More performance explorations

[Charles Oliver Nutter]

If you mean java.lang.invoke.GWT_FORCE_RICOCHET_FRAMES=true, then I
have more bad news...it's even slower :(

Still using 5/26 build. I will try to update later...

~/projects/jruby ➔ jruby
-J-Djava.lang.invoke.GWT_FORCE_RICOCHET_FRAMES=false --server
bench/bench_fib_recursive.rb 5 35
9227465
  1.333000   0.000000   1.333000 (  1.275000)
9227465
  1.109000   0.000000   1.109000 (  1.109000)
9227465
  1.077000   0.000000   1.077000 (  1.077000)
9227465
  1.072000   0.000000   1.072000 (  1.072000)
9227465
  1.087000   0.000000   1.087000 (  1.087000)

~/projects/jruby ➔ jruby
-J-Djava.lang.invoke.GWT_FORCE_RICOCHET_FRAMES=true --server
bench/bench_fib_recursive.rb 5 35
9227465
  3.315000   0.000000   3.315000 (  3.262000)
9227465
  3.130000   0.000000   3.130000 (  3.130000)
9227465
  3.270000   0.000000   3.270000 (  3.270000)
9227465
  3.299000   0.000000   3.299000 (  3.299000)
9227465
  3.328000   0.000000   3.328000 (  3.328000)

~/projects/jruby ➔ jruby -Xcompile.invokedynamic=false --server
bench/bench_fib_recursive.rb 5 359227465
  1.659000   0.000000   1.659000 (  1.606000)
9227465
  1.217000   0.000000   1.217000 (  1.217000)
9227465
  1.190000   0.000000   1.190000 (  1.190000)
9227465
  1.182000   0.000000   1.182000 (  1.182000)
9227465
  1.188000   0.000000   1.188000 (  1.188000)


On Sat, May 28, 2011 at 9:53 PM, John Rose <john.r.rose at oracle.com> wrote:
> It's in the guts of GWT, the old algorithm. Try disabling it with the magic property in meth-experiment.patch.
>
> -- John  (on my iPhone)
>
> On May 28, 2011, at 4:49 PM, Charles Oliver Nutter <headius at headius.com> wrote:
>
>> not sure how the logic asType logic could explain the extra stuff...
>>
> _______________________________________________
> mlvm-dev mailing list
> mlvm-dev at openjdk.java.net
> http://mail.openjdk.java.net/mailman/listinfo/mlvm-dev
>