RFR: 8015107: NPG: Use consistent naming for metaspace concepts

harold seigel harold.seigel at oracle.com
Wed Sep 4 11:13:59 PDT 2013

Hi Erik,

These changes look good.  Just a small nit.  In arguments.cpp, I think 
the new warning should say "has no affect" instead of "has no effect".

Thanks, Harold

On 9/4/2013 1:36 PM, Erik Helin wrote:
> Hi all,
> this patch changes the name of two flags:
> - -XX:+UseCompressedKlassPointers becomes
>    -XX:+UseCompressedClassPointers (note the 'K' instead of the 'C')
> - -XX:ClassMetaspaceSize becomes -XX:CompressedClassSpaceSize
> Background:
> These two flags were introduced when metaspace was merged into hs25.
> To use a smaller size for the Java object headers, the pointer in the
> object header that points to the Klass instance for the object can be 32
> bits on a 64-bit system. Such pointers has been called compressed klass
> pointers. This is the same concept as compressed oops.
> For the current implementation of compressed class pointers to work, the
> Klass instances need to be placed in a continuous virtual memory space.
> This virtual memory space was called ClassMetaspace but this patch
> introduces the name CompressedClassSpace to emphasize that the concept
> is closely related to compressed klass pointers.
> The 'K' in the CompressedKlassPointers and CompressedClassSpace was
> changed to a 'C' since the C++ class Klass is an implementation detail
> that we do not want to expose to the users.
> I've also added a check in Arguments::parse that checks if the user has
> set CompressedClassSpaceSize when compressed class pointers
> are _not_ used. If this is the case, then the VM will print a warning. I've
> also added a test that verifies this actually happens.
> The change touches a lot of files, but the change in each file is very
> small.
> Webrev:
> http://cr.openjdk.java.net/~ehelin/8015107/webrev.00/
> Bug:
> http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=8015107
> Testing:
> - JPRT
> - The jtreg tests in hotspot/test
> - Newly added jtreg test
> Thanks,
> Erik

More information about the hotspot-dev mailing list