Jitter seems to work just fine on my G4 with Mac OS 9.0.4, at least. 
(Although I did have a tough time trying to download it with Netscape...
I ended up downloading with Internet Explorer and then I had to use
ResEdit to change the ID's of the image before it would work.)

Anyway, it's pretty cool!  I mostly just played around with longish UI
operations, which were anywhere from 1.5 to 2.5 as fast when using
Jitter comparied with the regular 2.8 VM/image.  (For example expanding
a 5000-item collection in an object explorer.)  The 1.5 to 2.5 mix
probably depended on how much of the task was
message-send/bytecode-bound versus primitive-bound...

The only behavior difference I've noticed so far is that sometimes it
doesn't redraw its display when switching from a different Mac
application to Squeak.  It's worth an inconvenience or two to see the
speed of Squeak instantly double. :)

- Doug Way
  dway at riskmetrics.com


Karl Ramberg wrote:
> 
> Well... I get a stdout :
> 
> out of memory
> 
> classes:     266 bkts,    5320 byts,   272 ents,  1.0 ents/bkt,   6.5% coverage
> selectors:   759 bkts,   15180 byts,   836 ents,  1.1 ents/bkt,  18.5% coverage
> methods:     957 bkts,   19140 byts,  1103 ents,  1.2 ents/bkt,  23.4% coverage
> mapCache:      1 blks,   77884 byts,     0 lost,   61 byts/ent,   0.0% fragment
> codeCache:     7 blks,  798248 byts, 36184 lost,  676 byts/ent,   4.6% fragment
> iclCache:      1 blks,    2232 byts,     0 lost,    8 byts/ent,   0.0% fragment
> memory: 88816 heap, 878364 cached (227088+461208 = 22.8 bytes/bytecode)
> GC: 0 major, 0 minor,   0.00%   0.00%   0.00% CPU;  i-cache:   0.00%
> 0.00 NAN(000)
> traps: 0 SI/sec (avg 0), 0 TF (avg 0); trans: 0 (0/sec)
> 
> morituri te salutant
> 
> And thats it. (Trying to change the memory allocation given j3 gives a
> strange message about not being able to change the min value to higher
> than max etc.)
> This is on mac os 8.6 swedish on a 8500 /150, 604 ppc
> 
> Karl