RFR: 8257145: Performance regression with -XX:-ResizePLAB after JDK-8079555 [v2]
sjohanss at openjdk.java.net
Mon Nov 30 10:02:00 UTC 2020
On Sun, 29 Nov 2020 15:21:09 GMT, Dongbo He <dongbohe at openjdk.org> wrote:
>> this is the continuation of the review of the implementation for:
> Dongbo He has updated the pull request incrementally with one additional commit since the last revision:
> store the "default size" for the PLAB in the PLABStats
Looks good, just a few small things.
src/hotspot/share/gc/g1/g1CollectedHeap.cpp line 1410:
> 1408: _archive_allocator(NULL),
> 1409: _survivor_evac_stats("Young", YoungPLABSize, YoungPLABSize * ParallelGCThreads, PLABWeight),
> 1410: _old_evac_stats("Old", OldPLABSize, OldPLABSize * ParallelGCThreads, PLABWeight),
I think you could leave those two unchanged and in the constructor for `G1EvacStats` do the multiplication with `ParallelGCThreads` for the desired value.
src/hotspot/share/gc/shared/plab.hpp line 173:
> 172: public:
> 173: PLABStats(const char* description, size_t default_plab_sz, size_t desired_net_plab_sz_, unsigned wt) :
Since you are touching this line it would be nice to remove the trailing `_` in `desired_net_plab_sz_`.
Changes requested by sjohanss (Reviewer).
More information about the hotspot-gc-dev