Hi everyone!
Do you know or think that Excelsior JET is really much more performing than Hotspot VM as they stand on their website?
http://www.excelsior-usa.com/jet.html
See benchmarks as it shows JET as being by far more performing than any VM.
I can understand that the JNI benchmark is more performing…
Funny you should think that. All the stuff that matters to me JET comes out slower, save for one annoying-yet-non-essential feature: start-up time. A pain in the butt, yet far less important than general app performance (to me, at any rate).
You did check what the different benchmarks actually do, didn’t you?
EDIT: missed off a bit - I was going to add “JNI excepted” ;).
EDIT: e.g. http://www.excelsior-usa.com/jetbenchman.html - JET is 50% slower than the server JVM!!!
I see they’re also making some dodgy marketing BS statements, like claiming that native-compiling your code “protects” you from reverse-engineering.
They scare-monger that decompiled native code is really hard to understand because it’s all native assembler…as if they weren’t aware that compilers have existed for many years that compile FROM assembler TO higher-level languages (e.g. C, C++, … Java). :(.
[quote]A pain in the butt, yet far less important than general app performance (to me, at any rate).
[/quote]
100% agree.
yes so how do you explain these results while Sun has always declared that their VM runs at almost (95%-100%) the same speed of a C program?
[quote]I see they’re also making some dodgy marketing BS statements, like claiming that native-compiling your code “protects” you from reverse-engineering.
They scare-monger that decompiled native code is really hard to understand because it’s all native assembler…as if they weren’t aware that compilers have existed for many years that compile FROM assembler TO higher-level languages (e.g. C, C++, … Java). :(.
[/quote]
I don’t know if decompiling class files is so a problem when you distribute obfuscated code with a good obfuscator (like mighty bubbles jar).
I’ve tried JET 3-4 years ago and my conclusion was that it never ran faster then Hotspot and even worst, it ran slower in certain contexts.
For my usage, Jet has produced executables of approximately similar size, which start up in the blink of an eye, and run incredibly fast - typically at around the speed of -server (which is fast indeed).
Cas 
That is a good point! What is the difference of performance between -client and -server in percentage in your case?
And is it worth buying it because if I remember correctly it’s quite expensive isn’t?
If you need it, nothing else will do.
Server can be up to twice as fast as client. Jet pretty consistently is as fast as the server VM for my game code.
Cas
Looking at the benchmarks at http://www.excelsior-usa.com/jetbench.html I would say JRockit looks damn good. :o
Has any1 tried using it for games and compared its performance to Java 5 -server?