On Tue, 28 Sep 1999, Burkhart,Kelly wrote:
> <showing my ignorance>
>
> What performance does a JIT improve? Are you talking about the time to run
> an uncompiled application with this JIT vs. another JIT (i.e. performance
> improvement is in the JIT compilation not in running the application)? Or
> are you saying that the compiled code from this JIT is faster than the
> comiled code from javac or jikes?
>
> > I've tested Borland javacomp with my 35K lines application ....
> > ...
If I understand your question ...
The most perceptible improvement was visible in the *execution* of my
application, which code was already compiled. Notice that I've tested the
*same* bytecode with different jit's.
I think there's a misunderstanding here:
Bytecode optimizations are performed by the compiler (javac). The jit
plays no difference in the 'level' of optimization of the compiled byte
code (*.class). Maybe you will spend less time to generate the code...
JIT's improve **only** the execution time, given that the byte code is
available. But they do not change the way bytecodes are generated -
That's JAVAC's job.
Once the byte code is available it can be:
- interpreted (there's nothing else to do in terms of performance that
JAVAC hasn't done before)
- compiled just-in-time (now a better jit will really make the
difference - that's may case!)
--Antonio
DCA/FEEC/UNICAMP
----------------------------------------------------------------------
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
