Scientific Computing

Kyle Pierce on SNI kylep at csn.net
Wed Sep 2 16:15:56 UTC 1998


-----Original Message-----
From: Stephen Travis Pope <stp at limbo.create.ucsb.edu>
To: squeak at cs.uiuc.edu <squeak at cs.uiuc.edu>
Date: Wednesday, September 02, 1998 9:55 AM
Subject: Re: Scientific Computing


>
>Let's not forget the Smalltalk-to-C translator in Squeak. In fact, this is
>the perfect medium between performance and malleability. High-level
>algorithms for any domain can be debugged in Smalltalk and then translated
>and batch compiled (and optimized) by your favorite C compiler. This is
>preferable to optimizing single floating-point math operations.
>
>There are examples of this in John Maloney's sound synthesis classes and in
>the FFT class. I'm working on a bunch more for Siren; it's great! (and it's
>something that Java just can't do!)


I had the impression that the Smalltalk-to-C translator in Squeak is limited
in terms of the breadth of possible Smalltalk functions that it can
translate.  That is, it does a great job of translating the essential Squeak
VM into C, but was never intended for much more than that.  I thought I
would have to write extensions in C in order to get that kind of performance
improvement in any kind of app-specific code.  Can anyone identify what the
practical limits are to this translation facility?  (I posted the same
question awhile back, but got no responses).

Thanks,

Kyle Pierce





More information about the Squeak-dev mailing list