On 01/05/2013 08:08 PM, Charles Oliver Nutter wrote: > On Sat, Jan 5, 2013 at 12:37 PM, Remi Forax <fo...@univ-mlv.fr> wrote: >> No. >> invalidating a SwitchPoint that is JITed cost your an harm, >> invalidating a SwitchPoint which is not JITed is just a volatile write >> but you still have to update the callsite to use a new SwitchPoint. > I think I missed some details. In the case where I've "failed" a call > site, I no longer bind via SwitchPoint *at all*. The site is bound to > a separate mechanism. But on the invalidation side, I continue to > construct and invalidate new SwitchPoints due to the way the logic is > set up.
Ok, get it. I've dug a little bit in the code of jdk8, when you invalid a SwitchPoint several times, there is no short circuit that say, this switch point is already invalidated, so you have a call from Java to C to set the underlying target each time :( so don't do that :) > I'll try to come up with a test case. For now I have also set up the > invalidation side to eventually fail permanently and not create or > invalidate any new SwitchPoints. > > - Charlie Rémi _______________________________________________ mlvm-dev mailing list mlvm-dev@openjdk.java.net http://mail.openjdk.java.net/mailman/listinfo/mlvm-dev
