On 05/18/2011 09:03 PM, John Rose wrote: > On May 18, 2011, at 11:59 AM, Rémi Forax wrote: > >> All others are hard because the value of the method handle is not >> statically known >> i.e the method handle cannot be considered as constant. > It can be optimistically inlined, which is what we do for invokedynamic > instructions.
This is exactly what the backport does now, it inlines a method handle if it's the same method handle during the profiling phase but thei strategy doesn't work well. I've taken some benchmarks that use lambdas (some from JRuby, some from a proprietary language) and translate them in PHP.reboot. In only one bench, the method handle was stable. The solution proposed by Cliff Click, a dispatch table at the start of the method/loop, seems better > -- John Rémi _______________________________________________ mlvm-dev mailing list mlvm-dev@openjdk.java.net http://mail.openjdk.java.net/mailman/listinfo/mlvm-dev
