More performance explorations
Charles Oliver Nutter
headius at headius.com
Sun Jun 5 01:45:41 PDT 2011
On Fri, Jun 3, 2011 at 6:15 PM, Tom Rodriguez <tom.rodriguez at oracle.com> wrote:
> So I put in a little hack to fold repeated interface checkcasts and that gets back a lot of the performance. With fib on my machine dynopt=true reports 1.005000, invokedynamic=true reports 1.293000 and turning on my checkcast hack gets it down to 1.112000. Unfortunately what I've got right now isn't really suitable for inclusion in the JDK7.
I assume by "not suitable" you mean "too invasive". I'm quite alright
with that; I think the next step for those of us interested in pushing
performance forward is to find a fast-track way to get such
experiments into buildable codebases on our target platforms (e.g.
maybe we need a patch repo against bsd-port or macosx?) I'm keen to
try out any patches you want to throw at these issues, and maybe it
will help me start understanding Hotspot internals better myself.
> Just blindly skipping checkcast method handles for interface types bring the time on fib down to 1.071000.
Definitely within shooting distance of dynopt :)
More information about the mlvm-dev