Life without C [was: VM improvement: speeding up failingcalls...]

Marcel Weiher marcel at metaobject.com
Mon Feb 14 21:07:02 UTC 2000


> From: Dan Ingalls <Dan.Ingalls at disney.com>
>
> >Hmm, I just don't see how Andrew's concern about code-generation   
> >quality is addressed.
> >
> >Maybe I am too deep inside the box...
>
> You and me both ;-).
>
> I was just reflecting at a higher level about our current intermediate 
> generation of C code, which Andrew argues, and I agree, is a Good  
Thing.  And my
> reflection was simply that there might be other Good Things in our  
future.

Definitely.  Lots!  Anyway, my view on C is really pretty much that  
of a portable assembly language, and lots of a CPU manufacturer's  
know-how today goes into making the compilers move.  Now these people  
built the CPUs, and it takes them quite a while...

> Regarding code generation, I think that any decent code generator  
for a decent
> native code translator will beat any C compiler's translation of  
any bytecode
> interpreter.

Ahh, yes, certainly.  I was hoping to translate some of those  
methods to native code via C, instead of interpreting them (similar  
to ST/X, Objective-C, ... )

That being said, there aren't that many different things that  
Smalltalk code does (mostly just send messages), and it *might* even  
be true that a native code generator tuned to that might beat ST -> C  
-> native code.

Marcel





More information about the Squeak-dev mailing list