> Yes, and with a little bit more profiling just now I brought 
> it down to 40s...

That's more like it ;-) I would've been surprised if VW's overall speed
improvement would outperform Squeak's float array support. Incidentally, how
much time is spent in allocation/GC? I wouldn't be surprised if that's a
large percentage - selectively optimizing the numerical operations to avoid
unneeded GCs (e.g., using += and friends) should make another difference.

Oh - and has someone tried this on StrongTalk? I would guess that it should
be _extremely_ fast with StrongTalk's optimized float support.

Cheers,
  - Andreas