This is the mail archive of the
mailing list for the Java project.
Re: Speed Impact experiment on GCJ
Holger Hoffstaette writes:
> On Wed, 15 Feb 2006 18:22:02 +0000, Andrew Haley wrote:
> > Holger Hoffstaette writes:
> > > 1.6 build, because that will make the 1.5 test look slow. :-) The one
> > > thing where gcj used to have a very clear advantage was
> > > interface-method invocation since before 1.5 that used to be up to 10
> > > (!) times slower than direct method invocations; this was reduced to a
> > > factor of ~1.8 in 1.5 and completely optimized away in Mustang.
> > Err, how do they do that, then? Is there a paper somewhere?
> Sorry, not that I know. I more or less discovered this by accident too..
Hmm. I wonder if they're devirtualizing interface calls. Can you let
me see the benchmark?