Skip to main content



      Home
Home » Archived » Memory Analyzer (MAT) » Eden v Old
Eden v Old [message #10229] Sun, 03 May 2009 18:57 Go to next message
Eclipse UserFriend
Is there any way to identify which objects are in Eden versus Survivor
spaces v Tenured space?

And/or get the object copy count? (i.e the counter of the number of times
the object has survived GC in the young generation).

I'm attempting to identify what's causing the generation-0 garbage
collection to take so long.
Re: Eden v Old [message #10295 is a reply to message #10229] Tue, 05 May 2009 15:56 Go to previous message
Eclipse UserFriend
Hi Michael,

sorry, bad news, the heap dump neither contains the info about the space
the object is in (also not if the String is in perm) nor an object copy
count. With a heap dump only you will have problems to figure this out.

Maybe take a look at the number of weak and soft references. Many of those
references increase the stop-the-world time, because the VM has to decide
at one point or another whether to mark the objects "behind" the refrence
or not. From what I understand, this has to be done in sync with all gc
threads.

In the OQL pane, you could run "SELECT * FROM INSTANCEOF
java.lang.Reference" to see all those references currently in the heap.
Also the "Java Basics" -> Reference statistics queries can give you an
overview of what those references currently keep alive.

The other point to look into are finalizers. As you probably, finalizers
always require 2 full gc to actually free up the space. Again, the "Java
Basics" category also contains a query which displays finalizers and
what's in the queue.



Hopefully I could give you some hints,


Andreas.
Previous Topic:New inspector feature: Show SWT widget and whether they're disposed
Next Topic:hprof stacktrace info support
Goto Forum:
  


Current Time: Sun Apr 20 16:22:07 EDT 2025

Powered by FUDForum. Page generated in 0.04175 seconds
.:: Contact :: Home ::.

Powered by: FUDforum 3.0.2.
Copyright ©2001-2010 FUDforum Bulletin Board Software

Back to the top