Request for review: 8000968: NPG UseCompressedKlassPointers asserts withObjectAlignmentInBytes for > 32G Compressed Oops

harold seigel harold.seigel at oracle.com
Tue Jan 22 10:22:15 PST 2013


HI Roland,

Thanks again for your comments.  I incorporated them in this new webrev: 
http://cr.openjdk.java.net/~hseigel/bug_8000968_4/ 
<http://cr.openjdk.java.net/%7Ehseigel/bug_8000968_4/>

The only changes from the previous webrev are to module universe.cpp.

Could you take another look when you have a chance?

Thanks, Harold

On 1/18/2013 6:44 AM, Roland Westrelin wrote:
> Hi Harold,
>
>> I updated the webrev at http://cr.openjdk.java.net/~hseigel/bug_8000968_3/<http://cr.openjdk.java.net/%7Ehseigel/bug_8000968_3/>  with this change.
>   696     } else if ((total_size<= OopEncodingHeapMax)&&  (mode != HeapBasedNarrowOop)&&
>   697         (!UseCompressedKlassPointers ||
>   698           (((OopEncodingHeapMax - heap_size) + Universe::class_metaspace_size())<= KlassEncodingMetaspaceMax))) {
>
> heap_size<  OopEncodingHeapMax - KlassEncodingMetaspaceMax is possible, right? Then compressed klass pointers are off with this code. So wouldn't you also want to check for:
>
> KlassEncodingMetaspaceMax + heap_size - Universe::class_metaspace_size()<= OopEncodingHeapMax
>
> ?
>
> and then use KlassEncodingMetaspaceMax - Universe::class_metaspace_size() as base.
>
> Roland.


More information about the hotspot-runtime-dev mailing list