Just to help people find info about J3 in the future, I added a "J3" page on the Swiki ( http://minnow.cc.gatech.edu/squeak/j3 ), with some links from other relevant Swiki pages.  It could be a place to add more documentation as it becomes available.

- Doug Way
  dway at riskmetrics.com

Marcus Denker wrote:
> 
> On Sat, Jul 28, 2001 at 12:55:01AM -0700, Eric Arseneau wrote:
> > Can anyone point me to information, or provide information, on the
> > performance gains achieved by doing the dynamic translation with Squeak ?  I
> > would like to know some basics on the technique and how much performance was
> > actually gained by doing it.
> >
> Sorry for not answering earlier... we really need to have some good
> documentation for the Jitter.
> 
> There is a paper about J3 by Ian:
> http://www-sor.inria.fr/~piumarta/squeak/unix/zip/j3-2.6.0/doc/j3/
> Warning: it´s somewhat out of date, but it can give some good hints
> about the overall architecture...
> 
> Performance:
> 
> interp:     '43805612 bytecodes/sec; 1325959 sends/sec'
> J3:        '135665076 bytecodes/sec; 8100691 sends/sec'
> ------------------------------------------------------
>               3.1                     6.11
> 
> Mac G4/450 Mhz.
> 
> Of course, you won´t see such a speed-up in real-world code.
> Squeak spends *much* time in primitives. Dan´s "macroBenchmarks"
> show a speedup of about a factor 2.
> 
>       Marcus