Re: constant folding at parse time

drw@euclid.mit.edu (Dale R. Worley)
Mon, 24 Aug 1992 19:03:36 GMT

          From comp.compilers

Related articles
constant folding at parse time wjw@eb.ele.tue.nl (1992-08-17)
Re: constant folding at parse time markh@csd4.csd.uwm.edu (1992-08-17)
Re: constant folding at parse time twpierce@amhux1.amherst.EDU (Tim Pierce) (1992-08-19)
Re: constant folding at parse time scott@bbx.basis.com (1992-08-20)
Re: constant folding at parse time wjw@eb.ele.tue.nl (1992-08-21)
Re: constant folding at parse time buehlman@iwf.mabp.ethz.ch (1992-08-21)
Re: constant folding at parse time drw@euclid.mit.edu (1992-08-24)
Semantics Tools eifrig@beanworld.cs.jhu.edu (1992-08-25)
| List of all articles for this month |

Newsgroups: comp.compilers
From: drw@euclid.mit.edu (Dale R. Worley)
Organization: MIT Dept. of Tetrapilotomy, Cambridge, MA, USA
Date: Mon, 24 Aug 1992 19:03:36 GMT
Keywords: parse, optimize, comment
References: 92-08-114 92-08-132

wjw@eb.ele.tue.nl (Willem Jan Withagen) writes:
      Now I don't want to open another tarpit again, but this certainly is going
      to have effects on the semantics of the used program.


Yes, indeed. That's why you need an exacting specification of what the
language permits and forbids (as far as the visible consequences of
expression rearrangements go), and a thorough knowledge of the properties
of the arithmetic of the target computer.


Sadly, most language specifications don't discuss these issues, or discuss
them in such vague terms that it is very difficult to resolve ambiguous
cases.


I suppose that a really good compiler-generating system would be
parameterizable for all these matters, but I've never heard of such a
thing.


Dale Worley Dept. of Math., MIT drw@math.mit.edu
[Are there specification languages that would allow you do define your
language precisely enough to tell you mechanically whether a proposed
optimization was valid? -John]
--


Post a followup to this message

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