Question about the PrintMethodStatistics report

[Francis Rangel]

The programs are benchmarks. I'm using some programs of Dacapo, Java
Grande Forum and Shootout benchmarks to do some tests with the inline
optimization.
If one class is loaded and it has a implementation of a method named
"A", and there's just this implementation loaded at the moment, this
method will be a static method, right? Then, after the garbage
collector cleaned this class, another class with other implementation
of "A" method is loaded, the method would still be static?
I'm asking because I was wondering that this changes occured when the
garbage collector cleaned some implementations and the method become
static. Then, in other execution, the garbage collector didn't clean
the implementations and the method was considered virtual. Do you
think this might happening?

Regards.



2009/11/25 Christian Thalinger <Christian.Thalinger at sun.com>:
> On Tue, 2009-11-24 at 17:20 -0200, Francis Rangel wrote:
>> Hi there!
>>
>> I'm new here and I want to introduce myself. My name is Francis Rangel
>> and I'm a computing student. My final paper subject is the inline
>> optimization of the openJDK JVM.
>> Well, what I need to know is why the PrintMethodStatistics report
>> changes the amount of each kind of method using diferent compilers.
>> For example, when the programa run with the -server option, the static
>> methods are 100. Then, when the -client compiler is used, the static
>> methods come do 90.
>> If the program wasn't modified, why does this numbers change?
>> Is that because the devirtualization applied on virtual methods?
>
> It does not show different numbers for me.  One possibility could be
> that your environment is different (e.g. LANG) and so different classes
> are loaded (given you're not using the same terminal to run the
> program).
>
> What program is this?  You can try -XX:+PrintSystemDictionaryAtExit and
> compare the output of a C1 and C2 run.
>
> -- Christian
>
>



-- 
Att.

Francis Rangel