RFR: 8171238: Unify cleanup code used in G1 Remark and Full GC marking
thomas.schatzl at oracle.com
Tue Feb 21 11:11:01 UTC 2017
On Tue, 2017-02-21 at 11:17 +0100, Stefan Johansson wrote:
> Please review this re-factoring for:
> Some small parts of the code in the G1 Mark Sweep Full GC is
> one of those is the string and symbol table cleaning as well as the
> string deduplication cleanup. They are currently run as two separate
> tasks which means the worker threads have to be started and stopped
> times. The two tasks can be joined together and with the cleanup in
> place the Remark phase of the concurrent cycle can also be unified
> use the same code paths.
The following comments are mostly about naming, since these are very
much opinions, feel free to ignore what you do not like. The change
itself looks good.
- not really happy with "G1StringSymbolTableUnlinkTask" not showing
that it also does dedup now.
- I would kind of think "complete" as the matching word for "partial",
- I would prefer if the comments for full/partial cleaning would
explain what the method does, not (only) when/what it is used for?
- maybe improve the "cleaning" to indicate what is cleaned (or what
cleaning means), i.e. something like
"full/partial_clean_dead_weak_references" or so - that is however a bit
long for me too. Probably you can find a better name.
Some better comments might remove the need for this.
More information about the hotspot-gc-dev