Re: Making Squeak Fly ( Was: Re: Where Squeak is Headed [was: Module discussion] )

On Friday, November 8, 2002, at 08:00  AM, Dan Ingalls wrote:

> > Um, I think I'm inventing my own notation.  By "so that A + B ~ B" ,   
> >  I meant "would the speed improvement from adding J5 and  
> > Strongtalk-types to Squeak, be approximately equal to the speed  
> > improvement from just adding J5 to Squeak".
>
> I know.  Really, Ian would be the best person to answer this, and the  
> StrongTalk folks would have a good sense too.
>
>         - D

I had the pleasure of attending the strongtalk talk at oopsla and heard:

David Griswold said

a) The vm is not released.
     (perhaps people could whimper to Sun about that)

b) The entire fully typed, refactored (and he really means refactored)  
smalltalk class library is released under a bsd like license. He begged  
the community to grab it. The people refactoring collections should  
look at it etc.

b) Someone asked: "since you can define methods/vars typed or untyped  
how does this affect optimization?"

David said optimization is performed without using ANY of the typing  
information supplied. The optimizing logic doesn't care and the typing  
doesn't need to exist and the resulting code will run the same!

On reflecting on that and I'm sure David will correct me if needbe I  
think it was a poke at the folks that say a language needs to be typed  
for optimization. But Strongtalk is NOT typed for optimization, it is  
typed for other reasons, mm maybe to make those need strong typed  
language folks happy I'd guess.

Thus to peruse typing in the language solely for performance reasons is  
false.


--
======================================================================== 
===
John M. McIntosh <johnmci@xxxxxxxxxxxxxxxxxxxxxxx> 1-800-477-2659
Corporate Smalltalk Consulting Ltd.  http://www.smalltalkconsulting.com
======================================================================== 
===