[jdk11u-dev] RFR: 8228609: G1 copy cost prediction uses used vs. actual copied bytes [v7]

[Yude Lin]

On Wed, 1 Jun 2022 23:13:21 GMT, Roman Kennke <rkennke at openjdk.org> wrote:

>> Yude Lin has updated the pull request incrementally with one additional commit since the last revision:
>> 
>>   remove unused variables and their dependency
>
> src/hotspot/share/gc/g1/g1ParScanThreadState.cpp line 94:
> 
>> 92:   uint length = _g1h->collection_set()->young_region_length();
>> 93:   for (uint region_index = 0; region_index < length; region_index++) {
>> 94:     surviving_young_words[region_index] += G1ParScanThreadState::surviving_young_words()[region_index];
> 
> Was that code using the wrong index before? IOW, it used the surviving_young_words of the wrong regions?

Yes, that's what I think.

So this pr is achieving two things (sorry if it were not clear):

1. fix the index issue: this causes calculation of the total surviving young words to be incorrect. The correct stats are stored in _surviving_young_words[1, .., n] instead of [0, .., n-1]
3. use a precise "copied bytes" instead of inferring from (used() - freed())

-------------

PR: https://git.openjdk.org/jdk11u-dev/pull/927