RFC (M) 8058968: Compiler time traces should be improved

[John Rose]

On Sep 24, 2014, at 7:32 AM, Aleksey Shipilev <aleksey.shipilev at oracle.com> wrote:

> Thanks, Vladimir!
> 
> More reviews, please :)

Reviewed.  I like infrastructure that gives us insight into JVM behavior.

— John

P.S.  Random idea:  There is an ad-hoc "max" operation in one place.

It would be a mostly localized change to collect some statistical moments at this line of code:
  _accum->add(_t);

The accumulator (or all elapsedTimers) could hold sample count, sample sum, and sum of squared samples.  Also max sample, to cover the little bit of computation above.  Verbose mode could use these moments to provide dispersion information.  The collected dispersion numbers would be useless, except for top-level categories, so this idea is only half-baked.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20140924/7b532076/attachment-0001.html>