Michael van der Gulik writes:
 > Bryce Kampjes wrote:
 > > The plans are roughly: 
 > > 1) Finish a very simple optimiser. This should provide better bytecode
 > > performance than VW.
 > > 2) Write a basic PIC implementation. This will improve send
 > > performance
 > > 3) Release 0.05
 > > 4) Convince some people to use it. I'll probably need to provide
 > > precompiled VMs. That'll involve learning how to make Debian packages.
 > 
 > Debian packages? Nah. Just tar or zip it up. I usually prefer to run all 
 > of Squeak locally anyway; it makes keeping different builds of VM's 
 > apart easier.

If you're interested in trying a tar ball. I could tar one up and
email it to you. If it works then I can place it on the web. I really
haven't looked at how the dynamic linking works inside the Squeak VM
so I don't know how easy it is to move VMs around the file system.

 > > 5) Either port to the ARM or make it fast enough to be interesting.
 > 
 > Could I perhaps encourage the latter option more?

The real question is what is fast enough to be interesting. At the
moment, I'd say faster than VisualWorks for bytecodes is fast enough
for bytecodes. How much faster sends need to be, I don't know.

0.05 should be faster than VisualWorks for bytecodes and provide a
nice speedup for sends (sorry, I'm not going to estimate). That's
probably fast enough to be worth using.

Bryce