RFR: 8076265: Simplify deal_with_reference

Kim Barrett kim.barrett at oracle.com
Thu Apr 2 23:17:16 UTC 2015


On Apr 1, 2015, at 7:01 AM, Thomas Schatzl <thomas.schatzl at oracle.com> wrote:
> Some comments:

Thanks for making me look at this a little bit harder; I found some
further simplifications and improvements.

Note that in the sequence of tests to decide whether to push the
object onto the mark queue or not, I'm preserving the existing order
of tests.  I didn't want to make a possibly performance impactful
change to that order without data to motivate such a change.

> - what is the "mark bit processor/processing" mentioned in the
> comment?

I changed to consistently use "mark bitmap scan", but the
surrounding context in the comments has been substantially rewritten.
Hopefully it is more clear now.

> - would it be possible to make it explicit that we check whether we
> have a current region first?

I don't really want to introduce a new trivial helper function for
this purpose, but I've left a comment indicating why _finger is being
used for the check.

> passed -> past (I think).

Fixed.

> - I do not really like an extra parameter ... just for logging ...

Restructuring has eliminated the need for that.

CR:
https://bugs.openjdk.java.net/browse/JDK-8076265

New full webrev:
http://cr.openjdk.java.net/~kbarrett/8076265/webrev.01/

New incremental webrev:
http://cr.openjdk.java.net/~kbarrett/8076265/webrev.01.incr/

Testing:
JPRT
local RefWorkload with -XX:+UseG1GC
local JTREG with -XX:+UseG1GC



More information about the hotspot-gc-dev mailing list