RFR (XXS): 8143220: Fix documentation of InitiatingHeapOccupancyPercent

Jon Masamitsu jon.masamitsu at oracle.com
Fri Feb 5 21:26:37 UTC 2016


Thomas,

I would leave out the part about InitiatingHeapOccupancyPercent being 
used as
the initial value for IHOP because not enough data has been gathered to 
make a
better estimate.  While true today it may change and I don't think it is 
needed as
part of the description.  Explaining what happens when it is 100 would 
be good.
I added the explanation about fragmentation (is it correct) because I 
thought
that it would come up sooner or later but you can drop it, if you think 
it is
too much.  So I would suggest.

Percentage of the old generation heap occupancy (IHOP) after which a 
concurrent
GC cycle is started at the end of the next garbage collection pause.
If adaptive IHOP is not being used (-XX:-G1UseAdaptiveIHOP), the IHOP is 
fixed at
this value of InitiatingHeapOccupancyPercent.  If adaptive IHOP is being 
used,
InitiatingHeapOccupancyPercent is the starting value for IHOP.  A value 
of 0 causes
the concurrent GC cycle to run as much as possible.  A value of 100 
causes the
concurrent GC cycle to be delayed until the the old generation is 
completely full.
Fragmentation waste in the old generation is not considered to be free 
space.
InitiatingHeapOccupancyPercent is being used by G1.

Jon

On 2/5/2016 2:52 AM, Thomas Schatzl wrote:
> Hi all,
>
>    can I have reviews for this change to the documentation of the
>   InitiatingHeapOccupancyPercent global? I think due to recent changes
> to the behavior it is warranted that the (internal) documentation
> reflects it.
>
> Other suggestions for the text welcome.
>
> CR:
> https://bugs.openjdk.java.net/browse/JDK-8143220
> Webrev:
> http://cr.openjdk.java.net/~tschatzl/8143220/webrev/
> Testing:
> local compilation
>
> Thanks,
>    Thomas
>



More information about the hotspot-gc-dev mailing list