[kaffe] memory traces from Kaffe

Chris Gray chris at kiffer.eunet.be
Thu Sep 4 01:20:03 PDT 2003


On Wednesday 03 September 2003 23:03, who knows wrote:
> Hi Chris;
>
> I mean they are exactly same only the trace file sizes
> are different depending on the program. I would expect
>  different memory requests for each Java program.

Any Java VM has a certain amount of initialisation to do which is independent 
of the task to be performed; for example certain key classes such as 
java.lang.Object/String/Class/Throwable need to be created. So all your 
traces will probably have a common initial part which it is not interesting 
to analyse.

After that there should be some differences to be found; for example IIRC the 
mtrt test allocates millions of tiny objects, whereas compress works with 
larger arrays.  These differences tend to be obscured, however, by the fact 
that Java (like many OO languages) does tend to create and destroy a lot of 
objects.  Java's automatic garbage collection also introduces a quasi-random 
delay between the end of an object's useful life and the freeing of its 
memory.

Good luck,

Chris

> For example this is the first lines of memory trace
> file for program "compress" (taken out from Spec
> JVM98, "a" means malloc)
>
> a 56
> a 56
> a 224
> a 40
> a 4104
> a 224
> a 32
> a 56
> a 64
> a 64
> a 64
> a 64
> a 56
> a 224
> a 64
> a 40
> a 40
> a 16
> a 56
> a 56
> a 40
> a 24
> a 2016
> a 32
> a 32
> a 40
> .
> .
> .
>
>
> And the following is for "check" (Spec JVM98)
>
>
> a 56
> a 56
> a 224
> a 40
> a 4104
> a 224
> a 32
> a 56
> a 64
> a 64
> a 64
> a 64
> a 56
> a 224
> a 64
> a 40
> a 40
> a 16
> a 56
> a 56
> a 40
> a 24
> a 2016
> a 32
> a 32
> a 40
> .
> .
-- 
Chris Gray                                /k/ Embedded Java Solutions
Embedded & Mobile Java, OSGi              http://www.kiffer.be/k/
chris.gray at kiffer.be                      +32 477 599 703




More information about the kaffe mailing list