Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!rutgers!caip.rutgers.edu!peskin
From: peskin@caip.rutgers.edu (R. L. Peskin)
Newsgroups: comp.lang.smalltalk
Subject: Re: NOOL's (not OOL)
Message-ID: 
Date: 14 Aug 89 15:42:48 GMT
Organization: Rutgers Univ., New Brunswick, N.J.
Lines: 37


>I, for one, would greatly miss three major features in Smalltalk if I went 
back
>to straight C.
>
>(1) Polymorphism. It really slows one down to have explicit routines that do
>the same thing for different "objects".
>
>(2) The shortness of the Compile/Link/Execute cycle in Smaltalk.
>
>(3) All the great stuff already there in the Collection Classses.
>
>However, I am currently struggling with a project that requires more speeed
>and a different graphic imaging environment (3D color). I am debating the
>pros and cons involved in doing it in Straight C or linking in C segments
>into the Virtual Machine. If anyone else has any additional feelings on this
>subject, I could use your input.
>
Our route has been to link C code graphics segments into the VM. We've done 
this using PPS 2.4 on the Ardent Titan, and with Tek's Color Smalltalk. There 
are some interesting questions to deal with, i.e. how to best handle Form 
caches, etc. but, all in all., we think that in htose cases where the "control 
and access" aspects of Smalltalk are important, enhance the VM with primitives; 
don't throw the baby out with the bath water.
--dick peskin


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 Richard L. Peskin     CAIP Parallel Computing Lab                      
 CAIP Center CN - 1390  Rutgers University       
 Piscataway, N. J. 08855-1390             
         net: peskin@caip.rutgers.edu              
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

 
-- 
goodby