Re: [rvm-research] Why more heap required to finish Dacapo iteration runs?
Junjie Qian <[email protected]>
| Newsgroups | gmane.comp.java.jikes.rvm.devel |
|---|---|
| Message-ID | <[email protected]> |
Hi Andreas, Sorry for late response! Yes, the question is why the minimum total heap size growing with more iterations. Some links online saying there are ways to dump heap in Jikes, but I do not have time to try them yet. Will let you know, if any work. For the second question, I rethink it and need some time to make it clear. Thanks! Best Junjie On Monday, August 11, 2014 1:53 PM, Andreas Sewe <[email protected]> wrote: Hi, > Thank you for the explanation first! > This becomes interesting, since I tested it with Hotspot 1.7 and Jikes > 3.1.3. Dump the heap and compare what is changing is a good way. not sure whether there is a way to dump the heap with Jikes RVM. Maybe rdb [1] can do it or others on this list know of a way how. > One question is will the compiled code after warmups be in the heap > dumps? Or we can just compare real heap usage between two GCs (like, GC > size_x -> size_y, and so size_x is the real heap usage for nursery)? HotSpot doesn't include compiled code in its heap dumps. That being said, I am not sure what you mean by the second question: Are you referring to some kind of debug output like the one obtained with -verbose:gc? And how does the nursery fit into this? So far, we have been talking about the minimum total heap size growing over time, right? Best wishes, Andreas [1] <http://sape.inf.usi.ch/rdb> ------------------------------------------------------------------------------ _______________________________________________ Jikesrvm-researchers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers ------------------------------------------------------------------------------ _______________________________________________ Jikesrvm-researchers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers