RFR(S) 8206115 - Use shared macros for JavaClasses::compute_offsets and MetaspaceShared::serialize_well_known_classes
jiangli.zhou at oracle.com
Wed Aug 15 17:54:10 UTC 2018
On 8/15/18 10:46 AM, Ioi Lam wrote:
> Hi Jiangli,
> Thanks for the review. How about changing the comment to
> // We have already called the compute_offsets() of the
> // BASIC_JAVA_CLASSES_DO_PART1 classes (java_lang_String and
> // earlier inside SystemDictionary::resolve_preloaded_classes()
> - Ioi
> On 8/15/18 9:10 AM, Jiangli Zhou wrote:
>> Hi Ioi,
>> Looks good!
>> I have a very minor suggestion. Could you please add some details in
>> the following comment and describe where/when the compute_offsets()
>> calls for java_lang_String and java_lang_Class are done.
>> 4538 // We have already called the compute_offsets() of the
>> 4539 // BASIC_JAVA_CLASSES_DO_PART1 classes (java_lang_String and
>> On 8/14/18 9:37 PM, Ioi Lam wrote:
>>> As I noted in the comments on this RFE, we already have a bug where
>>> java_lang_Integer_IntegerCache was missing from
>>> To avoid such problems in the future, I added a new macro in
>>> BASIC_JAVA_CLASSES_DO, which is used by JavaClasses::compute_offsets
>>> For clean up, I also renamed all the <Class>::serialize() functions to
>>> <Class>::serialize_offsets() to reflect what these functions really do.
>>> The old MetaspaceShared::serialize_well_known_classes function is now
>>> Testing with hs tiers 1/2.
>>> - Ioi
More information about the hotspot-runtime-dev