Re: Intermediate code design

Charles Fiterman <cef@geodesic.com>
8 Jan 1998 00:31:57 -0500

          From comp.compilers
Related articles
INTERMEDIATE CODE DESIGN stephen.suryaputra@icon.singnet.com.sg (1995-06-24)
Intermediate code design david.stone@cambridge.simoco.com (David Stone 5887) (1998-01-05)
Re: Intermediate code design gclind01@spd.louisville.edu (1998-01-06)
Re: Intermediate code design chase@world.std.com (David Chase) (1998-01-06)
Re: Intermediate code design cef@geodesic.com (Charles Fiterman) (1998-01-08)
Re: Intermediate code design roques@scalar.pond.sub.org (Christian von Roques) (1998-01-08)
| List of all articles for this month |
From: Charles Fiterman <cef@geodesic.com>
Newsgroups: comp.compilers
Date: 8 Jan 1998 00:31:57 -0500
Organization: Geodesic Systems
References: 98-01-019 98-01-022
Keywords: design

George C. Lindauer wrote:
>
> David Stone (david.stone@cambridge.simoco.com) wrote:
> > (1) Updating operations.
>
> > Suppose the source language supports updating operations, for example
> > x += y in C.
>
> > Should this become
>
> > x = x + y


No. There is one absolute difference between x += y; and x = x + y;
In the former x is only evaluated once. Consider
  x[functionWithSideEffects()] += y;


Stepping outside C to the world of partial evaluation the big
optimizations are on the large scale not the small one. Suppose
x and y are objects and + is a method.


x + y is really clone(x) += y


x = x + y; is really


tmp is clone(x)
tmp += y
x = tmp
delete tmp


If there is notation for saying the code improvment
x += y
is legal that's a major win.
[Hey, in Fortran it is, and we all know that's the only language worth
optimizing. I felt way cool when I got i=i+1 down to one instruction on
my PDP-11 Fortran 77 compiler 20 years ago. -John]


--

Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.