|Re: Memory Profiling : Object exclusive Deep size [message #40769 is a reply to message #40678]
||Wed, 16 November 2005 04:32
| Nirav Thaker
Registered: July 2009
Hi Marius & Ali,|
I'm profiling an application server with lot of frameworks and libraries
with nearly 12k+ classes loaded
and generated on the fly, the fact that i've to run profiling with no-filter
seems prohibitive, not only
it will generate lot of unnecessary data it will be difficult to extract the
information, is there any programming
solution I can jump in, I'm relatively new to TPTP.
I've opened this feature in bugzilla.
"Marius Slavescu" <email@example.com> wrote in message
> Hi Nirav,
> With the information that you can currently get in the TPTP trace model in
> order to get the deep (cumulative) size for an object I see two methods
> you could use (currently TPTP doesn't compute/provide the deep size) on
> client side (you may also look into the JVMPI agent and do the
> - track all method invocations then look at the call chain to see
> constructors where called from all method invocations (defined in a class)
> against a specific object (which is an instance of the previously
> class), then you need to cumulate the object's size allocated by those
> - collect a heap dump and then try to figure out (avoiding cycles
> be tricky) from the resulting object reference graph which are all the
> objects that hangs under a specific object and cumulate their size.
> Both methods will required to run with no filters if you want to get
> accurate results, so you should expect a lot of profiling data being
> generated and processed (you should also give a lot of Java heap space to
> TPTP in that case).
> As Ali suggested please open a feature in Bugzilla.
> "Nirav Thaker" <firstname.lastname@example.org> wrote in message
> > Hi,
> > Is there any way i can determine deep size of an object with TPTP, what
> > gives is the size which will be freed upon GC. Is there any easier way
> > getting deep size?
> > thanks,
> > Nirav Thaker
Powered by FUDForum
. Page generated in 0.01702 seconds