Ian Wilson ha scritto:
> FAQ
> Excelsior Jet
[quoted text clipped - 7 lines]
> http://mindprod.com/jgloss/nativecompiler.html
> http://www.bearcave.com/software/java/comp_java.html
Thank you.
I have already considered other ways, but i think a compiled EXE is what
i need (because i need fast code).
I saw those links, but i need a Win32 compiler... exists something
(fine) but for free?
thanks
Andrew Thompson - 15 May 2007 11:08 GMT
>Ian Wilson ha scritto:
(mentioned web start)
Here are some examples..
<http://www.physci.org/jws>
>I have already considered other ways, but i think a compiled EXE is what
>i need (because i need fast code).
1st error in your assumptions.
A slow Java app. will end up a slow exe.
A fast Java app. might not only become faster on
future JVM's, but also using different hardware
configurations.
When you compile to an exe, you must generally make
decisions about what to optimise for (speed, memory
footprint etc.), whereas the JVM can optimise for each
different PC and its configuration.
Some further considerations..
<http://www.physci.org/codes/javafaq.html#exe>

Signature
Andrew Thompson
http://www.athompson.info/andrew/
Petomarmitta - 15 May 2007 12:24 GMT
Andrew Thompson ha scritto:
>> Ian Wilson ha scritto:
>
[quoted text clipped - 8 lines]
>
> A slow Java app. will end up a slow exe.
I found this: http://schmidt.devlib.org/java/native-compilers.html
Seems a compiled runs faster than an VM-pass-through...
Lew - 15 May 2007 13:38 GMT
> Seems a compiled runs faster than an VM-pass-through...
I wouldn't bet the farm on that.

Signature
Lew
Andrew Thompson - 15 May 2007 14:53 GMT
>Andrew Thompson ha scritto:
...
>I found this: http://schmidt.devlib.org/java/native-compilers.html
Perhaps you should read it again.
>Seems a compiled runs faster than an VM-pass-through...
Seems (to quote from the page)..
"Higher execution speed. Actually, that is not guaranteed, it
depends largely on the application type, the *system* *environment*,
the way a particular solution was coded etc. You must do a lot
of testing to find out if your code benefits from native execution."
OK - so let's imagine you've completed that process for a
single configuration of one system environment. You might
end up with native code that is 5% faster in that environment.
What about the next PC (with a slightly different config.), or
the one after that, or the ten others down the hall, or..
It is best left to the JVM to do optimisations, because,
as I have mentioned before in this thread, it can do
them at run-time - depending on the machine and the
execution environment as it exists at that moment.

Signature
Andrew Thompson
http://www.athompson.info/andrew/
bugbear - 15 May 2007 16:18 GMT
> Thank you.
> I have already considered other ways, but i think a compiled EXE is what
> i need (because i need fast code).
What is your application?
BugBear
Petomarmitta - 15 May 2007 16:30 GMT
bugbear ha scritto:
>> Thank you.
>> I have already considered other ways, but i think a compiled EXE is
[quoted text clipped - 3 lines]
>
> BugBear
Pathfinding (Astar) on nodes/edges graphs.
Ian Wilson - 15 May 2007 17:36 GMT
> bugbear ha scritto:
>
[quoted text clipped - 7 lines]
>
> Pathfinding (Astar) on nodes/edges graphs.
I suspect that a good algorithm and careful coding will make far more
difference than ahead-of-time compilation vs Just-In-Time compilation.
bugbear - 16 May 2007 09:34 GMT
>>> BugBear
>>
>> Pathfinding (Astar) on nodes/edges graphs.
>
> I suspect that a good algorithm and careful coding will make far more
> difference than ahead-of-time compilation vs Just-In-Time compilation.
Heartily agreed.
Why not develop under JVM and IF FORCED, move to
EXE ?
BugBear
Arne Vajhøj - 19 May 2007 21:29 GMT
> Ian Wilson ha scritto:
>> Excelsior Jet
[quoted text clipped - 7 lines]
>> http://mindprod.com/jgloss/nativecompiler.html
>> http://www.bearcave.com/software/java/comp_java.html
> I have already considered other ways, but i think a compiled EXE is what
> i need (because i need fast code).
Why are you expecting EXE to be faster than the Java code ?
I see no basis for that assumption.
> I saw those links, but i need a Win32 compiler... exists something
> (fine) but for free?
The above is available for Win32.
Arne