We did have optional instrumentation to maintain the PIC counts, and used that to guide our choice of ordering, but we didn¹t use it on a per PIC level to do anything at run time, it was just a case of gathering a lot of data and printing out the stats. It did add some overhead, but I think Vlad¹s work for profiling of GWTs show that it could be done with an efficient API. How much real benefit it would offer though, I don¹t know.
Duncan. On 11/03/2015 17:48, "Marrows, George A (GE Energy Management)" <george.marr...@ge.com> wrote: >> so the order of the handles is never changed to for example trying the >>last one first or the one with the most hits recently. Is it not worth >>the trouble? > >We found a useful gain at one point from reordering GWT chains so that >the new MH was put on the end of the chain and the chain as a whole was >therefore sorted chronologically, with first seen at the top. That >obviously requires a bit of book-keeping to arrange. We haven't >experimented with maintaining our own hit counts, and in fact I think we >last checked the perf gain from reversing the chain on Java 7, so results >may no longer be relevant. _______________________________________________ mlvm-dev mailing list mlvm-dev@openjdk.java.net http://mail.openjdk.java.net/mailman/listinfo/mlvm-dev
