Idea for a possibly better Collection occurrencesOf method.

Klaus D. Witzel klaus.witzel at cobss.com
Tue Sep 12 16:33:27 UTC 2006


Hi J J,

on Tue, 12 Sep 2006 16:44:24 +0200, you wrote:

>> From: "Klaus D. Witzel" <klaus.witzel at cobss.com>
...
>> But I think this is something for a 1.0 Exupery release; if I  
>> understand  Bryce correctly he aims to outperform a certain commercial  
>> VM and  MethodContext>>#apply:from:to: looks like it's worth the effort  
>> (of  course, with fall back to good old #do: code-in case the new  
>> primitive is  missing). My CHF 0.05.
>>
>
> Exupery is a JIT, no?

Yup, a JIT.

> JIT is certainly important and very needed, but will
> future squeaks use it by default?

I expect cross-fertilization (yes, I'm here in squeak-dev and I use the  
word expect ;-). Imagine that #apply:from:to: runs in Exupery, is tested  
and does what is expected. Then a handful of copy&paste's (or  
fileOut/fileIn)s later the code can be in a non-JIT VM...

> If we put in the #apply:from:to: in 3.9 or 3.10 or whatever, then we  
> could
> go and clean up all the "rolll your own" count:, etc.'s out there.

Smalltalkers usually don't wait for x.y or whatever: the primitive  
associated with #apply:from:to: must be backed by Smalltalk code because  
it (the primitive) may fail or may only be available together with a  
JITter. In both cases the Smalltalk code must already exist and run  
(hopefully) bug free.

Of course, what'd also be needed is a commitment from the VM folks to  
include the new primitive.

/Klaus




More information about the Squeak-dev mailing list