RFR 8005698 : Handle Frequent HashMap Collisions with Balanced Trees

Alan Bateman Alan.Bateman at oracle.com
Mon May 27 08:57:21 UTC 2013

On 26/05/2013 13:46, Doug Lea wrote:
> On 05/26/13 08:08, Peter Levart wrote:
>> Clever idea. So your common TreeNode would extend LinkedHashMap.Entry.
> Right. (Or renamings/refactorings of these).
> And that begets other improvements in part by guaranteeing root
> is first node of bin, so doesn't need TreeBins (which I cannot do in
> CHM because links are pinned.) So far, I haven't gotten
> this to the point of realistically comparing performance etc,
> and wasn't going to post until I did, but then figured I'd
> save you some effort.
Any objection if we move ahead with what Brent has now, at least as 
first step? There are are other follow-on changes, the removal of String 
hash32 field/method particular, that we are anxious to get in.


More information about the core-libs-dev mailing list