 Aug 2002, Russell Standish wrote:
>  
>
> > Problems with Java.
> >
> > 1) Its slow
> >    
>
> OK. Compared to, exactly what? It's much slower than
> C, of course, but I am not sure about Objective-C (with
> TCL/TK, JavaSwarm, etc.). Especially not on a Windows
> computer that acutally has to emulate Unix to start it up.
Run a Hello World with Sun JVM on a Windows box and compare
it to the runtime of a Cygwin executable.  In real time, the latter
is about six times faster.  

Now compile with -mno-cygwin and generate a pure Windows
executable with none of that scary "emulate Unix" stuff and you'll see 
that the
difference is about seven times faster.   Not exactly what I would call 
painful.

Note that the JVM DLL itself is even bigger than the Cygwin, and it is 
just the
first of several DLLs to get loaded in practice, never mind the 
countless bytecode classes.

Of course, this has nothing to do with Java per se, just a common result 
of using a common implementation of Java.  There are other options, such 
as using native code compilers for Java like GCJ. GCJ-compiled Java 
classes are binary-compatible with G++ compiled C++ classes, and except 
for memory allocation overhead of the conservative garbage collector, 
have the same performance: the performance of C.

When it comes to screen widgetry the story of tradeoffs is the same. 
For example, with JavaSwarm  there are the Tcl/Tk libraries to be 
loaded.  With a pure Java there's the AWT DLL, and tons of Swing 
bytecode.   There's no free lunch.



                 ==================================
  Swarm-Modelling is for discussion of Simulation and Modelling techniques
  esp. using Swarm.  For list administration needs (esp. [un]subscribing),
  please send a message to <address@hidden> with "help" in the
  body of the message.
                 ==================================