RFR: JDK-8086056: ParNew: auto-tune ParGCCardsPerStrideChunk

Bengt Rutisson bengt.rutisson at oracle.com
Thu Jun 18 07:02:43 UTC 2015


Hi Tony,

NIce to hear from you!


On 18/06/15 00:30, Tony Printezis wrote:
> Hi all,
>
> A small patch for your consideration:
>
> http://cr.openjdk.java.net/~tonyp/8086056/webrev.0/ 
> <http://cr.openjdk.java.net/%7Etonyp/8086056/webrev.0/>

A couple of style comments:

  891   const size_t MinOldGenCapacity =      G;
  892   const size_t MaxOldGenCapacity = 16 * G;
  893   assert(MinOldGenCapacity <= MaxOldGenCapacity, "sanity");

Local variables are normmaly named using lower case and underscore. So, 
min_old_gen_capacity and max_old_gen_capacity. Same for the rest of the 
local variables in ParNewGeneration::adjust_cards_per_stride().

Having said that, I think it would be good to turn some of these 
constants into flags (which means that the naming should be as it is now 
;) ). Especially since you say that you haven't done much performance 
testing of these values. I would prefer that we could adjust them at 
runtime. Best would of course be if you could take the time to do the 
performance testing.

I realize that the assert on line 893 might have been useful if you were 
playing around with the numbers when you were experimenting with the 
patch. But now it looks mostly like line noise to me. Same with the 
assert on line 898.

I would also consider factoring the actual computation out into a 
separate helper method.

I also second Jon's comment on that it would be good with a more 
explicit way of turning this feature off.

Thanks,
Bengt

>
> (BTW, for some reason some of the webrev output is a bit messed up. 
> Not sure why, maybe some hg incompatibility I guess. The diffs look OK 
> though. I also attached the patch to this e-mail.)
>
> There’s a bit of info on the JIRA on the rationale for the patch:
>
> https://bugs.openjdk.java.net/browse/JDK-8086056
>
> The min / max values for the old gen capacity 
> and ParGCCardsPerStrideChunk were chosen empircally after running a 
> few (mostly synthetic tests) on Linux x64. If someone has the cycles 
> to do a more extensive performance study, I’d be happy to revise them 
> accordingly.
>
> Regards,
>
> Tony
>
> -----
>
> Tony Printezis | JVM/GC Engineer / VM Team | Twitter
>
> @TonyPrintezis
> tprintezis at twitter.com <mailto:tprintezis at twitter.com>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/hotspot-gc-dev/attachments/20150618/84e8fb87/attachment-0001.html>


More information about the hotspot-gc-dev mailing list