Other |
---|
tbd_majorResolved |
Duplicate :
|
G1 full gc visits free regions during the various phases of the MSC algorithm. This is completely unnecessary, and just slows down performance. This mostly affects system.gc calls by the user called on almost empty heaps for whatever reason. Anecdotally, system.gc duration at startup for specjvm benchmarks can be decreased by ~15% already by only applying this optimization to the compaction phase (this is a very special case).