[9] RFR (S): 8177963: Parallel GC fails fast when per-thread task log overflows

Thomas Schatzl thomas.schatzl at oracle.com
Thu Apr 6 07:56:03 UTC 2017


Hi Ramki,

On Wed, 2017-04-05 at 16:38 -0700, Srinivas Ramakrishna wrote:
> Re webrev.1: Good, but I think you shouldn't print the overflow
> message each time overflow occurs while writing, but only at the end 

The message appears only once during writing (per thread).

> when the timestamp entries are printed. That should suffice to let
> user know an overflow occurred and how big a size they may need (that
> shd be part of the overflow message).

The _task_time_stamp counter always contains the total number of
entries that the GC wanted to write.

Now Kim suggested to explicitly print how many messages too much were
written at the end, but then the user would need to manually add the
two values (GCTaskTimeStampEntries plus the number of overflowed
message) together manually.

I can remove the single overflow message while writing if you want, I
see it as a courtesy to tell the user what to change.

Please suggest your preferences.

Thanks,
  Thomas

> 
> -- ramki
> 
> On Wed, Apr 5, 2017 at 11:16 AM, Thomas Schatzl <thomas.schatzl at oracl
> e.com> wrote:
> > Hi Sangheon,
> > 
> > On Wed, 2017-04-05 at 10:47 -0700, sangheon wrote:
> > > Hi Thomas,
> > >
> > [...]
> > > > New webrevs at:
> > > > http://cr.openjdk.java.net/~tschatzl/8177963/webrev.0_to_1
> > (diff)
> > > > http://cr.openjdk.java.net/~tschatzl/8177963/webrev.1 (full)
> > > webrev.1 seems good to me.
> > 
> >   thanks for your review.
> > 
> > Thomas
> > 
> > 


More information about the hotspot-gc-dev mailing list