New allocation profiler

Aleksey Shipilev aleksey.shipilev at oracle.com
Wed Apr 15 21:22:39 UTC 2015


On 04/16/2015 12:04 AM, Vladimir Sitnikov wrote:
>> (but still closest to mathematically correct!) :)
> 
> I would alter that to NaN if you don't mind.

Okay, NaN is also good as the special result in the face of division to
zero.

> And just one more thing: "gc.churn.{" + space + "}.norm" is broken for
> non 1sec iterations since it uses "churnRate/getAllOps()" while it
> should use "churn.count(space)/getAllOps()"
> 
> Should I heal that or it will come in a separate commit?

Oh wow! See, that's why I want allocation stats besides the churn
counters -- cross-checks. Please start a separate thread.

Thanks,
-Aleksey.




More information about the jmh-dev mailing list