RFR(S) 8193184: NMT: Report array class count in NMT summary

[Zhengyu Gu]

Hi Coleen,

On 02/05/2018 01:15 PM, coleen.phillimore at oracle.com wrote:
> 
> Hi Zhengyu,
> 
> This looks really good.  The instanceKlass count does seem to belong to 
> ClassLoaderData.

You want to move both counters to ClassLoaderData, or just instanceKlass?

To me, CLD is per-classloader metadata, CLDG is where aggregates all 
class loader metadata, and the counters aggregate all class loaders, 
that why I put them there.

I am okay to move to CLD if you insist.

> 
> http://cr.openjdk.java.net/~zgu/8193184/webrev.00/src/hotspot/share/classfile/classLoaderData.hpp.udiff.html 
> 
> 
> + static void inc_instance_classes(size_t count) {
> + Atomic::add(count, &_num_instance_classes);
> + }
> +
> + static void dec_instance_classes(size_t count) {
> + assert(count <= _num_instance_classes, "Sanity");
> + Atomic::sub(count, &_num_instance_classes);
> + }
> +
> + static void inc_array_classes(size_t count) {
> + Atomic::add(count, &_num_array_classes);
> + }
> +
> + static void dec_array_classes(size_t count) {
> + assert(count <= _num_array_classes, "Sanity");
> + Atomic::sub(count, &_num_array_classes);
> + }
> 
> 
> Can you move these functions to classLoaderData.inline.hpp?
> 
Will do

> Maybe you should add an NMT test for this to verify the output?
Will do.

Thanks,

-Zhengyu

> 
> Thanks,
> Coleen
> 
> 
> On 2/2/18 3:28 PM, Zhengyu Gu wrote:
>> Please review this patch that adds array class count to NMT report, so 
>> the class count can closely match the output from GC.class_stats.
>>
>> For example:
>> Summary:
>> -                     Class (reserved=1069781KB, committed=1069269KB)
>>                             (classes #3362)
>>                             (  instance classes# 3105, array classes#257)
>>
>> Summary diff:
>> -                     Class (reserved=1071843KB +2061KB, 
>> committed=1070051KB +781KB)
>>                             (classes #3405 +43)
>>                             (  instance classes# 3143 +38, array 
>> classes#262 +5)
>>
>> Bug: https://bugs.openjdk.java.net/browse/JDK-8193184
>> Webrev: http://cr.openjdk.java.net/~zgu/8193184/webrev.00/
>>
>> Test:
>>   hotspot_tier1_runtime on Linux 64 (fastdebug and release)
>>
>> Thanks,
>>
>> -Zhengyu
>>
>