RFR: 8155678: ClassLoader::initialize_module_loader_map should only be called when dumping CDS archive
Jiangli Zhou
jiangli.zhou at Oracle.COM
Wed May 4 16:45:56 UTC 2016
Hi Lois,
Thanks for the review!
Thanks,
Jiangli
> On May 4, 2016, at 5:53 AM, Lois Foltan <lois.foltan at oracle.com> wrote:
>
> Looks good Jiangli.
> Lois
>
> On 5/3/2016 9:16 PM, Jiangli Zhou wrote:
>> Hi Calvin,
>>
>>> On May 3, 2016, at 4:42 PM, Jiangli Zhou <jiangli.zhou at Oracle.COM> wrote:
>>>
>>> Hi Calvin,
>>>
>>>> On May 3, 2016, at 4:32 PM, Calvin Cheung <calvin.cheung at oracle.com> wrote:
>>>>
>>>> Hi Jiangli,
>>>>
>>>> Just a minor comment in classLoaderExt.hpp:
>>>>
>>>> 58 jshort classloader_type = ClassLoader::classloader_type(
>>>>
>>>> How about u2 instead of jshort?
>>> I agree. u2 is a better choice here. Will change it.
>> I spoke too soon. InstanceKlass::set_class_loader_type() sets the loader type bits in '_misc_flags’ based on the ‘loader_type’ argument value. InstanceKlass has a method named loader_type(), which returns a u2 value. Note the u2 value returned by InstanceKlass::loader_type() is different from the ClassLoader::ClassLoaderType. The name of the InstanceKlass::loader_type() is missing leading, it’s better to be renamed. I’ll handle that separately.
>>
>> ‘jshort’ is the same a ‘s2’’. To be consistent, I changed above ‘jshort’ to be ‘s2’, since ClassLoader.classloader_type() returns s2.
>>
>> Here is the updated webrev: http://cr.openjdk.java.net/~jiangli/8155678/webrev.01/
>>
>> Thanks,
>> Jiangli
>>
>>> Thanks,
>>> Jiangli
>>>
>>>> It is because in set_class_loader_type(), we are setting the _misc_flags which is of type u2.
>>>>
>>>> If you make the above change, you'll need to change set_class_loader_type() as well.
>>>>
>>>> thanks,
>>>> Calvin
>>>>
>>>> On 5/3/16, 3:59 PM, Jiangli Zhou wrote:
>>>>> Please review the following changes that make ClassLoader::initialize_module_loader_map() CDS dump time only. The module to loader map is not used by CDS runtime, so this avoids the unnecessary overhead for runtime. As part of the changes, I moved the ClassLoader::class loader_type() call from ClassLoader::load_class() into ClassLoaderExt::Context::record_result(), since ‘class loader_type’ is only needed by record_result(). That allows ClassLoader::class loader_type() become a CDS only API.
>>>>>
>>>>> webrev: http://cr.openjdk.java.net/~jiangli/8155678/webrev.00/
>>>>> bug: JDK-8155678<https://bugs.openjdk.java.net/browse/JDK-8155678>
>>>>>
>>>>> Thanks,
>>>>> Jiangli
>
More information about the hotspot-runtime-dev
mailing list