More performance explorations

[John Rose]

On Jun 10, 2011, at 3:37 PM, Charles Oliver Nutter wrote:

> Sorry, the logic for this never actually landed on JRuby master, and I
> lost the commit somehow...so the inlining you're seeing is the old
> logic with a custom JRuby CallSite subclass.
> 
> I'll get the new logic in place this weekend. Been kicking myself for
> losing it :(

I just rebased to b145.

Here is a meth.jar which includes b145:
  http://cr.openjdk.java.net/~jrose/pres/indy-javadoc-mlvm/meth-0612.jar

It also includes these changes beyond b145:

meth-rot-7052202.patch  (various point fixes)
meth-unittests.patch  (upgraded unit tests)
meth-deprecate.patch  (deprecated methods; this is going away soon)
meth-experiment.patch  (a hack to disable RFs for some GWTs; going away)

As of now, the default behavior for GWT is always to use ricochet frames.  Let me know if this causes regressions.  To turn back the clock temporarily, try this:
  java ... -Djava.lang.invoke.GWT_FORCE_RICOCHET_FRAMES=false

-- John