RFR: 8156548: gc/gctests/StringInternSyncWithGC2 fails with Test level exit status: 151

[Kim Barrett]

> On May 22, 2016, at 10:01 PM, Jon Masamitsu <jon.masamitsu at oracle.com> wrote:
> On 5/20/2016 2:24 PM, Kim Barrett wrote:
>> Please review this fix for a serious performance bug in non-product
>> verification code in CMS precleaning.
>> 
> 
> Instead of eliminate the verification can you leave it in for non-object arrays?
> You might be able to do the verification for the object array once if you did it
> only when an object header was in the dirty range but do it for the length of
> the entire object array.

It is possible, and even fairly likely, that the objArray's header is
not in any relevant dirty range.

> If the verification were to consistently fail here and never fail when done during
> the sweep, I think that bounds where in the code we need to look for a problem.

That doesn't seem like a very likely scenario to me; timing
differences between runs could easily result in problems being missed
until the sweep phase.

OTOH, having spent some time studying old bug reports, I now agree the
Printezis-bit consistency checks there have real value, and will
reinstate them.

I'm less convinced about the utility of the interior mark bit
verification there.  However, there's a way to mostly keep it, while
strongly reducing the performance impact.  The idea is to record the
last verified object and only perform the verification if the new
object differs from that last verified.  So I'll do that too.

New webrev:
http://cr.openjdk.java.net/~kbarrett/8156548/webrev.01/

(No incremental webrev, since this version is pretty much completely
different.)

Testing:
Redid local testing to verify performance benefit.
RBT GC testing with CMS as the collector.